APPENDIX A
PBPK Modeling of TCE and
Metabolites—Detailed Methods and Results

-------
CONTENTS—Appendix: PBPK Modeling of TCE and Metabolites—Detailed Methods
and Results
LIST OF TABLES	A-vi
LIST OF FIGURES	A-viii
A. 1. THE HIERARCHICAL BAYESIAN APPROACH TO CHARACTERIZING
PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK) MODEL
UNCERTAINTY AM) VARIABILITY	A-l
A. 1.1. EVALUATION OF THE HACK ET AL. (2006)
PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK)
MODEL	A-4
A.2. Convergence	A-4
A.2.1. Evaluation of Posterior Distributions for Population Parameters	A-6
A.2.2. Comparison of Model Predictions With Data	A-7
A.2.2.1. Mouse Model	A-9
A.2.2.2. Subject-specific and population-based predictions	A-9
A.2.2.3. Rat Model	A-16
A.2.2.4. Human Model	A-26
A 3. PRELIMINARY ANALYSIS OF MOUSE GAS UPTAKE DATA:
MOTIVATION FOR MODIFICATION OF RESPIRATORY
METABOLISM	A-36
A 3. DETAILS OF THE UPDATED PHYSIOLOGICALLY BASED
PHARMACOKINETIC (PBPK) MODEL FOR TRICHLOROETHYLENE
(TCE) AM) ITS METABOLITES	A-50
A.3.1. Physiologically Based Pharmacokinetic (PBPK) Model Structure and
Equations	A-50
A.3.1.1. Trichloroethylene (TCE) Sub-Model	A-65
A.3.1.2. Trichloroethanol (TCOH) Sub-Model	A-73
A.3.1.3. Trichloroethanol-Glucuronide Conjugate (TCOG) Sub-Model... A-75
A.3.1.4. Trichloroacetic Acid (TCA) Sub-Model	A-79
A.3.1.5. Glutathione (GSH) Conjugation Sub-Model	A-84
A.3.2. Model Parameters and Baseline Values	A-85
A.3.3. Statistical Distributions for Parameter Uncertainty and Variability	A-85
A.3.3.1. Initial Prior Uncertainty in Population Mean Parameters	A-85
A.3.3.2. Interspecies Scaling to Update Selected Prior Distributions in
the Rat and Human	A-85
A.3.3.3. Population Variance: Prior Central Estimates and Uncertainty.... A-98
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-iii DRAFT—DO NOT CITE OR QUOTE

-------
A.3.3.4. Likelihood Function and Prior distributions for Residual Error
Estimates	A-98
A.3.4. Summary of Bayesian Posterior Distribution Function	A-105
A.4. RESULTS OF UPDATED PHYSIOLOGICALLY BASED
PHARMACOKINETIC (PBPK) MODEL	A-108
A.4.1. Convergence and Posterior Distributions of Sampled Parameters	A-108
A.4.2. Comparison of Model Predictions with Data	A-148
A.4.2.1. Mouse Data and Model Predictions	A-149
A.4.2.2. Rat Data and Model Predictions	A-162
A.4.2.3. Human Data and Model Predictions	A-177
A.4. EVALUATION OF RECENTLY PUBLISHED TOXICOKINETIC DATA	A-211
A.4.3. Trichloroethylene (TCE) Metabolite Toxicokinetics in Mice: Kim et
al. (2009)	A-212
A.4.4. Trichloroethylene (TCE) Toxicokinetics in Rats: Liu et al. (2009)	A-217
A.4.5. Trichloroacetic Acid (TCA) Toxicokinetics in Mice and Rats: Mahle
et al. (2001) and Green (2003a, 2003b)	A-217
A.4.5.1. Analysis Using Evans et al. (2009) and Chiu et al. (2009)
Physiologically Based Pharmacokinetic (PBPK) Model	A-217
A.4.5.2. Summary of Results From Chiu of Bayesian Updating of
Evans et al. (2009) and Chiu et al. (2009) Model Using
Trichloroacetic Acid (TCA) Drinking Water Data	A-221
A. 5. UPDATED PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK)
MODEL CODE	A-229
A.6. REFERENCES	A-254
5/12/11
This document is a draft for review purposes only and does not constitute Agency policy.
A-iv DRAFT—DO NOT CITE OR QUOTE

-------
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-v DRAFT—DO NOT CITE OR QUOTE

-------
LIST OF TABLES
Table A-l. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in mice A-
11
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats .. A-
18
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans
	A-27
Table A-4. PBPK model parameters, baseline values, and scaling relationships	A-51
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters . A-
89
Table A-6. Updated prior distributions for selected parameters in the rat and human	A-95
Table A-7. Uncertainty distributions for the population variance of the PBPK model parameters
	A-100
Table A-8. Measurements used for calibration	A-104
• Table A-9. Posterior distributions for mouse PBPK model population parameters	A-l 10
Table A-10. Posterior distributions for mouse residual errors	A-l 12
Table A-l 1. Posterior correlations for mouse population mean parameters	A-l 13
Table A-12. Posterior distributions for rat PBPK model population parameters	A-120
Table A-13. Posterior distributions for rat residual errors	A-123
Table A-14. Posterior correlations for rat population mean parameters	A-127
Table A-15. Posterior distributions for human PBPK model population parameters	A-134
Table A-16. Posterior distributions for human residual errors	A-138
Table A-17. Posterior correlations for human population mean parameters	A-140
Table A-18. Summary characteristics of model runs	A-224
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-vi DRAFT—DO NOT CITE OR QUOTE

-------
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-vii DRAFT—DO NOT CITE OR QUOTE

-------
LIST OF FIGURES
Figure A-l. Hierarchical population statistical model for PBPK model parameter uncertainty
and variability (see Gelman et al., 1996). Square nodes denote fixed or observed quantities;
circle notes represent uncertain or unobserved quantities, and the nonlinear model outputs are
denoted by the inverted triangle. Solid arrows denote a stochastic relationship represented by a
conditional distribution [A^B means B ~ P{B\A)\ while dashed arrows represent a function
relationship [B =f(A)~\. The population consists of subjects each of which undergoes one or
more experiments j with exposure parameters Ey with data y,:lki collected at times where k
denotes different types of outputs and / denotes the different time points. The PBPK model
produces outputs /^/ for comparison with the datay^. The difference between them
("measurement error") has variance a & with a fixed prior distribution Pr, which in this case is
the same for the entire population. The PBPK model also depends on measured covariates fa
(e.g., body weight) and unobserved model parameters 9, (e.g., Vmax)- The parameters 9, are
drawn from a population with mean |j, and variance Z , each of which is uncertain and has a prior
distribution assigned to it	A-2
Figure A-2. Schematic of how posterior predictions were generated for comparison with
experimental data. Two sets of posterior predictions were generated: population predictions
(diagonal hashing) and subject-specific predictions (vertical hashing)	A-8
Figure A-3. Limited optimization results for male closed-chamber data from Fisher et al. (1991)
without (top) and with (bottom) respiratory metabolism	A-45
Figure A-4. Limited optimization results for female closed-chamber data from Fisher et al.
(1991) without (top) and with (bottom) respiratory metabolism	A-46
Figure A-5. Respiratory metabolism model for updated PBPK model	A-49
Figure A-6. Sub-model for TCE gas exchange, respiratory metabolism, and arterial blood
concentration	A-67
Figure A-7 Sub-model for TCE oral absorption, tissue distribution, and metabolism	A-69
Figure A-8. Submodel for TCOH	A-76
Figure A-9. Submodel for TCOG	A-78
Figure A-10. Submodel for TCA	A-79
Figure A-l 1. Submodel for TCE GSH conjugation metabolites	A-84
Figure A-12. Updated hierarchical structure for rat and human models. Symbols have the same
meaning as Figure A-l, with modifications for the rat and human. In particular, in the rat, each
"subject" consists of animals (usually comprising multiple dose groups) of the same sex, species,
and strain within a study (possibly reported in more than one publication, but reasonably
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-viii DRAFT—DO NOT CITE OR QUOTE

-------
presumed to be of animals in the same "lot"). Animals within each subject are presumed to be
"identical," with the same PBPK model parameters, and each such subject is assigned its own set
of "residual" error variances o In humans, each "subject" is a single person, possibly exposed
in multiple experiments, and each subject is assigned a set of PBPK model parameters drawn
from the population. However, in humans, "residual" error variances are assigned at an
intermediate level of the hierarchy—the "study" level, o km—rather than the subject or the
population level	A-107
Figure A-13. Prior and posterior mouse population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 15
Figure A-14. Prior and posterior mouse population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 16
Figure A-l5. Prior and posterior mouse population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 16
Figure A-16. Prior and posterior mouse population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 17
Figure A-17. Prior and posterior mouse population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 18
Figure A-18. Prior and posterior mouse population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 19
Figure A-19. Prior and posterior rat population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions,, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-128
Figure A-20. Prior and posterior rat population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-129
Figure A-21. Prior and posterior rat population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-130
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-ix DRAFT—DO NOT CITE OR QUOTE

-------
Figure A-22. Prior and posterior rat population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-131
Figure A-23. Prior and posterior rat population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-132
Figure A-24. Prior and posterior rat population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-133
Figure A-25. Prior and posterior human population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-142
Figure A-26. Prior and posterior human population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-143
Figure A-27. Prior and posterior human population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-144
Figure A-28. Prior and posterior human population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-145
Figure A-29. Prior and posterior human population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-146
Figure A-30. Prior and posterior human population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-147
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK model predictions (red
line: using the posterior mean of the subject-specific parameters; + with error bars: single data
points; or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-149
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model predictions (red line:
using the posterior mean of the subject-specific parameters; + with error bars: single data points;
or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-162
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-x DRAFT—DO NOT CITE OR QUOTE

-------
Figure A-33. Comparison of rat evaluation data (boxes) and PBPK model predictions (+ with
error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-174
Figure A-34. Comparison of human calibration data (boxes) and PBPK model predictions (red
line: using the posterior mean of the subject-specific parameters; + with error bars: single data
points; or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-178
Figure A-35. Comparison of human evaluation data (boxes) and PBPK model predictions (+
with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-203
Figure A-36. Comparison of Kim et al. (2009) mouse data (boxes) and PBPK model predictions
(+ with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and
97.5% population-based predictions)	A-212
Figure A-37. Comparison of best-fitting (out of 50,000 posterior samples) PBPK model
prediction and Kim et al. (2009) TCA blood concentration data for mice gavaged with
2,140 mg/kg TCE	A-213
Figure A-38. Comparison of best-fitting (out of 50,000 posterior samples) PBPK model
prediction and Kim et al. (2009) DCVG blood concentration data for mice gavaged with
2,140 mg/kg TCE	A-214
Figure A-39. PBPK model predictions for the fraction of intake undergoing GSH conjugation in
mice continuously exposed orally to TCE. Lines and error bars represent the median and 95th
percentile confidence interval for the posterior predictions, respectively (also reported in
Section 3.5.7.2.1). Filled circles represent the predictions from the sample (out of 50,000 total
posterior samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG blood
concentration data for mice gavaged with 2,140 mg/kg TCE	A-215
Figure A-40. PBPK model predictions for the fraction of intake undergoing GSH conjugation in
mice continuously exposed via inhalation to TCE. Lines and error bars represent the median and
95th percentile confidence interval for the posterior predictions, respectively (also reported in
Section 3.5.7.2.1). Filled circles represent the predictions from the sample (out of 50,000 total
posterior samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG blood
concentration data for mice gavaged with 2,140 mg/kg TCE	A-216
Figure A-41. Comparison of Liu et al. (2009) rat data (boxes) and PBPK model predictions (+
with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-219
Figure A-42. Assumed drinking water patterns as a function of time since beginning of
exposure. The upper left panel (LH) assumes that t = 0 is at the beginning of the "light" part of
This document is a draft for review purposes only and does not constitute Agency policy.
5/12/11	A-xi DRAFT—DO NOT CITE OR QUOTE

-------
the "light/dark" cycle (light is dashed grey line at the bottom, dark is thick black line at the
bottom). The upper right panel (LHL) assumes that t = 0 is in the middle of the "light" part of
the cycle. The lower left panel (HL) assumes that t = 0 is at the end of the "light" part of the
Figure A-43. PBPK model predictions for TCA in blood and liver of male B6C3Fi mice from
Mahle et al. (2001). Three- and 14-day exposures to 0.08 (data: open circles, predictions: solid
line), 0.8 (data: open triangle, predictions: dashed line), and 2 g/L TCA in drinking water (data:
crosses, predictions: dotted line). Predictions use a representative parameter sample from the
Figure A-44. PBPK model predictions for TCA in blood and liver of male B6C3F1 mice from
Green (2003a, 2003b). Green (2003a): 5-day drinking water exposures to 0.5 (data: open circle;
predictions: solid line), 1 (data: open triangle; predictions: dashed line), and 2.5 g/L TCA (data:
crosses; predictions: dotted lines). Green (2003b): 5- and 14-day drinking water exposures to 1
(data: open circle; predictions: solid line) and 2.5 g/L TCA (data: open triangle; predictions:
dashed line). Predictions use a representative parameter sample from the converged MCMC
chain for the LHL drinking water intake pattern	A-226
Figure A-45. Distribution of fractional absorption fit to each TCA drinking water kinetic study
group in mice, using LHL drinking water intake patterns. Fits are to a Michaelis-Menten
function for "effective" concentration Ceff = Cmax x C/(Cy2 + C), so that the fractional absorption
Fabs = Ceff/C = Cmax/(Ci/2 + C). Sweeney et al. (2009) estimates of Fabs, along with a Michaelis-
Menten fit, are included for comparison. The ratio Cmax/Cu gives the fractional uptake at low
concentrations	A-228
cycle
A-223
converged MCMC chain for the LHL drinking water intake pattern.
A-225
5/12/11
This document is a draft for review purposes only and does not constitute Agency policy.
A-xii DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
A.l. THE HIERARCHICAL B A YE SI AN APPROACH TO CHARACTERIZING
PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK) MODEL
UNCERTAINTY AND VARIABILITY
The Bayesian approach for characterizing uncertainty and variability in physiologically
based pharmacokinetic (PBPK) model parameters, used previously for trichloroethylene (TCE)
in Bois (2000a, 2000b) and Hack et al. (2006), is briefly described here as background. Once a
PBPK model structure is specified, characterizing the model reduces to calibrating and making
inferences about model parameters. The use of least-squares point estimators is limited by the
large number of parameters and small amounts of data. The use of least-squares estimation is
reported after imposing constraints for several parameters (Clewell, Gentry, Covington, &
Gearhart, 2000; J. Fisher, 2000). This is reasonable for a first estimate, but it is important to
follow-up with a more refined treatment. This is implemented by a Bayesian approach to
estimate posterior distributions on the unknown parameters, a natural choice, and almost a
compulsory consequence given the large number of parameters and relatively small amount of
data, and given the difficulties of frequentist estimation in this setting.
As described by Gelman et al. (1996), the Bayesian approach to population PBPK
modeling involves setting up the overall model in several stages. A nonlinear PBPK model, with
predictions denoted/ describes the absorption, distribution, metabolism, and excretion of a
compound and its metabolites in the body. This model depends on several, usually known,
parameters such as measurement times t, exposure E, and measured covariates (p. Additionally,
each subject i in a population has a set of unmeasured parameters 9,. A random effects model
2	2
describes their population variability /J(0, | (j,, E ), and a prior distribution E ) on the
population mean [j, and covariance E (often assumed to be diagonal) incorporates existing
scientific knowledge about them. Finally, a "measurement error" model P(y |_/[9, (p, l\ /], o )
describes deviations (with variance o ) between the dataj' and model predictions/(which of
course depends on the unmeasured parameters 0, and the measured parameters i, E, and (p). This
"measurement error" level of the hierarchical model typically also encompasses intrasubject
variability as well as model misspecification, but for notational convenience we refer to it here as
"measurement error." Because these other sources of variance are lumped into a single
"measurement error," a prior distribution of its variance o must be specified even if the actual
analytic measurement error is known. All these components are illustrated graphically in
Figure A-l.
This document is a draft for review purposes only and does not constitute Agency policy.
A-1 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Experiment /
Subject i
Population
Prc
Figure A-l. Hierarchical population statistical model for PBPK model
parameter uncertainty and variability (see Gelman et al., 1996). Square nodes
denote fixed or observed quantities; circle notes represent uncertain or unobserved
quantities, and the nonlinear model outputs are denoted by the inverted triangle.
Solid arrows denote a stochastic relationship represented by a conditional
distribution [A—>B means B ~ P{B\A)\ while dashed arrows represent a function
relationship [B =j{A)~\. The population consists of subjects each of which
undergoes one or more experiments j with exposure parameters Ey with data
collected at times where k denotes different types of outputs and / denotes the
different time points. The PBPK model produces outputs /,/;/ for comparison with
the data^yi-/. The difference between them ("measurement error") has variance
a k, with a fixed prior distribution Pr, which in this case is the same for the entire
population. The PBPK model also depends on measured covariates (j), (e.g., body
weight) and unobserved model parameters 9, (e.g., Vmax)- The parameters 9, are
drawn from a population with mean |j, and variance Z , each of which is uncertain
and has a prior distribution assigned to it.
This document is a draft for review purposes only and does not constitute Agency policy.
A-2 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
The posterior distribution for the unknown parameters is obtained in the usual manner by
multiplying (1) the prior distribution for the population mean and variance and the
2 2
"measurement" error E ) P(c ), (2) the population distribution for the subject parameters
2	2
P(Q (i, S ), and (3) the likelihood P(y | 0, o ), where for notational convenience, the dependence
on/ cp, E, and t (which are taken as fixed for a given data set) is dropped:
P(Q, |i, S2, o2 | y)cc P(^ I2) P(c2) P(Q | |i, I2) P(y | 9, g2)	(Eq. A-l)
Here, each subject's parameters 0, have the same sampling distribution (i.e., they are
independently and identically distributed), so their joint prior distribution is
P(Q | |i, E2) = P(®i I H, S2)	(Eq. A-2)
Different experiments j = 1 ...n, may have different exposure and different data collected and
different time points. In addition, different types of measurements k = (e.g., TCE blood,
TCE breath, trichloroacetic acid [TCA] blood, etc.) may have different errors, but errors are
otherwise assumed to be iid. Since the subjects are treated as independent given the total
likelihood function is simply
P(y | 9, o2) - ri/= i...« 11 i .. Wk l...m Yh=\..MjkP{yijki | 0i, Ok2, tijki)	(Eq. A-3)
where n is the number of subjects, riy is the number of experiments in that subject, m is the
number of different types of measurements, Nyk is the number (possibly 0) of measurements
(e.g., time points) for subject i of type k in experiment /, and 1,,/d are the times at which
measurements for subject / of type & were made in experiment j.
Given the large number of parameters, complex likelihood functions, and nonlinear
PBPK model, Markov chain Monte Carlo (MCMC) simulation was used to generate samples
from the posterior distribution. An important practical advantage of MCMC sampling is the
ability to implement inference in nearly any probability model and the possibility to report
inference on any event of interest. MCMC simulation was introduced by Gelfand and Smith
(1990) as a generic tool for posterior inference. See Gilks et al. (1996) for a review. In addition,
because many parameters are allowed to vary simultaneously, the local parameter sensitivity
analyses often performed with PBPK models (in which the changes in model predictions are
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
assessed with each parameter varied by a small amount) are unnecessary. 1 In the context of
PBPK models, the MCMC simulation can be carried out as described by Hack et al. (2006). The
simulation program MCSim (version 5.0.0) was used to implement MCMC posterior simulation,
with analysis of the results performed using the R statistical package. Simulation-based
parameter estimation with MCMC posterior simulation gives rise to an additional source of
uncertainty. For instance, averages computed from the MCMC simulation output represent the
desired posterior means only asymptotically, in the limit as the number of iterations goes to
infinity. Any implementation needs to include a convergence diagnostic to judge practical
convergence. The potential scale-reduction-factor convergence diagnostic R of Gelman et al.
(1996) was used here, as it was in Hack et al. (2006).
All EVALUATION OF THE HACK ET AL. (2006) PHYSIOLOGICALLY BASED
PHARMACOKINETIC (PBPK) MODEL
U.S. Environmental Protection Agency (EPA) obtained the original model code for the
version of the TCE PBPK model published in Hack et al. (2006) and conducted a detailed
evaluation of the model, focusing on the following areas: convergence, posterior estimates for
model parameters, and comparison of model predictions with in vivo data.
A.2. Convergence
As noted in Hack et al. (2006), the diagnostics for the MCMC simulations (3 chains of
length 20,000-25,000 for each species) indicated that additional samples might further improve
convergence. A recent analysis of tetrachloroethylene pharmacokinetics indicated the need to be
especially careful in ensuring convergence (W. A. Chiu & Bois, 2007). Therefore, the number of
MCMC samples per chain was increased to 75,000 for rats (first 25,000 discarded) and 175,000
for mice and humans (first 75,000 discarded). Using these chain lengths, the vast majority of the
parameters had potential scale reduction factors R < 1.01, and all population parameters had
R < 1.05, indicating that longer chains would be expected to reduce the standard deviation (or
other measure of scale, such as a confidence interval) of the posterior distribution by less than
this factor (Gelman, Carlin, Stern, & Rubin, 2004).
1 In particular, local sensitivity analyses are typically used to assess the impact of alternative parameter estimates on
model predictions, inform experimental design, or assist prioritizing risk assessment research. Only the first purpose
is relevant here; however, the full uncertainty and variability analysis allows for a more comprehensive assessment
than can be done with sensitivity analyses. Separately, such analyses could be done to design experiments and
prioritize research that would be most likely to help reduce the remaining uncertainties in TCE toxicokinetics, but
that is beyond the scope of this assessment.
This document is a draft for review purposes only and does not constitute Agency policy.
A-4 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
In addition, analysis of autocorrelation within chains using the R-CODA package
(Plumber et al., 2008) indicated that there was significant serial correlation, so additional
"thinning" of the chains was performed in order to reduce serial correlations. In particular, for
rats, for each of three chains, every 100th sample from the last 50,000 samples was used; and for
mice and humans, for each of three chains, every 200^ sample from the last 100,000 samples
was used. This thinning resulted in a total of 1,500 samples for each species available for use for
posterior inference.
Finally, an evaluation was made of the "convergence" of dose metric predictions—that is,
the extent to which the standard deviation or confidence intervals for these predictions would be
reduced with additional samples. This is analogous to a "sensitivity analysis" performed so that
most effort is spent on parameters that are most influential in the result. In this case, the purpose
is to evaluate whether one can sample chains only long enough to ensure convergence of
predictions of interest, even if certain more poorly identified parameters take longer chains to
converge. The motivation for this analysis is that for a more complex model, running chains
until all parameters have R< 1.01 or 1.05 may be infeasible given the available time and
resource. In addition, as some of the model parameters had prior distributions derived from
"visual fitting" to the same data, replacing those distributions with less informative distributions
(in order to reduce bias from "using the same data twice") may require even longer chains for
convergence.
Indeed, it was found that R-values for dose metric predictions approached one more
quickly than PBPK model input parameters. The most informative simulations were for mice,
which converged the slowest and, thus, had the most potential for convergence-related error.
Results for rats could not be assessed because the model converged so rapidly, and results for
humans were similar to those in mice, though the deviations were all less because of the more
rapid convergence. In the mouse model, after 25,000 iterations, many PBPK model parameters
had /^-values >2, with more than 25% greater than 1.2. However, all dose metric predictions had
R < 1.4, with the more than 96% of then <1.2 and the majority of them <1.01. In addition, when
compared to the results of the last 100,000 iterations (after the total of 175,000 iterations), more
than 90% of the medians estimates shifted by less than 20%, with the largest shifts less than 40%
(for glutathione [GSH] metabolism dose metrics, which had no relevant calibration data). Tail
quantiles had somewhat larger shifts, which was expected given the limited number of samples
in the tail, but still more than 90% of the 2.5 and 97.5 percentile quantiles had shifts of less than
40%). Again, the largest shifts, on order of 2-fold, were for GSH-related dose metrics that had
high uncertainty, so the relative impact of limited sample size is small.
This document is a draft for review purposes only and does not constitute Agency policy.
A-5 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Therefore, the additional simulations performed in this evaluation, with three- to
sevenfold longer chains, did not result in much change in risk assessment predictions from the
original Hack et al. (2006) results. Thus, assessing prediction convergence appears sufficient for
assessing convergence of the TCE PBPK model for the purposes of risk assessment prediction.
A.2.1. Evaluation of Posterior Distributions for Population Parameters
Posterior distributions for the population parameters were first checked for whether they
appeared reasonable given the prior distributions. Inconsistency between the prior and posterior
distributions may indicate an insufficiently broad prior distribution (i.e., overconfidence in their
specification), a mis-specification of the model structure, or an error in the data. Parameters that
were flagged for further investigation were those for which the interquartile ranges (intervals
bounded by the 25th and 75th percentiles) of the prior and posterior distributions did not overlap.
In addition, lumped metabolism and clearance parameters for TCA, trichloroethanol (TCOH),
and trichloroethanol-glucuronide conjugate (TCOG) were checked to make sure that they
remained physiological—e.g., metabolic clearance was not more than hepatic blood flow and
urinary clearance not more than kidney blood flow (constraints that were not present in the Hack
et al. (2006) priors)..
In mice, population mean parameters that had lack of overlap between priors and
posteriors included the affinity of oxidative metabolism (lnKfvi), the TCA plasma-blood
concentration ratio (TCAPlas), the TCE stomach to duodenum transfer coefficient (InKTSD),
and the urinary excretion rates of TCA and TCOG (InkUrnTCAC and InkUrnTCOGC). For KM,
this is not unexpected, as previous investigators have noted inconsistency in the KM values
between in vitro values (upon which the prior distribution was based) and in vivo values derived
from oral and inhalation exposures in mice (Abbas & Fisher, 1997; Greenberg, Burton, & Fisher,
1999). For the other mean parameters, the central estimates were based on visual fits, without
any other a priori data, so it is reasonable to assume that the inconsistency is due to insufficiently
broad prior distributions. In addition, the population variance for the TCE absorption coefficient
from the duodenum (kAD) was rather large compared to the prior distribution, likely due to the
fact that oral studies included TCE in both oil and aqueous solutions, which are known to have
very different absorption properties. Thus, the larger population variance was required to
accommodate both of them. Finally, the estimated clearance rate for glucurondiation of TCOH
was substantially greater than hepatic blood flow. This is an artifact of the one-compartment
model used for TCOH and TCOG, and suggests that first pass effects are important for TCOH
glucurondiation. Therefore, the model would benefit from the additional of a separate liver
This document is a draft for review purposes only and does not constitute Agency policy.
A-6 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
compartment so that first pass effects can be accounted for, particularly when comparing across
dose-routes.
In rats, the only population mean or variance parameter for which the posterior
distribution was somewhat inconsistent with the prior distribution was the population mean for
the InKM. While the interquartile regions did not overlap, the 95th percentile regions did, so the
discordance was relatively minor. However, as with mice, the estimated clearance rate for
glucurondiation of TCOH was substantially greater than hepatic blood flow.
In humans, some of the chemical-specific parameters for which priors were established
using visual fits had posterior distributions that were somewhat inconsistent, including the
oxidative split between TCA and TCOH, biliary excretion of TCOG (InkBileC), and the TCOH
distribution volume (VBodC). More concerning was the fact that the posterior distributions for
several physiological volumes and flows were rather strongly discordant with the priors and/or
near their truncation limits, including gut, liver, and slowly perfused blood flow, the volumes of
the liver and rapidly perfused compartments. In addition, a number of tissue partition
coefficients were somewhat inconsistent with their priors, including those for TCE in the gut,
rapidly perfused, and slowly perfused tissues, and TCA in the body and liver. Finally, a number
of population variances (for TCOH clearance [InClTCOHC], urinary excretion of TCOG
[InkUrnTCOGC], ventilation-perfusion ratio [InVPRC], cardiac output [InQCC], fat blood flow
and volume [QFatC and VFatC], and TCE blood-air partition coefficient [PBC]) were somewhat
high compared to their prior distributions, indicating much greater population variability than
expected.
A.2.2. Comparison of Model Predictions With Data
A schematic of the comparisons between model predictions and data are shown in
Figure A-2. In the hierarchical population model, subject-specific parameters were estimated for
each data set used in calibrating the model (posterior subject-specific 0, in Figure A-2). Because
these parameters are in a sense "optimized" to the experimental data themselves, the
subject-specific predictions (posterior subject-specific^ in Figure A-2) using these parameters
should be accurate by design. Poor fits to the data using these subject-parameters may indicate a
misspecification of the model structure, prior parameter distributions, or an error in the data. In
addition, it is useful to generate "population-based" parameters (posterior population 9) using
only the posterior distributions for the population means ([j.) and variances (E ), instead of the
estimated subject-specific parameters. These population predictions provide a sense as to
This document is a draft for review purposes only and does not constitute Agency policy.
A-7 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
MCMC outputs
Posterior
Posterior population
Posterior I2
Posterior population
prediction/\
PBPK
model.
Posterior subject-
specific A
Posterior group-specific
prediction A
Experiment j
Group/
Individual i
Figure A-2. Schematic of how posterior predictions were generated for
comparison with experimental data. Two sets of posterior predictions were
generated: population predictions (diagonal hashing) and subject-specific
predictions (vertical hashing).
whether the model and the predicted degree of population uncertainty and variability adequately
account for the range of heterogeneity in the experimental data. Furthermore, assuming the
subject-specific predictions are accurate, the population-based predictions are useful to identify
whether one or more if the data sets are "outliers" with respect to the predicted population. In
addition, a substantial number of in vivo data sets was available in all three species that were not
previously used for calibration. Thus, it is informative to compare the population-based model
predictions, discussed above, to these additional "validation" data in order to assess the
predictive power of the PBPK model.
This document is a draft for review purposes only and does not constitute Agency policy.
A-8 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
A.2.2.1. Mouse Model
A.2.2.2. Subject-specific and population-based predictions
Initially, the sampled subject-specific parameters were used to generate predictions for
comparison to the calibration data. Because these parameters were "optimized" for each subject,
these "subject-specific" predictions should be accurate by design. However, unlike for the rat
(see below), this was not the case for some experiments (this is partially responsible for the
slower convergence). In particular, the predictions for TCE and TCOH concentrations for the
Abbas and Fisher (1997) data were poor. In addition, TCE blood concentrations for the
Greenberg et al. (1999) data were consistently overpredicted. These data are discussed further in
Table A-1.
Next, only samples of the population parameters (means and variances) were used, and
"new subjects" were sampled from appropriate distributions using these population means and
variances. These "new subjects" then represent the predicted population distribution,
incorporating both variability in the population as well as uncertainty in the population means
and variances. These "population-based" predictions were then compared to both the data used
in calibration, as well as the additional data identified that was not used in calibration. The
PBPK model was modified to accommodate some of the different outputs (e.g., tissue
concentrations) and exposure routes (TCE, TCA, and TCOH intravenous [i.v.]) used in the
"noncalibration" data, but otherwise it is unchanged.
A.2.2.2.1.1. Subject-specific predictions and calibration data
(See "Supplementary data for TCE assessment: Hack mouse subject calibration," 2011)
A.2.2.2.1.2. Population-based predictions and calibration and additional evaluation data
(See "Supplementary data for TCE assessment: Hack mouse population calibration evaluation,"
2011)
This document is a draft for review purposes only and does not constitute Agency policy.
A-9 DRAFT—DO NOT CITE OR QUOTE

-------
A.2.2.2.2. Conclusions regarding mouse model
A.2.2.2.2.1. Trichloroethylene (TCE) concentrations in blood and tissues not well-
predicted
1	The PBPK model for the parent compound does not appear to be robust. Even
2	subject-specific fits to data sets used for calibration were not always accurate. For oral dosing
3	data, there is clearly high variability in oral uptake parameters, and the addition of uptake
4	through the first (stomach) compartment should improve the fit. Unfortunately, inaccurate TCE
5	uptake parameters may lead to inaccurately estimated kinetic parameters for metabolites TCA
6	and TCOH, even if current fits are adequate.
This document is a draft for review purposes only and does not constitute Agency policy.
A-10 DRAFT—DO NOT CITE OR QUOTE

-------
1
Table A-l. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in mice
S?
>!
cs
s;
>1
£
>
a
&
H
1
o
o
2
o
H
"T3
S
O
>!
>1
O
a
a
§•
>!
o
Cb
O
5
>!
, ^
I &Q
oS
H
ffl.
o
p
o
c
o
H
W
Reference
Simulation #
Calibration
data
Discussion
Abbas et al. (1997)
41-42

These data are only published as an abstract. They consist of TCA and TCOH blood and urine data from
TCA and TCOH i.v. dosing. Blood levels of TCA and TCOH are fairly accurately predicted. From
TCOH dosing, urinary TCOG excretion is substantially overpredicted, and from TCA dosing, urinary
TCA excretion is substantially overpredicted.
Abbas and Fisher
(1997)
3-6
a/
Results for these data were mixed. TCA levels were the best fit. The calibration data included TCA blood
and liver data, which were well predicted except at the earliest time-point. In addition, TCA
concentrations in the kidney were fairly consistent with the surrogate TCA body concentrations predicted
by the model. Urinary TCA was well predicted at the lower two and highest doses, but somewhat
underpredicted (though still in the 95% confidence region) at 1,200 mg/kg.
TCE levels were in general not well fit. Calibration data included blood, fat, and liver concentrations,
which were predicted poorly particularly at early and late times. One reason for this is probably the
representation of oral uptake. Although both the current model and the original Abbas and Fisher (1997)
model had two-compartments representing oral absorption, in the current model uptake can only occur
from the second compartment. By contrast, the Abbas and Fisher (1997) model had uptake from both
compartments, with the majority occurring from the first compartment. Thus, the explanation for the poor
fit, particularly of blood and liver concentrations, at early times is probably simply due to differences in
modeling oral uptake. This is also supported by the fact that the oral uptake parameters tended to be
among those that took the longest to converge.
Subject-specific blood TCOH predictions were poor, with under-prediction at early times and
overprediction at late times. Population-based blood TCOH predictions tended to be underpredicted,
though generally within the 95% confidence region. Subject-specific urinary TCOG predictions were
fairly accurate except at the highest dose. These predictions are also probably affected by the apparent
misrepresentation of oral uptake. In addition, a problem as found in the calibration data in that data on
free TCOH was calibrated against predictions of total TCOH (TCOH+TCOG).
A number of TCOH and TCOG measurements were not included in the calibration—among them
tissue concentrations of TCOH and tissue and blood concentrations of TCOG. Blood concentrations (the
only available surrogate) were poor predictors of tissue concentrations of TCOH and TCOG (model
generally under-predicted). For TCOG, this may be due in part to the model assumption that the
distribution volume of TCOG is equal to that of TCOH.

-------
Table A-l. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in mice (continued)
S?
>!
cs
s;
>1
£
>
to
o
H
1
o
o
2
o
H
"T3
S
O
>!
>1
O
a
a
§•
>!
o
Cb
O
5
>!
, ^
I &Q
oS
H
ffl.
o
o
c
o
H
W
Reference
Simulation #
Calibration
data
Discussion
Fisher et al. (1991)
1-2
(open-
chamber)
a/
Venous blood TCE concentrations were somewhat underpredicted (a common issue with inhalation
exposures in mice below) (Greenberg et al., 1999), but within the 95% confidence region of both
subject-specific and population-based predictions. Plasma TCA levels were well predicted, with most of
the data near the interquartile region of both subject-specific and population-based predictions (but with
substantial scatter in the male mice). However, it should be noted that only a single exposure
concentration for each sex was used in calibration, with six additional exposures (three for each sex) not
included (see simulations 21-26, below).

7-16 (closed-
chamber)
a/
Good posterior fits were obtained for these data—closed-chamber data with initial concentrations from
300 - 10,000 ppm. Some variability in Vmax, however, was noted in the posterior distributions for that
parameter. Using subject-specific Vmax values resulted in better fits to these data. However, there
appears to be a systematic trend of lower estimated apparent Vmax at higher exposures. Similarly,
posterior estimates of cardiac output and the ventilation-perfusion ratio declined (slightly) with higher
exposures. These could be related to documented physiological changes (e.g., reduced ventilation rate and
body temperature) in mice when exposed to some volatile organics.

21-26 (open-
chamber,
additional
exposures)

Data from three additional exposures for each sex were available for comparison to model predictions.
Plasma TCA levels were generally well predicted, though the predictions for female mice data showed
some systematic over-prediction, particularly at late times (i.e., data showed shorter apparent half-life).
Blood TCE concentrations were consistently overpredicted, sometimes by almost an order of magnitude,
except in the case of female mice at 236 ppm, for which predictions were fairly accurate.
Fisher and Allen
(1993)
31-36

Predictions for these gavage data were generally fairly accurate. There was a slight tendency to
overpredict TCA plasma concentrations, with predictions tending to be worse in the female mice. Blood
levels of TCE were adequately predicted, though there was some systematic underprediction at 2-6 h after
dosing.
Green and Prout
(1985)
40

This datum consists of a single measurement of urinary excretion of TCA at 24 h as a fraction of dose,
from TCA i.v. dosing. The model substantially over-predicts the amount excreted. Whereas Green and
Prout (1985) measured 35% excreted at 24 h, the model predicts virtually complete excretion at 24 h.

-------
Table A-l. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in mice (continued)
S?
>!
cs
s;
>1
£
>
a
&
H
1
o
o
2
o
H
"T3
S
O
>!
>1
O
a
a
§•
>!
o
Cb
O
5
>!
, ^
I &Q
oS
H
ffl.
o
p
o
c
o
H
W
Reference
Simulation #
Calibration
data
Discussion
Greenberg et al.
(1999)
17-18
a/
The calibration data included blood TCE, TCOH, and TCA data. Fits to blood TCA and TCOH were
adequate, but as with the Fisher et al. (1991) inhalation data, TCE levels were overpredicted (outside the
95% confidence region during and shortly after exposure).
As with Abbas and Fisher (1997), there were additional data in the study that was not used in
calibration, including blood levels of TCOG and tissue levels of TCE, TCA, TCOH, and TCOG. Tissue
levels of TCE were somewhat overpredicted, but generally within the 95% confidence region. TCA levels
were adequately predicted, and mostly in or near the interquartile region. TCOH levels were somewhat
underpredicted, though within the 95% confidence region. TCOG levels, for which blood served as a
surrogate for all tissues, were well predicted in blood and the lung, generally within the interquartile
region. However, blood TCOG predictions underpredicted liver and kidney concentrations.
Larson and Bull
(1992)
37-39

Blood TCA predictions were fairly accurate for these data. However, TCE and TCOH blood
concentrations were underpredicted by up to an order of magnitude (outside the 95% confidence region).
Part of this may be due to uncertain oral dosing parameters. Urinary TCA and TCOG were also generally
underpredicted, in some cases outside of the 95% confidence region.
Prout et al. (1985)
19
a/
Fits to these data were generally adequate—within or near the interquartile region.

27-30 (urinary
excretion at
different
doses)

These data consisted of mass balance studies of the amount excreted in urine and exhaled unchanged at
doses from 10-2,000 mg/kg. TCA excretion was consistently overpredicted, except at the highest dose.
TCOG excretion was generally well predicted—within the interquartile range. The amount exhaled was
somewhat overpredicted, with a fourfold difference (but still within 95% confidence) at the highest dose.
Templinetal. (1993)
20
V
Blood TCA levels from these data were well predicted by the model. Blood TCE and TCOH levels were
well predicted using subject-specific parameters, but did not appear representative using
population-derived parameters. However, this is probably a result of the subject-specific oral absorption
parameter, which was substantially different than the population mean.

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
The TCE data from inhalation experiments also are not well estimated, particularly blood
levels of TCE. While fractional uptake has been hypothesized, direct evidence for this is
lacking. In addition, physiologic responses to TCE vapors (reduced ventilation rates, lowered
body temperature) are a possibility. These are weakly supported by the closed-chamber data, but
the amount of the changes is not sufficient to account for the low blood levels of TCE observed
in the open-chamber experiments. It is also not clear what role presystemic elimination due to
local metabolism in the lung may play. It is known that the mouse lung has a high capacity to
metabolize TCE (T. Green, Mainwaring, & Foster, 1997). However, in the Hack et al. (2006)
model, lung metabolism is limited by flow to the tracheobronchial region. An alternative
formulation for lung metabolism in which TCE is available for metabolism directly from inhaled
air (similar to that used for styrene) (Sarangapani, Gentry, Covington, Teeguarden, & 3rd, 2003),
may allow for greater presystemic elimination of TCE, as well as for evaluating the possibility of
wash-in/wash-out effects. Furthermore, the potential impact of other extrahepatic metabolism
has not been evaluated. Curiously, predictions for the tissue concentrations of TCE observed by
Greenberg et al. (1999) were not as discrepant as those for blood. A number of these hypotheses
could be tested; however, the existing data may not be sufficient to distinguish them. The
Merdink et al. (1998) study, in which TCE was given by i.v. (thereby avoiding both first pass in
the liver and any fractional uptake issue in the lung), may be somewhat helpful, but
unfortunately only oxidative metabolite concentrations were reported, not TCE concentrations.
A.2.2.2.2.2. Trichloroacetic acid (TCA) blood concentrations well predicted following
trichloroethylene (TCE) exposures, but TCA flux and disposition may not be accurate
TCA blood and plasma concentrations following TCE exposure are consistently well
predicted. However, the total flux of TCA may not be correct, as evidenced by the varying
degrees of consistency with urinary excretion data. Of particular importance are TCA dosing
studies, none of which were included in the calibration. In these studies, total recovery of
urinary TCA was found to be substantially less than the administered dose. However, the current
model assumes that urinary excretion is the only source of clearance of TCA, leading to
overestimation of urinary excretion. This fact, combined with the observation that under TCE
dosing, the model appears to give accurate predictions of TCA urinary excretion for several data
sets, strongly suggests a discrepancy in the amount of TCA formed from TCE. That is, since the
model appears to overpredict the fraction of TCA that appears in urine, it may be reducing TCA
production to compensate. Inclusion of the TCA dosing studies (including some oral dosing
studies), along with inclusion of a nonrenal clearance pathway, would probably be helpful in
This document is a draft for review purposes only and does not constitute Agency policy.
A-14 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
reducing these discrepancies. Finally, improvements in the TCOH/TCOG submodel, below,
should also help to ensure accurate estimates of TCA kinetics.
A.2.2.2.2.3. Trichloroethanol-trichloroethanol-glucuronide conjugate (TCOH/TCOG)
submodel requires revision and recalibration
Blood levels of TCOH and TCOG were inconsistently predicted. Part of this is due to the
problems with oral uptake, as discussed above. In addition, the problems identified with the use
of the Abbas and Fisher (1997) data (i.e., free TCOH vs. total TCOH), mean that this submodel
is not likely to be robust.
An additional concern is the over-prediction of urinary TCOG from the Abbas et al.
(1997) TCOH i.v. data. Like the case of TCA, this indicates that some other source of TCOH
clearance (not to TCA or urine—e.g., to dichloroacetic acid [DCA] or some other untracked
metabolite) is possible. This pathway can be considered for inclusion, and limits can be placed
on it using the available data.
Also, like for TCA, the fact that blood and urine are relatively well predicted from TCE
dosing strongly suggests a discrepancy in the amount of TCOH formed from TCE. That is, since
the model appears to overpredict the fraction of TCOH that appears in urine, it may be reducing
TCOH production to compensate. Including the TCOH dosing data would likely be helpful in
reducing these discrepancies.
Finally, as with the rat, the model needs to ensure that any first pass effect is accounted
for appropriately. Importantly, the estimated clearance rate for glucuronidation of TCOH is
substantially greater than hepatic blood flow. As was shown in Okino et al. (2005), in such a
situation, the use of a single compartment model across dose routes will be misleading because it
implies a substantial first-pass effect in the liver that cannot be modeled in a single compartment
model. That is, since TCOH is formed in the liver from TCE, and TCOH is also glucuronidated
in the liver to TCOG, a substantial portion of the TCOH may be glucuronidated before reaching
systemic circulation. This suggests that a liver compartment for TCOH is necessary.
Furthermore, because substantial TCOG can be excreted in bile from the liver prior to systemic
circulation, a liver compartment for TCOG may also be necessary to address that first pass
effect.
The addition of the liver compartment will necessitate several changes to model
parameters. The distribution volume for TCOH will be replaced by two parameters: the
liver:blood and body:blood partition coefficients. Similarly for TCOG, livenblood and
body:blood partition coefficients will need to be added. Clearance of TCOH to TCA and TCOG
This document is a draft for review purposes only and does not constitute Agency policy.
A-15 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
can be redefined as occurring in the liver, and urinary clearance can be redefined as coming from
the rest of the body. Fortunately, there are substantial data on circulating TCOG that has not
been included in the calibration. These data should be extremely informative in better estimating
the TCOH/TCOG submodel parameters.
A.2.2.2.2.4. Uncertainty in estimates of total metabolism
Closed-chamber data are generally thought to provide a good indicator of total
metabolism. Both subject-specific and population-based predictions of the only available
closed-chamber data (J. W. Fisher et al., 1991) were fairly accurate. Unfortunately, no additional
closed-chamber data were available. In addition, the discrepancies in observed and predicted
TCE blood concentrations following inhalation exposures remain unresolved. Hypothesized
explanations such as fractional uptake or presystemic elimination could have a substantial impact
on estimates of total metabolism.
In addition, no data are directly informative as to the fraction of total metabolism in the
lung, the amount of "untracked" hepatic oxidative metabolism (parameterized as "FracDCA"), or
any other extrahepatic metabolism. The lung metabolism as currently modeled could just as well
be located in other extrahepatic tissues, with little change in calibration. In addition, it is
difficult to distinguish between untracked hepatic oxidative metabolism and GSH conjugation,
particularly at low doses.
A.2.2.3. Rat Model
A.2.2.3.1. Subject-specific and population-based predictions
As with the mouse mode, initially, the sampled subject-specific parameters were used to
generate predictions for comparison to the calibration data. Because these parameters were
"optimized" for each subject, these "subject-specific" predictions should be accurate by design,
and indeed they were, as discussed in more detail in Table A-2.
Next, as with the mouse, only samples of the population parameters (means and
variances) were used, and "new subjects" were sampled from appropriate distribution using these
population means and variances. These "new subjects" then represent the predicted population
distribution, incorporating both variability in the population as well as uncertainty in the
population means and variances. These "population-based" predictions were then compared to
This document is a draft for review purposes only and does not constitute Agency policy.
A-16 DRAFT—DO NOT CITE OR QUOTE

-------
1	both the data used in calibration, as well as the additional data identified that was not used in
2	calibration. The Hack et al. (2006) PBPK model used for prediction was modified to
3	accommodate some of the different outputs (e.g., tissue concentrations) and exposure routes (i.v.,
4	intra-arterial [i.a.], and intraperivenous [p.v.]) used in the "noncalibration" data, but otherwise
5	unchanged.
This document is a draft for review purposes only and does not constitute Agency policy.
A-17 DRAFT—DO NOT CITE OR QUOTE

-------
1
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats
S?
>!
rs
s;
>1
£
>
a
&
H
1
o
o
2
o
H
O
HH
H
W
O
P
o
c
o
H
W
"T3
S
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Andersen et al.
(1987)
7-11
a/
Good posterior fits were obtained for these data—closed-chamber data with initial concentrations from
100 -4,640 ppm.
Barton et al.
(1995)
17-20

It was assumed that the closed-chamber volume was the same as for Andersen et al. (1987). However, the
initial chamber concentrations are not clear in the paper. The values that were used in the simulations do not
appear to be correct, since in many cases the time-course is inaccurately predicted even at the earliest
time-points. Conclusions as to these data need to await definitive values for the initial chamber concentrations,
which were not available.
Bernauer et al.
(1996)
1-3
a/
Urinary time-course data (see Figure 6-7) for TCA, TCOG, and NAcDCVC was given in concentration units
(mg/mg creat-h), whereas total excretion at 48 h (see Table 2) was given in molar units (mmol excreted). In
the original calibration files, the conversion from concentration to cumulative excretion was not
consistent—i.e., the amount excreted at 48 h was different. The data were revised using a conversion that
forced consistency. One concern, however, is that this conversion amounts to 6.2 mg creatinine over 48 h, or
1.14 micromol/h. This seems very low for rats; Trevisan et al. (2001), in samples from 195 male control rats,
found a median value of 4.95 micromol/h, a mean of 5.39 micromol/h, and a 1—99th percentile range of
2.56-10.46 micromol/h.
In addition, the NAcDCVC data were revised in include both 1,2- and 2,2-isomers, since the goal of the
GSH pathway is primarily to constrain the total flux. Furthermore, because of the extensive interorgan
processing of GSH conjugates, and the fact that excretion was still ongoing at the end of the study (48 h), the
amount of NAcDCVC recovered can only be a lower bound on the amount ultimately excreted in urine.
However, the model does not attempt to represent the excretion time-course of GSH conjugates—it merely
models the total flux. This is evinced by the fact that the model predicts complete excretion by the first time
point of 12 h, whereas in the data, there is still substantial excretion occurring at 48 h.
Posterior fits to these data were poor in all cases except urinary TCA at the highest dose. In all other
cases, TCOH/TCOG and TCA excretion was substantially overpredicted, though this is due to the revision of
the data (i.e., the different assumptions about creatinine excretion). Unfortunately, of the original calibration
data, this is the only one with TCA and TCOH/TCOG urinary excretion. Therefore, that part of the model is
poorly calibrated. On the other hand, NAcDCVC was underpredicted for a number of reasons, as noted above.
Because of the incomplete capture of NAcDCVC in urine, unless the model can accurately portray the
time-course of NAcDCVC in urine, it should probably not be used for calibration of the GSH pathway, except
perhaps as a lower bound.

-------
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats (continued)
S?
>!
rs
s;
>1
£
>
v,O
a
&
H
1
o
o
2
o
H
O
HH
H
W
O
P
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Birner et al.
(1993)
21-22

These data only showed urine concentrations, so a conversion was made to cumulative excretion based on an
assumed urine flow rate of 22.5 mL/day. Based on this, urinary NAcDCVC was underestimated by 100- to
1,000-fold. Urinary TCA was underestimated by about twofold in females (barely within the 95% confidence
interval), and was accurately estimated in males. Note that data on urinary flow rate from Trevisan et al.
(2001) in samples from 195 male control rats showed high variability, with a geometric standard deviation of
1.75, so this may explain the discrepancy in urinary TCA. However, the underestimation of urinary
NAcDCVC cannot be explained this way.
Dallas et al.
(1991)
23-24

At the lower (50 ppm) exposure, arterial blood concentrations were consistently overpredicted by about
2.5-fold, while at the higher (500 ppm) exposure, arterial blood was overpredicted by 1.5- to 2-fold, but within
the range of variability. Exhaled breath concentrations were in the middle of the predicted range of variability
at both exposure levels. The ratio of exhaled breath and arterial blood should depend largely on the blood-air
partition coefficient, with minor dependence on the assumed dead space. This suggests the possibility of some
unaccounted-for variability in the partition coefficient (e.g., posterior mean estimated to be 15.7; in vitro
measured values from the literature are as follows: 25.82 (Sato, Nakajima, Fujiwara, & Murayama, 1977), 21.9
(Gargas, Burgess, Voisard, Cason, & Andersen, 1989), 25.8 (Koizumi, 1989), 13.2 (J. W. Fisher, Whittaker,
Taylor, 3rd, & Andersen, 1989), posterior). Alternatively, there may be a systematic error in these data, since,
as discussed below, the fit of the model to the arterial blood data of Keys et al. (2003) was highly accurate.
Fisher et al.
(1989)
25-28

Good posterior fits were obtained for these data (in females)—closed-chamber data with initial concentrations
from 300-5,100 ppm. There was some slight overprediction of chamber concentrations (i.e., data showed
more uptake/metabolism) at the lower doses, but still within the 95% confidence interval.
Fisher et al.
(1991)
4-6
a/
Good posterior fits were obtained from these data—plasma levels of TCA and venous blood levels of TCE.
Green and Prout
(1985)
29-30

In naive rats at 500 mg/kg, urinary excretion of TCOH/TCOG and TCA at 24 h was underpredicted (twofold),
although within the 95% confidence interval. With bile-cannulated rats at the same dose, the amount of TCOG
in bile was well within the 95% confidence interval. Urinary TCOH/TCOG was still underpredicted by about
twofold, but again still within the 95% confidence interval.
Jakobson et al.
(1986)
31

The only data from the experiment (500 ppm in female rats) were venous blood concentrations during
exposure. There were somewhat overpredicted at early times (outside of 95% confidence interval for
first 30 min) but was well predicted at the termination of exposure. This suggests some discrepancies in
uptake to tissues that reach equilibrium quickly—the model approaches the peak concentration at a faster rate
than the data suggest.

-------
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats (continued)
S?
>!
rs
s;
>1
£
to
o
o
H
1
o
o
2
o
H
O
HH
H
W
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Kaneko et al.
(1994)
32-35

In these inhalation experiments (50-1,000 ppm), urinary excretion of TCOH/TCOG and TCA are consistently
overpredicted, particularly at lower doses. The discrepancy decreases systematically as dose increases, with
TCA excretion accurately predicted at 1,000 ppm (TCOH/TCOG excretion slightly below near the lower
95% confidence interval at this dose). This suggests a discrepancy in the dose-dependence of TCOH, TCOG,
and TCA formation and excretion.
On the other hand, venous blood TCE concentrations postexposure are well predicted. TCE blood
concentrations right at the end of the exposure are overpredicted; however, concentrations are rapidly declining
at this point, so even a few minutes delay in obtaining the blood sample could explain the discrepancy.
Keys et al.
(2003)
36-39

These experiments collected extensive data on TCE in blood and tissues following i.a., oral, and inhalation
exposures. For the i.a. exposure, blood and tissue concentrations were very well predicted by the model, even
with the use of the rapidly perfused tissue concentration as a surrogate for brain, heart, kidney, liver, lung, and
spleen concentrations. Similarly accurate predictions were found with the higher (500 ppm) inhalation
exposure. At the lower inhalation exposure (50 ppm), there was some minor overprediction of concentrations
(twofold), particularly in fat, but values were still within the 95% confidence intervals.
For oral exposure, the GI absorption parameters needed to be revised substantially to obtain a good fit.
When the values reported by Keys et al. (2003) were used, the model generally had accurate predictions.
Two exceptions were the values in the gut and fat in the first 30 min after exposure. In addition, the liver
concentration was over-predicted in the first 30 min, and under-predicted at 2-4 h, but still within the
95% confidence interval during the entire period.
Kimmerle and
Eben (1973a)
40-44

In these inhalation experiments (49-3,160 ppm), urinary excretion of TCOH/TCOG was systematically
overpredicted (>twofold; outside 95% confidence interval), while excretion of TCA was accurately predicted.
In addition, elimination by exhaled breath was substantially overpredicted at the lowest exposure. Blood
TCOH levels were accurately predicted, but blood TCE levels were overpredicted at the 55 ppm. Part of the
discrepancies may be due to limited analytic sensitivities at the lower exposures.
Larson and Bull
(1992)
12-14
a/
The digitization in the calibration file did not appear to be accurate, as there was a 10-fold discrepancy with the
original paper in the TCOH data. The data were replaced this those used by Clewell et al. (2000) and Bois
(2000a). Except for the TCOH data, differences between the digitizations were 20% or less.
Adequate posterior predictions were obtained for these data (oral dosing from 200 mg/kg to 3,000 mg/kg).
All predictions were within the 95% confidence interval of posterior predictions. Better fits were obtained
using subject-specific posterior parameters, for which gut absorption and TCA urinary excretion parameters
were more highly identified.

-------
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats (continued)
S?
>!
rs
s;
>1
£
to
o
H
1
o
o
2
o
H
O
HH
H
ffl
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Lash et al. (2006)
45-46

In these corn-oil gavage experiments, almost all of the measurements appeared to be systematically low,
sometimes by many orders of magnitude. For example, at the lowest dose (263 mg/kg), urinary excretion of
TCOH/TCOG and TCA, and blood concentrations of TCOH were overpredicted by the model by around
>105-fold. TCE concentrations in blood and tissues at 2, 4, and 8 h were underpredicted by 103- to 104-fold.
Many studies, including those using the corn oil gavage (T. Green & Prout, 1985; Hissink et al., 2002), with
similar ranges of oral doses show good agreement with the model, it seems likely that these data are aberrant.
Lee et al. (1996)
47-61

This extensive set of experiments involved multiroute administration of TCE (oral, i.v., i.a., or portal vein),
with serial measurements of arterial blood concentrations. For the oral route (8 mg/kg-64 mg/kg), the GI
absorption parameters had to be modified. The values from Keys et al. (2003) were used, and the resulting
predictions were quite accurate, albeit a more prominent peak was predicted. Predictions >30 min after dosing
were highly accurate.
For the i.v. route (0.71 mg/kg-64 mg/kg), predictions were also highly accurate in almost all cases. At the
lower doses (0.71 mg/kg and 2 mg/kg), there was slight overprediction in the first 30 min after dosing. At
highest dose (64 mg/kg), there was slight underprediction between 1 and 2 h after dosing. In all cases, the
values were within the 95% confidence interval.
For the i.a. route (0.71 mg/kg-16 mg/kg), all predictions were very accurate.
For the p.v. route (0.71 mg/kg-64 mg/kg), predictions still remained in the 95% confidence interval,
although there was more variation. At the lowest dose, there was overprediction in the first 30 min after
dosing. At the highest two doses (16 mg/kg and 64 mg/kg), there was slight underprediction between 1 and 5 h
after dosing. This may in part be because a pharmacodynamic change in metabolism (e.g., via direct solvent
injury proposed by K. Lee, Muralidhara, Schnellmann, & Bruckner, 2000).
Lee et al. (2000)
62-69

In the p.v. and i.v. exposures, blood and liver concentrations were accurately predicted. For oral exposures, the
GI absorption parameters needed to be changed. While the values from Keys et al. (2003) led to accurate
predictions for lower doses (2 mg/kg-16 mg/kg), at the higher doses (48 mg/kg-432 mg/kg), much slower
absorption was evident. Comparisons at these higher dose are not meaningful without calibration of absorption
parameters.
Prout et al.
(1985)
15
a/
Adequate posterior fits were obtained for these data—rat dosing at 1,000 mg/kg in corn oil. All predictions
were within the 95% confidence interval of posterior predictions. Better fits were obtained using
subject-specific posterior parameters, for which gut absorption and TCA urinary excretion parameters were
more highly identified.

-------
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats (continued)
Reference
Simulation #
Calibration
data
Discussion
Stenner et al.
(1997)
70

As with other oral exposures, different GI absorption parameters were necessary. Again, the values from Keys
et al. (2003) were used, with some success. Blood TCA levels were accurately predicted, while TCOH blood
levels were systematically under-predicted (up to 10-fold).
Additional data with TCOH and TCA dosing, including naive and bile-cannulated rats, can be added when
those exposure routes are added to the model. These could be useful in better calibrating the enterohepatic
recirculation parameters.
Templin et al.
(1995)
16
a/
Adequate posterior fits were obtained for blood TCA from these data—oral dosing at 100 mg/kg in Tween.
Blood levels of TCOH were underpredicted, while the time-course of TCE in blood exhibited an earlier peak.
Better fits were obtained using subject-specific posterior parameters, for which gut absorption and TCA
urinary excretion parameters (and to a lesser extent glucuronidation of TCOH and biliary excretion of TCOG)
were more highly identified.
GI = gastrointestinal, NAc-l,2-DCVC = N-acetyl-S-(l,2-dichlrovinyl)-L-cysteine, NAc-2,2-DCVC = N-acetyl-S-(2,2-dichlrovinyl)-L-cysteine,
NAcDCVC = NAc-l,2-DCVC and NAc-2,2-DCVC.

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
A.2.2.3.1.1. Subject-specific predictions and calibration data
(See "Supplementary data for TCE assessment: Hack mouse subject calibration," 2011)
A.2.2.3.1.2. Population-based predictions and calibration and additional evaluation data
(See "Supplementary data for TCE assessment: Hack mouse subject calibration," 2011)
A.2.2.3.2. Conclusions regarding rat model
A.2.2.3.2.1. Trichloroethylene (TCE) concentrations in blood and tissues generally
well-predicted
The PBPK model for the parent compound appears to be robust. Multiple data sets not
used for calibration with TCE measurements in blood and tissues were simulated, and overall the
model gave very accurate predictions. A few data sets seemed somewhat anomalous—Dallas
et al. (1991), Kimmerle and Eben (1973a), Lash et al. (2006). However, data from Kaneko et al.
(1994), Keys et al. (2003), and Lee et al. (1996; 2000) were all well simulated, and corroborated
the data used for calibration (J. W. Fisher et al., 1991; J. Larson & R. Bull, 1992; Prout et al.,
1985; Templin et al., 1995). Particularly important is the fact that tissue concentrations from
Keys et al. (2003) were well simulated.
A.2.2.3.2.2. Total metabolism probably well simulated, but ultimate disposition is less
certain
Closed-chamber data are generally thought to provide a good indicator of total
metabolism. Two closed-chamber studies not used for calibration were available—Barton et al.
(1995) and Fisher et al. (1989). Additional experimental information is required to analyze the
Barton et al. (1995) data, but the predictions for the Fisher et al. (1989) data were quite accurate.
However, the ultimate disposition of metabolized TCE is much less certain. Clearly, the
flux through the GSH pathway is not well constrained, with apparent discrepancies between the
N-acetyl-S-(l,2-dichlorovinyl)-L-cysteine (NAc-l,2-DCVC) data of Bernauer et al. (1996) and
Birner et al. (1993). Moreover, each of these data has limitations—in particular, the Bernauer
et al. (1996) data show that excretion is still substantial at the end of the reporting period, so that
the total flux of mercapturates has not been collected. Moreover, there is some question as to the
This document is a draft for review purposes only and does not constitute Agency policy.
A-23 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
consistency of the Bernauer et al. (1996) data (see Table 2 vs. Figures 6 and 7), since a direct
comparison seems to imply a very low creatinine excretion rate. The Birner et al. (1993) data
only report concentrations—not total excretion—so a urinary flow rate needs to be assumed.
In addition, no data are directly informative as to the fraction of total metabolism in the
lung or the amount of "untracked" hepatic oxidative metabolism (parameterized as "FracDCA").
The lung metabolism could just as well be located in other extrahepatic tissues, with little change
in calibration. In addition, there is a degeneracy between untracked hepatic oxidative
metabolism and GSH conjugation, particularly at low doses.
The ultimate disposition of TCE as excreted TCOH/TCOG or TCA is also poorly
estimated in some cases, as discussed in more detail below.
A.2.2.3.2.3. Trichloroethanol-trichlorethanol-glucuronide conjugate (TCOH/TCOG)
submodel requires revision and recalibration
TCOH blood levels of TCOH were inconsistently predicted in noncalibration data sets
[well predicted for Larson and Bull (1992); Kimmerle and Eben (1973a); but not Stenner et al.
(1997) or Lash et al. (2006)], and the amount of TCE ultimately excreted as TCOG/TCOH also
appeared to be poorly predicted. The model generally underpredicted TCOG/TCOH urinary
excretion (underpredicted Green and Prout (1985), overpredicted Kaneko et al. (1994),
Kimmerle and Eben (1973a), and Lash et al. (2006)). This may in part be due to discrepancies in
the Bernauer et al. (1996) data as to the conversion of excretion relative to creatinine.
Moreover, there are relatively sparse data on TCOH in combination with a relatively
complex model, so the identifiability of various pathways—conversion to TCA, enterohepatic
recirculation, and excretion in urine—is questionable.
This could be improved by the ability to incorporate TCOH dosing data from Merdink
et al. (1999) and Stenner et al. (1997), the latter of which included bile duct cannulation to better
estimate enterohepatic recirculation parameters. However, the TCOH dosing in these studies is
by the intravenous route, whereas with TCE dosing, TCOH first appears in the liver. Thus, the
model needs to ensure that any first pass effect is accounted for appropriately. Importantly, the
estimated clearance rate for glucuronidation of TCOH is substantially greater than hepatic blood
flow. That is, since TCOH is formed in the liver from TCE, and TCOH is also glucuronidated in
the liver to TCOG, a substantial portion of the TCOH may be glucuroni dated before reaching
systemic circulation. Thus, suggests that a liver compartment for TCOH is necessary.
Furthermore, because substantial TCOG can be excreted in bile from the liver prior to systemic
This document is a draft for review purposes only and does not constitute Agency policy.
A-24 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
circulation, a liver compartment for TCOG may also be necessary to address that first pass
effect.
The addition of the liver compartment will necessitate several changes to model
parameters. The distribution volume for TCOH will be replaced by two parameters: the
liver:blood and body:blood partition coefficients. Similarly for TCOG, livenblood and
body:blood partition coefficients will need to be added. Clearance of TCOH to TCA and TCOG
can be redefined as occurring in the liver, and urinary clearance can be redefined as coming from
the rest of the body.
Finally, additional clearance of TCOH (not to TCA or urine—e.g., to DCA or some other
untracked metabolite) is possible. This may in part explain the discrepancy between the accurate
predictions to blood data along with poor predictions to urinary excretion (i.e., there is a missing
pathway). This pathway can be considered for inclusion, and limits can be placed on it using the
available data.
A.2.2.3.2.4. Trichloroacetic acid (TCA) submodel would benefit from revised
trichloroethanol/trichloroethanol-glucuronide conjugate (TCOH/TCOG) submodel and
incorporating TCA dosing studies
While blood levels of TCA were well predicted in the one noncalibration data set
(Stenner et al., 1997), the urinary excretion of TCA was inconsistently predicted [underpredicted
in Green and Prout (1985); overpredicted in Kaneko et al. (1994) and Lash et al. (2006);
accurately predicted in Kimmerle and Eben (1973a)]. Because TCA is in part derived from
TCOH, a more accurate TCOH/TCOG submodel would probably improve the TCA submodel.
In addition, there are a number of TCA dosing studies that could be used to isolate the
TCA kinetics from the complexities of TCE and TCOH. These could be readily incorporated
into the TCA submodel.
Finally, as with TCOH, additional clearance of TCA (not to urine—e.g., to DCA or some
other untracked metabolite) is possible. This may in part explain the discrepancy between the
accurate predictions to blood data along with poor predictions to urinary excretion (i.e., there is a
missing pathway). As with TCOH, this pathway can be considered for inclusion, and limits can
be placed on it using the available data.
This document is a draft for review purposes only and does not constitute Agency policy.
A-25 DRAFT—DO NOT CITE OR QUOTE

-------
A.2.2.4. Human Model
A.2.2.4.1. Subject-specific and population-based predictions
1	As with the mouse and rat models, initially, the sampled subject-specific parameters were
2	used to generate predictions for comparison to the calibration data. Because these parameters
3	were "optimized" for each subject, these "subject-specific" predictions should be accurate by
4	design. However, unlike for the rat, this was not the case for some experiments (this is partially
5	responsible for the slower convergence), although the inaccuracies were generally less than those
6	in the mouse. For example, alveolar air concentrations were systematically overpredicted for
7	several data sets. There was also variability in the ability to predict the precise time-course of
8	TCA and TCOH blood levels, with a few data sets more difficult for the model to accommodate.
9	These data are discussed further in Table A-3. Next, only samples of the population parameters
10 (means and variances) were used, and "new subjects" were sampled from appropriate
This document is a draft for review purposes only and does not constitute Agency policy.
A-26 DRAFT—DO NOT CITE OR QUOTE

-------
1
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans
S?
>!
rs
s;
>1
£
to
<1
o
H
1
o
o
2
o
H
O
HH
H
W
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Bartonicek
(1962)
38-45

The measured minute-volume was multiplied by a factor of 0.7 to obtain an estimate for alveolar ventilation
rate, which was fixed for each subject. These data are difficult to interpret because they consist of many single
data points. It is easiest to go through the measurements one at a time:
Alveolar retention (1—exhaled dose/inhaled dose during exposure) and Retained dose (inhaled dose—exhaled
dose during exposure): Curiously, retention was generally under-predicted, which in many cases retained dose
was accurately predicted. However, alveolar retention was an adjustment of the observed total retention:
TotRet = (CInh - CExh)/CInh = QAlv x (CInh - CAlv)/(MV x CInh), so that
AlvRet = TotRet x (QAlv/MV), with QAlv/MV assumed to be 0.7.
Because retained dose is the more relevant quantity, and is less sensitive to assumptions about QAlv/MV, then
this is the better quantity to use for calibration.
Urinary TCOG: This was generally underpredicted, although generally within the 95% confidence
interval. Thus, these data will be informative as to intersubject variability.
Urinary TCA: Total collection (at 528 h) was accurately predicted, although the amount collected at 72 h
was generally under-predicted, sometimes substantially so.
Plasma TCA: Generally well predicted.
Bernauer et al.
(1996)
1-3
a/
Subject-specific predictions were good for the time-courses of urinary TCOG and TCA, but poor for total
urinary TCOG+TCA and for urinary NAc-l,2-DCVC. One reason for the discrepancy in urinary excretion of
TCA and TCOG is that the urinary time-course data (see Figures 4-5 in the manuscript) for TCA, TCOG, and
NAc-l,2-DCVC was given in concentration units (mg/mg creat-h), whereas total excretion at 48 h (see Table 2
in the manuscript) was given in molar units (mmol excreted). In the original calibration files, the conversion
from concentration to cumulative excretion was not consistent—i.e., the amount excreted at 48 h was different.
For population-based predictions, the data were revised using a conversion that forced consistency.
One concern, however, is that this conversion amounts to 400-500 mg creatinine over 48 h, or
200-250 mg/day, which seems rather low. For instance, Araki (1978) reported creatinine excretion of
11.5 ± 1.8 mmol/24 h (mean ± SD) in nine subjects, corresponding to 1,300 ± 200 mg/day.
In addition, for population-based predictions, the data were revised include both the NAc-l,2-DCVC and
the N acetyl-S-(2,2-dichlorovinyl)-L-cysteine isomer (the combination denoted NAcDCVC), since the goal of
the GSH pathway is primarily to constrain the total flux. Furthermore, because of the extensive interorgan
processing of GSH conjugates, and the fact that excretion was still ongoing at the end of the study (48 h), the
amount of NAcDCVC recovered can only be a lower bound on the amount ultimately excreted in urine.
However, the model does not attempt to represent the excretion time-course of GSH conjugates—it merely
models the total flux. This is evinced by the fact that the model predicts complete excretion by the first time
point of 12 h, whereas in the data, there is still substantial excretion occurring at 48 h.

-------
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans (continued)
S?
>!
rs
s;
>1
£
to
00
o
H
1
o
o
2
o
H
o
HH
H
W
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>i
sT
0q
O
t
Reference
Simulation #
Calibration
data
Discussion
Bernauer et al.
(1996)
(continued)
1-3
(continued)

Population-based posterior fits to these data were quite good for urinary TCA and TCOH, but not for
NAcDCVC in urine. Because of the incomplete capture of NAcDCVC in urine, unless the model can
accurately portray the time-course of NAcDCVC in urine, it should probably not be used for calibration of the
GSH pathway, except perhaps as a lower bound.
Bloemen et al.
(2001)
72-75

Like Bartonicek (1962), these data are more difficult to interpret due to their being single data points for each
subject and exposure. However, in general, posterior population-based estimates of retained dose, urinary
TCOG, and urinary TCA were fairly accurate, staying within the 95% confidence interval, and mostly inside
the interquartile range. The data on GSH mercapturates are limited—first they are all nondetects. In addition,
because of the 48-56 h collection period, excretion of GSH mercapturates is probably incomplete, as noted
above in the discussion of Bernauer et al. (1996).
Chiu et al.
(2007)
66-71

The measured minute-volume was multiplied by a factor of 0.7 to obtain an estimate for alveolar ventilation
rate, which was fixed for each subject. Alveolar air concentrations of TCE were generally well predicted,
especially during the exposure period. Postexposure, the initial drop in TCE concentration was generally
further than predicted, but the slope of the terminal phase was similar. Blood concentrations of TCE were
consistently overpredicted for all subjects and occasions.
Blood concentrations of TCA were consistently over-predicted, though mostly staying in the lower
95% confidence region. Blood TCOH (free) levels were generally over-predicted, in many cases falling below
the 95% confidence region, though in some cases the predictions were accurate. On the other hand, total
TCOH (free+glucuronidated) was well predicted (or even under-predicted) in most cases—in the cases where
free TCOH was accurately predicted, total TCOH was underpredicted. The free and total TCOH data reflect
the higher fraction of TCOH as TCOG than previously reported (e.g., Fisher et al. (1998) reported no
detectable TCOG in blood).
Data on urinary TCA and TCOG were complicated by some measurements being saturated, as well as the
intermittent nature of urine collection after Day 3. Thus, only the nonsaturated measurements for which the
time since the last voiding was known were included for direct comparison to the model predictions. Saturated
measurements were kept track of separately for comparison, but were considered only rough lower bounds.
TCA excretion was generally over-predicted, whether looking at unsaturated or saturated measurements (the
latter, would of course, be expected). Urinary excretion of TCOG generally stayed within the 95% confidence
range.
Fernandez et al.
(1977)


Alveolar air concentrations are somewhat overestimated. Other measurements are fairly well predicted.

-------
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans (continued)
S?
>!
rs
s;
>1
£
to
VO
o
H
1
o
o
2
o
H
O
HH
H
W
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Fisher et al.
(1998)
13-33
a/
The majority of the data used in the calibration (both in terms of experiments and data points) came from this
study. In general, the subject-specific fits to these data were good, with the exception of alveolar air
concentrations, which were consistently over-predicted. In addition, for some subjects, the shape of the TCOH
time-course deviated from the predictions (#14, 24, 29, and 30)—the predicted peak was too "sharp," with
underprediction at early times. Simulation #23 showed the most deviation from predictions, with substantial
inaccuracies in blood TCA, TCOH, and urinary TCA.
Interestingly, in the population-based predictions, in same cases the predictions were not very
accurate—indicating that the full range of population variability is not accounted for in the posterior
simulations. This is particularly the case with venous blood TCE concentrations, which are generally
under-predicted in population estimates (although in some cases the predictions are accurate).
One issue with the way in which these data were utilized in the calibration is that in some cases, the same
subject was exposed to two different concentrations, but in the calibration, they were treated as separate
"subjects." Thus, parameters were allowed to vary between exposures, mixing intersubject and interoccasion
variability. It is recommended that in subsequent calibrations, the different occasions with the same subject be
modeled together. This will also allow identification of any dose-related changes in parameters (e.g.,
saturation).
Kimmerle and
Eben (1973b)
46-57

Blood TCE levels are generally over-predicted for both single and multiexposure experiments. However,
levels at the end of exposure are rapidly changing, so some of those values may be better predicted if the
"exact" time after cessation of exposure were known.
Blood TCOH levels are fairly accurately predicted, although in some subjects in single exposure
experiments, there is a tendency to overpredict at early times and underpredict at late times. In multiexposure
experiments, the decline after the last exposure was somewhat steeper than predicted. Urinary excretion of
TCA and TCOH was well predicted.
Only grouped data on alveolar air concentrations were available, so they were not used.
Lapare et al.
(1995)
34
a/
Predictions for these data were not accurate. However, there was an error in some of the exposure
concentrations used in the original calibration. In addition, the last exposure "occasion" in these experiments
involved exercise/workload, and so should be excluded. Finally, subject data are available for these
experiments.

62-65
(individual
data)

Taking into account these changes, population-based predictions were somewhat more accurate. However,
alveolar air concentrations and venous blood TCE concentrations were still over-predicted.

-------
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans (continued)
S?
>!
rs
s;
>1
£
U>
o
o
H
1
o
o
2
o
H
O
HH
H
W
O
V
o
c
o
H
W
"T3
S
^3
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>i
sT
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Monster et al.
(1976)
5-6
(summary
data)
a/
Subject-specific predictions were quite good, except that for blood TCA concentrations exhibited a higher
peak that predicted. However, TCOH values were entered as free TCOH, whereas the TCOH data were
actually total (free + glucuronidated) TCOH. Therefore, for population-based predictions, this change was
made. In addition, as with the Monster et al. (1979) data, minute-volume and exhaled air concentrations were
measured and incorporated for population-based predictions. Finally, subject-specific data are available, so, in
this case, those data should replace the grouped data in any revised calibration. These individual data also
included estimates of retained dose based on complete inhaled and exhaled air samples during exposure.
For population-based predictions, as with the Monster et al. (1979) data, grouped urinary and blood
TCOH/TCOG was somewhat under-predicted in the population-based predictions, and grouped alveolar and
blood TCE concentrations were somewhat over-predicted.

58-61
(individual
data)

The results for the individual data were similar, but exhibited substantially greater variability that predicted.
For instance, in subject A, blood TCOH levels were generally greater than the 95% confidence interval at both
70 and 140 ppm, whereas predictions for blood TCOH in subject D were quite good. In another example, for
blood TCE levels, predictions for subject B were quite good, but those for subject D were poor (substantially
overpredicted). Thus, it is anticipated that adding these individual data will be substantially informative as to
intersubject variability, especially since all four individuals were exposed at two different doses.
Monster et al.
(1979)
4
a/
Subject-specific predictions for these data were quite good. However, TCA values were entered as plasma,
whereas the TCA data were actually in whole blood. Therefore, for population-based predictions, this change
was made. In addition, two additional time-courses were available that were not used in calibration: exhaled
air concentrations and total TCOH blood concentrations. These were added for population-based predictions.
In addition, the original article had data on ventilation rate, which as incorporated into the model. The
minute volume needed to be converted to alveolar ventilation rate for the model, but this required adjusted for
an extra dead space volume of 0.15 L due to use of a mask, as suggested in the article. The measured mean
minute volume was 11 L/min, and with a breathing rate of 14 breaths/min (assumed in the article), this
corresponding to a total volume of 0.79 L. Subtracting the 0.15 L of mask dead space and 0.15 L of
physiological dead space (suggested in the article) gives 0.49 L of total physiological dead space. Thus, the
minute volume of 11 L/min was adjusted by the factor 0.49/0.79 to give an alveolar ventilation rate of
6.8 L/min, which is a reasonably typical value at rest.
Due to extra nonphysiological dead space issue, some adjustment to the exhaled air predictions also
needed to be made. The alveolar air concentration CAlv was, therefore, estimated based on the formula

-------
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans (continued)
S?
>!
rs
s;
>1
£
>
a
&
H
1
o
o
2
o
H
O
HH
H
W
O
P
o
c
o
H
W
"T3
S
O
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1
Oq
O
t
Reference
Simulation #
Calibration
data
Discussion
Monster et al.
(1979)
(continued)
4 (continued)

CAlv = (CExh x VTot - CInh x VDs)/VAlv
where CExh is the measured exhaled air concentration, VTot is the total volume (alveolar space VAlv of
0.49 L, physiological dead space of 0.15 L, and mask dead space of 0.15 L), VDs is the total dead space of
0.3 L, and CInh is the inhaled concentration.
Population-based predictions for these data lead to slight underestimation urinary TCOG and blood TCOH
levels, as well as some over-prediction of alveolar air and venous blood concentrations by factors of 3~10-fold.
Muller et al.
(1972, 1974;
1975)
7-10
a/
Subject-specific predictions for these data were good, except for alveolar air concentrations. However, several
problems were found with these data as utilized in the original calibration:
•	Digitization problems, particular with the time axis in the multiday exposure study (Simulation 9) that led
to measurements taken prior to an exposure modeled as occurring during the exposure. The original
digitization from Bois (2000a) and Clewell et al. (2000) was used for population-based estimates.
•	Original article showed TCA as measured in plasma, not blood as was assumed in the calibration.
•	Blood was taken from the earlobe, which is thought to be indicative of arterial blood concentrations, rather
than venous blood concentrations.
•	TCOH in blood was free, not total, as Ertle et al. (1972) (cited in Methods) had no use of p-glucuronidase
in analyzing blood samples. Separate free and total measurements were done in plasma (not whole blood),
but these data were not included.
•	Simulation 9, contiguous data on urinary excretion were only available out to 6 day, so only that data
should be included.
•	Simulation 10, is actually the same as the first day of simulation 9, from Muller et al. (1972; 1975) (the
data were reported in both papers), and, thus, should be deleted.
These were corrected in the population-based estimates. Alveolar air concentration measurements remained
over-predicted, while the change to arterial blood led to over-prediction of those measurements during
exposure (but postexposure predictions were accurate).

-------
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans (continued)
Reference
Simulation #
Calibration
data
Discussion
Muller et al.
(1974)
81-82 (TCA
and TCOH
dosing)

The experiment with TCA showed somewhat more rapid decline in plasma levels than predicted, but still well
within the 95% confidence range. Urinary excretion was well predicted, but only accounted for 60% of the
administered dose—this is not consistent with the rapid decline in TCA plasma levels (10-fold lower than peak
at the end of exposure), which would seem to suggest the majority of TCA has been eliminated. With TCOH
dosing, blood levels of TCOH were over-predicted in the first 5 h, perhaps due to slower oral absorption (the
augmented model used instantaneous and complete absorption). TCA plasma and urinary excretion levels
were fairly well predicted. However, urinary excretion of TCOG was near the bottom of the 95% confidence
interval; while, in the same individuals with TCE dosing (Simulation 7), urinary excretion of TCOG was
substantially greater (near slightly above the interquartile region). Furthermore, total TCA and TCOG urinary
excretion accounted for <40% of the administered dose.
Paykoc and
Powell (1945)
35-37

Population-based fits were good, within the inner quartile region.
Sato et al. (1977)
76

Both alveolar air and blood concentrations are over-predicted in this model. Urinary TCA and TCOG, on the
other hand, are well predicted.
Stewart et al.
(1970)
11
a/
Subject-specific predictions for these data were good, except for some alveolar air concentrations. However, a
couple of problems were found with these data as utilized in the original calibration:
•	The original article noted that individuals took a lunch break during which there was no exposure. This
was not accounted for in the calibration runs, which a assumed a continuous 7-h exposure. The exposures
were, therefore, revised with a 3-h morning exposure (9-12), a 1 h lunch break (12-1), and 4-h afternoon
exposure (1-5), to mimic a typical workday. The times of the measurements had to be revised as well,
since the article gave "relative" rather than "absolute" times (e.g., x h postexposure).
•	Contiguous data on urinary excretion were only available out to 11 day, so only that data should be
included (see Table 2).
With these changes, population-based predictions of urinary TCA and TCOG were still accurate, but alveolar
air concentrations were over-predicted.
Triebig et al.
(1976)
12
a/
Only two data points are available for alveolar air, and blood TCA and TCOH. Only one data point is
available on blood TCE. Alveolar air was underpredicted at 24 h. Blood TCA and TCOH were within the
95% confidence ranges. Blood TCE was over-predicted substantially (outside 95% confidence range).

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
distribution using these population means and variances. These "new subjects" then represent
the predicted population distribution, incorporating both variability as well as uncertainty in the
population means and variances. These "population-based" predictions were then compared to
both the data used in calibration, as well as the additional data identified that was not used in
calibration. The Hack et al. (2006) PBPK model was modified to accommodate some of the
different outputs (e.g., arterial blood, intermittently collected urine, retained dose) and exposure
routes (TCA i.v., oral TCA, and TCOH) used in the "noncalibration" data, but otherwise
unchanged.
A.2.2.4.1.1. Subject-specific predictions and calibration data
(See "Supplementary data for TCE assessment: Hack mouse subject calibration," 2011)
A.2.2.4.1.2. Population-based predictions and calibration and additional evaluation data
(See "Supplementary data for TCE assessment: Hack mouse subject calibration," 2011)
A.2.2.4.2. Conclusions regarding human model
A.2.2.4.2.1. Trichloroethylene (TCE) concentrations in blood and air are often not
well-predicted
Except for the Chiu et al. (2007) during exposure, TCE alveolar air levels were
consistently overpredicted. Even in Chiu et al. (2007), TCE levels postexposure were
over-predicted, as the drop-off after the end of exposure was further than predicted. Because
predictions for retained dose appear to be fairly accurate, this implies that less clearance is
occurring via exhalation than predicted by the model. This could be the result of additional
metabolism or storage not accounted for by the model.
Except for the Fisher et al. (1998) data, TCE blood levels were consistently
overpredicted. Because the majority of the data used for calibration was from Fisher et al.
(1998), this implies that the Fisher et al. (1998) data had blood concentrations that were
consistently higher than the other studies. This could be due to differences in metabolism and/or
distribution among studies.
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 3 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Interestingly, the mouse inhalation data also exhibited inaccurate prediction of blood
TCE levels. Hypotheses such as fractional uptake or presystemic elimination due to local
metabolism in the lung have not been tested experimentally, nor is it clear that they can explain
the discrepancies.
Due to the difficulty in accurately predicted blood and air concentrations, there may be
substantial uncertainty in tissue concentrations of TCE. However, such potential model errors
can be characterized estimated and estimated as part of a revised calibration.
A.2.2.4.2.2. Trichloroacetic acid (TCA) blood concentrations well predicted following
trichloroethylene (TCE) exposures, but some uncertainty in TCA flux and disposition
TCA blood and plasma concentrations and urinary excretion, following TCE exposure,
are generally well predicted. Even though the model's central estimates over-predicted the Chiu
et al. (2007) TCA data, the confidence intervals were still wide enough to encompass those data.
However, the total flux of TCA may not be correct, as evidenced by TCA dosing studies,
none of which were included in the calibration. In these studies, total recovery of urinary TCA
was found to be substantially less than the administered dose. However, the current model
assumes that urinary excretion is the only source of clearance of TCA. This leads to
overestimation of urinary excretion. This fact, combined with the observation that under TCE
dosing, the model appears to give accurate predictions of TCA urinary excretion for several data
sets, strongly suggests a discrepancy in the amount of TCA formed from TCE. That is, since the
model appears to overpredict the fraction of TCA that appears in urine, it may be reducing TCA
production to compensate. Inclusion of the TCA dosing studies, along with inclusion of a
nonrenal clearance pathway, would probably be helpful in reducing these discrepancies. Finally,
improvements in the TCOH/TCOG submodel, below, should also help to insure accurate
estimates of TCA kinetics.
A.2.2.4.2.3. Trichloroethanol-trichlorethanol-glucuronide conjugate (TCOH/TCOG)
submodel requires revision and recalibration
Blood levels of TCOH and urinary excretion of TCOG were generally well predicted.
Additional individual data show substantial intersubject variability than can be incorporated into
the calibration. Several errors as to the measurement of free or total TCOH in blood need to be
corrected.
This document is a draft for review purposes only and does not constitute Agency policy.
A-34 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
A few inconsistencies with noncalibration data sets stand out. The presence of
substantial TCOG in blood in the Chiu et al. (2007) data are not predicted by the model.
Interestingly, only two studies that included measurements of TCOG in blood (rather than just
total TCOH or just free TCOH)—Muller et al. (1975), which found about 17% of total TCOH to
be TCOG, and Fisher et al. (1998), who could not detect TCOG. Both of these studies had
exposures at 100 ppm. Interestingly Muller et al. (1975) reported increased TCOG (as fraction
of total TCOH) with ethanol consumption, hypothesizing the inhibition of a glucuronyl
transferase that slowed glucuronidation. This also would result in a greater half-life for TCOH in
blood with ethanol consumptions, which was observed.
An additional concern is the over-prediction of urinary TCOG following TCOH
administration from the Muller et al. (1974) data. Like the case of TCA, this indicates that some
other source of TCOH clearance (not to TCA or urine—e.g., to DCA or some other untracked
metabolite) is possible. This pathway can be considered for inclusion, and limits can be placed
on it using the available data.
Also, as for TCA, the fact that blood and urine are relatively well predicted from TCE
dosing strongly suggests a discrepancy in the amount of TCOH formed from TCE. That is, since
the model appears to overpredict the fraction of TCOH that appears in urine, it may be reducing
TCOH production to compensate.
Finally, as with the rat and mice, the model needs to ensure that any first pass effect is
accounted for appropriately. Particularly for the Chiu et al. (2007) data, in which substantial
TCOG appears in blood, since TCOH is formed in the liver from TCE, and TCOH is also
glucuronidated in the liver to TCOG, a substantial portion of the TCOH may be glucuronidated
before reaching systemic circulation. Thus, suggests that a liver compartment for TCOH is
necessary. Furthermore, because substantial TCOG can be excreted in bile from the liver prior
to systemic circulation, a liver compartment for TCOG may also be necessary to address that
first pass effect. In addition, in light of the Chiu et al. (2007) data, it may be useful to expand the
prior range for the KM of TCOH glucuronidation.
The addition of the liver compartment will necessitate several changes to model
parameters. The distribution volume for TCOH will be replaced by two parameters: the
liver:blood and body:blood partition coefficients. Similarly for TCOG, livenblood and
body:blood partition coefficients will need to be added. Clearance of TCOH to TCA and TCOG
can be redefined as occurring in the liver, and urinary clearance can be redefined as coming from
the rest of the body. Fortunately, there are in vitro partition coefficients for TCOH. It may be
important to incorporate the fact that Fisher et al. (1998) found no TCOG in blood. This can be
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 5 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
included by having the TCOH data be used for both free and total TCOH (particularly since that
is how the estimation of TCOG was made—by taking the difference between total and free).
A.2.2.4.2.4. Uncertainty in estimates of total metabolism
Estimates of total recovery after TCE exposure (TCE in exhaled air, TCA and TCOG in
urine) have been found to be only 60-70% (W. A. Chiu et al., 2007; Monster et al., 1976, 1979).
Even estimates of total recovery after TCA and TCOH dosing have found 25-50% unaccounted
for in urinary excretion (Muller et al., 1974; Paykoc & Powell, 1945). Bartonicek (1962) found
some TCOH and TCA in feces, but this was about 10-fold less than that found in urine, so this
cannot account for the discrepancy. Therefore, it is likely that additional metabolism of TCE,
TCOH, and/or TCA are occurring. Additional metabolism of TCE could account for the
consistent overestimation of TCE in blood and exhaled breath found in many studies. However,
no data are directly informative as to the fraction of total metabolism in the lung, the amount of
"untracked" hepatic oxidative metabolism (parameterized as "FracDCA"), or any other
extrahepatic metabolism. The lung metabolism as currently modeled could just as well be
located in other extrahepatic tissues, with little change in calibration. In addition, it is difficult to
distinguish between untracked hepatic oxidative metabolism and GSH conjugation, particularly
at low doses.
A.3. PRELIMINARY ANALYSIS OF MOUSE GAS UPTAKE DATA: MOTIVATION
FOR MODIFICATION OF RESPIRATORY METABOLISM
Potential different model structures can be investigated using the core PBPK model
containing averaged input parameters, since this approach saves computational time and is more
efficient when testing different structural hypotheses. This approach is particularly helpful for
quick comparisons of data with model predictions. During the calibration process, this approach
was used for different routes of exposure and across all three species. For both mice and rats, the
closed-chamber inhalation data resulted in fits that were considered not optimal when visually
examined. Although closed-chamber inhalation usually combines multiple animals per
experiment, and may not be as useful in differentiating between individual and experimental
uncertainty (Hack et al., 2006), closed-chamber data do describe in vivo metabolism and have
been historically used to quantify averaged in vivo Michaelis-Menten kinetics in rodents.
There are several assumptions used when combining PBPK modeling and
closed-chamber data to estimate metabolism via regression. The key experimental principles
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 6 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
require a tight, sealed, or air-closed system where all chamber variables are controlled to known
set points or monitored, that is all except for metabolism. For example, the inhalation chamber is
calibrated without an animal, to determine normal absorption to the empty system. This empty
chamber calibration is then followed with a dead animal experiment, identical in every way to
the in vivo exposure, and is meant to account for every factor other than metabolism, which is
zero in the dead animal. When the live animal(s) are placed in the chamber, oxygen is provided
for, and carbon dioxide accumulated during breathing is removed by absorption with a chemical
scrubber. A bolus injection of the parent chemical, TCE, is given and this injection time starts
the inhalation exposure. The chemical inside the chamber will decrease with time, as it is
absorbed by the system and the metabolic process inside the rodent. Since all known processes
contributing to the decline are quantified, except for metabolism, the metabolic parameters can
be extracted from the total chamber concentration decline using regression techniques.
The basic structure for the PBPK model that is linked to closed-chamber inhalation data
has the same basic structure as described before. The one major difference is the inclusion of
one additional equation that accounts for mass balance changes inside the inhalation chamber or
system, and connects the chamber with the inhaled and exhaled concentrations breathed in and
out by the animal:
where
^ =RATS(Or)(Cr -^)-Kioss4,	(Eq A-4)
dt	•' Vct
RATS	= number of animals in the chamber
Qp	= alveolar ventilation rate
Cx	= exhaled concentration
Ach	= net amount of chemical inside chamber
Vch	= volume of chamber
Kf/jss	= loss rate constant to glassware.
An updated model was developed that included updated physiological and
chemical-specific parameters as well as GSH metabolism in the liver and kidney, as discussed in
©
Chapter 3. The PBPK model code was translated from MCSim to use in Matlab
(version 7.2.0.232, R2006a, Natick, MA) using their m language. This PBPK model made use of
fixed or constant, averaged values for physiological, chemical and other input parameters; there
were no statistical distributions attached to each average value. As an additional step in quality
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 7 DRAFT—DO NOT CITE OR QUOTE

-------
1	control, mass balance was checked for the MCSim code, and comparisons across both sets of
2	code were made to ensure that both sets of predictions were the same.
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 8 DRAFT—DO NOT CITE OR QUOTE

-------
1
The resulting simulations were compared to mice gas uptake data (J. W. Fisher et al.,
This document is a draft for review purposes only and does not constitute Agency policy.
A-3 9 DRAFT—DO NOT CITE OR QUOTE

-------
1 1991) after some adjustments of the fat compartment volumes and flows based on visual fits, and
This document is a draft for review purposes only and does not constitute Agency policy.
A-40 DRAFT—DO NOT CITE OR QUOTE

-------
1 limited least-squares optimization of just Vmax (different for males and females) and KM (same
This document is a draft for review purposes only and does not constitute Agency policy.
A-41 DRAFT—DO NOT CITE OR QUOTE

-------
1 for males and females). The results are shown in the top panels of Figures A-3 and A-4, which
This document is a draft for review purposes only and does not constitute Agency policy.
A-42 DRAFT—DO NOT CITE OR QUOTE

-------
1 showed poor fits particularly at lower chamber concentrations. In particular, metabolism is
This document is a draft for review purposes only and does not constitute Agency policy.
A-43 DRAFT—DO NOT CITE OR QUOTE

-------
time (h)
1	observed to be faster than predicted by simulation. This is directly related to metabolism of TCE
2	being limited by hepatic blood flow at these exposures. Indeed, Fisher et al. (1991) was able to
3	obtain adequate fits to these data by using cardiac output and ventilation rates that were about
time (h)
4 twofold higher than is typical for mice. Although their later publication reporting inhalation
This document is a draft for review purposes only and does not constitute Agency policy.
A-44 DRAFT—DO NOT CITE OR QUOTE

-------
4
10
3
10
Q.
Q.
2
10
1
10
o
10
0
time (h)
1
2
3
4	Figure A-3. Limited optimization results for male closed-chamber data from
5	Fisher et al. (1991) without (top) and with (bottom) respiratory metabolism.
4
10
3
10
Cl
Cl
2
10
1
10
o
10
0
time (h)
This document is a draft for review purposes only and does not constitute Agency policy.
A-45 DRAFT—DO NOT CITE OR QUOTE

-------
1	Figure A-4. Limited optimization results for female closed-chamber data
2	from Fisher et al. (1991) without (top) and with (bottom) respiratory
3	metabolism.
4
This document is a draft for review purposes only and does not constitute Agency policy.
A-46 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
experiments (Greenberg et al., 1999) used the lower values from Brown et al. (1997) for these
parameters, they did not revisit the Fisher et al. (1991) data with the updated model. In addition,
the Hack et al. (2006) model estimated the cardiac output and ventilation rate and for these
experiments to be about twofold higher than typical. However, it seems unlikely that
cardiacoutput and ventilation rate were really as high as used in these models, since TCE and
other solvents typically have central nervous system-depressing effects. In the mouse, after the
liver, the lung has the highest rate of oxidative metabolism, as assessed by in vitro methods (see
footnote in Section 3.5.4.2 for a discussion of why kidney oxidative metabolism is likely to be
minor quantitatively). In addition, TCE administered via inhalation is available to the lung
directly, as well as through blood flow. Therefore, it was hypothesized that a more refined
treatment of respiratory metabolism may be necessary to account for the additional metabolism.
The structure of the updated respiratory metabolism model is shown in Figure A-5, with
the mathematical formulation shown in the model code in Section A.6, where the "D" is the
diffusion rate, "concentrations" and "amounts" are related by the compartment volume, and the
other symbols have their standard meanings in the context of PBPK modeling. In brief, this is a
more highly "lumped" version of the Sarangapani et al. (2003) respiratory metabolism model for
styrene combined with a "continuous breathing" model to account for a possible
wash-in/wash-out effect. In brief, upon inhalation (at a rate equal to the full minute volume, not
just the alveolar ventilation), TCE can either (1) diffuse between the respiratory tract lumen and
the respiratory tract tissue; (2) remain in the dead space, or (3) enter the gas exchange region. In
the respiratory tract tissue, TCE can either be "stored" temporarily until exhalation, during which
it diffuses to the "exhalation" respiratory tract lumen, or be metabolized. In the dead space, TCE
is transferred directly to the "exhalation" respiratory tract lumen at a rate equal to the
minute-volume minus the alveolar ventilation rate, where it mixes with the other sources. In the
gas exchange region, it undergoes transfer to and from blood, as is standard for PBPK models of
volatile organics. Therefore, if respiratory metabolism is absent (VMAxClara = 0), then the
model reduces to a wash-in/wash-out effect where TCE is temporarily adsorbed to the
respiratory tract tissue, the amount of which depends on the diffusion rate, the volume of the
tissue, and the partition coefficients.
The results of the same limited optimization, now with additional parameters VMAxClara,
KMClara, and D being estimated simultaneously with the hepatic Vmax and KM, are shown in the
bottom panels of Figures A-2 and A-3. The improvement in the model fits is obvious, and these
results served as a motivation to include this respiratory metabolism model for analysis by the
more formal Bayesian methods.
This document is a draft for review purposes only and does not constitute Agency policy.
A-47 DRAFT—DO NOT CITE OR QUOTE

-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-48 DRAFT—DO NOT CITE OR QUOTE

-------
QM*CInh	QM*CExhResp
D*CResp
D*CResp
D*CInhResp
D*CExhResp
VMaxClara*CResp/(KMClara + CResp)
(QM - QP)*CInhResp
QP*CInhResp
QP*CArt/PB
,QC*CArt
QC*CVen
Respiratory
Tract Tissue
(AResp)
Respiratory
Tract During
Inhalation
(AInhResp)
Respiratory
Tract During
Exhalation
(AExhResp)
Alveolar (Gas Exchange) Region
1	Figure A-5. Respiratory metabolism model for updated PBPK model.
This document is a draft for review purposes only and does not constitute Agency policy.
A-49 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
A.3. DETAILS OF I II I UPDATED PHYSIOLOGICALLY BASED
PHARMACOKINETIC (PBPK) MODEL FOR TRICHLOROETHYLENE (TCE)
AND ITS METABOLITES
The structure of the updated PBPK model and the statistical population model are shown
graphically in Chapter 3, with the model code shown below in Section A.6. Details as to the
model structure, equations, and parameter values and prior distributions are given below.
A.3 .1. Physiologically Based Pharmacokinetic (PBPK) Model Structure and Equations
The equations below, along with the parameters defined in Table A-4, specify the PBPK
model. The ordinary differential equations are shown in bold, with the remaining equations
being algebraic definitions. The same equations are in the PBPK model code, with some
additional provisions for unit conversions (e.g., ppm to mg/L) or numerical stability (e.g.,
truncating small values at 10 l5, so states are never negative). For reference, the stoichiometric
adjustments for molecular weights are given by the following:
# Molecular Weights
NAcDCVC: MWNADCVC = 258.8
# Stoichiometry
StochTCATCE = MWTCA/MWTCE;
StochTCATCOH = MWTCA/MWTCOH;
StochTCOHTCE = MWTCOH/MWTCE;
StochGlucTCOH = MWTCOHGluc/MWTCOH;
StochTCOHGluc = MWTCOH/MWTCOHGluc;
StochTCEGluc = MWTCE/MWTCOHGluc;
StochDCVCTCE = MWDCVC/MWTCE;
StochN = MWNADCVC/MWDCVC;
TCE:
DCVC:
TCA:
TCOH:
TCOG:
MWTCE = 131.39
MWDCVC = 216.1
MWTCA= 163.5
MWTCOH= 149.5
MWTCOHGluc = 325.53
This document is a draft for review purposes only and does not constitute Agency policy.
A-50 DRAFT—DO NOT CITE OR QUOTE

-------
1
Table A-4. PBPK model parameters, baseline values, and scaling relationships
3^
rs
sj
3
a
>
a
&
H
1
o
o
2
o
H
O
HH
H
W
O
P
o
c
o
H
W
>!
ca
*¦
fs*
si
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
Oq
a
o
t
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
BW
Body weight
(kg)
-
BW0
Standard body
weight
0.03
0.3
60/70
-
a
Flows
QC
Cardiac Output
(L/h)
QC = QCC0 x exp(lnQCC)
X BW0-75
QCC0
Cardiac output
allometrically
scaled
11.6
13.3
16/16
InQCC
b
QP
Alveolar
ventilation
(L/h)
QP = QC x VPR0
x exp(lnVPR)
VPR0
Ventilation-
perfusion ratio
2.5
1.9
0.96/0.96
InVPRC
c
DResp
Diffusion
clearance rate
(L/h)
DResp = QP
x exp(lnDRespC)
—
—
—
—
—
lnDRespC
d
Physiological blood flows to tissues
QFat
Blood flow to
fat (L/h)
QFat = QC x QFatC0
x QFatC
QFatC0
Fraction of
blood flow to
fat
0.07
0.07
0.085/0.05
QFatC
e
QGut
Blood flow to
gut(L/h)
QGut = QC x QGutCo
x QGutC
QGutCo
Fraction of
blood flow to
gut
0.141
0.153
0.21/0.19
QGutC
e
QLiv
Hepatic artery
blood flow
(L/h)
QLiv = QC x QLivC0
x QLivC
QLivC0
Fraction of
blood flow to
hepatic artery
0.02
0.021
0.065/0.065
QLivC
e
QSlw
Blood flow to
slowly
perfused
tissues (L/h)
QSlw = QC x QSlwC0
x QSlwC
QSlwC0
Fraction of
blood flow to
slowly
perfused
tissues
0.217
0.336
0.17/0.22
QSlwC
e
QKid
Blood flow to
kidney (L/h)
QKid = QC x QKidCo
x QKidC
QKidCo
Fraction of
blood flow to
kidney
0.091
0.141
0.085/0.05
QKidC
e

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a
to
o
&
H
1
o
o
2
o
H
O
HH
H
W
O
V
o
c
o
H
W
>!
&
*¦
fs*
si
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
Oq
a
o
t
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
QRap
Blood flow to
rapidly
perfused
tissues (L/h)
QRap = QC-(QFat
+ QGut + QLiv + QSlw
+ QKid)




0.21/0.19

e
FracPlas
Fraction of
blood that is
plasma
FracPlas = FracPlas0
x FracPlasC
FracPlaso
Fraction of
blood that is
plasma
0.52
0.53
0.065/0.065
FracPlasC
f
Physiological volumes
Wat
Volume of fat
(L)
VFat = BW x VFatCo
x VFatC
VFatCo
Fraction of
body weight
that is fat
0.07
0.07
0.317/0.199
VFatC
g
VGut
Volume of gut
(L)
VGut = BW x VGutCo
x VGutC
VGutCo
Fraction of
body weight
that is gut
0.049
0.032
0.022/0.02
VGutC
g
VLiv
Volume of
liver (L)
VLiv = BW x VLivC0
x VLivC
VLivC0
Fraction of
body weight
that is liver
0.055
0.034
0.023/0.025
VLivC
g
VRap
Volume of
rapidly
perfused
tissues (L)
VRap = BW x VRapC0
x VRapC
VRapCo
Fraction of
body weight
that is rapidly
perfused
0.1
0.088
0.093/0.088
VRapC
g
VRespLum
Volume of
respiratory
tract lumen (L)
VRespLum = BW
x VRespLumCo
x VRespLumC
VRespLumCo
Respiratory
lumen volume
as fraction
body weight
0.004667
0.004667
0.002386/0.002386
VRespLum
C
g
VResp
Volume of
respiratory
tract tissue (L)
VResp = BW x VRespC0 x
VRespC
VRespC0
Fraction of
body weight
that is
respiratory
tract
0.0007
0.0005
0.00018/0.00018
VRespC
g

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a
Lti
u>
o
H
1
o
o
2
o
H
o
HH
H
W
O
V
o
c
o
H
W
>!
&
*¦
fs*
si
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
Oq
a
o
t
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
VRespEff
Effective air
volume of
respiratory
tract tissue
VRespEff = VResp
x PResp x PB






g
VKid
Volume of
kidney (L)
VKid = BW
x VKidC0 x VKidC
VKidCo
Fraction of
body weight
that is kidney
0.017
0.007
0.0046/0.0043
VKidC
g
VBld
Volume of
blood (L)
VBld = BW x VBldCo
x VBldC
VBldCo
Fraction of
body weight
that is blood
0.049
0.074
0.068/0.077
VBldC
g
VSlw
Volume of
slowly
perfused tissue
(L)
VSlw = BW x VperfC0
- (VFat + VGut + VLiv
+ Wap + VResp + VKid
+ VBld)
VperfC0
Fraction of
body weight
that is blood
perfused
0.8897
0.8995
0.85778/0.8560

g
VPlas
Volume of
plasma (L)
VPlas = FracPlas x VBld
-
-
-
-
-
-
h
VBod
Volume body
for TCA
submodel (L)
VBod = VFat + VGut
+ VRap + VResp + VKid
+ VSlw
—
—
—
—
—
—
i
VBodTCOH
Volume body
for TCOH and
TCOG
submodels (L)
VBodTCOH = VBod
+ VBld






j
TCE distribution/partitioning
PB
TCE blood-air
PC
PB=PB0xPBC
PB0
TCE blood-air
PC
15
22
9.5
PBC
k
PFat
TCE fat-blood
PC
PFat=PFatC0x
exp(PFatC)
PFatCo
TCE fat-blood
PC
36
27
67
PFatC
1
PGut
TCE gut-blood
PC
PGut=(PGutC0)x
exp(lnPGutC)
PGutC0
TCE gut-blood
PC
1.9
1.4
2.6
lnPGutC
m

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a
4^
o
&
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
PLiv
TCE liver-
blood PC
PLiv = (PLivC0)
x exp(lnPLivC)
PLivC0
TCE liver-
blood PC
1.7
1.5
4.1
lnPLivC
n
PRap
TCE rapidly
perfused-blood
PC
PRap = (PRapC0)
x exp(lnPRapC)
PRapC0
TCE rapidly
perfused-blood
PC
1.9
1.3
2.6
lnPRapC
0
PResp
TCE
respiratory
tract tissue-
blood PC
Presp = (PRespC0)
x exp(lnPRespC)
PRespC0
TCE
respiratory
tract tissue-
blood PC
2.6
1.0
1.3
InPRespC
P
PKid
TCE kidney-
blood PC
PKid = (PKidC0)
x exp(lnPKidC)
PKidC0
TCE kidney-
blood PC
2.1
1.3
1.6
lnPKidC
q
PSlw
TCE slowly
perfused-blood
PC
PSlw = (PSlwCo) x
exp(lnPSlwC)
PSlwC0
TCE slowly
perfused-blood
PC
2.4
0.58
2.1
lnPSlwC
r
TCA distribution/partitioning
TCAPlas
TCA blood-
plasma
concentration
ratio
TCAPlas = FracPlas
+ (1 - FracPlas)
x PRBCPlasTCAo
x exp(lnPRBCPlasTCAC)
PRBCPlasTCAo
TCA red blood
cell-plasma
partition
coefficient
0.5
0.5
0.5/0.5
lnPRBCPla
sTCAC
s
PBodTCA
Free TCA
body-plasma
PC
PBodTCA = TCAPlas
x PBodTCAC0
x exp(lnPBodTCAC)
PBodTCACo
Free TCA
body-blood PC
0.88
0.88
0.52
InPBodTC
AC
t
PLivTCA
Free TCA
liver-plasma
PC
PLivTCA = TCAPlas
x PLivTCAC0
x exp(lnPLivTCAC)
PLivTCACo
Free TCA
liver-blood PC
1.18
1.18
0.66
InPLivTC
AC
t
TCA plasma binding
kDissoc
Protein TCA
dissociation
constant
(microM)
kDissoc = kDissoco x
exp(lnkDissocC)
kDissoco
Protein TCA
dissociation
constant
(microM)
107
275
182
InkDissocC
u

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a

a
&
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
o;
H
I
w
z; ^
¦ a
"T3
H S.
W|
O •
p
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
BMax
Protein
concentration
(microM)
BMax = BMaxkD0
x kDissoc
x exp(lnBMaxkDC)
BMaxkD0
BMax/kDissoc
ratio
0.88
1.22
4.62
lnBMaxkD
C
u
TCOH and TCOG distribution/partitioning
PBodTCOH
TCOH body-
blood PC
PBodTCOH
= PBodTCOH0
x exp(lnPBodTCOHC)
PBodTCOH0
TCOH body-
blood PC
1.11
1.11
0.91
InPBodTC
OHC
V
PLivTCOH
TCOH liver-
blood PC
PBodTCOH
= PLivTCOHo
x exp(lnPLivTCOHC)
PLivTCOHo
TCOH liver-
blood PC
1.3
1.3
0.59
InPLivTC
OHC
V
PBodTCOG
TCOG body-
blood PC
PBodTCOG =
PBodTCOGo
x exp(lnPBodTCOGC)
PBodTCOG0
TCOG body-
blood PC
1.11
1.11
0.91
InPBodTC
OGC
w
PLivTCOG
TCOG liver-
blood PC
PBodTCOG = PLivTCOG0
x exp(lnPLivTCOGC)
PLivTCOGo
TCOG liver-
blood PC
1.3
1.3
0.59
InPLivTC
OGC
w
DCVG distribution/partitioning
VDCVG
DCVG
distribution
volume (L)
VDCVG = VBld
+ (VBod+VLiv)
x exp(lnPeffDCVG)
—
—
—
—
—
lnPeffDCV
G
X
TCE metabolism
Vmax
VMAx for TCE
hepatic
oxidation
(mg/h)
Vmax- VMAXo x VLiv
x exp(lnVMAXC)
Vmaxo
VMAx per kg
liver for TCE
hepatic
oxidation
(mg/h/kg liver)
2700
600
255
lnVMAXC
y

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a
o\
O
&
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
KM
KM for TCE
hepatic
oxidation
(mg/L blood)
KM = KM„ x exp(lnKMC)
[Mouse and Rat]
KM0
KM for TCE
hepatic
oxidation
(mg/L)
36
21

InKMC
y


KM=Vmax/(C1C0
x VLiv x exp(lnClC))
[Human]
C1C0
VMAx/KMper
kg liver for
TCE hepatic
oxidation (L
blood/h/kg
liver)


66
InCIC
y
FracOther
Fraction of
TCE oxidation
not to TCA or
TCOH
FracOther
= exp(lnFracOtherC)/
(l+exp(lnFracOtherC))





lnFracOthe
rC
z
FracTCA
Fraction of
TCE oxidation
to TCA
FracTCA = (1-FracOther) x
logitFracTCAo
x exp(lnFracTCAC)/
(1 + logitFracTCA0
x exp(lnFracTCAC))
logitFracTCAo
Log of ratio of
fraction to
TCA to
fraction not to
TCA
0.32
0.32
0.32
lnFracTCA
C
aa
VmaxDCVG
VMAx for TCE
hepatic GSH
conjugation
(mg/h)
VmaxDCVG
= VmaxDCVGq x VLiv
x exp(lnVMAXDCVGC)
[Mouse and Rat]
VmaxDCVGo
VMAx per kg
liver for TCE
GSH
conjugation
(mg/h/kg liver)
300
66

lnVMAxDC
VGC
bb


VMAXDCVG = VLiv
x CIDCVGo
x exp(lnClDCVGC)
x KMDCVGo
x exp(lnKMDCVGC)
[Human]
CIDCVGo
VMAX/KMper
kg liver for
TCE GSH
conjugation (L
blood/h/kg
liver)


19
InClDCVG
C
bb



KMDCVGo
KM for TCE
GSH
conjugation
(mg/L blood)


2.9
InKMDCV
GC
bb

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a

o
&
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
KMDCVG
KM for TCE
hepatic GSH
conjugation
(mg/L blood)
KMDCVG = VmaxDCVG/
(C1DCVG0
x exp(lnClDCVGC)
[Mouse and Rat]
C1DCVG0
VMAX/KMper
kg liver for
TCE hepatic
GSH
conjugation (L
blood/h/kg
liver)
1.53
0.25

InClDCVG
C
bb


KMDCVG = KMDCVG0
x exp(lnKMDCVGC)
[Human]
KMDCVG0
KM for TCE
GSH
conjugation
(mg/L blood)


2.9
lnKMDCV
GC
bb
V MAxKidDC V G
VMAx for TCE
kidney GSH
conjugation
(mg/h)
V MAxKidDC V G
= V MAXKidDC V G0
x VKid
x exp(lnVMAXKidDC VGC)
[Mouse and Rat]
V MAXKidDC V G0
VMAxPerkg
kidney for TCE
GSH
conjugation
(mg/h/kg
kidney)
60
6.0

lnVMAXKid
DCVGC
bb


VMAXKidDCVG = VKid
x ClKidDCVG0
x exp(lnClKidDCVGC)
x KMKidDCVG0
x exp(lnKMKidDCVGC)
[Human]
ClKidDCVG0
VMAX/KMper
kg kidney for
TCE GSH
conjugation (L
blood/h/kg
liver)


230
InClKidDC
VGC
bb



KMKidDCVG0
KM for TCE
GSH
conjugation
(mg/L blood)


2.7
lnKMKidD
CVGC
bb

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a
oo
o
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
KMKidDCVG
KM for TCE
kidney GSH
conjugation
(mg/L blood)
KMKidDCVG
= V MAXKidDC V G/
(ClKidDCVG0
x exp(lnClKidDCVGC)
[Mouse and Rat]
ClKidDCVG0
V MAx/KM per
kg kidney for
TCE kidney
GSH
conjugation (L
blood/h/kg
liver)
0.34
0.026

InClDCVG
C
bb
KMKidDCVG
= KMKidDCVG0
x exp(lnKMKidDCVGC)
[Eluman]
KMKidDCVG0
KM for TCE
GSH
conjugation
(mg/L blood)


2.7
lnKMKidD
CVGC
bb
TCE metabolism (respiratory tract)
KMClara
KM for TCE
lung oxidation
(mg/L air)
KMClara
= exp(lnKMClara)
—
—
—
—
—
—
cc
^MAxClara
VMAx for TCE
lung oxidation
(mg/h)
^MAxClara = VMAX
x VMAXLungLiv0
x exp(lnVMAXLungLivC)
VMAXLungLiv0
Ratio of lung to
liver total
VMax (mg/h
per mg/h)
0.07
0.0144
0.0138/
0.0128
lnVMAXLun
gLivC
cc
FracLungSys
Fraction of
respiratory
oxidation
entering
systemic
circulation
FracLungSys
= exp(lnFracLungSysC)/
(1+exp(lnFracLung Sy sC))





lnFracLung
SysC
dd
TCOH metabolism
vmaxtcoh
Vmax f°r
TCOH
oxidation to
TCA (mg/h)
VmaxTCOH=BW%
x exp(lnVMAXTCOHC)
[Mouse and Rat]
—
—
—
—
—
IkVmaxTC
OHC

VmaxTCOH = BW*
x exp(lnClTCOHC
+ InKMTCOHC)
[Eluman]





InClTCOH
C
InKMTCO
HC

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
rs
sj
3
a

O
&
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
KMTCOH
KM for TCOH
oxidation to
TCA (mg/L
air)
KMTCOH
= exp(lnKMTCOHC)





lnKMTCO
HC

VmaxGIuc
Vmax for
TCOH
glucuroni-
dation (mg/h)
VmaxGIuc = BW*
x exp(lnVMAXGlucC)
[Mouse and Rat]
—
—
—
—
—
lnVMAXGlu
cC

VmaxGIuc = BW%
x exp(lnClGlucC
+ lnKMGlucC)
[Human]





InClGlucC
lnKMGluc
C

KMGluc
KM for TCOH
glucuroni-
dation (mg/L
air)
KMGluc
= exp(lnKMGlucC)





lnKMGluc
C

kMetTCOH
Rate constant
for TCOH
other clearance
(/h)
kMetTCOH = BWJ/4
x exp(lnkMetTCOHC)





InkMetTC
OHC

TCA metabolism/clearance
kUrnTCA
Rate constant
for TCA
excretion to
urine (/h)
kUrnTCA = GFR_BW
exp(lnkUrnT C AC)
x BW/VPlas
GFR_BW
Glomerular
filtration rate
per kg body
weight (L/h/kg)
0.6
0.522
0.108
InkUrnTC
AC
ee
kMetTCA
Rate constant
for other TCA
clearance (/h)
kMetTCA = BWJ/*
x exp(lnkMetTCAC)
—
—
—
—
—
InkMetTC
AC

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
3^
§•
cs
s
3
TO
a
On
O
O
>!
&
*¦
fs*
. si
*3
O
Grj
>!
o
a
a
§•
>!
o
rs
o
a
>!
H
I
o
o
2!

H
o;
a
"T3
H S.
W|
O •
o
c
o
H
W
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
TCOG metabolism/clearance
kBile
Rate constant
for other
TCOG
excretion to
bile (/h)
kBile = BWJ/4
x exp(lnkBileC)





InkBileC

kEHR
Rate constant
for other bile
TCOG
reaborption as
TCOH (/h)
kEHR = BWJ/*
x exp(lnkEHRC)





InkEHRC

kUrnTCOG
Rate constant
for TCOH
excretion to
urine (/h)
kUrnTCOG = GFR BW
exp(lnkUrnTCOGC)
x BW/(VBodTCOH
x PBodTCOG)
GFR_BW
Glomerular
filtration rate
per kg body
weight
(L/h/kg)
0.6
0.522
0.108
InkUrnTC
OGC
ee
DCVG metabolism
kDCVG
Rate constant
for DCVC
formation
from DCVG
(/h)
kDCVG = BWJ/<
x exp(lnkDCVGC)





InkDCVG
C
ff
kNAT
Rate constant
for urinary
excretion of
NAcDCVC
(/h)
kNAT = BWJ/<
x exp(lnkNATC)





InkNATC
gg
kBioact
Rate constant
for other bio-
activation of
DCVC (/h)
kKidBioact = BWJ/4
x exp(lnkKidBioactC)





InkKidBioa
ctC
gg

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
Parameter
Description
(units)
Formula
Baseline value or
parameter
Description
Mouse
Rat
Human F/M
Scaling
parameter
Sources(s)
Oral uptake/transfer coefficients
kTSD
TCE gavage
stomach-
duodenum
transfer
coefficient (/h)
kTSD = exp(lnkTSD)
1.4




InkTSD
hh
kAS
TCE gavage
stomach-
absorption
coefficient (/h)
kAS = exp(lnkAS)
1.4




InkAS
hh
kAD
TCE gavage
duodenum-
absorption
coefficient (/h)
kAD = exp(lnkAD)
0.75




InkAD
hh
kASTCA
TCA stomach
absorption
coefficient (/h)
kASTCA
= exp(lnkASTCA)
0.75
—
—
—
—
InkASTCA
hh
kASTCOH
TCOH
stomach
absorption
coefficient (/h)
kASTCOH
= exp(lnkASTCOH)
0.75




InkASTCO
H
hh
Explanatory note. Unless otherwise noted, the model parameter is obtained by multiplying (1) the "baseline value" (equals one if not specified) times (2) the
scaling parameter (or for those beginning with "In," which are natural-log transformed, exp[lnXX]) times (3) any additional scaling as noted in the second to
last column. Unless otherwise noted, all log-transformed scaling parameters have baseline value of 0 (i.e., exp[lnXX] has baseline value of 1) and all other
scaling parameters have baseline parameters of 1.
aUse measured value if available.
bIf QP is measured, then scale by QP using VPR. Baseline values are from Brown et al. (1997) (mouse and rat) and International Commission on Radiological
Protection (ICRP) Publication 89 (2003) (human).
°Use measured QP, if available; otherwise scale by QC using alveolar VPR. Baseline values are from Brown et al. (1997) (mouse and rat) and ICRP
Publication 89 (2003) (human).
dScaling parameter is relative to alveolar ventilation rate.

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
eFat represents adipose tissue only. Gut is the gastro-intestinal tract, pancreas, and spleen (all drain to the portal vein). Slowly perfused tissue is the muscle and
skin. Rapidly perfused tissue is the rest of the organs, plus the bone marrow and lymph nodes, the blood flow for which is calculated as the difference between
the cardiac output (QC) and the sum of the other blood flows. Baseline values are from Brown et al. (1997) (mouse and rat) and ICRP Publication 89 (2003)
(human).
fThis is equal to 1 minus the hematocrit (measured value used if available). Baseline values from control animals in (Hejtmancik et al., 2002) (mouse and rat)
and ICRP Publication 89 (2003) (human).
8Fat represents adipose tissue only, and the measured value is used, if available. Gut is the gastro-intestinal tract, pancreas, and spleen (all drain to the portal
vein). Rapidly perfused tissue is the rest of the organs, plus the bone marrow and lymph nodes, minus the tracheobronchial region. The respiratory tissue
volume is tracheobronchial region, with an effective air volume given by multiplying by its tissue:air partition coefficient (= tissue:blood times blood:air). The
slowly perfused tissue is the muscle and skin. This leaves a small (10-15% of body weight [BW]) unperfused volume that consists mostly of bone (minus
marrow) and the gastro-intestinal tract contents. Baseline values are from Brown et al. (1997) (mouse and rat) and ICRP Publication 89 (2003) (human), except
for volumes of the respiratory lumen, which are from Sarangapani et al. (2003).
hDerived from blood volume using FracPlas.
'Sum of all compartments except the blood and liver.
JSum of all compartments except the liver.
kMouse value is from pooling Abbas and Fisher (1997) and Fisher et al. (1991). Rat value is from pooling Sato et al. (1977), Gargas et al. (1989), Barton et al.
(1995), Simmons et al. (2002), Koizumi (1989), and Fisher et al. (1989). Human value is from pooling Sato and Nakajima (1979), Sato et al. (1977), Gargas
et al. (1989), Fiserova-Bergerova et al. (1984), Fisher et al. (1998), and Koizumi (1989).
'Mouse value is from Abbas and Fisher (1997). Rat value is from pooling Barton et al. (1995), Sato et al. (1977), and Fisher et al. (1989). Human value is from
pooling Fiserova-Bergerova et al. (1984), Fisher et al. (1998), and Sato et al. (1977).
"Value is the geometric mean of liver and kidney (relatively high uncertainty) values.
"Mouse value is from Fisher et al. (1991). Rat value is from pooling Barton et al. (1995), Sato et al. (1977), and Fisher et al. (1989). Human value is from
pooling Fiserova-Bergerova et al. (1984) and Fisher et al. (1998).
"Mouse value is geometric mean of liver and kidney values. Rat value is the brain value from Sato et al. (1977). Human value is the brain value from Fiserova-
Bergerova et al. (1984).
pMouse value is the lung value from Abbas and Fisher (1997). Rat value is the lung value from Sato et al. (1977). Human value is from pooling lung values
from Fiserova-Bergerova et al. (1984) and Fisher et al. (1998).
qMouse value is from Abbas and Fisher (1997). Rat value is from pooling Barton et al. (1995) and Sato et al. (1977). Human value is from pooling Fiserova-
Bergerova et al. (1984) and Fisher et al. (1998).
'Mouse value is the muscle value from Abbas and Fisher (1997). Rat value is the muscle value from pooling Barton et al. (1995), Sato et al. (1977), and Fisher et
al. (1989). Human value is the muscle value from pooling Fiserova-Bergerova et al. (1984) and Fisher et al. (1998).
"Scaling parameter is the effective partition coefficient between red blood cells and plasma. Thus, the TCA blood-plasma concentration ratio depends on the
plasma fraction. Baseline value is based on the blood-plasma concentration ratio of 0.76 in rats (Schultz, Merdink, Gonzalez-Leon, & Bull, 1999).
'in vitro partition coefficients were determined at high concentration, when plasma binding is saturated, so should reflect the free blood:tissue partition
coefficient. To get the plasma partition coefficient, the partition coefficient is multiplied by the blood:plasma concentration ratio (TCAPlas). In vitro values
were from Abbas and Fisher (1997) in the mouse (used for both mouse and rat) and from Fisher et al. (1998). Body values based on measurements in muscle.

-------
Table A-4. PBPK model parameters, baseline values, and scaling relationships (continued)
"Values are based on the geometric mean of estimates based on data from Lumpkin et al. (2003), Schultz et al. (1999), Templin et al. (1993; 1995), and Yu et al.
(2000). Scaling parameter for BMAX is actually the ratio of BV|AX/kD. which determines the binding at low concentrations.
vData are from Abbas and Fisher (1997) in the mouse (used for the mouse and rat) and Fisher et al. (1998) (human).
wUsed in vitro measurements in TCOH as a proxy, but higher uncertainty is noted.
xThe scaling parameter (only used in the human model) is the effective partition coefficient for the "body" (nonblood) compartment, so that the distribution
volume VDCVG is given by VBld + exp(lnPeffDCVG) x (VBod + VLiv).
yBaseline values have the following units: for VMax, mg/h/kg liver; for KM, mg/L blood; and for clearance (CI), L/h/kg liver (in humans, KM is calculated from
Km = VMAx/(exp(lnClC) x Vliv). Values are based on in vitro (microsomal and hepatocellular preparations) from Elfarra et al. (1998), Lipscomb et al. (1998;
1997, 1998). Scaling from in vitro data based on 32 mg microsomal protein/g liver and 99 x 106 hepatocytes/g liver (Barter et al., 2007). Scaling of KM from
microsomes were based on two methods: (1) assuming microsomal concentrations equal to liver tissue concentrations and (2) using the measured microsome:air
partition coefficient and a central estimate of the blood:air partition coefficient. For KM from human hepatocyte preparations, the measured hepatocyte:air
partition coefficient and a central estimate of the blood:air partition coefficient was used.
zScaling parameter is ratio of "DCA" to "non-DCA" oxidative pathway (where DCA is a proxy for oxidative metabolism not producing TCA or TCOH).
Fraction of "other" oxidation is exp(lnFracOtherC)/(l + exp[lnFracOtherC]).
aaScaling parameter is ratio of TCA to TCOH pathways. Baseline value based on geometric mean of Lipscomb et al. (1998) using fresh hepatocytes and
Bronley-DeLancey et al. (2006) using cryogenically-preserved hepatocytes. Fraction of oxidation to TCA is
(1 -FracOther) x exp(lnFracTCAC)/(l + exp[lnFracTCAC]).
bbBaseline values are based on in vitro data. In the mouse and rat, the only in vitro data are at 1 or 2 mM (L. H. Lash, Visarius, Sail, Qian, & Tokarz, 1998; L. H.
Lash, Xu, Elfarra, Duescher, & Parker, 1995). In most cases, rates at 2 mM were increased over the same sex/species at 1 mM, indicating VMax has not yet
been reached. These data therefore put lower bounds on both Vmax (in units of mg/h/kg tissue) and clearance (in units of L/h/kg tissue), so those are the scaling
parameters used, with those bounds used as baseline values. For humans, data from Lash et al. (1999) in the liver (hepatocytes) and the kidney (cytosol) and
Green et al. (1997) (liver cytosol) was used to estimate the clearance in units of L/h/kg tissue and KM in units of mg/L in blood.
C0Scaling parameter is the ratio of the lung to liver VMax (each in units of mg/h), with baseline values based on microsomal preparations (mg/h/mg protein)
assayed at ~1 mM (T. Green, Mainwaring, et al., 1997), further adjusted by the ratio of lung to liver tissue masses (Brown et al., 1997; Publication 89, ICRP,
2003).
ddScaling parameter is the ratio of respiratory oxidation entering systemic circulation (translocated to the liver) to that locally cleared in the lung. Fraction of
respiratory oxidation entering systemic circulation is exp(lnFracLungSysC)/(l + exp[lnFracLungSysC]).
eeBaseline parameters for urinary clearance (L/h) were based on glomular filtration rate per unit body weight (L/h/kg B W) from Lin (1995), multiplied by the
body weights cited in the study. For TCA, these were scaled by plasma volume to obtain the rate constant (/h), since the model clears TCA from plasma. For
TCOG, these were scaled by the effective distribution volume of the body (VBodTCOH x PBodTCOG) to obtain the rate constant (/h), since the model clears
TCOG from the body compartment.
ffHuman model only.
88Rat and human models only.
^Baseline value for oral absorption scaling parameter are as follows: kTSD and kAS, 1.4/h, based on human stomach half time of 0.5 h; kAD, kASTCA, and
kASTCOH, 0.75/h, based on human small intestine transit time of 4 h (Publication 89, ICRP, 2003). These are noted to have very high uncertainty.
DCVG = S-dichlorovinyl glutathione.

-------

§•
cs
s
3
TO
a
On
o
>!
&
fs*
. si
O
Grj
>!
o
a
a
§•
>!
5
o
rs
o
a
>!
H
I
o
o
2;
o^

H
o;
a
,3
"13
H S.
W|
O •
p
o
c
o
H
W

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
A 3 .1.1. Trichloroethylene (TCE) Sub-Model
The TCE sub-model is a whole-body, flow-limited PBPK model, with gas respiratory
exchange, oral absorption, and metabolizing and non-metabolizing tissues (see Figures A-6 and
A-7).
A.3 .1.1.1. Gas exchange, respiratory metabolism, arterial blood concentration, and
closed-chamber concentrations
For an open-chamber concentrationand a closed-chamber concentration of ACh/VCh, the
rates of change for the amount in the respiratory lumen during inhalation (AInhResp, in mg), the
amount in the respiratory tract tissue (AResp, in mg), and the respiratory lumen during
exhalation (AExhResp, in mg) are given by the following:
d(AInhResp)/dt = (QM x CInh + DResp x (CResp-CInhResp)	(Eq. A-5)
- QM x CInhResp)
d(AResp)/dt = (DResp x (CInhResp + CExhResp - 2	(Eq. A-6)
x CResp) - RAMetLng)
d(AExhResp)/dt = (QM x (CInhResp - CExhResp) + QP	(Eq. A-7)
x (CArt_tmp/PB-CInhResp) + DResp
x (CResp-CExhResp))
where
CInh
QM
CInhResp
CResp
CExhResp
RAMetLng
CArt_tmp
= inhaled concentration (mg/L) = ACh/VCh + Cone
= minute volume (L/h) = QP/0.7
= concentration in respiratory lumen during inhalation (mg/L)
= AInhResp/VRespLum
= concentration in respiratory tract tissue (mg/L)
= AResp/VRespEff
= concentration in respiratory lumen during exhalation (mg/L)
= AExhResp/VRespLum
= rate of metabolism in respiratory tract tissue
= (VMAxClara x CResp)/(KMClara + CResp)
= arterial blood concentration after gas exchange
= (QCxCVen + QPxCInhResp)/(QC + (QP/PB))
This document is a draft for review purposes only and does not constitute Agency policy.
A-65 DRAFT—DO NOT CITE OR QUOTE

-------
1	Because alveolar breath concentrations can include desorption from the respiratory tract tissue,
2	the concentration at the alveolae (CArt_tmp/PB) may not equal the measured concentration in
3	end-exhaled breath. It is therefore assumed that the ratio of the measured end-exhaled breath
4
This document is a draft for review purposes only and does not constitute Agency policy.
A-66 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
Inhaled air
(Clnh)
l
Exhaled air
(CMixEXh)
" QIVTCInh
~	l~)Resp*(r,Resp-
ClnhResp)
Respiratory
Tract Lumen
Inhalation
(AlnhResp)
Respiratory
Tract Tissue
QIVTCMixExh
DResp*(CResp-
CExhResp)
	(AFtesp)	
i Oxidation *
I (VMaxClara, 1
\ _ KM Clara} _ J
I
Respiratory
Tract Lumen
Exhalation
(AExhResp)

	~
Dead space
QP*ClnhResp
r
(QM-QP)*ClnhResp QP*CArt_tmp/PB
' Gas Exchange
QC*CArt_tmp
Intra-
arterial
l _dose_(k!A)_ ^
QC*CVen
QC*CArt
From venous blood	To rest of body
Figure A-6. Sub-model for TCE gas exchange, respiratory metabolism, and
arterial blood concentration.
This document is a draft for review purposes only and does not constitute Agency policy.
A-67 DRAFT—DO NOT CITE OR QUOTE

-------
To gas exchange
From gas exchange
QC*CVen
Venous
Blood
(ABId)
QRap*CVRap
*	
Rapidly
Perfused
(ARap)
QC*CArt
QRap*CArt
¦4	
QSIw*CVSIw
<	
QFat*CVFat
4
Slowly
Perfused
(ASIw)
QSIw*CArt
<	
kAS*
AStom
''wad*
Intra
venous
dose
kl
QFat CArt
*
Fat (AFat)
ADuod
Gut (AGut)
QGut CArt
QGut CVGut
QLiv CArt
(QGut+QLiv)
Liver (ALiv)
CVLiv
QKid CVKid
*
Kidney
(AKid)
Kd CArt
Oral
gavage
(kStom)
Stomach
(AStom)
kTSD*AStom
Duodenum
(ADuod)
Portal Vein
dose (kPV)
f Oxidation ^
v JVMax^KM)_ ,
Conjugation l
! (VMaxKidDCVG, I
j KMKidDCVG) J
Conjugation (
! (VMaxDCVG, i
t KMDCVG) I
This document is a draft for review purposes only and does not constitute Agency policy.
A-68 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Figure A-7 Sub-model for TCE oral absorption, tissue distribution, and
metabolism.
concentration to the concentration in the absence of desorption is the same as the ratio of the rate
of TCE leaving the lumen to the rate of TCE entering the lumen:
CAlv/(CArt_tmp/PB) = (QM x CMixExh)/{(QP x CArt_tmp/PB	(Eq. A-8)
+ (QM-QP) x CInhResp)}
That is, it is assumed that desorption occurs proportionally throughout the "breath." The
concentration of arterial blood entering circulation needs to add the contribution from the
intra-arterial dose (IADose in mg/kg, infused over a time period TChng):
CArt = CArt_tmp + klA/QC	(Eq. A-9)
where
klA = (IADose x BW)/TChng
For closed-chamber experiments, the additional differential equation for the amount in the
chamber (ACh, in mg) is
d(ACh)/dt = Rodents x (QM x CMixExh - QM x ACh/VCh) - kLoss x
Ach	(Eq. A-10)
where rodents is the number of animals in the chamber, and kLoss is the chamber loss rate
(per h).
A 3.1.1.2. Oral absorption to gut compartment
For oil-based gavage, the dose PDose is defined in terms of units of mg/kg, entering the
stomach during a time TChng, with rates of change in the stomach (AStom, in mg) and
duodenum (ADuod, in mg):
d(AStom)/dt = kStom - AStom x (kAS + kTSD) (Eq. A-ll)
This document is a draft for review purposes only and does not constitute Agency policy.
A-69 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
d(ADuod)/dt = (kTSD x AStom) - kAD x ADuod (Eq. A-12)
where
kStom = rate of TCE entering stomach (mg/h) = (PDose x BW)/TChng
Note that there is absorption to the gut from both the stomach and duodenal compartments.
Analogous equations are defined for aqueous gavage, with the expectation that absorption and
transfer coefficients would differ with the different vehicle. In particular, the aqueous gavage
dose PDoseAq is defined in terms of units of mg/kg, entering the stomach during a time TChng,
with rates of change in the stomach (AStomAq, in mg) and duodenum (ADuodAq, in mg):
d(AStomAq)/dt = kStomAq - AStomAq x (kASAq + kTSDAq)(Eq. A-13)
d(ADuodAq)/dt = (kTSDAq x AStomAq) - kADAq x
ADuodAq (Eq. A-14)
where
kStomAq = rate of TCE entering stomach (mg/h) = (PDoseAq x BW)/TChng
For drinking water, the rate Drink is defined in terms of mg/kg-day, and it is assumed that
absorption is direct to the gut:
kDrink = (Drink x BW)/24.0	(Eq. A-15)
Therefore, the total rate of absorption to the gut via oral exposure (RAO, in mg/h) is:
RAO = kDrink + (kAS x AStom) + (kAD x ADuod) + (kASAq	(Eq. A-16)
x AStomAq) + (kADAq x ADuodAq)
The differential equation for the gut compartment (AGut, in mg) is, therefore, given by
d(AGut)/dt = QGut x (CArt - CVGut) + RAO (Eq. A-17)
where
CVGut = concentration in the gut (mg/L) = AGut/VGut/PGut
This document is a draft for review purposes only and does not constitute Agency policy.
A-70 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
A 3.1.1.3. Nonmetabolizing tissues
The differential equations for nonmetabolizing tissues (rapidly perfused, ARap, in mg;
slowly perfused, ASlw, in mg; and fat, AFat, in mg) follow the standard flow-limited form:
d(ARap)/dt = QRap x (CArt - CVRap)
(Eq. A-18)
d(ASlw)/dt = QSlw x (CArt - CVSlw)
(Eq. A-19)
d(AFat)/dt = QFat x (CArt - CVFat)
(Eq. A-20)
where
CVRap = venous blood concentration leaving rapidly perfused issues
= ARap/VRap/PRap
CVSlw = venous blood concentration leaving slowly perfused issues
= ASlw/VSlw/PSlw
CVFat = venous blood concentration leaving fat
= AFat/VFat/PFat
A 3.1.1.4. Liver compartment
The liver has two metabolizing pathways:
Some experiments also had portal vein dosing (PVDose in mg/kg, infused over a time
period TChng), with a rate entering the liver of
RAMetLivl = Rate of TCE oxidation by P450 in liver (mg/h)
= (Vmax x CVLiv)/(KM + CVLiv)
(Eq. A-21)
RAMetLiv2 = Rate of TCE metabolized to S-dichlorovinyl glutathione
(DCVG_ in liver (mg/h)
= (VmaxDCVG x CVLiv) (KMDCVG + CVLiv)	(Eq. A-22)
kPV = (PVDose x BW)/TChng
(Eq. A-23)
This document is a draft for review purposes only and does not constitute Agency policy.
A-71 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
The differential equation for TCE in liver (ALiv, in mg) is thus
d(ALiv)/dt = (QLiv x (CArt - CVLiv)) + (QGut x (CVGut (Eq. A-24)
- CVLiv)) - RAMetLivl - RAMetLiv2 + kPV
where
CVLiv = venous blood concentration leaving liver
= ALiv/VLiv/PLiv
A.3 .1.1.5. Kidney compartment
The kidney has one metabolizing pathway, GSH conjugation:
RAMetKid = Rate of TCE metabolized to DCVG in kidney (mg/h)	(Eq. A-25)
= (VMAxKidDCVG x CVKid)/(KMKidDCVG + CVKid)
The differential equation for TCE in kidney (AKid, in mg) is thus
d(AKid)/dt = (QKid x (CArt - CVKid)) - RAMetKid (Eq. A-26)
where
CVKid = venous blood concentration leaving kidney = AKid/VKid/PKid
A 3.1.1.6. Venous blood compartment
The venous blood compartment (ABld, in mg) has inputs both from the venous blood
exiting tissues as well as from an IV dose (IVDose in mg/kg infused during a time TChng), and
output to the gas exchange region.
d(ABld)/dt = (QFat x CVFat + QGutLiv x CVLiv + QSlw (Eq. A-27)
x CVSlw + QRap x CVRap + QKid x CVKid)
+ klV - CVen x QC
where
klV = IV infusion rate
= (IVDose x BW)/TChng
CVen = concentration in mixed venous blood
= ABld/VBld
This document is a draft for review purposes only and does not constitute Agency policy.
A-72 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
A 3.1.2. Trichloroethanol (TCOH) Sub-Model
The TCOH sub-model is a simplified whole-body, flow-limited PBPK model, with only a
body (ABodTCOH, in mg) and liver (ALivTCOH, in mg) compartment (see Figure A-8).
A 3.1.2.1. Blood concentration
The venous blood concentration, including an IV dose (IVDoseTCOH in mg/kg infused
during a time TChng), is given by
CTCOH = (QBod x CVBodTCOH + QGutLiv	(Eq. A-28)
x CVLivTCOH + kIVTCOH)/QC
where
CVBodTCOH = ABodTCOH/VBodTCOH/PBodTCOH
CVLivTCOH = ALivTCOH/VLiv/PLivTCOH
klVTCOH	= IV infusion rate
= (IVDoseTCOH x BW)/TChng
and the partition coefficients for the body:blood and liver:blood are PBodTCOH and
PLivTCOH, respectively, QGutLiv is the sum of the portal vein and hepatic artery blood flows,
QBod is the remaining blood flow, VLiv is the liver volume, and VBodTCOH is the remaining
perfused volume.
A 3.1.2.2. Body compartment
The rate of change of the amount of TCOH in the body compartment is
d(ABodTCOH)/dt = QBod x (CTCOH - CVBodTCOH) (Eq. A-29)
A 3 .1.2.3. Liver compartment
The liver has three metabolizing pathways:
This document is a draft for review purposes only and does not constitute Agency policy.
A-73 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
RAMetTCOHTCA = Rate of oxidation of TCOH to TCA (mg/h)	(Eq. A-30)
= (VmaxTCOH x CVLivTCOH)/(KMTCOH
+ CVLivTCOH)
RAMetTCOHGluc = Amount of glucuronidation to TCOG (mg/h)	(Eq. A-31)
= (VmaxGIuc x CVLivTC OH)/(KMGluc
+ CVLivTCOH)
RAMetTCOH = Amount of TCOH metabolized to other (e.g., DC A) (Eq. A-3 2)
= kMetTCOH x ALivTCOH
Some experiments also had oral dosing (PODoseTCOH in mg/kg, entering the stomach over a
time TChng):
d(AStomTCOH)/dt = kStomTCOH - AStomTCOH x kASTCOH (Eq. A-
kStomTCOH = (PODoseTCOH x BW)/TChng; (Eq. A-
# TCOH PO dose rate into stomach
kPOTCOH = AStomTCOH x kASTCOH; # TCOH oral absorption rate (mg/h)
In addition, there are three additional sources of TCOH:
Production in the liver from TCE (a fraction of hepatic oxidation) (Eq. A-3 6)
= (1.0 - FracOther - FracTCA) x StochTCOHTCE x RAMetLivl
Production in the lung from TCE (a fraction of lung oxidation)	(Eq. A-3 7)
= (1.0 - FracOther - FracTCA) x StochTCOHTCE
x FracLungSys x RAMetLng
Enterohepatic recirculation (rate kEHR) from TCOG in the bile	(Eq. A-38)
(amount ABileTCOG) = StochTCOHGluc x RARecircTCOG
= StochTCOHGluc x kEHR x ABileTCOG
Note that StochTCOHTCE is the ratio of molecular weights of TCOH and TCE,
StochTCOHGluc is the ratio of molecular weights of TCOH and TCOG, FracOther is the
fraction of TCE oxidation not producing TCA or TCOH, FracTCA is the fraction of TCE
This document is a draft for review purposes only and does not constitute Agency policy.
A-74 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
oxidation producing TCA, and FracLungSys is the fraction of lung TCE oxidation that is
translocated to the liver and not locally cleared.
The differential equation for TCOH in liver (ALivTCOH, in mg) is thus
d(ALivTCOH)/dt = kPOTCOH + QGutLiv x (CTCOH -
CVLivTCOH)	(Eq. A-39)
- RAMetTCOH - RAMetTCOHTCA - RAMetTCOHGluc
+ ((1.0 - FracOther - FracTCA) x StochTCOHTCE
x (RAMetLivl + FracLungSysxRAMetLng))
+ (StochTCOHGluc x RARecircTCOG)
A.3.1.3. Trichloroethanol-Glucuronide Conjugate (TCOG) Sub-Model
The TCOG sub-model is a simplified whole-body, flow-limited PBPK model, with body
(ABodTCOG, in mg), liver (ALivTCOG, in mg), and bile (ABileTCOG) compartments (see
Figure A-9).
A. 3 .1.3 .1. Blood concentration
The venous blood concentration is given by
CTCOG = (QBod x CVBodTCOG + QGutLiv x CVLivTCOG)/QC (Eq. A-
where
CVBodTCOG = ABodTCOG/VBodTCOH/PBodTCOG
CVLivTCOG = ALivTCOG/VLiv/PLivTCOG
This document is a draft for review purposes only and does not constitute Agency policy.
A-75 DRAFT—DO NOT CITE OR QUOTE

-------
__CCTCOHj__J
Oral dose
PODoseTCOH
QBod*
CVBodTCOH
QBod*
CTCOH
(QGut+QLiv)*
CVLivTCOH
(QGut+QLiv)
CTCOH
Liver TCE
Oxidation* l
(1 - FracTCA - ,
_FracOther]	/
Oxidation to TCA
(VMaxTCOH,
KMTCOH)
Glucuronindation
to TCOG [
(VMaxGluc, |
KMGluc)	/
Lung TCE »
Oxidation* 1
FracLungSys !
(1-FracTCA- ,
_ FracOther]_ _ /
Clearance to
Other
(kMetTCOH)
Enterohepatic
Red rculation
(kEHR *
Body
(ABodTCOH)
Stomach
(AStomTCOH)
Liver
(ALivTCOH)
Liver TCE
Oxidation*
(1-FracTCA-
_FracOther]	/
f
1
2
3
4
Enterohepatic
Red rculation
(kEHR
L -ABileTCOG) _ /
Figure A-8. Submodel for TCOH,
This document is a draft for review purposes only and does not constitute Agency policy.
A-76 DRAFT—DO NOT CITE OR QUOTE

-------
Blood
(CJCOG)
QBod*
CVBodTCOG
I TCOG in
I urine
(QGut+QLiv)*
CVLivTCOG
(QGut+QLiv)
CTCOG
I Glucuronindation
I ofTCOH
Enterohepatic
Recirculation (
(kEHR * |
_ABileTCOG} _ /
Body
(ABodTCOG)
Liver
(ALivTCOG)
Bile
(ABileTCOG)
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-77 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Figure A-9. Submodel for TCOG.
and the partition coefficients for the body:blood and liver:blood are PBodTCOG and
PLivTCOG, respectively, QGutLiv is the sum of the portal vein and hepatic artery blood flows,
QBod is the remaining blood flow, VLiv is the liver volume, and VBodTCOH is the remaining
perfused volume.
A 3.1.3.2. Body compartment
The body compartment is flow limited, with urinary excretion rate (mg/h)
RUrnTCOG = kUrnTCOG x ABodTCOG	(Eq. A-41)
So the rate of change of the amount of TCOG in the body compartment is
d(ABodTCOG)/dt = QBod x (CTCOG - CVBodTCOG) -
RUrnTCOG	(Eq. A-42)
Thus, the amount excreted in urine (AUrnTCOG, mg) is given by
d(AUrnTCOG)/dt = RUrnTCOG	(Eq. A-43)
A 3 .1.3 .3. Liver compartment
The liver is flow limited, with one input, glucuronidation of TCOH (defined above in the
TCOH submodel):
StochGlucTCOH x RAMetTCOHGluc	(Eq. A-44)
and one additional output, excretion in bile:
This document is a draft for review purposes only and does not constitute Agency policy.
A-78 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
RBileTCOG = rate of excretion in bile (mg/h) = kBile x ALivTCOG (Eq. A-45)
The rate of change of the amount of TCOG in the liver is, therefore,
d(ALivTCOG)/dt = QGutLiv x (CTCOG - CVLivTCOG)
+ (StochGlucTCOH x RAMetTCOHGluc) - RBileTCOG
(Eq. A-46)
A 3.1.3.4. Bile compartment
The bile compartment has one input, excretion of TCOG in bile from the liver (defined
above) and one output, enterohepatic recirculation to TCOH in the liver (defined above in the
TCOH submodel), with rate of change
A 3.1.4. Trichloroacetic Acid (TCA) Sub-Model
The TCA sub-model is the same as that in Hack et al. (2006), with an error in the plasma
flow to the liver corrected (see Figure A-10). In brief, TCA in plasma is assumed to undergo
saturable plasma
Figure A-10. Submodel for TCA.
d(ABileTCOG)/dt = RBileTCOG - RARecircTCOG;
(Eq. A-47)
This document is a draft for review purposes only and does not constitute Agency policy.
A-79 DRAFT—DO NOT CITE OR QUOTE

-------
Intra-
venous
dose
klVTCA
kllrnTCA*
APIasTCAFree
=~1 TCAir
I urine
J L
(QBodPlas+QGutLivPlas)
CPIasTCABnd
QBodPlas*
CVBodTCAFree
QBodPlas*
CPIasTCAFree
Oral dose
PODoseTCA'
kASTCA*
AStomTCA
QGutLivPlas*
CVLivTCAFree
QGutLivPlas*
CPIasTCAFree
I Other Clearance
I (kMet*ALivTCA)
Liver TCE
Oxidation*
FracTCA
Lung TCE
Oxidation*
FracLungSys
FracTCA
Oxidation ofTCOH'
Body
(ABodTCA)
Stomach
(AStomTCA)
Plasma
(APIasTCA)
Liver
(ALivTCOG)
1
2
3	protein binding. TCA in tissues is assumed to be flow limited, but with the tissue partition
4	coefficient reflecting equilibrium with the free concentration of TCA in plasma.
5
A 3.1.4.1. Plasma binding and concentrations
6	For an IV dose of TCA given by IVDoseTCA (mg/kg during an infusion period of
7	TChng), the rate of the change of the amount of total TCA in plasma (APIasTCA, in mg) is
8
9
This document is a draft for review purposes only and does not constitute Agency policy.
A-80 DRAFT—DO NOT CITE OR QUOTE

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
d(APlasTCA)/dt = klVTCA + (QBodPlas x CVBodTCA)	(Eq. A-48)
+ (QGutLivPlas x CVLivTCA) - (QCPlas x CPlasTCA) - RUrnTCAplas
where
klVTCA
= rate of IV infusion of TCA = (IVDoseTCA x BW)/TChng
QBodPlas
= plasma flow from body = QBod x FracPlas
QGutLivPlas
= plasma flow from liver = (QGut + QLiv) x FracPlas
CVBodTCA
= venous concentration leaving body = CPlasTCABnd +

CVBodTCAFree
CVBodTCAFree
= free venous concentration leaving body

= (AB odTC A/VB od/PB odTC A)
CVLivTCA
= venous concentration leaving liver

= CPlasTCABnd + CVLivTCAFree
CVLivTCAFree
= free venous concentration leaving liver

= (ALivTCA/VLiv/PLivTCA)
QCPlas
= total plasma flow

= QC x FracPlas
RUrnTCAplas
= rate of urinary excretion of TCA from plasma

= kUrnTCA x APlasTCAFree
The free (CPlasTCAFree) and bound (CPlasTCABnd) concentrations are calculated from the
total concentration (CPlasTCA = APlasTCA/VPlas) by solving the equations:
CPlasTCABndMole = BMax x CPlasTCAFreeMole/(kDissoc	(Eq. A-49)
+ CP1 asTC AF reeMol e)
CPlasTCABndMole = CPlasTCAMole - CPlasTCAFreeMole	(Eq. A-50)
Here the suffix "Mole" means that all concentrations are in micromole/L, because BMax and
kDissoc in Table A-4 are given in those units. These lead to explicit solutions of
CPlasTCAFreeMole = (sqrt(a>
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
These concentrations are converted to mg/L (CPlasTCABnd, CPlasTCAFree) by multiplying by
the molecular weight in mg/|imoles. The amount of free TCA in plasma is, thus,
APlasTCAFree = CPlasTCAFree x VPlas.	(Eq. A-52)
Here, VPlas is derived from the blood volume and hematocrit (see Table A-4).
A 3.1.4.2. Urinary excretion
Urinary excretion is modeled as coming from the plasma compartment, so the rate of
change of TCA in urine (AUrnTCA, in mg) is
d(AUrnTCA)/dt = RUrnTCA	(Eq. A-53)
where
RUrnTCA = RUrnTCAplas
For some human data (W. A. Chiu et al., 2007), urinary excretion was only collected during
certain time periods, with data missing in other time periods. Thus, a switch UrnMissing was
defined, which equals 0 during times of urine collection, and one when urinary data are missing.
The total amount of urinary TCA "collected" (AUrnTCA collect, in mg) is, thus, given by
d(AUrnTCA_collect)/dt = (1-UrnMissing) x RUrnTCA	(Eq. A-54)
A 3.1.4.3. Body compartment
The body compartment is flow limited, with the rate of change for the amount of TCA in
the body (ABodTCA, in mg) given by
d(ABodTCA)/dt = QBodPlas x (CPlasTCAFree - CVBodTCAFree) (Eq. A-
A.3.1.4.4. Liver compartment
The rate of change for the amount of TCA in the liver (ALivTCA, in mg) is given by
This document is a draft for review purposes only and does not constitute Agency policy.
A-82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
d(ALivTCA)/dt = QGutLivPlas x (CPlasTCAFree - CVLivTCAFree) (Eq. A-56)
+ (FracTCA x StochTCATCE x (RAMetLivl + FracLungSysxRAMetLng))
+ (StochTCATCOH x RAMetTCOHTCA) - RAMetTCA + kPOTCA
The first term reflects the free TCA in plasma flowing into and out of the liver compartment, the
second term reflects production of TCA from liver (adjusted for molecular weights and fractional
yield of TCA) and lung (adjusted for molecular weights, fraction of lung metabolism
translocated to the liver, and fractional yield of TCA) metabolism of TCE, the third term reflects
production of TCA from TCOH, the fourth term reflects other clearance of TCA from the liver,
and the fifth term reflects absorption from the stomach of TCA. The contribution from liver
metabolism of TCE is adjusted for molecular weights and production of oxidative metabolites
other than TCA. The rate of clearance of TCA is given by
The oral intake rate of TCA (mg/h) includes a one-compartment stomach. So for an oral dose of
PODoseTCA (in mg/kg), occurring over a time TChng, the rate of change of TCA in the stomach
(AStomTCA, in mg) is given by
RAMetTCA = kMetTCA x ALivTCA
(Eq. A-57)
d(AStomTCA)/dt = kStomTCA - AStomTCA x kASTCA
(Eq. A-58)
where
kStomTCA = rate of input into stomach
= (PODoseTCA x BW)/TChng
The rate of absorption into the liver is, thus,
kPOTCA = AStomTCA x kASTCA
(Eq. A-59)
This document is a draft for review purposes only and does not constitute Agency policy.
A-83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
A 3.1.5. Glutathione (GSH) Conjugation Sub-Model
The GSH conjugation sub-model only tracks S-dichlorovinyl glutathione (DCVG),
DCVC, and urinary excretion of NAc-DCVC (see Figure A-l 1).
The rate of change for DCVG (ADCVGmol, in mmoles) depends on production from
TCE in the liver and metabolism to DCVC:
d(ADCVGmol)/dt = RAMetLiv2/MWTCE - RAMetDCVGmol (Eq. A-60)
where
RAMetDCVGmol = rate of metabolism of DCVG to DCVC
= kDCVG x ADCVGmol
Liver
Conjugation
kDCVG*
ADCVGmol
Kidney
Conjugation
kNAT*
ADCVC
Urine
kKidBioact
ADCVC
Bio-
DCVG
(ADCVGmol)
DCVC
(ADCVC)
v lNAc_DCVC}_ j
v _ activation _ j
Figure A-ll. Submodel for TCE GSH conjugation metabolites.
The rate of change of DCVC (ADCVC, in mg) depends on the production from TCE in the
kidney (adjusted for molecular weights), production from DCVG, urinary excretion as
NAc-DCVC (rate constant kNAT), and other bioactivation (rate constant kKidBioact):
d(ADCVC)/dt = RAMetDCVGmol x MWDCVC	(Eq. A-61)
+ RAMetKid x StochDCVCTCE - ((kNAT + kKidBioact) x ADCVC)
where
RAUrnDCVC = Rate of NAcDCVC excretion into urine
= kNAT x ADCVC
This document is a draft for review purposes only and does not constitute Agency policy.
A-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
The rate of change of the amount of NAc-DCVC excreted (AUrnNDCVC, in mg) is given
(adjusted for molecular weights) by
d(AUrnNDCVC)/dt = StochN x RAUrnDCVC	(Eq. A-62)
For the rat model, the DCVG compartment is "turned off by setting kDCVG to an arbitrarily
high value.
A 3.2. Model Parameters and Baseline Values
The multipage Table A-4 below describes all the parameters of the updated PBPK model,
their baseline values (which are used as central estimates in the prior distributions for the
Bayesian analysis), and any scaling relationship used in their calculation. More detailed notes
are included in the comments of the model code (see Section A.6).
A 3 .3. Statistical Distributions for Parameter Uncertainty and Variability
A.3 .3 .1. Initial Prior Uncertainty in Population Mean Parameters
The following multipage Table A-5 describes the initial prior distributions for the
population mean of the PBPK model parameters. For selected parameters, rat prior distributions
were subsequently updated using the mouse posterior distributions, and human prior distributions
were then updated using mouse and rat posterior distributions (see Section A.4.2.2).
A.3.3.2. Interspecies Scaling to Update Selected Prior Distributions in the Rat and Human
As shown in Table A-5, for several parameters, there is little or no in vitro or other prior
information available to develop informative prior distributions, so many parameters had
lognormal or log-uniform priors that spanned a wide range. Initially, the PBPK model for each
species was run with the initial prior distributions in Table A-5, but, in the time available for
analysis (up to about 100,000 iterations), only for the mouse did all these parameters achieve
adequate convergence. Additional preliminary runs indicated replacing the log-uniform priors
with lognormal priors and/or requiring more consistency between species could lead to adequate
This document is a draft for review purposes only and does not constitute Agency policy.
A-85 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
convergence. However, an objective method of "centering" the lognormal distributions that did
not rely on the in vivo data (e.g., via visual fitting or limited optimization) being calibrated
against was necessary in order to minimize potential bias.
Therefore, the approach taken was to consider three species sequentially, from mouse to
rat to human, and to use a model for interspecies scaling to update the prior distributions across
species (the original prior distributions define the prior bounds). This sequence was chosen
because the models are essentially "nested" in this order—the rat model adds to the mouse model
the "downstream" GSH conjugation pathways, and the human model adds to the rat model the
intermediary DCVG compartment. Therefore, for those parameters with little or no independent
data only, the mouse posteriors were used to update the rat priors, and both the mouse and rat
posteriors were used to update the human priors. A list of the parameters for which this scaling
was used to update prior distributions is contained in Table A-6, with the updated prior
distributions. The correspondence between the "scaling parameters" and the physical parameters
generally follows standard practice, and were explicitly described in Table A-4. For instance,
Vmax and clearance rates are scaled by body weight to the 3/4 power, whereas KM values are
assumed to have no scaling, and rate constants (inverse time units) are scaled by body weight to
the -A power.
The scaling model is given explicitly as follows. If 9, are the "scaling" parameters
(usually also natural-log-transformed) that are actually estimated, and A is the "universal"
(species-independent) parameter, then 9, = A + s/, where e, is the species-specific "departure"
from the scaling relationship, assumed to be normally distributed with variance oe . This
"scatter" in the interspecies scaling relationship is assumed to have a standard deviation of
1.15 = ln(3.16), so that the unlogarithmically transformed 95% confidence interval spans about
100-fold (i.e., exp(2o) = 10). This implies that 95% of the time, the species-specific scaling
parameter is assumed be within 10-fold higher or lower than the "species-independent" value.
However, the prior bounds, which generally span a wider range, are maintained so that if the data
strongly imply an extreme species-specific value, it can be accommodated. In addition, the
model transfers the marginal distributions for each parameter across species, so correlations
between parameters are not retained. This is a restriction on the software used for conducting
MCMC analyses, however, assuming independence will lead to a "broader" joint distribution,
given the same marginal distributions. Thus, this assumption tends to reduce the weight of the
interspecies scaling as compared to the species-specific calibration data.
Therefore, the mouse model gives an initial estimate of "A," which is used to update the
prior distribution for 9r = A + sr in the rat (alternatively, since there is only one species at this
stage, one could think of this as estimating the rat parameter using the mouse parameter, but with
This document is a draft for review purposes only and does not constitute Agency policy.
A-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
a cross-species variance is twice the allometric scatter variance). The rat and mouse together
then give a "better" estimate of A, which is used to update the prior distribution for 0/, = A + s/, in
the human, with the assumed distribution for s/,. This approach is implemented by
approximating the posterior distributions by normal distributions, deriving heuristic "data" for
the specific-specific parameters, and then using these pseudo-data to derive updated prior
distributions for the other species parameters. Specifically, the procedure is as follows:
1.	Run the mouse model.
2.	Use the mouse posterior to derive the mouse "pseudo-data" Dm (equal to the posterior
mean) and its uncertainty cm (equal to the posterior variance).
2 2
3.	Use the Dm as the prior mean for the rat. The prior variance for the rat is 2oe + cm ,
which accounts for two components of species-specific departure from
"species-independence" (one each for mouse and rat), and the mouse posterior
uncertainty.
4.	Match the rat posterior mean and variance to the values derived from the normal
2 2	2	2 2	2
approximation (posterior mean = (Dm/(2oe + cm) + Dr/cr }/{l/(2oe + cm ) + l/or };
posterior variance = {l/(2oe2 + o„, ) + l/o,2} ' ), and solve for the rat "data" Dr and its
uncertainty or2.
2	2
5.	Use, cm , and or to derive the updated prior mean and variance for the human model.
For the mean (={Dm/(oe2 + om2) + Dr/(oe2 + or2)}/{ l/(oe2 + cm2) + l/(oe2 + or2)}), it is the
weighted average of the mouse and rat, with each weight including both posterior
uncertainty and departure from "species-independence." For the variance (={ l/(oe
+ <3m) + l/(oe2 + o,2)} 1 + oe2), it is the variance in the weighted average of the mouse
and rat plus an additional component of species-specific departure from
"species-independence."
Formally, then, the probability of 0, given A can be written as
P(0i | A) = (p(0, -A, oe2)	(Eq. A-63)
2	2
where cp(x, o ) is the normal density centered on 0 with variance o . Let D, be a heuristic
"datum" for species /, so the likelihood given 0, is adequately approximated by
P(Pi I 00 = (p(D, - 0,, G 2)	(Eq. A-64)
This document is a draft for review purposes only and does not constitute Agency policy.
A-87 DRAFT—DO NOT CITE OR QUOTE
-------
1	Therefore, considering A to have a uniform prior distribution, then running the mouse model
2	gives a posterior of the form
This document is a draft for review purposes only and does not constitute Agency policy.
A-88 DRAFT—DO NOT CITE OR QUOTE
-------
1
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters
S?
>!
rs
s;
>1
£
>g
00
VO
o
*5
o
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1

H
I
o
o
z;. _
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/
Source
Distribution"
SI) or Mitt
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
Flows
InQCC
TruncNormal
0.2
4
TruncNormal
0.14
4
TruncNormal
0.2
4
a
InVPRC
TruncNormal
0.2
4
TruncNormal
0.3
4
TruncNormal
0.2
4
a
InDRespC
Uniform
-11.513
2.303
Uniform
-11.513
2.303
Uniform
-11.513
2.303
b
Physiological blood flows to tissues
QFatC
TruncNormal
0.46
2
TruncNormal
0.46
2
TruncNormal
0.46
2
a
QGutC
TruncNormal
0.17
2
TruncNormal
0.17
2
TruncNormal
0.18
2
a
QLivC
TruncNormal
0.17
2
TruncNormal
0.17
2
TruncNormal
0.45
2
a
QSlwC
TruncNormal
0.29
2
TruncNormal
0.3
2
TruncNormal
0.32
2
a
QKidC
TruncNormal
0.32
2
TruncNormal
0.13
2
TruncNormal
0.12
2
a
FracPlasC
TruncNormal
0.2
3
TruncNormal
0.2
3
TruncNormal
0.05
3
c
Physiological volumes
WatC
TruncNormal
0.45
2
TruncNormal
0.45
2
TruncNormal
0.45
2
a
VGutC
TruncNormal
0.13
2
TruncNormal
0.13
2
TruncNormal
0.08
2
a
VLivC
TruncNormal
0.24
2
TruncNormal
0.18
2
TruncNormal
0.23
2
a
VRapC
TruncNormal
0.1
2
TruncNormal
0.12
2
TruncNormal
0.08
2
a
VRespLumC
TruncNormal
0.11
2
TruncNormal
0.18
2
TruncNormal
0.2
2
a
VRespEffC
TruncNormal
0.11
2
TruncNormal
0.18
2
TruncNormal
0.2
2
a
VKidC
TruncNormal
0.1
2
TruncNormal
0.15
2
TruncNormal
0.17
2
a
VBldC
TruncNormal
0.12
2
TruncNormal
0.12
2
TruncNormal
0.12
2
a
-------
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters (continued)
S?
>!
rs
s;
>1
£
>g
VO
O
O
*5
o
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1

H
I
o
o
2; . -
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/
Source
Distribution"
SI) or Mitt
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
TCE distribution/partitioning
InPBC
TruncNormal
0.25
3
TruncNormal
0.25
3
TruncNormal
0.2
3
d
InPFatC
TruncNormal
0.3
3
TruncNormal
0.3
3
TruncNormal
0.2
3
InPGutC
TruncNormal
0.4
3
TruncNormal
0.4
3
TruncNormal
0.4
3
InPLivC
TruncNormal
0.4
3
TruncNormal
0.15
3
TruncNormal
0.4
3
InPRapC
TruncNormal
0.4
3
TruncNormal
0.4
3
TruncNormal
0.4
3
InPRespC
TruncNormal
0.4
3
TruncNormal
0.4
3
TruncNormal
0.4
3
InPKidC
TruncNormal
0.4
3
TruncNormal
0.3
3
TruncNormal
0.2
3
InPSlwC
TruncNormal
0.4
3
TruncNormal
0.3
3
TruncNormal
0.3
3
TCA distribution/partitioning
InPRB CPlasTC AC
Uniform
-4.605
4.605
TruncNormal
0.336
3
Uniform
-4.605
4.605
e
InPBodTCAC
TruncNormal
0.336
3
TruncNormal
0.693
3
TruncNormal
0.336
3
f
InPLivTCAC
TruncNormal
0.336
3
TruncNormal
0.693
3
TruncNormal
0.336
3
TCA plasma binding
InkDissocC
TruncNormal
1.191
3
TruncNormal
0.61
3
TruncNormal
0.06
3
g
InBMaxkDC
TruncNormal
0.495
3
TruncNormal
0.47
3
TruncNormal
0.182
3
TCOH and TCOG distribution/partitioning
InPBodTCOHC
TruncNormal
0.336
3
TruncNormal
0.693
3
TruncNormal
0.336
3

InPLivTCOHC
TruncNormal
0.336
3
TruncNormal
0.693
3
TruncNormal
0.336
3

InPBodTCOGC
Uniform
-4.605
4.605
Uniform
-4.605
4.605
Uniform
-4.605
4.605

-------
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters (continued)
S?
>!
rs
s;
>1
£
>g
VO
o
*5
o
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1

H
I
o
o
2; . -
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/
Source
Distribution"
SI) or Mitt
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
InPLivTCOGC
Uniform
-4.605
4.605
Uniform
-4.605
4.605
Uniform
-4.605
4.605

DCVG distribution/partitioning
InPeffDCVG
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
TCE Metabolism
IiiVmaxC
TruncNormal
0.693
3
TruncNormal
0.693
3
TruncNormal
0.693
3
1
lnKMC
TruncNormal
1.386
3
TruncNormal
1.386
3



1
InCIC






TruncNormal
1.386
3
1
InFracOtherC
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
InFracTCAC
TruncNormal
1.163
3
TruncNormal
1.163
3
TruncNormal
1.163
3
J
InVMAxDCVGC
Uniform
-4.605
9.21
Uniform
-4.605
9.21



k
InClDCVGC
Uniform
-4.605
9.21
Uniform
-4.605
9.21
TruncNormal
4.605
3
k
lnKMDCVGC






TruncNormal
1.386
3
k
InV ,AvKidDCVGC
Uniform
-4.605
9.21
Uniform
-4.605
9.21



k
InClKidD C VGC
Uniform
-4.605
9.21
Uniform
-4.605
9.21
TruncNormal
4.605
3
k
lnKMKidDCVGC






TruncNormal
1.386
3
k
InVMAxLungLivC
TruncNormal
1.099
3
TruncNormal
1.099
3
TruncNormal
1.099
3
1
lnKMClara
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
InFracLungSysC
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
-------
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters (continued)
S?
>!
rs
s;
>1
£
>g
VO
to
o
*5
o
>!
>!
O
a
a
a
§•
>!
o
rs
o
a
>1

H
I
o
o
2; . -
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/
Source
Distribution"
SI) or Mitt
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
Distribution
SI) or Min
Truncation
(±nxSD) or
Max
TCOH metabolism
IiiVmaxTCOHC
Uniform
-9.21
9.21
Uniform
-9.21
9.21



h
InClTCOHC






Uniform
-11.513
6.908
MCmTCOH
Uniform
-9.21
9.21
Uniform
-9.21
9.21
Uniform
-9.21
9.21
In V\i\vG1licC
Uniform
-9.21
9.21
Uniform
-9.21
9.21



InClGlucC






Uniform
-9.21
4.605
lnKMGluc
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
InkMetTCOHC
Uniform
-11.513
6.908
Uniform
-11.513
6.908
Uniform
-11.513
6.908
TCA metabolism/clearance
InkUrnTCAC
Uniform
-4.605
4.605
Uniform
-4.605
4.605
Uniform
-4.605
4.605
h
InkMetTCAC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
TCOG metabolism/clearance
InkBileC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
h
InkEHRC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
InkUrnTCOGC
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
DCVG metabolism
InFracKidDCVCC
Uniform
-6.908
6.908
Uniform
-6.908
6.908
Uniform
-6.908
6.908
h
InkDCVGC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
DCVC metabolism/clearance
InkNATC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
h
InkKidBioactC
Uniform
-9.21
4.605
Uniform
-9.21
4.605
Uniform
-9.21
4.605
-------
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters (continued)
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/
Source
Distribution"
SD or Mitt
Truncation
(±nxSD) or
Max
Distribution
SD or Min
Truncation
(±nxSD) or
Max
Distribution
SD or Min
Truncation
(±nxSD) or
Max
Oral uptake/transfer coefficients
InkTSD
Uniform
-4.269
4.942
Uniform
-4.269
4.942
Uniform
-4.269
4.942
h
Ink AS
Uniform
-6.571
7.244
Uniform
-6.571
7.244
Uniform
-6.571
7.244
InkTD
Uniform
-4.605
0
Uniform
-4.605
0
Uniform
-4.605
0
InkAD
Uniform
-7.195
6.62
Uniform
-7.195
6.62
Uniform
-7.195
6.62
InkASTCA
Uniform
-7.195
6.62
Uniform
-7.195
6.62
Uniform
-7.195
6.62
h
InkASTCOH
Uniform
-7.195
6.62
Uniform
-7.195
6.62
Uniform
-7.195
6.62
Explanatory note. All population mean parameters have either truncated normal (TruncNormal) or uniform distributions. For those with TruncNormal
distributions, the mean for the population mean is 0 for natural-log transformed parameters (parameter name starting with "In") and one for untransformed
parameters, with the truncation at the specified number (n) of standard deviations (SD). All uniformly distributed parameters are natural-log transformed, so
their untransformed minimum and maximum are exp(Min) and exp(Max), respectively.
"Uncertainty based on coefficient of variation (CV) or range of values in Brown et al. (1997) (mouse and rat) and a comparison of values from ICRP
Publication 89 (2003), Brown et al. (1997), and Price et al. (2003) (human).
bNoninformative prior distribution intended to span a wide range of possibilities because no independent data are available on these parameters. These priors for
the rat and human were subsequently updated (see Section A.4.2.2).
"Because of potential strain differences, uncertainty in mice and rat assumed to be 20%. In humans, Price et al. (2003) reported variability of about 5%, and this
is also used for the uncertainty in the mean.
dFor partition coefficients, it is not clear whether interstudy variability is due to intersubject or assay variability, so uncertainty in the mean is based on interstudy
variability among in vitro measurements. For single measurements, uncertainty SD of 0.3 was used for fat (mouse) and 0.4 for other tissues was used. In
addition, where measurements were from a surrogate tissue (e.g., gut was based on liver and kidney), an uncertainty SD 0.4 was used.
"Single in vitro data point available in rats, so a geometric standard deviation (GSD) of 1.4 was used. In mice and humans, where no in vitro data was available,
a noninformative prior was used.
fSingle in vitro data points available in mice and humans, so a GSD of 1.4 was used. In rats, where the mouse data was used as a surrogate, a GSD of 2.0 was
used, based on the difference between mice and rats in vitro.
8GSD for uncertainty based on different estimates from different in vitro studies.
hNoninformative prior distribution.
-------
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters (continued)
'Assume twofold uncertainty GSD in VmaX, based on observed variability and uncertainties of in vitro-to-in vivo scaling. For KM and C1C, the uncertainty is
assumed to be fourfold, due to the different methods for scaling of concentrations from TCE in the in vitro medium to TCE in blood.
Uncertainty GSD of 3.2-fold reflects difference between in vitro measurements from Lipscomb et al. (1998) and Bronley-DeLancey et al. (2006).
kIn mice and rats, the baseline values are notional lower-limits on VMax and clearance, however, the lower bound of the prior distribution is set to 100-fold less
because of uncertainty in in vitro-in vivo extrapolation, and because Green et al. (1997) reported values 100-fold smaller than Lash et al. (1998; 1995). In
humans, the uncertainty GSD in clearance is assumed to be 100-fold, due to the difference between Lash et al. (1998) and Green et al. (1997). For KM, the
uncertainty GSD of 4-fold is based on differences between concentrations in cells and cytosol.
'Uncertainty GSD of threefold was assumed due to possible differences in microsomal protein content, the fact that measurements were at a single concentration,
and the fact that the human baseline values was based on the limit of detection.
DCVG = S-dichlorovinyl glutathione; SD = standard deviation.
-------
1	Table A-6. Updated prior distributions for selected parameters in the rat
2	and human
3
Scaling parameter
Initial prior bounds
Updated rat prior
Updated human prior
exp(min)
exp(max)
exp(jt)
exp(o)
exp(n)
exp(o)
InDRespC
1.00E-05
1.00E+01
1.22
5.21
1.84
4.18
InPBodTCOGC
1.00E-02
1.00E+02
0.42
5.47
0.81
5.10
InPLivTCOGC
1.00E-02
1.00E+02
1.01
5.31
2.92
4.31
InFracOtherC
1.00E-03
1.00E+03
0.02
6.82
0.14
4.76
lnVMAxDCVGC
1.00E-02
1.00E+04
2.61
42.52


InClDCVGC
1.00E-02
1.00E+04
0.36
15.03


lnVMAXKidDCVGC
1.00E-02
1.00E+04
2.56
22.65


InClKidD C VGC
1.00E-02
1.00E+04
1.22
15.03


lnVMAxLungLivC
3.70E-02
2.70E+01
2.77
6.17
2.80
4.71
lnKMClara
1.00E-03
1.00E+03
0.01
6.69
0.02
4.85
InFracLungSysC
1.00E-03
1.00E+03
4.39
11.13
3.10
8.08
lnVMAxTCOHC
1.00E-04
1.00E+04
1.65
5.42


InClTCOHC
1.00E-05
1.00E+03


0.37
4.44
lnKMTCOH
1.00E-04
1.00E+04
0.93
5.64
4.81
4.53
lnVMAxGlucC
1.00E-04
1.00E+04
69.41
5.58


InClGlucC
1.00E-04
1.00E+02


3.39
4.35
lnKMGluc
1.00E-03
1.00E+03
30.57
6.11
11.13
4.57
InkMetTCOHC
1.00E-05
1.00E+03
3.35
5.87
2.39
4.62
InkUrnTCAC
1.00E-02
1.00E+02
0.11
5.42
0.09
4.22
InkMetTCAC
1.00E-04
1.00E+02
0.61
5.37
0.45
4.26
InkBileC
1.00E-04
1.00E+02
1.01
5.70
3.39
4.44
InkEHRC
1.00E-04
1.00E+02
0.01
6.62
0.22
4.71
InkUrnTCOGC
1.00E-03
1.00E+03
8.58
6.05
16.12
4.81
InkNATC
1.00E-04
1.00E+02


0.00
6.11
InkKidBioactC
1.00E-04
1.00E+02


0.01
6.49
4
5	Notes: updated rat prior is based on the mouse posterior; and the updated human priors are based on combining the
6	mouse and rat posteriors, except in the case of InkNATC and InkKidBioactC, which are unidentified in the mouse
7	model. Columns labeled exp(min) and exp(max) are the exponentiated prior bounds; columns labeled cxp(|i) and
8	exp(c) are the exponentiated mean and standard deviation of the updated prior distributions, which are normal
9	distributions truncated at the prior bounds.
10
11
This document is a draft for review purposes only and does not constitute Agency policy.
A-95 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
P(A, Qm | Dm ) oc P(0OT | P(Dm | Qm) oc cp(9m - A, ge2) cp(Dm - 0m, cm2) (Eq. A-
From the MCMC posterior, the values of Dm and o,„ are simply the mean and variance of the
scaled parameter 9m.
Now, adding the rat data gives
P(A, em, er I Dm, Dr) oc P(A) P(0m I A) P(Dm | 0„) P(6r | A) P(Dr | 9r) (Eq. A-
OC (p(0m — A, oe ) (p(Dm — 0m, Om ) (p(0r — A, Ge ) cp(Dr - 0r, )
Dr and cr can be derived by marginalizing first over 0„, and then over ^:
J P(A, 0m, 0r | Dm, Dr) d0m cL4	(Eq. A-67)
oc [J P(A) {J P(Qm | P(Dm | 0„) d0m} P(0r | A) dA ]P(Dr | 0r)
= [J P(A) P(Dm | A) P(0r | A) dA] P(Dr | 0r)
oc [J P(A | Dm) P(Qr | ^4) cL4] .P(Dr | 0r)
= P(0r | Dm) P(Dr | 0r)
So P(9r | Dm) can be identified as the prior for 0r based on the mouse data, and P(Dr | 0r) as the
rat-specific likelihood. The updated prior for the rats is then
P(6r I Dm) oc J (p(0m -A, Ge2) (p(D m ~ Qm, Cm) (p(6r " A, Og2) d0m 
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
This distribution is also normal with
E(Qr) = {Dm/(2oe2 + cm2) + Dr/or2}/{ l/(2oe2 + cm2) + l/or2}	(Eq. A-70)
= weighted mean of Dr
VAR(6r) = {l/(2oe2 + om2) + l/or2} 1 (Eq. A-
= harmonic mean of variances
Thus, using the mean and variance of the posterior distribution from the MCMC analysis,
Dr and or can be derived.
2	2
Now, Dm, g,„ , Dr, and cr are known, so the analogous "mouse + rat" based prior used in
the human model can be derived. As with the rat prior, the human prior is based on a normal
approximation of the posterior for A, and then incorporates a random term for cross-species
variation (allometric scatter).
P(A, Qm, 9r, 9/j | Dm, Dr, Dh)	(Eq. A-72)
oc P(A) P(Qm | A) P(Dm | 9m) P(9r | A) P(Dr | 9r) P(9/, | A) P(Dh | 9/,)
qc cp(9m - A, oe) (p(Dm — Qm, <5m ) cp(9r - A, oe) cp(Dr — 9r, or)
cp(9h - A, ge2) cp(D/, - 9/2, g/,2)
Consider marginalizing first over 9m, then over 9r, and then over A:
J P(A, 9m, 9r, 9a | Dm, Dr, Dh) d9m d9r cL4	(Eq. A-73)
oc [J P(A) {J P(Qm | P(Dm | Qm) d9m} {J P(9r | A) P(Dr \ 9r) d9r} P(Qh \ A) dA
P(D/21 9h)
= [J P(A) P(Dm | A) P(Dr | A) P(9/21 A) dA ] P(D/21 9/,)
« [J P(A | DmDr) P(Qh | A) dA] P(Bh \ 9/0
= P(Qh | Dm Dr) P(Dh | 9/,)
So P(Qh I Dm Dr) is the prior for 9/, based on the mouse and rat data, and P(D/, | 9/,) as the
human-specific likelihood. The prior is used in the MCMC analysis for the humans, and it is
derived to be
P(Qh | Dm Dr) oc J (p(Qm - A, Ge2) cp(Dm - 9m, Gm) cp(0r - A, g£2) cp(Dr - 9r, Gr2) (Eq. A-74)
cp(9 h -A, ge ) d9m d9r dA
= J [(p(Dm -A, Ge2 + Qm) (p(D r - A, Gg2 + Gr2)] (p(9/, Gg2) (L4
QC J (p(Dm+r A7 Om+r ) ^p(9h A, Gg ) (L4
(p(Dm+r — 9/2, Gm+r ~l~ Ge )
This document is a draft for review purposes only and does not constitute Agency policy.
A-97 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
2
where Dm+r and cm+r are the weighted mean and variances of A under the density
[cp(D~ A, ge2 + cm2) cp(Dr -A, ge2 + or2)]	(Eq. A-75)
which is given by
Dm+r = E(A\Dm Dr) = {Dm/(ae2 + cm2) + Dr/(ce2 + cr2)}/{ 1/(ge2 + cm2) + 1/(ge2 + Gr2)}
= weighted mean of Dm and Dr
Gm+r2 = VAR(y4| D,„ Dr) = { 1/(ge2 + Gm2) + 1/(ge2 + Gr2)}-1
= harmonic mean of variances
At this point, these values are used in the normal approximation of the combined rodent
posterior, which will be incorporated into the cross-species extrapolation as described in Step 5
above.
The results of these calculations for the updated prior distributions, are shown in
Table A-6. With this methodology for updating the prior distributions, adequate convergence
was achieved for the rat and human after 110,000 ~ 140,000 iterations.
A.3.3.3. Population Variance: Prior Central Estimates and Uncertainty
The following multipage Table A-7 describes the uncertainty distributions used for the
population variability in the PBPK model parameters.
A.3.3.4. Likelihood Function and Prior distributions for Residual Error Estimates
From Equation A-3 for the total likelihood function, different measurement types may
have different partial likelihoods. In all cases except one, the likelihood was assumed to be
lognormal, with probability density for a particular measurement^/ at time tyki given by
P(yijki | 0i, Gp2 , tijki) = (2no2y% exp[{- In ypi - In fyki(0i, tijki)} 2/(2Gyk2)] - (Eq. A-76)
As before, the subject is labeled z, the study is labeled /, the type of measurement is labeled &,
and the different time points are labeled /. The parameters 0, are the "scaling parameters" at the
subject-level, shown in Table A-4, whereas the parameters g^ represent the "residual error"
variance g . This error term may include variability due to measurement error, intrasubject and
intrastudy heterogeneity, as well as model misspecification. The available in vivo measurements
This document is a draft for review purposes only and does not constitute Agency policy.
A-98 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
to which the model was calibrated are listed in Table A-8. The variances for each of the
corresponding residual errors were given log-uniform distributions. For all measurements, the
bounds on the log-uniform distribution was 0.01 and 3.3, corresponding to geometric standard
deviations bounded by 1.11 and 6.15. The lower bound was set to prevent "over-fitting," as was
done in Bois (2000b) and Hack et al. (2006).
Nondetects (ND) of DCVG from Lash et al. (1999) were also included in the data, at it
was found that these data were needed to place constraints on the clearance rate of DCVG from
blood. The detection limit reported in the study was LD = 0.05 pmol/mL= 5 x 10~5 mmol/L. It
was assumed, as is standard in analytical chemistry, that the detection limit represents a response
from a blank sample at 3-standard deviations. Because detector responses near the detection
limit are generally normally distributed, the likelihood for observing a nondetect given a
model-predicted value off,/JO,, tyu) is equal to
P(= ND| e, tyki) = 0(3 x {1 -fyk!(e, tiJkl)/LD),	(Eq. A-77)
This document is a draft for review purposes only and does not constitute Agency policy.
A-99 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-7. Uncertainty distributions for the population variance of the
2	PBPK model parameters
3
Scaling (sampled)
parameter
Mouse
Rat
Human

CV
cu
CV
CU
CV
CU
Notes/source
Flows
InQCC
0.2
2
0.14
2
0.2
2
a
InVPRC
0.2
2
0.3
2
0.2
2

InDRespC
0.2
0.5
0.2
0.5
0.2
0.5

Physiological blood flows to tissues
QFatC
0.46
0.5
0.46
0.5
0.46
0.5
a
QGutC
0.17
0.5
0.17
0.5
0.18
0.5

QLivC
0.17
0.5
0.17
0.5
0.45
0.5

QSlwC
0.29
0.5
0.3
0.5
0.32
0.5

QKidC
0.32
0.5
0.13
0.5
0.12
0.5

FracPlasC
0.2
0.5
0.2
0.5
0.05
0.5

Physiological volumes
VFatC
0.45
0.5
0.45
0.5
0.45
0.5
a
VGutC
0.13
0.5
0.13
0.5
0.08
0.5

VLivC
0.24
0.5
0.18
0.5
0.23
0.5

VRapC
0.1
0.5
0.12
0.5
0.08
0.5

VRespLumC
0.11
0.5
0.18
0.5
0.2
0.5

VRespEffC
0.11
0.5
0.18
0.5
0.2
0.5

VKidC
0.1
0.5
0.15
0.5
0.17
0.5

VBldC
0.12
0.5
0.12
0.5
0.12
0.5

TCE distribution/partitioning
InPBC
0.25
2
0.25
0.333
0.185
0.333
b
InPFatC
0.3
2
0.3
0.333
0.2
1

InPGutC
0.4
2
0.4
2
0.4
2

InPLivC
0.4
2
0.15
0.333
0.4
1.414

InPRapC
0.4
2
0.4
2
0.4
2

InPRespC
0.4
2
0.4
2
0.4
2

InPKidC
0.4
2
0.3
0.577
0.2
1.414

InPSlwC
0.4
2
0.3
0.333
0.3
1.414

4
This document is a draft for review purposes only and does not constitute Agency policy.
A-100 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-7. Uncertainty distributions for the population variance of the PBPK
model parameters (continued)
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/sour
ce
CV
cu
CV
CU
CV
CU
TCA distribution/partitioning
InPRB CPlasT C AC
0.336
2
0.336
2
0.336
2
c
InPBodTCAC
0.336
2
0.693
2
0.336
2
b
InPLivTCAC
0.336
2
0.693
2
0.336
2
TCA plasma binding
InkDissocC
1.191
2
0.61
2
0.06
2
b
InBMaxkDC
0.495
2
0.47
2
0.182
2
TCOH and TCOG distribution/partitioning
InPBodTCOHC
0.336
2
0.693
2
0.336
2
b
InPLivTCOHC
0.336
2
0.693
2
0.336
2
b
InPBodTCOGC
0.4
2
0.4
2
0.4
2
d
InPLivTCOGC
0.4
2
0.4
2
0.4
2
d
DCVG distribution/partitioning
InPeffDCVG
0.4
2
0.4
2
0.4
2
b
TCE metabolism
lnVMAxC
0.824
1
0.806
1
0.708
0.26
e
lnKMC
0.270
1
1.200
1


InCIC




0.944
1.41
InFracOtherC
0.5
2
0.5
2
0.5
2
f
InFracTCAC
0.5
2
0.5
2
1.8
2
g
lnVMAxDCVGC
0.5
2
0.5
2


f
InClDCVGC
0.5
2
0.5
2
0.5
2
lnKMDCVGC




0.5
2
lnVMAxKidDCVGC
0.5
2
0.5
2


InClKidD C VGC
0.5
2
0.5
2
0.5
2
lnKMKidDCVGC




0.5
2
lnVMAxLungLivC
0.5
2
0.5
2
0.5
2
lnKMClara
0.5
2
0.5
2
0.5
2
InFracLungSysC
0.5
2
0.5
2
0.5
2
TCOH metabolism
lnVMAxTCOHC
0.5
2
0.5
2


f
InClTCOHC




0.5
2
lnKMTCOH
0.5
2
0.5
2
0.5
2
lnVMAxGlucC
0.5
2
0.5
2


InClGlucC




0.5
2
lnKMGluc
0.5
2
0.5
2
0.5
2
This document is a draft for review purposes only and does not constitute Agency policy.
A-101 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Table A-7. Uncertainty distributions for the population variance of the PBPK
model parameters (continued)
Scaling (sampled)
parameter
Mouse
Rat
Human
Notes/sour
ce
CV
CU
CV
CU
CV
CU
InkMetTCOHC
0.5
2
0.5
2
0.5
2


TCA metabolism/clearance
InkUrnTCAC
0.5
2
0.5
2
0.5
2
f
InkMetTCAC
0.5
2
0.5
2
0.5
2
TCOG metabolism/clearance
InkBileC
0.5
2
0.5
2
0.5
2
f
InkEHRC
0.5
2
0.5
2
0.5
2
InkUrnTCOGC
0.5
2
0.5
2
0.5
2
f
DCVG metabolism/clearance
InFracKidDCVCC
0.5
2
0.5
2
0.5
2
f
InkDCVGC
0.5
2
0.5
2
0.5
2
DCVC metabolism/clearance
InkNATC
0.5
2
0.5
2
0.5
2
f
InkKidBioactC
0.5
2
0.5
2
0.5
2
Oral uptake/transfer coefficients
InkTSD
2
2
2
2
2
2
h
InkAS
2
2
2
2
2
2
InkTD
2
2
2
2
2
2
Ink AD
2
2
2
2
2
2
InkASTCA
2
2
2
2
2
2
InkASTCOH
2
2
2
2
2
2
Explanatory note. All population variance parameters (Vpname, for parameter "pname") have Inverse-Gamma
distributions, with the expected value given by CV and coefficient of uncertainty given by CU (i.e., standard
deviation of V_pname divided by expected value of V_pname) (notation the same as Hack et al. (2006)). Under
these conditions, the Inverse-Gamma distribution has a shape parameter is given by a = 2 + 1/CU2 and scale
parameter p = (al - 1) CV2. In addition, it should be noted that, under a normal distribution and a uniform prior
distribution on the population variance, the posterior distribution for the variance given n data points with a sample
variance s2 is given by and Inverse-Gamma distribution with a = (n - l)/2 and p = a s2. Therefore, the "effective"
number of data points is given by n = 5 + 2/CU2 and the "effective" sample variance is s2 = CV2 acoriax/(a - 1).
aFor physiological parameters, CV values generally taken to be equal to the uncertainty SD in the population mean,
most of which were based on variability between studies (i.e., not clear whether variability represents uncertainty
or variability). Given this uncertainty, CU of 2 assigned to cardiac output and ventilation-perfusion, while CU of
0.5 assigned to the remaining physiological parameters.
This document is a draft for review purposes only and does not constitute Agency policy.
A-102 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
Table A-7. Uncertainty distributions for the population variance of the
PBPK model parameters (continued)
bAs discussed above, it is not clear whether interstudy variability is due to intersubject or assay variability, so the
same central were assigned to the uncertainty in the population mean as to the central estimate of the population
variance. In the cases were direct measurements were available, the CU for the uncertainty in the population
variance is based on the actual sample n. with the derivation discussed in the notes preceding this table. Otherwise,
a CU of 2 was assigned, reflecting high uncertainty.
cUsed value from uncertainty in population in mean in rats for all species with high uncertainty.
dNo data, so assumed CV of 0.4 with high uncertainty.
eFor mice and rats, based on variability in results from Lipscomb et al. (1998) and Elfarra et al. (1998) in
microsomes. Since only pooled or mean values are available, CU of one was assigned (moderate uncertainty). For
humans, based on variability in individual samples from Lipscomb et al. (1997) (microsomes), Elfarra et al. (1998)
(microsomes) and Lipscomb et al. (1998) (freshly isolated hepatocytes). High uncertainty in clearance (InCIC)
reflects two different methods for scaling concentrations in microsomal preparations to blood concentrations:
(1) assuming microsomal concentration equals liver concentration and then using the measured livenblood partition
coefficient to convert to blood and (2) using the measured microsome:air partition coefficient and then using the
measured blood:air partition coefficient to convert to blood.
fNo data on variability, so a CV of 0.5 was assigned, with a CU of 2.
gFor mice and rats, no data on variability, so a CV of 0.5 was assigned, with a CU of 2. For humans, sixfold
variability based on in vitro data from Bronley-DeLancy et al. (2006), but with high uncertainty.
hNo data on variability, so a CV of two was assigned (larger than assumed for metabolism due to possible vehicle
effects), with a CU of two.
This document is a draft for review purposes only and does not constitute Agency policy.
A-103 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
Table A-8. Measurements used for calibration
Measurement
abbreviation
Mouse
Rat
Human
Measurement description
RetDose


V
Retained TCE dose (mg)
CAlvPPM


V
TCE concentration in alveolar air (ppm)
CInhPPM
V
V

TCE concentration in closed-chamber (ppm)
Cart

V

TCE concentration in arterial blood (mg/L)
CVen
V
V
V
TCE concentration in venous blood (mg/L)
CBldMix
V
V

TCE concentration in mixed arterial and venous blood (mg/L)
CFat
V
V

TCE concentration in fat (mg/L)
CGut

V

TCE concentration in gut (mg/L)
CKid
V
V

TCE concentration in kidney (mg/L)
CLiv
V
V

TCE concentration in liver (mg/L)
CMus

V

TCE concentration in muscle (mg/L)
AExhpost
V
V

Amount of TCE exhaled postexposure (mg)
CTCOH
V
V
V
Free TCOH concentration in blood (mg/L)
CLivTCOH
V


Free TCOH concentration in liver (mg/L)
CPlasTCA
V
V
V
TCA concentration in plasma (mg/L)
CBldTCA
V
V
V
TCA concentration in blood (mg/L)
CLivTCA
V
V

TCA concentration in liver (mg/L)
AUrnTCA
V
V
V
Cumulative amount of TCA excreted in urine (mg)
AUrnTCA_collect


V
Cumulative amount of TCA collected in urine (noncontinuous sampling)
(mg)
ABileTCOG

V

Cumulative amount of bound TCOH excreted in bile (mg)
CTCOG

V

Bound TCOH concentration in blood (mg/L)
CTCOGTCOH
V


Bound TCOH concentration in blood in free TCOH equivalents (mg/L)
CLivTCOGTCOH
V


Bound TCOH concentration in liver in free TCOH equivalents (mg/L)
AUrnTCOGTCOH
V
V
V
Cumulative amount of total TCOH excreted in urine (mg)
AUrnTCOGTCOH_col
lect


V
Cumulative amount of total TCOH collected in urine (noncontinuous
sampling) (mg)
CDCVGmol


V
DCVG concentration in blood (mmol/L)
CDCVG_ND


V
DCVG nondetects from Lash et al. (1999)
AUrnNDCVC

V
V
Cumulative amount of NAcDCVC excreted in urine (mg)
AUrnTCT otMole

V

Cumulative amount of TCA+total TCOH excreted in urine (mmol)
TotCTCOH
V
V
V
Total TCOH concentration in blood (mg/L)
3
4
This document is a draft for review purposes only and does not constitute Agency policy.
A-104 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
where
O(y) is the cumulative standard normal distribution.
The rat and human models differed from mouse model in terms of the hierarchical
structure of the residual errors. In the mouse model, all the studies were assumed to have the
same residual error, as shown in Figure A-l, so that the residual error is only indexed by k, the
type of measurement: Ck • This appeared reasonable because there were fewer studies, and there
appeared to be less variation between studies. In the rat and human models, each of which used
a much larger database of in vivo studies, residual errors were assumed to be the same within a
study, but may differ between studies, and so are labeled by study j and the type of measurement
k: Gjk . The updated hierarchical structures are shown in Figure A-12. Initial attempts to use a
single set of residual errors led to large residual errors for some measurements, even though fits
to many studies appeared reasonable. Residual errors were generally reduced when
study-specific errors were used, except for some data sets that appeared to be outliers (discussed
below).
A 3.4. Summary of Bayesian Posterior Distribution Function
As described in Section A. 1, the posterior distribution for the unknown parameters is
obtained in the usual Bayesian manner by multiplying:
(1)	The prior distributions for the population mean of the scaling parameter^), (see
Sections A.4.3.1-A.4.3.2), the population variance of the scaling parameters(E ), (see
Section A.4.3.3), and the "residual" error (o2), (see Section A.4.3.4);
(2)	The population distribution, assumed to be a truncated normal distribution, for the subject
parameters (9 | E2); and
(3)	The likelihood functions (y | 9, o2), (see Section A.4.3.4)
as follows:
(9, m, I2, o2 | y) x (n)(S2) (a2) (9 | |i, I2) (y | 9, a2) (Eq. A-78)
Each subject's parameters 9, have the same sampling distribution (i.e., they are independently
and identically distributed), so their joint prior distribution is
(9 I n, S2) = n (0i I H, 22)	(Eq. A-79)
This document is a draft for review purposes only and does not constitute Agency policy.
A-105 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
This document is a draft for review purposes only and does not constitute Agency policy.
A-106 DRAFT—DO NOT CITE OR QUOTE
-------
1
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Rat
Human
	~y ijkl/M	
Yijki
Experiment j
Subject i
Population
—~ y ijkl/<	
Experiment j
Subject i
Study m
km
Population
Figure A-12. Updated hierarchical structure for rat and human models.
Symbols have the same meaning as Figure A-l, with modifications for the rat and
human. In particular, in the rat, each "subject" consists of animals (usually
comprising multiple dose groups) of the same sex, species, and strain within a
study (possibly reported in more than one publication, but reasonably presumed to
be of animals in the same "lot"). Animals within each subject are presumed to be
"identical," with the same PBPK model parameters, and each such subject is
assigned its own set of "residual" error variances o In humans, each "subject"
is a single person, possibly exposed in multiple experiments, and each subject is
assigned a set of PBPK model parameters drawn from the population. However,
in humans, "residual" error variances are assigned at an intermediate level of the
hierarchy—the "study" level, o km—rather than the subject or the population
level.
This document is a draft for review purposes only and does not constitute Agency policy.
A-107 DRAFT—DO NOT CITE OR QUOTE
-------
1	Different experiments j = 1 may have different exposure and different data collected and
2	different time points. In addition, different types of measurements k = l...rik (e.g., TCE blood,
3	TCE breath, TCA blood, etc.) may have different errors, but errors are otherwise assumed to be
4	independently and identically distributed. Because the subjects are treated as independent given
5	9ithe likelihood function is simply
6
7
8	y | 0, O ) PX=1...h PL 1 ¦¦¦/*»/ X\l 1.../?? n/ ' ..Mjk(}'ilkl | 0*s ®ijk tijkl)	(EQ- A-80)
9
10	where n is the number of subjects, riy is the number of experiments in that subject, m is the
11	number of different types of measurements, Nyk is the number (possibly 0) of measurements
12	(e.g., time points) for subject i of type k in experiment /, and t^u are the times at which
13	measurements for subject i of type k were made in experiment j.
14
15
16	The MCSim software (version 5.0.0) was used to sample from this distribution.
17
A.4. RESULTS OF UPDATED PHYSIOLOGICALLY BASED PHARMACOKINETIC
(PBPK) MODEL
18	The evaluation of the updated PBPK model was discussed in Chapter 3. Detailed results
19	in the form of tables and figures are provided in this section.
20
A.4.1. Convergence and Posterior Distributions of Sampled Parameters
21	For each sampled parameter (population mean and variance and the variance for residual
22	errors), summary statistics (median, [2.5, 97.5%] confidence interval) for the posterior
23	distribution are tabulated in Tables A-9 through A-14 below. In addition, the potential scale
24	reduction factor R, calculated from comparing four independent chains, is given. For each
25	species, graphs of the prior and posterior distributions for the population mean and variance
26	parameters are shown in Figures A-13 to A-18 for mice, A-19 to A-24 for rats, and A-25 to A-30
27	for humans. Finally, posterior correlations between population mean parameters are given in
28	Tables A-l 1, A-14, and A-17, which show parameter pairs with correlation coefficients >0.25.
29
30	In addition, posterior distributions for the subject-specific parameters are summarized in
31	supplementary figures accessible here:
32
33
This document is a draft for review purposes only and does not constitute Agency policy.
A-108 DRAFT—DO NOT CITE OR QUOTE
-------
Mouse: ("Supplementary data for TCE assessment: Mouse posteriors by subject," 2011)
Rat: ("Supplementary data for TCE assessment: Rat posterior by subject," 2011)
Human: ("Supplementary data for TCE assessment: Human posteriors by subject," 2011)
This document is a draft for review purposes only and does not constitute Agency policy.
A-109 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-9. Posterior distributions for mouse PBPK model population parameters
2
Sampled parameter"
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InQCC
1.237(0.8972, 1.602)
1
1.402 (1.183,2.283)
1
InVPRC
0.8076 (0.6434, 1.022)
1
1.224 (1.108, 1.63)
1.001
QFatC
1.034 (0.5235, 1.55)
1
0.436 (0.3057, 0.6935)
1
QGutC
1.183 (1.002, 1.322)
1
0.1548 (0.1101,0.2421)
1
QLivC
1.035 (0.8002, 1.256)
1
0.1593 (0.1107,0.2581)
1
QSlwC
0.9828 (0.6043, 1.378)
1
0.275 (0.1915,0.4425)
1
InDRespC
1.214(0.7167, 2.149)
1.002
1.215 (1.143, 1.375)
1
QKidC
0.995 (0.5642, 1.425)
1
0.3001 (0.21, 0.48)
1
FracPlasC
0.8707 (0.5979, 1.152)
1.001
0.1903 (0.1327,0.3039)
1
VFatC
1.329 (0.8537, 1.784)
1.002
0.4123 (0.2928, 0.6414)
1
VGutC
0.9871 (0.817, 1.162)
1
0.1219 (0.085,0.1965)
1
VLivC
0.8035 (0.5609, 1.093)
1.013
0.2216 (0.1552,0.3488)
1
VRapC
0.997 (0.8627, 1.131)
1
0.09384 (0.06519,0.1512)
1
VRespLumC
0.9995 (0.8536, 1.145)
1
0.1027 (0.07172,0.1639)
1
VRespEffC
1 (0.8537, 1.148)
1.001
0.1032(0.07176,0.1652)
1
VKidC
1.001 (0.8676, 1.134)
1
0.09365 (0.06523,0.1494)
1
VBldC
0.9916 (0.8341, 1.153)
1.001
0.1126 (0.07835,0.1817)
1
InPBC
0.9259 (0.647, 1.369)
1
1.644 (1.278, 3.682)
1
InPFatC
0.9828 (0.7039, 1.431)
1.001
1.321 (1.16,2.002)
1.001
InPGutC
0.805 (0.4735, 1.418)
1
1.375 (1.198,2.062)
1
InPLivC
1.297 (0.7687, 2.039)
1
1.415 (1.21,2.342)
1
InPRapC
0.9529 (0.5336, 1.721)
1
1.378 (1.203,2.141)
1
InPRespC
0.9918 (0.5566, 1.773)
1.001
1.378 (1.2, 2.066)
1
InPKidC
1.277 (0.7274, 2.089)
1
1.554 (1.265, 2.872)
1
InPSlwC
0.92 (0.5585, 1.586)
1.001
1.411 (1.209,2.3)
1.001
InPRB CPlasT C AC
2.495 (1.144, 5.138)
1.001
1.398 (1.178,2.623)
1.001
InPBodTCAC
0.8816 (0.6219, 1.29)
1.003
1.27 (1.158, 1.609)
1
InPLivTCAC
0.8003 (0.5696, 1.15)
1.003
1.278(1.157, 1.641)
1.001
InkDissocC
1.214(0.2527, 4.896)
1.003
2.71 (1.765,8.973)
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-110 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-9. Posterior distributions for mouse PBPK model population
parameters (continued)
Sampled parameter"
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InBMaxkDC
1.25 (0.6793,2.162)
1.002
1.474 (1.253,2.383)
1
InPBodTCOHC
0.8025 (0.5607, 1.174)
1
1.314(1.17, 1.85)
1.001
InPLivTCOHC
1.526 (0.9099, 2.245)
1
1.399 (1.194,2.352)
1
InPBodTCOGC
0.4241 (0.1555, 1.053)
1.004
1.398 (1.207,2.156)
1
InPLivTCOGC
1.013 (0.492, 2.025)
1.002
1.554 (1.279, 2.526)
1
InPeffDCVG
0.9807 (0.008098, 149.6)
1.041
1.406 (1.206, 2.379)
1
InkTSD
5.187 (0.3909, 69.34)
1.001
5.858 (2.614, 80)
1
InkAS
1.711 (0.3729, 11.23)
1.001
4.203 (2.379, 18.15)
1
InkTD
0.1002(0.01304,0.7688)
1
5.16(2.478, 60.24)
1
Ink AD
0.2665 (0.05143, 1.483)
1.003
4.282 (2.378, 20.21)
1
InkASTCA
3.986 (0.1048, 141.9)
1
5.187 (2.516, 58.72)
1
InkASTCOH
0.7308 (0.006338, 89.75)
1.001
5.047 (2.496, 54.8)
1
IwVmaxC
0.6693 (0.4093, 1.106)
1.005
1.793 (1.49,2.675)
1
IhKmC
0.07148 (0.0323,0.1882)
1
2.203 (1.535,4.536)
1.001
InFracOtherC
0.02384 (0.003244,0.1611)
1.006
1.532 (1.265,2.971)
1
InFracTCAC
0.4875 (0.2764, 0.8444)
1.002
1.474(1.258,2.111)
1
lnVMAxDCVGC
1.517(0.02376, 1,421)
1.001
1.53 (1.263,2.795)
1
InClDCVGC
0.1794 (0.02333,79.69)
1.013
1.528(1.261,2.922)
1
lnVMAxKidDCVGC
1.424 (0.04313,704.9)
1.014
1.533 (1.262,2.854)
1
InClKidD C VGC
0.827 (0.04059, 167.2)
1.019
1.527 (1.263, 2.874)
1
lnVMAxLungLivC
2.903 (0.487, 12.1)
1.001
4.157(1.778, 29.01)
1.018
lnKMClara
0.01123 (0.001983,0.09537)
1.012
1.629 (1.278, 5.955)
1.003
InFracLungSysC
3.304 (0.2619, 182.1)
1.011
1.543 (1.266, 3.102)
1.001
lnVMAxTCOHC
1.645 (0.6986, 3.915)
1.005
1.603 (1.28,2.918)
1
lnKMTCOH
0.9594 (0.2867, 2.778)
1.007
1.521 (1.264, 2.626)
1
lnVMAxGlucC
65.59(27.58, 232.5)
1.018
1.487 (1.254,2.335)
1
lnKMGluc
31.16(6.122, 137.3)
1.015
1.781 (1.299, 5.667)
1.002
InkMetTCOHC
3.629 (0.7248, 9.535)
1.009
1.527 (1.265, 2.626)
1
InkUrnTCAC
0.1126(0.04083,0.2423)
1.012
1.757 (1.318,3.281)
1.003
InkMetTCAC
0.6175 (0.2702, 1.305)
1.027
1.508 (1.262, 2.352)
1.002
InkBileC
0.9954 (0.316,3.952)
1.003
1.502 (1.26,2.453)
1
InkEHRC
0.01553 (0.001001,0.0432)
1.008
1.534 (1.264, 2.767)
1
InkUrnTCOGC
7.874 (2.408, 50.28)
1
3.156 (1.783, 12.18)
1.001
InFracKidDCVCC
1.931 (0.01084, 113.7)
1.018
1.53 (1.264, 2.77)
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-111 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-9. Posterior distributions for mouse PBPK model population
parameters (continued)
Sampled parameter"
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InkDCVGC
0.2266(0.001104, 16.46)
1.011
1.525 (1.263,2.855)
1
InkNATC
0.1175 (0.0008506, 14.34)
1.024
1.528 (1.264,2.851)
1
InkKidBioactC
0.07506 (0.0009418, 12.35)
1.035
1.527 (1.263, 2.84)
1.001
1	aThese "sampled parameters" are scaled one or more times (see Table A-4) to obtain a biologically-meaningful
2	parameter, posterior distributions of which are summarized in Tables 3-36 through 3-40). For natural log
3	transformed parameters (name starting with "In"), values are for the population geometric means and standard
4	deviations.
5
6	Table A-10. Posterior distributions for mouse residual errors
7
Measurement
Residual error geometric standard deviation
Median (2.5,97.5%)
R
CInhPPM
1.177(1.16, 1.198)
1.001
CVen
2.678 (2.354,3.146)
1.001
CBldMix
1.606(1.415, 1.96)
1.001
CFat
2.486 (2.08, 3.195)
1
CKid
2.23 (1.908, 2.796)
1
CLiv
1.712 (1.543, 1.993)
1
AExhpost
1.234 (1.159, 1.359)
1
CTCOH
1.543 (1.424, 1.725)
1
CLivTCOH
1.591 (1.454, 1.818)
1
CPlasTCA
1.396 (1.338, 1.467)
1.001
CBldTCA
1.488 (1.423, 1.572)
1.001
CLivTCA
1.337(1.271, 1.43)
1
AUrnTCA
1.338 (1.259, 1.467)
1
CTCOGTCOH
1.493 (1.38, 1.674)
1.001
CLivTCOGTCOH
1.63 (1.457, 1.924)
1
AUrnTCOGTCOH
1.263 (1.203, 1.355)
1
TotCTCOH
1.846 (1.506, 2.509)
1.002
8	Note: the hierarchical statistical model for residual errors did not separate by subject.
9
This document is a draft for review purposes only and does not constitute Agency policy.
A-112 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-ll. Posterior correlations for mouse population mean parameters
2
Mouse
Corr. Coeff.
Parameter 1
Parameter 2
InKMGluc
InVMAxGlucC
0.765
InClDCVGC
InVMAxDCVGC
-0.553
InkMetTCAC
InkUrnTCAC
-0.488
InKMTCOH
InVMAxTCOHC
0.464
InClKidD C VGC
lnVMAvKidDCVGC
-0.394
InkUrnTCAC
InPRBCPlasTCAC
0.358
InkDissocC
InPBodTCAC
0.328
InkEHRC
InkMetTCOHC
0.314
IuVmaxC
VLivC
-0.305
InKMClara
InVMAxLungLivC
0.302
InBMaxkDC
InPLivTCAC
0.299
InKMGluc
InKMTCOH
0.293
InkBileC
InkEHRC
-0.280
InkEHRC
InKMTCOH
-0.273
InPBodTCOGC
InVMAxGlucC
0.269
InFracTCAC
InVMAxTCOHC
-0.267
InkMetTCAC
InPBodTCAC
0.264
InkDissocC
InPLivTCAC
0.253
InPSlwC
QFatC
-0.252
3
4	Note: only parameter pairs with correlation coefficient >0.25 are listed.
This document is a draft for review purposes only and does not constitute Agency policy.
A-113 DRAFT—DO NOT CITE OR QUOTE
-------
M InQCC
Mouse
M_lnVPRC *	 	M QFatC
M QGutC *
Prior Posterior
M_QLivC
Prior Posterior
M_QSIwC
—I	1	
Prior Posterior
M FracPlasC
—I	1	
Prior Posterior
M VFatC
Prior Posterior
MJnDRespC *
Prior Posterior
M_QKidC
—I	1	
Prior Posterior
M VGutC
—I	1	
Prior Posterior
M VLivC
Prior Posterior
M_VRapC
Prior Posterior
M_VRespLumC
Prior Posterior
—I	1	
Prior Posterior
Prior Posterior
M_VRespEffC
Prior Posterior
M VKidC
I	1	
Prior Posterior
—I	1	
Prior Posterior




~i	r
Prior Posterior
M InPLivC
MJnPRapC
Prior Posterior
MJnPRespC	
Prior Posterior
M InPKidC
This document is a draft for review purposes only and does not constitute Agency policy.
A-l 14 DRAFT—DO NOT CITE OR QUOTE
-------
Mouse
M InPRBCPIasTCAC
M InPBodTCAC
M InPLivTCAC

Prior Posterior
Prior
Posterior
Prior Posterior
Prior Posterior

MJnBMaxkDC


~l	T
Prior Posterior
M InPBodTCOGC
~l	T
Prior Posterior
M InPLivTCOGC
T	T
Prior Posterior
M InPeffDCVG
~i	r
Prior Posterior
M InkTSD

—I	1	
Prior Posterior
M InkAS

—I	1	
Prior Posterior
M InkTD
—I	1	
Prior Posterior
M InkAD
Prior Posterior
M InkASTCA
#
Prior Posterior
M InkASTCOH
—I	1	
Prior Posterior
M InVMaxC
#
—I	1	
Prior Posterior
M InKMC*
Prior Posterior
M InFracOtherC
—I	1	
Prior Posterior
M InFracTCAC

Prior Posterior
M InVMaxDCVGC
—I	1	
Prior Posterior
M InCIDCVGC
Prior Posterior
M InVMaxKidDCVGC
#
Figure A-13. Prior and posterior mouse population mean parameters
(Part 1). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-115 DRAFT—DO NOT CITE OR QUOTE
-------
M InCIKidDCVGC
—I	1	
Prior Posterior
M InVMaxTCOHC
*
—I	1	
Prior Posterior
M InkMetTCOHC
—I	1	
Prior Posterior
M InkEHRC
Prior Posterior
M InkNATC
—I	1	
Prior Posterior
Mouse
MJnVMaxLungLivC
—I	1	
Prior Posterior
M InKMTCOH


*
—I	1	
Prior Posterior
M InkUrnTCAC
—I	1	
Prior Posterior
M InkUrnTCOGC
—I	1	
Prior Posterior
M InkKidBioactC
M InKMCIara*
—I	1	
Prior Posterior
M InVMaxGlucC
MJnFracLungSysC
—I	1	
Prior Posterior
M InKMGluc
—I	1	
Prior Posterior
M InkMetTCAC
#
Prior Posterior
M InkBileC
	1	1	
Prior Posterior
M InFracKidDCVCC
—I	1	
Prior Posterior
M InkDCVGC
1	Figure A-14. Prior and posterior mouse population mean parameters
2	(Part 2). Thick lines are medians, boxes are interquartile regions, and error bars
3	are (2.5, 97.5%) confidence intervals. Parameters labeled with have
4	nonoverlapping interquartile regions.
5
6	Figure A-15. Prior and posterior mouse population mean parameters
7	(Part 3). Thick lines are medians, boxes are interquartile regions, and error bars
8	are (2.5, 97.5%) confidence intervals. Parameters labeled with have
9	nonoverlapping interquartile regions.
10
This document is a draft for review purposes only and does not constitute Agency policy.
A-116 DRAFT—DO NOT CITE OR QUOTE
-------
V JnQCC *
Mouse
VJnVPRC	 	V QFatC
V_QGutC

—I	1	
Prior Posterior
V_QLivC
Prior Posterior
V_QSIwC
—I	1	
Prior Posterior
—I	1	
Prior Posterior
Prior Posterior
VJnDRespC	
—I	1	
Prior Posterior
V_QKidC
I	1	
Prior Posterior
—I	1	
Prior Posterior




Prior Posterior
V VRapC	
Prior Posterior
V_VRespLumC
Prior Posterior
V VBIdC
—I	1	
Prior Posterior
V InPBC *
Prior Posterior
V VRespEffC
Prior Posterior
V VKidC
—I	1	
Prior Posterior
V InPGutC
—I	1	
Prior Posterior
V InPLivC
~i	r
Prior Posterior
VJnPRapC
—I	1	
Prior Posterior
VJnPRespC
—I	1	
Prior Posterior
V InPKidC
Prior
Posterior
Prior
Posterior
Prior
Posterior
Prior
Posterior
Figure A-16. Prior and posterior mouse population variance parameters
(Part 1). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-117 DRAFT—DO NOT CITE OR QUOTE
-------
V InPSIwC
Mouse
V InPRBCPIasTCAC	V InPBodTCAC
VJnPLivTCAC
—I	1	
Prior Posterior
V InkDissocC
—I	1	
Prior Posterior
V InBMaxkDC
—I	1	
Prior Posterior
V InPBodTCOHC
—I	1	
Prior Posterior
V InPLivTCOHC
—I	1	
Prior Posterior
V InPBodTCOGC
—I	1	
Prior Posterior
V InPLivTCOGC
—I	1	
Prior Posterior
V InPeffDCVG
I	I
Prior Posterior
V InkTSD
—I	F	
Prior Posterior
V InkAS
—I	1	
Prior Posterior
V InkTD
—I	1	
Prior Posterior
V InkAD
	F	F	
Prior Posterior
V InkASTCA
Prior Posterior
V InkASTCOH
	F	1	
Prior Posterior
V InVMaxC
—I	1	
Prior Posterior
V In KMC*
	F	F	
Prior Posterior
V InFracOtherC
—I	1	
Prior Posterior
V InFracTCAC
	F	I	
Prior Posterior
V InVMaxDCVGC
~i	r
Prior Posterior
V InCIDCVGC
	1	1	
Prior Posterior
V InVMaxKidDCVGC
Prior
Posterior
Prior
Posterior
Prior
Posterior
Prior
Posterior
Figure A-17. Prior and posterior mouse population variance parameters
(Part 2). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-118 DRAFT—DO NOT CITE OR QUOTE
-------
Mouse
V InCIKidDCVGC
—I	1	
Prior Posterior
V InVMaxTCOHC
VJnVMaxLungLivC *
V InKMCIara
—I	1	
Prior Posterior
V InKMTCOH
—I	1	
Prior Posterior
V InVMaxGlucC
VJnFracLungSysC
—I	1	
Prior Posterior
V InKMGluc
—I	1	
Prior Posterior
—I	1	
Prior Posterior
~~I	1	
Prior Posterior
Prior Posterior




~l	T
Prior Posterior
V InkEHRC
~l	T
Prior Posterior
V InkUrnTCOGC *
t	r
Prior Posterior
V InFracKidDCVCC
~i	r
Prior Posterior
V InkDCVGC
E3
Prior Posterior
V InkNATC
—I	1	
Prior Posterior
V InkKidBioactC
Figure A-18. Prior and posterior mouse population variance parameters
(Part 3). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-119 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-12. Posterior distributions for rat PBPK model population
2	parameters
3
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InQCC
1.195 (0.9285, 1.448)
1.034
1.298 (1.123,2.041)
1.031
InVPRC
0.6304 (0.4788, 0.8607)
1.012
1.446(1.247,2.011)
1.005
QFatC
1.167 (0.8321, 1.561)
1
0.4119 (0.2934,0.6438)
1
QGutC
1.154 (0.988, 1.306)
1
0.1613 (0.1132,0.2542)
1
QLivC
1.029 (0.8322, 1.223)
1.002
0.1551 (0.1092,0.2483)
1
QSlwC
0.9086 (0.5738, 1.251)
1.001
0.2817 (0.1968,0.4493)
1
InDRespC
2.765 (1.391, 5.262)
1.018
1.21 (1.142, 1.358)
1.001
QKidC
1.002 (0.8519, 1.152)
1.001
0.1185 (0.08284,0.1871)
1
FracPlasC
1.037 (0.8071, 1.259)
1.002
0.1785 (0.1272,0.2723)
1
VFatC
0.9728 (0.593, 1.378)
1
0.4139 (0.2924,0.6552)
1.002
VGutC
0.9826 (0.8321, 1.137)
1
0.1187 (0.08296,0.1873)
1
VLivC
0.9608 (0.7493, 1.19)
1.015
0.1682 (0.1168,0.2718)
1.001
VRapC
0.9929 (0.8563, 1.133)
1.001
0.1093 (0.07693,0.175)
1
VRespLumC
1.001 (0.7924, 1.21)
1
0.1636 (0.116,0.2601)
1
VRespEffC
0.999 (0.7921, 1.208)
1.001
0.1635 (0.1161,0.2598)
1
VKidC
0.999 (0.8263, 1.169)
1
0.1361 (0.09617,0.2167)
1
VBldC
1.002 (0.8617, 1.141)
1
0.1096 (0.07755,0.176)
1
InPBC
0.8551 (0.6854, 1.065)
1.001
1.317 (1.232, 1.462)
1.001
InPFatC
1.17 (0.8705, 1.595)
1.003
1.333 (1.247, 1.481)
1.001
InPGutC
0.8197 (0.5649, 1.227)
1
1.362 (1.198, 1.895)
1
InPLivC
1.046 (0.8886, 1.234)
1.001
1.152(1.115, 1.214)
1
InPRapC
1.021 (0.6239, 1.675)
1.002
1.373 (1.201, 1.988)
1
InPRespC
0.993 (0.5964, 1.645)
1.001
1.356 (1.197, 1.948)
1
InPKidC
0.9209 (0.6728, 1.281)
1
1.304 (1.201, 1.536)
1
InPSlwC
1.258 (0.9228, 1.711)
1.001
1.364 (1.263, 1.544)
1
InPRB CPlasT C AC
0.9763 (0.6761, 1.353)
1
1.276 (1.159, 1.634)
1
InPBodTCAC
1.136 (0.6737, 1.953)
1.008
1.631 (1.364,2.351)
1.003
InPLivTCAC
1.283 (0.6425, 2.491)
1.008
1.651 (1.356,2.658)
1
InkDissocC
1.01 (0.5052, 2.017)
1.002
1.596 (1.315,2.774)
1
4
This document is a draft for review purposes only and does not constitute Agency policy.
A-120 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-12. Posterior distributions for rat PBPK model population
parameters (continued)
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InBMaxkDC
0.9654 (0.5716, 1.733)
1.02
1.412(1.234,2.01)
1
InPBodTCOHC
0.9454 (0.4533, 1.884)
1.045
1.734 (1.39,3.151)
1.002
InPLivTCOHC
0.926 (0.3916, 2.196)
1.013
1.785 (1.382,4.142)
1.003
InPBodTCOGC
1.968 (0.09185, 14.44)
1.031
1.414(1.208, 2.571)
1
InPLivTCOGC
7.484 (2.389, 26.92)
1.017
1.41 (1.208,2.108)
1
InkTSD
3.747 (0.2263, 62.58)
1.01
6.777 (2.844, 87.29)
1
InkAS
2.474 (0.2542, 28.35)
1.004
10.16 (4.085, 143.7)
1
Ink AD
0.1731 (0.04001,0.7841)
1.018
4.069 (2.373, 14.19)
1.009
InkASTCA
1.513 (0.1401, 17.19)
1.002
4.376 (2.43, 22.83)
1
InkASTCOH
0.6896 (0.01534, 25.81)
1.001
4.734 (2.444, 35.2)
1.001
lnVMAxC
0.8948 (0.6377, 1.293)
1.028
1.646(1.424,2.146)
1.021
lnKMC
0.0239 (0.01602, 0.04993)
1.001
2.402 (1.812, 4.056)
1.001
InFracOtherC
0.344 (0.0206, 1.228)
1.442
3 (1.332, 10.04)
1.353
InFracTCAC
0.2348 (0.122,0.4616)
1.028
1.517 (1.264,2.393)
1.001
lnVMAxDCVGC
7.749 (0.2332, 458.8)
1.088
1.534 (1.262, 2.804)
1.001
InClDCVGC
0.3556 (0.06631,2.242)
1.018
1.509 (1.261,2.553)
1
lnVMAxKidDCVGC
0.2089 (0.04229, 1.14)
1.011
1.542 (1.263, 2.923)
1.001
InClKidD C VGC
184 (26.29, 1312)
1.02
1.527 (1.265, 2.873)
1.001
lnVMAxLungLivC
2.673 (0.4019, 14.16)
1.002
4.833 (1.599, 48.32)
1.002
lnKMClara
0.02563 (0.005231,0.197)
1.01
1.66 (1.279, 18.74)
1.002
InFracLungSysC
2.729 (0.04124, 63.27)
1.027
1.536 (1.267, 2.868)
1.001
lnVMAxTCOHC
1.832 (0.6673, 6.885)
1.041
1.667 (1.292,3.148)
1.002
lnKMTCOH
22.09 (3.075, 131.9)
1.186
1.629 (1.276, 3.773)
1.017
1iiV\i\vG1licC
28.72 (10.02, 86.33)
1.225
2.331 (1.364, 5.891)
1.126
lnKMGluc
6.579 (1.378, 23.57)
1.119
2.046 (1.309, 10.3)
1.125
InkMetTCOHC
2.354 (0.3445, 15.83)
1.287
1.876 (1.283, 11.82)
1.182
InkUrnTCAC
0.07112 (0.03934,0.1329)
1.076
1.513 (1.27,2.327)
1.003
InkMetTCAC
0.3554 (0.1195,0.8715)
1.036
1.528 (1.263, 2.444)
1.001
InkBileC
8.7 (1.939, 26.71)
1.05
1.65 (1.282, 5.494)
1.017
InkEHRC
1.396 (0.2711,6.624)
1.091
1.647 (1.277, 5.582)
1.005
This document is a draft for review purposes only and does not constitute Agency policy.
A-121 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-12. Posterior distributions for rat PBPK model population
parameters (continued)
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InkUrnTCOGC
20.65 (2.437, 138)
1.041
1.595 (1.269, 5.257)
1.026
InkNATC
0.002035 (0.0004799,
0.01019)
1.01
1.523 (1.261,2.593)
1.001
InkKidBioactC
0.006618 (0.0009409,
0.0367)
1.039
1.52(1.261,2.674)
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-122 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-13. Posterior distributions for rat residual errors
2


Residual error geometric standard deviation
Measurement
Subject
Median (2.5,97.5%)
R
CInhPPM
Subject 3
1.124(1.108, 1.147)
1

Subject 16
1.106(1.105, 1.111)
1
CMixExh
Subject 2
1.501 (1.398, 1.65)
1
Cart
Subject 2
1.174(1.142, 1.222)
1

Subject 6
1.523 (1.321, 1.918)
1.002
CVen
Subject 4
1.22(1.111, 1.877)
1

Subject 7
1.668 (1.489, 1.986)
1.001

Subject 8
1.45 (1.234, 2.065)
1.014

Subject 9
1.571 (1.426, 1.811)
1

Subject 10
4.459 (2.754, 6.009)
1

Subject 11
1.587 (1.347, 2.296)
1.002

Subject 16
1.874 (1.466, 2.964)
1.011

Subject 18
1.676 (1.188,3.486)
1.003
CBldMix
Subject 12
1.498 (1.268,2.189)
1
CFat
Subject 9
1.846 (1.635,2.184)
1

Subject 16
2.658 (1.861, 4.728)
1.001
CGut
Subject 9
1.855 (1.622, 2.243)
1
CKid
Subject 9
1.469 (1.354, 1.648)
1
CLiv
Subject 9
1.783 (1.554,2.157)
1

Subject 12
1.744 (1.401, 2.892)
1

Subject 16
1.665 (1.376,2.411)
1.001
CMus
Subject 9
1.653 (1.494, 1.919)
1
AExhpost
Subject 6
1.142 (1.108, 1.239)
1.003

Subject 10
1.117(1.106, 1.184)
1.004

Subject 14
1.166(1.107, 1.475)
1

Subject 15
1.125 (1.106, 1.237)
1
CTCOH
Subject 6
1.635 (1.455, 1.983)
1.002

Subject 10
1.259 (1.122, 1.868)
1.009

Subject 11
1.497 (1.299, 1.923)
1.01

Subject 13
1.611 (1.216, 3.556)
1.001

Subject 17
1.45 (1.213,2.208)
1.004

Subject 18
1.142(1.107, 1.268)
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-123 DRAFT—DO NOT CITE OR QUOTE
-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-124 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-13. Posterior distributions for rat residual errors (continued)


Residual error geometric standard deviation
Measurement
Subject
Median (2.5,97.5%)
R
CPlasTCA
Subject 4
1.134 (1.106, 1.254)
1

Subject 5
1.141 (1.107, 1.291)
1

Subject 11
1.213 (1.136, 1.381)
1

Subject 19
1.201 (1.145, 1.305)
1
CBldTCA
Subject 4
1.134 (1.106, 1.258)
1

Subject 5
1.14(1.107, 1.289)
1

Subject 6
1.59 (1.431, 1.878)
1.001

Subject 11
1.429(1.292, 1.701)
1.001

Subject 17
1.432 (1.282, 1.675)
1.03

Subject 18
1.193 (1.12, 1.358)
1.004

Subject 19
1.214 (1.153, 1.327)
1
CLivTCA
Subject 19
1.666(1.443,2.104)
1
AUrnTCA
Subject 1
1.498 (1.125,2.18)
1.135

Subject 6
1.95 (1.124, 5.264)
1.003

Subject 8
1.221 (1.146, 1.375)
1.003

Subject 10
1.18(1.108, 1.444)
1.007

Subject 17
1.753 (1.163,4.337)
1.001

Subject 19
1.333 (1.201, 1.707)
1
ABileTCOG
Subject 6
2.129 (1.128, 5.363)
1.003
CTCOG
Subject 17
2.758 (1.664, 5.734)
1.028
AUrnTCOGTCOH
Subject 1
1.129 (1.106, 1.232)
1.004

Subject 6
1.483 (1.113,4.791)
1.002

Subject 8
1.115 (1.106, 1.162)
1

Subject 10
1.145 (1.107, 1.305)
1

Subject 17
2.27 (1.53,4.956)
1.009
AUrnNDCVC
Subject 1
1.168 (1.11, 1.33)
1.002
AUrnTCTotMole
Subject 6
1.538 (1.182,3.868)
1.002

Subject 7
1.117 (1.106, 1.153)
1.001

Subject 14
1.121 (1.106, 1.207)
1

Subject 15
1.162 (1.108, 1.358)
1
TotCTCOH
Subject 17
1.488 (1.172,2.366)
1.015
1	Table A-13. Posterior distributions for rat residual errors (continued)
This document is a draft for review purposes only and does not constitute Agency policy.
A-125 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3	The nineteen subjects are (1) Bernauer et al. (1996) (2) Dallas et al. (1991); (3) Fisher et al. (1989)
4	females; (4) Fisher et al. (1991) females; (5) Fisher et al. (1991) males; (6) Green and Prout (1985),
5	Prout et al. (1985), male OA rats; (7) Hissink et al. (2002); (8) Kaneko et al. (1994) (9) Keys et al.
6	(2003); (10) Kimmerle and Eben (1973a); (11) Larson and Bull (1992; 1992); (12) Lee et al. (2000);
7	(13) Merdink et al. (1999); (14) Prout et al. (1985) AP rats; (15) Prout et al. (1985) OM rats;
8	(16) Simmons et al. (2002); (17) Stenner et al. (1997); (18) Templin et al. (1995); (19) Yu et al.
9	(2000).
10
This document is a draft for review purposes only and does not constitute Agency policy.
A-126 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-14. Posterior correlations for rat population mean parameters
2
Rat
Corr. Coeff.
Parameter 1
Parameter 2
InkNATC
lnVMAxKidDCVGC
-0.599
InkBileC
InPLivTCOGC
-0.587
InKMTCOH
lnVMAxTCOHC
0.567
InKMGluc
lriV\|» vGlucC
0.506
InClKidDCVGC
InkNATC
-0.497
InkUrnTCAC
InPBodTCAC
0.421
IiiVmaxC
VLivC
-0.417
InBMaxkDC
InkUrnTCAC
0.397
InkUrnTCOGC
InPBodTCOGC
-0.389
InPFatC
VFatC
-0.385
InClKidDCVGC
lnVMAxKidDCVGC
0.384
InKMGluc
InKMTCOH
0.383
InPLivTCOGC
lnVMAxGlucC
0.358
InBMaxkDC
InPBodTCAC
0.352
InClDCVGC
InClKidDCVGC
0.343
FracPlasC
InPRB CPlasTC AC
-0.337
InClDCVGC
InkNATC
-0.331
InkEHRC
1iiV\i\vG1licC
0.322
InkBileC
InkUrnTCOGC
0.307
InFracLungSysC
InFracOtherC
0.304
InFracOtherC
InkMetTCOHC
-0.296
InFracLungSysC
InKMTCOH
-0.271
InkMetTCAC
InPBodTCAC
0.264
InkMetTCAC
VLivC
-0.261
InKMTCOH
InPBodTCOGC
-0.260
InFracTCAC
InKMTCOH
0.258
InDRespC
InVPRC
0.254
InFracOtherC
InKMTCOH
-0.252
3
4	Note: only parameter pairs with correlation coefficient >0.25 are listed.
5
6
7
This document is a draft for review purposes only and does not constitute Agency policy.
A-127 DRAFT—DO NOT CITE OR QUOTE
-------
Rat.seqpriors.test4
MJnQCC*
M QFatC
M QGutC
Prior Posterior
M_QLivC
Prior Posterior
M_QSIwC
Prior Posterior
M_FracPlasC
—I	1	
Prior Posterior
M VFatC	
Prior Posterior
MJnDRespC	
Prior Posterior
M_QKidC
#
Prior Posterior
M VGutC	
Prior Posterior
M_VLivC	
Prior Posterior
M VRapC
Prior Posterior
M_VRespLumC
Prior Posterior
M VBIdC	
—I	1	
Prior Posterior
M InPBC	
Prior Posterior
M VRespEffC
Prior Posterior
M VKidC
Prior Posterior
M InPFatC
MJnPLivC
—I	1	
Prior Posterior
MJnPRapC	
Prior Posterior
M InPGutC
Prior Posterior
MJnPRespC	
—I	1	
Prior Posterior
MJnPKidC
Figure A-19. Prior and posterior rat population mean parameters (Part 1).
Thick lines are medians, boxes are interquartile regions,, and error bars are (2.5,
97.5%) confidence intervals. Parameters labeled with have nonoverlapping
interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-128 DRAFT—DO NOT CITE OR QUOTE
-------

Prior Posterior
M InkDissocC
Prior Posterior
M InPBodTCOGC
—I	1	
Prior Posterior
M InkTD
Prior Posterior
M InVMaxC
Prior Posterior
M_lnVMaxDCVGC
Rat.seqpriors.test4
M InPRBCPIasTCAC
#
M InPBodTCAC
—I	1	
Prior Posterior
M InBMaxkDC
Prior Posterior
M InPBodTCOHC
—I	1	
Prior Posterior
M InPLivTCOGC *
—I	1	
Prior Posterior
M InkTSD
—I	1	
Prior Posterior
M InkAD
Prior Posterior
M InkASTCA
—I	1	
Prior Posterior
M InKMC *
Prior Posterior
MJnFracOtherC *
~i	r
Prior Posterior
M InCIDCVGC
Prior Posterior
M InVMaxKidDCVGC *
M InPLivTCAC
—I	1	
Prior Posterior
M InPLivTCOHC
Prior Posterior
M InkAS	
Prior Posterior
M InkASTCOH
Prior Posterior
M InFracTCAC*
	1	1	
Prior Posterior
M InCIKidDCVGC *
Figure A-20. Prior and posterior rat population mean parameters (Part 2).
Thick lines are medians, boxes are interquartile regions, and error bars are (2.5,
97.5%) confidence intervals. Parameters labeled with have nonoverlapping
interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-129 DRAFT—DO NOT CITE OR QUOTE
-------
MJnVMaxLungLivC
#
Prior Posterior
M InKMTCOH *
—I	1	
Prior Posterior
M InkUrnTCAC
Prior Posterior
M InkUrnTCOGC
Rat.seqpriors.test4
MJnKMCIara	M_lnFracLungSysC
—I	1	
Prior Posterior
M InVMaxGlucC

—I	1	
Prior Posterior
M InkMetTCAC
—I	1	
Prior Posterior
M InkNATC
—I	1	
Prior Posterior
#
Prior Posterior
M InKMGluc
Prior Posterior
M InkBileC*
—I	1	
Prior Posterior
M InkKidBioactC
M InVMaxTCOHC
—I	1	
Prior Posterior
M InkMetTCOHC
—I	1	
Prior Posterior
M InkEHRC *
Figure A-21. Prior and posterior rat population mean parameters (Part 3).
Thick lines are medians, boxes are interquartile regions, and error bars are (2.5,
97.5%) confidence intervals. Parameters labeled with have nonoverlapping
interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-130 DRAFT—DO NOT CITE OR QUOTE
-------
V InQCC *
Rat.seqpriors.test4
V InVPRC	V QFatC
V_QGutC
Prior Posterior
Prior Posterior
Prior Posterior
V FracPlasC
Prior Posterior
V VFatC
Prior Posterior
Prior Posterior




Prior Posterior
V VGutC
1	T
Prior Posterior
V VLivC
Prior Posterior
V VRapC	
Prior Posterior
V_VRespLumC
Prior Posterior
V VBIdC	
—I	1	
Prior Posterior
VJnPBC	
Prior Posterior
V VRespEffC
Prior Posterior
V VKidC
Prior Posterior
V_lnPFatC
Prior Posterior
V InPLivC
—I	1	
Prior Posterior
VJnPRapC
Prior Posterior
V InPGutC
Prior Posterior
VJnPRespC	
—I	1	
Prior Posterior
V InPKidC
Figure A-22. Prior and posterior rat population variance parameters
(Part 1). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-131 DRAFT—DO NOT CITE OR QUOTE
-------
Rat.seqpriors.test4
V InPRBCPIasTCAC
V InPBodTCAC
V InPLivTCAC
Prior Posterior
V InkDissocC
—I	1	
Prior Posterior
V InBMaxkDC
—I	1	
Prior Posterior
V InPBodTCOHC
—I	1	
Prior Posterior
V InPLivTCOHC
Prior Posterior
V InPBodTCOGC
—I	1	
Prior Posterior
V InPLivTCOGC
—I	1	
Prior Posterior
V InkTSD
Prior Posterior
V InkTD
—I	1	
Prior Posterior
V InkAD
Prior Posterior
V InkAS	
Prior Posterior
V InkASTCA
I	I
Prior Posterior
V InkASTCOH
Prior Posterior
VJnVMaxC
—I	1	
Prior Posterior
V In KMC
Prior Posterior
VJnFracOtherC *
Prior Posterior
VJnVMaxDCVGC
—I	1	
Prior Posterior
VJnCIDCVGC
Prior Posterior
VJnFracTCAC
Prior Posterior
VJnVMaxKidDCVGC
—I	1	
Prior Posterior
VJnCIKidDCVGC
Figure A-23. Prior and posterior rat population variance parameters
(Part 2). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-132 DRAFT—DO NOT CITE OR QUOTE
-------
VJnVMaxLungLivC *
—I	1	
Prior Posterior
VJnKMTCOH
Rat.seqpriors.test4
VJnKMCIara	V_lnFracLungSysC
Prior
VJnVMaxGlucC *
—I	1	
Prior Posterior
V InkUrnTCAC
—I	1	
Prior Posterior
V InkUrnTCOGC
Posterior
—I	1	
Prior Posterior
V InkMetTCAC
—I	1	
Prior Posterior
V InkNATC
V InVMaxTCOHC
Prior Posterior
VJnKMGluc
—I	1	
Prior Posterior
VJnkMetTCOHC
i
Prior Posterior
V InkBileC
Prior Posterior
V InkEHRC
—I	1	
Prior Posterior
V InkKidBioactC
i
Prior
Posterior
Prior
Posterior
Prior
Posterior
Figure A-24. Prior and posterior rat population variance parameters
(Part 3). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-133 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
Table A-15. Posterior distributions for human PBPK model population
parameters
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InQCC
0.837(0.6761, 1.022)
1.038
1.457(1.271, 1.996)
1.036
InVPRC
1.519 (1.261, 1.884)
1.007
1.497 (1.317, 1.851)
1.008
QFatC
0.7781 (0.405, 1.143)
1.014
0.6272 (0.4431,0.9773)
1
QGutC
0.7917 (0.6631,0.925)
1.017
0.1693 (0.1199,0.2559)
1.019
QLivC
0.5099 (0.1737,0.8386)
1.031
0.4167 (0.2943,0.6324)
1.009
QSlwC
0.7261 (0.4864, 0.9234)
1.011
0.3166 (0.2254,0.4802)
1.005
InDRespC
0.626 (0.3063, 1.013)
1.197
1.291 (1.158,2.006)
1.083
QKidC
1.007 (0.9137, 1.103)
1.009
0.1004 (0.07307,0.1545)
1
FracPlasC
1.001 (0.9544, 1.047)
1.01
0.04275 (0.03155,0.06305)
1
VFatC
0.788 (0.48, 1.056)
1.005
0.3666 (0.2696, 0.5542)
1
VGutC
1 (0.937, 1.067)
1.007
0.06745 (0.04923,0.1038)
1
VLivC
1.043 (0.8683, 1.23)
1.047
0.1959 (0.1424,0.3017)
1.003
VRapC
0.9959 (0.9311, 1.06)
1.006
0.06692(0.04843,0.1027)
1
VRespLumC
1.003 (0.8461, 1.164)
1.001
0.1671 (0.1209,0.255)
1
VRespEffC
1 (0.8383, 1.159)
1.001
0.1672 (0.1215,0.259)
1
VKidC
0.9965 (0.8551, 1.14)
1.007
0.1425 (0.1037,0.2183)
1
VBldC
1.013 (0.9177, 1.108)
1.003
0.1005 (0.07265,0.1564)
1
InPBC
0.9704(0.8529, 1.101)
1.001
1.216(1.161, 1.307)
1.002
InPFatC
0.8498 (0.7334, 0.9976)
1.002
1.188 (1.113, 1.366)
1.002
InPGutC
1.095 (0.7377, 1.585)
1.029
1.413 (1.214,2.05)
1.002
InPLivC
0.9907 (0.6679, 1.441)
1.01
1.338 (1.203, 1.683)
1
InPRapC
0.93 (0.6589, 1.28)
1.003
1.528 (1.248, 2.472)
1.001
InPRespC
1.018(0.6773, 1.5)
1.015
1.32 (1.192, 1.656)
1
InPKidC
0.9993 (0.8236, 1.219)
1.003
1.155 (1.097, 1.287)
1
InPSlwC
1.157 (0.8468, 1.59)
1.018
1.69 (1.383,3.157)
1.008
InPRB CPlasT C AC
0.3223 (0.04876, 0.8378)
1.007
5.507 (3.047, 19.88)
1.003
InPBodTCAC
1.194 (0.929, 1.481)
1.043
1.327 (1.185, 1.67)
1.018
InPLivTCAC
1.202 (0.8429, 1.634)
1.046
1.285 (1.162, 1.648)
1.007
InkDissocC
0.9932 (0.9387, 1.053)
1.012
1.043 (1.026, 1.076)
1.003
This document is a draft for review purposes only and does not constitute Agency policy.
A-134 DRAFT—DO NOT CITE OR QUOTE
-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-13 5 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-15. Posterior distributions for human PBPK model population parameters
(continued)
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InBMaxkDC
0.8806 (0.7492, 1.047)
1.038
1.157 (1.085, 1.37)
1.012
InPBodTCOHC
1.703 (1.439,2.172)
1.019
1.409 (1.267, 1.678)
1.011
InPLivTCOHC
1.069 (0.7643, 1.485)
1.028
1.288 (1.165, 1.629)
1.002
InPBodTCOGC
0.7264 (0.1237,2.54)
1.003
11.98 (5.037, 185.3)
1.017
InPLivTCOGC
6.671 (1.545, 24.87)
1.225
5.954 (2.653, 23.68)
1.052
InPeffDCVG
0.01007 (0.003264, 0.03264)
1.004
1.385 (1.201,2.03)
1.001
InkASTCA
4.511 (0.04731,465.7)
1
5.467 (2.523, 71.06)
1
InkASTCOH
8.262 (0.0677, 347.9)
1
5.481 (2.513,67.86)
1
lnVMAxC
0.3759 (0.2218,0.5882)
1.026
2.21 (1.862, 2.848)
1.003
InCIC
12.64 (5.207, 39.96)
1.028
4.325 (2.672, 9.003)
1.016
InFracOtherC
0.1186 (0.02298,0.2989)
1.061
3.449 (1.392, 9.146)
1.102
InFracTCAC
0.1315 (0.07115,0.197)
1.026
2.467 (1.916, 3.778)
1.01
InClDCVGC
2.786 (1.326, 5.769)
1.08
2.789 (1.867, 4.877)
1.02
lnKMDCVGC
1.213 (0.3908, 4.707)
1.029
4.43 (2.396, 18.56)
1.035
InClKidD C VGC
0.04538 (0.001311,0.1945)
1.204
3.338 (1.295, 30.46)
1.095
lnKMKidDCVGC
0.2802(0.1096, 1.778)
1.097
1.496 (1.263,2.317)
1.001
lnVMAxLungLivC
3.772 (0.8319, 9.157)
1.035
2.228 (1.335,21.89)
1.014
lnKMClara
0.2726 (0.02144, 1.411)
1.041
11.63 (1.877, 682.7)
1.041
InFracLungSysC
24.08 (6.276,81.14)
1.016
1.496 (1.263, 2.439)
1.001
InClTCOHC
0.1767 (0.1374,0.2257)
1.011
1.888 (1.624, 2.307)
1.01
lnKMTCOH
2.221 (1.296, 4.575)
1.02
2.578 (1.782, 4.584)
1.015
InClGlucC
0.2796 (0.2132,0.3807)
1.056
1.955 (1.583,2.418)
1.079
lnKMGluc
133.4 (51.56, 277.2)
1.02
1.573 (1.266, 4.968)
1.011
InkMetTCOHC
0.7546 (0.1427,2.13)
1.007
5.011 (2.668, 15.71)
1.002
lnkUrnTCAC
0.04565 (0.0324, 0.06029)
1.005
1.878 (1.589,2.48)
1.006
InkMetTCAC
0.2812 (0.1293,0.5359)
1.004
2.529(1.78,4.211)
1.002
InkBileC
6.855 (3.016, 20.69)
1.464
1.589 (1.27,3.358)
1.015
InkEHRC
0.1561 (0.09511,0.2608)
1.1
1.699 (1.348, 2.498)
1.015
InkUrnTCOGC
15.78 (6.135,72.5)
1.007
9.351 (4.93,29.96)
1.003
This document is a draft for review purposes only and does not constitute Agency policy.
A-13 6 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-15. Posterior distributions for human PBPK model population parameters
(continued)
Sampled parameter
Posterior distributions reflecting uncertainty in population distribution
Population (geometric) mean
Population (geometric) standard
deviation
Median (2.5,97.5%)
R
Median (2.5,97.5%)
R
InkDCVGC
7.123 (5.429, 9.702)
1.026
1.507 (1.311, 1.897)
1.008
InkNATC
0.0003157 (0.0001087, 0.002305)
1.008
1.54 (1.261,3.306)
1
InkKidBioactC
0.06516 (0.01763,0.1743)
1.001
1.523 (1.262, 2.987)
1
1
2
This document is a draft for review purposes only and does not constitute Agency policy.
A-13 7 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-16. Posterior distributions for human residual errors
2


Residual error geometric standard deviation
Measurement
Subject
Median (2.5,97.5%)
R
RetDose
Subject 4
1.131 (1.106, 1.25)
1.001
CAlvPPM
Subject 1
1.832(1.509,2.376)
1.015

Subject 4
1.515 (1.378, 1.738)
1

Subject 5
1.44(1.413, 1.471)
1
CVen
Subject 1
1.875 (1.683,2.129)
1.018

Subject 3
1.618(1.462, 1.862)
1

Subject 4
1.716(1.513,2.057)
1.001

Subject 5
2.948 (2.423, 3.8)
1.007
CTCOH
Subject 1
1.205 (1.185, 1.227)
1.012

Subject 3
1.213 (1.187, 1.247)
1

Subject 5
2.101 (1.826,2.571)
1.001

Subject 7
1.144(1.106,2.887)
1.123
CPlasTCA
Subject 2
1.117(1.106, 1.17)
1.001

Subject 7
1.168(1.123, 1.242)
1
CBldTCA
Subject 1
1.138 (1.126, 1.152)
1.003

Subject 2
1.119(1.106, 1.178)
1

Subject 4
1.488(1.351, 1.646)
1.018

Subject 5
1.438(1.367, 1.537)
1.002
zAUrnTCA
Subject 1
1.448(1.414, 1.485)
1.001

Subject 2
1.113 (1.105, 1.149)
1.001

Subject 3
1.242(1.197, 1.301)
1.001

Subject 4
1.538 (1.441, 1.67)
1

Subject 6
1.158(1.118, 1.228)
1

Subject 7
1.119(1.106, 1.181)
1
z AU rnT C Acollect
Subject 3
1.999(1.178, 3.903)
1.003

Subject 5
2.787 (2.134,4.23)
1.001
AUrnTCOGTCOH
Subject 1
1.106(1.105, 1.112)
1.001

Subject 3
1.11 (1.105, 1.125)
1

Subject 4
1.124(1.107, 1.151)
1.001

Subject 6
1.117(1.106, 1.157)
1.001

Subject 7
1.134(1.106, 1.348)
1.003
This document is a draft for review purposes only and does not constitute Agency policy.
A-13 8 DRAFT—DO NOT CITE OR QUOTE
-------
Table A-16. Posterior distributions for human residual errors
(continued)
Measurement
Subject
Residual error geometric standard deviation
Median (2.5,97.5%)
R
AUrnTCOGTCOHcollect
Subject 3
1.3 (1.111,2.333)
1.004
Subject 5
1.626(1.524, 1.767)
1
CDCVGmol
Subject 1
1.53 (1.436, 1.656)
1.009
zAUrnNDCVC
Subject 6
1.167(1.124, 1.244)
1
TotCTCOH
Subject 1
1.204(1.185, 1.226)
1.011
Subject 4
1.247(1.177, 1.366)
1.009
Subject 5
1.689(1.552, 1.9)
1.001
1
2	The seven subjects are (1) Fisher et al. (1998); (2) Paycok and Powell (1945); (3) Kimmerle and Eben
3	(1973b); (4) Monster et al. (1976); (5) Chiu et al. (2007); (6) Bernauer et al. (1996); (7) Muller et al.
4	(1974).
5
6
This document is a draft for review purposes only and does not constitute Agency policy.
A-13 9 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-17. Posterior correlations for human population mean parameters
2
Human
Corr. coeff.
Parameter 1
Parameter 2
InkBileC
InPLivTCOGC
-0.649
InClKidDCVGC
InKMKidDCVGC
-0.567
InClGlucC
InkEHRC
0.438
InkMetTCAC
InPLivTCAC
-0.392
InClKidDCVGC
InDRespC
-0.324
InClKidDCVGC
InkEHRC
-0.301
InKMTCOH
InPBodTCAC
0.289
InkMetTCAC
InPBodTCAC
0.283
InClKidDCVGC
InkBileC
-0.277
InkEHRC
InPBodTCOHC
-0.277
InClDCVGC
InkDCVGC
0.269
InBMaxkDC
InPBodTCAC
0.267
InFracOtherC
InQCC
0.260
InFracOtherC
InkDCVGC
-0.258
InFracOtherC
VLivC
0.257
InFracOtherC
InPLivTCOGC
-0.256
InClDCVGC
InFracOtherC
-0.256
InClDCVGC
VLivC
-0.252
3
4	Note: only parameter pairs with correlation coefficient >0.25 are listed.
This document is a draft for review purposes only and does not constitute Agency policy.
A-140 DRAFT—DO NOT CITE OR QUOTE
-------
M InOCC
Human.seqpriors.vl


M...QGutC
Prior Posterior
Prior Posterior
Prior Posterior
Pnor Posterior



Pnor Posterior
M FracPfasC
Prior Posterior
M VFatC
Prior Posterior
M VGutC
Pnor Posterior
M VLivC
Pnor Posterior
M. VRapC
Prior Posterior
MJ/RespLumC
Pnor Posterior
M....VRespEffC
Prior Posterior
M VKtdC
Prior Posterior
M VBIdC
Prior Posterior
M fnPBC
Prior
Posterior
£
Prior Posterior
MJnPGutC
Prior Posterior
MJnPLivC
Prior Posterior
MJnPRapC
Prior Posterior
MJnPRespC
Pnor Posterior
MJnPKidC


Prior
Prior
Posterior
Prior
Posterior
Prior
Posterior
This document is a draft for review purposes only and does not constitute Agency policy.
A-141 DRAFT—DO NOT CITE OR QUOTE
-------
Human
1	Figure A-25. Prior and poster.*,.	FuFU.ation mean parameters
2	(Part 1). Thick lines are medians, boxes are interquartile regions, and error bars
3	are (2.5, 97.5%) confidence intervals. Parameters labeled with have
4	nonoverlapping interquartile regions.
5
This document is a draft for review purposes only and does not constitute Agency policy.
A-142 DRAFT—DO NOT CITE OR QUOTE
-------
MJnPSIwC
Human, seqpriors.vl
MJnPRBCPIasTCAC	M_lnPBodTCAC
MJnPLivTCAC
—I	1	
Prior Posterior
M InkDissocC
—I	1	
Prior Posterior
M InPBodTCOGC
—I	1	
Prior Posterior
M InkASTCOH
—I	1	
Prior Posterior
M InFracTCAC '
	1	1	
Prior Posterior
M In KMKidDCVGC
v-
f
—I	1	
Prior Posterior
M InBMaxkDC
—I	1	
Prior Posterior
M InPLIvTCOGC	
—I	1	
Prior Posterior
M InVMaxC *
Prior Posterior
M inCIDCVGC
Prior Posterior
MJnVMaxLungLivC

Prior Posterior
M InPBodTCOHC *
—I	1	
Prior Posterior
M InPeffDCVG *
I	I
Prior Posterior
M InCIC '
—I	1	
Prior Posterior
M InKMDCVGC
—I	1	
Prior Posterior
M InKMCIara "
—I	1	
Prior Posterior
M InPLivTCOHC
—I	1	
Prior Posterior
M InkASTCA	
Prior Posterior
M JnFracOtherC
—I	1	
Prior Posterior
M InCIKidPCVGC

Prior Posterior
MJnFracLungSysC '
*
Prior Posterior
Figure A-26. Prior and posterior human population mean parameters
(Part 2). Thick lines are medi ans, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-143 DRAFT—DO NOT CITE OR QUOTE
-------
H u ma iVseqpriors.v 1
i
1	1	
Prior Posterior
V QLivC

Prior Posterior
V QSIwC
—I	1	
Prior Posterior
	V FracPlasC	
—I	1	
Prior Posterior
V VFatC	
—I	1	
Prior Posterior
VJnDRespC
—I	1	
Prior Posterior
V_QKidC
Prior Posterior
V VGutC	
—I	1	
Prior Posterior
V VLivC	
—I—
Prior
Posterior
V VRapC
—I	1	
Prior Posterior
V VBidC
Prior Posterior
V_VRespLumC
—r~
Prior
Posterior
V VRespEffC
Prior Posterior
V InPBC	
—I	1	
Prior Posterior
VJnPLivC
—I	1	
Prior Posterior
V InPFatC
Prior Posterior
V VKidC	
S_
—I	1	
Prior Posterior
V In PGutC
—I—
Prior
VJnPRapC
Prior Posterior
VJnPRespC
—I	1	
Prior Posterior
VJnPKidC
Figure A-27. Prior and posterior human population mean parameters
(Part 3). Thick lines are medi ans, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-144 DRAFT—DO NOT CITE OR QUOTE
-------
Mnman
Human.seqpriors.vl
V InPRBCPIasTCAC *
—I	1	
Prior Posterior
V InkDissocC
81
	1	1	
Prior Posterior
V InPBodTCOGC *
—I	1	
Prior Posterior
V InkASTCOH
Prior Posterior
VJnFracTCAC *
Prior Posterior
VJnKMKidDCVGC
V InPBodTCAC
V InPLivTCAC
—I	1	
Prior Posterior
V InBMaxKDC
—I	1	
Prior Posterior
V InPBodTCOHC
—I	1	
Prior Posterior
V InPLivTCOHC
—I	1	
Prior Posterior
V InPLivTCOGC '
—I	1	
Prior Posterior
V InPeffDCVG
—I	1	
Prior Posterior
V InVMaxC
—I	1	
Prior Posterior
V InkASTCA
—I	1	
Prior Posterior
V InCIDCVGC *
—I	1	
Prior Posterior
V InCIC *
—I	1	
Prior Posterior
V InFracOtherC *
—I	1	
Prior Posterior
VJnKMDCVGC*



|
T J

Prior Posterior
VJnCIKidPCVGC
Prior Posterior
V InKMCIara *



CO -


_


<£> ~


-






—1 ^
M
O -
T 1—

	1	1	
Prior Posterior
VJnFracLungSysC
Figure A-28. Prior and posterior human population variance parameters
(Part 1). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals,
nonoverlapping interquartile regions.
Parameters labeled with have
This document is a draft for review purposes only and does not constitute Agency policy.
A-145 DRAFT—DO NOT CITE OR QUOTE
-------
Nnman .
Human, seqpriors.vl
V InCITCOHC *
V InKMTCOH *
V InCIGIucC *
~1	T
Prior Posterior
V InkMetTCOHC *
Prior Posterior
V InkUrnTCAC *
~1	T
Prior Posterior
V InkMetTCAC *
Prior Posterior
V InkBileC

—I	1	
Prior Posterior
	V InkEHRC
	1	1	
Prior Posterior
V InkUrnTCOGC *
—I	1	
Prior Posterior
V InkKidBioactC
—I	1	
Prior Posterior
V InkDCVGC
Prior Posterior
V InkNATC
I
Prior
Figure A-29. Prior and posterior human population variance parameters
(Part 2). Thick lines are medians, boxes are interquartile regions, and error bars
are (2.5, 97.5%) confidence intervals. Parameters labeled with have
nonoverlapping interquartile regions.
This document is a draft for review purposes only and does not constitute Agency policy.
A-146 DRAFT—DO NOT CITE OR QUOTE
-------
Human
1	Figure A-30. Prior and poster.^.	^^.ation variance parameters
2	(Part 3). Thick lines are medians, boxes are interquartile regions, and error bars
3	are (2.5, 97.5%) confidence intervals. Parameters labeled with have
4	nonoverlapping interquartile regions.
5
6
This document is a draft for review purposes only and does not constitute Agency policy.
A-147 DRAFT—DO NOT CITE OR QUOTE
-------
A.4.2. Comparison of Model Predictions with Data
1	Time-course graphs of calibration and evaluation data compared to posterior predictions
2	are shown in Figures A-31 to A-35. For each panel, the boxes are the experimental data, the
3	solid red line is the prediction using the posterior mean of the subject-specific parameters (only
4	shown for calibration data), and the shaded regions (or + with error bars, for single data points)
5	are bounded by the 2.5, 25, 50, 75, and 97.5% population-based predictions.
6
This document is a draft for review purposes only and does not constitute Agency policy.
A-148 DRAFT—DO NOT CITE OR QUOTE
-------
A.4.2.1. Mouse Data and Model Predictions
Abbas et al. 97a Male Mouse
1200 mg/kg TCE Oral gavage (oil)
Abbas et al. 97a Male Mouse
1200 mg/kg TCE Oral gavage (oil)
Abbas et al. 97a Male Mouse
1200 mg/Kg TCE Oral gavage (oil)
5 o_T
O 
-------
Abbas at al. 97a Mala Mouse
2000 mg/kg TCE Oral gavage (oil)
Abbas et al. 97a Male Mouse
2000 mg/kg TCE Oral gavage (oM)
Abbas et al 97a Male Mouse
2000 mg/kg TCE Oral gavage (oH)
t(hr)
Abbas et al 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
~1	1	1	1	1	
0.5 1.0 2.0 5.0 10.0
t(hr)
Abbas et al 97a Male Mouse
2000 mg/kg TCE Oral gavage (oil)
—i	1	1	1	r~
0.5 1.0 2.0 5.0 10.0
t(hr)
Abbas et al 97a Male Mouse
2000 mg/kg TCE Oral gavage (oil)
~i	1	1	1	r~
0.5 1 0 2.0 5.0 10 0
t(hr)
Abbas et al 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
Abbas et al 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
i
«
x
O P.
o 
~i—i—i	1—i—i	r
0.5 1.0 2.0 5.0 20.0 50.0
t(hr)
Abbas et al. 97a Male Mouse
2000 mg/kg TCE Oral gavage (oil)
-|	1	1	1	r~
0.5 1.0 2.0 5.0 10.0
t 
-------
Abbas et al. 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
Abbas et al. 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
Abbas et al 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
I	I I
2.0 50 10.0
t(hr)
Abbas et al. 97a Male Mouse
300 mg/kg TCE Oral gavage (oil)
n	1	1	r
1 0 2.0 5.0 10 0
t(hr)
Abbas et al. 97a Male Mouse
300 mg/kg TCE Oral gavage (oif)

Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
-1	1	1	1	1	r~
40 60 80 100 120 140
«(hr)
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
i
a
So.
1.0 2.0 5.0 10.0
t(hr)
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
1-4
|
.5 CM-
""I	
40
—J—
60
-1	1	1	1-
80 100 1 20 140
Abbas et al 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
~i—i	1	1—i	r
0.5 1.0 2.0 5.0 20.0
t(hr)
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
i	1	1	1	r i
0.5 1.0 2.0 5.0 10.0
tflir)
Abbas et al 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
T	1	1	1	 I I
0.5 1.0 2.0 5.0 10.0
t(hr)
Abbas et al 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
U m
n	1	T
0.5 1.0 2.0 5.0 10.0
t(hr)
Abbas et al. 97a Male Mouse
600 mgAcg TCE Oral gavage (oil)
"I	1	r
0.5 1.0 2.0 5.0
t(hr)
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
Abbas et al 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
1.0 2.0 5.0 10.0
t(hr)
1.0 2.0 5.0 10.0
t(hr)
i	1	1	1	r
0.5 1.0 2.0 5.0 10.0
t(hr)
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-151 DRAFT—DO NOT CITE OR QUOTE
-------
1	Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
2	model predictions (red line: using the posterior mean of the subject-specific
3	parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
4	50, 75, and 97.5% population-based predictions) (continued).
5
This document is a draft for review purposes only and does not constitute Agency policy.
A-152 DRAFT—DO NOT CITE OR QUOTE
-------
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
~i	1	1	r
0.5 1.0 2.0 5.0
t(hr)
Abbas et al 97b Male Mouse
100 mg/kg TCA iv
® o
E tN"
Abbas et al. 97a Male Mouse
600 mg/kg TCE Oral gavage (oil)
—r~
60
-1	1—
80 100
t(hr)
Abbas et al. 97b Male Mouse
100 mg/kg TCOH iv
~~l	
120
Abbas et al 97b Male Mouse
100 mg/kg TCA iv

2.0 50 20.0
Abbas et al. 97b Male Mouse
100 mg/kg TCOH iv
I
i	1	1	1	1	r
20 40 60 80 100
I—
140
5 5-
—i	1—i—i—i—i—
20 40 60 80 100 140
I
i	1	1	1	1	r
0.05 0 10 0.20 0.50 1.00 2.00
Abbas et al. 97b Male Mouse
100 mg/kg TCOH iv
Fisher et a! 91 Female Mouse
42 ppm TCE 4 hr inhalation
Fisher et al. 91 Female Mouse
42 ppm TCE 4 hr inhalation
I I	I I r
0.05 0.10 0.20 0.50 1.00 2.00
t (hr)
Fisher et al. 91 Female Mouse
236 ppm TCE 4 hr inhalation
Fisher et al 91 Female Mouse
236 ppm TCE 4 hr inhalation

n	1	1	1	1	r
2 3 4 5 6 7
t (hr)
Fisher et al 91 Female Mouse
368 ppm TCE 4 hr inhalation
i	1	1	r~
2.0 2.5 3.0 3.5
t (hr)
Fisher et al. 91 Female Mouse
368 ppm TCE 4 hr inhalation
1	1	1	1	1	r
2.0 2.5 3.0 3.5 4.0 4.5 5.0
t (hr)
Fisher et al 91 Female Mouse
889 ppm TCE 4 hr inhalation
Fisher et al. 91 Female Mouse
889 ppm TCE 4 hr inhalation
i	1	1	1	1	1	r
2.0 2.5 3.0 3.5 4.0 4.5 5.0
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-153 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al 91 Female Mouse
1100 ppm TCE Closed Chamber
Fisher et al. 91 Female Mouse
300 ppm TCE Closed Chamber
Fisher et al. 91 Female Mouse
3700 ppm TCE Closed Chamber
Fisher et al 91 Female Mouse
700 ppm TCE Closed Chamber
Fisher et al 91 Female Mouse
7000 ppm TCE Closed Chamber
Fisher et al 91 Male Mouse
110 ppm TCE 4 hr inhalation
Lil 
-------
Fisher et al 91 Male Mouse
10000 ppm TCE Closed Chamber
Fisher et al 91 Male Mouse
3800 ppm TCE Closed Chamber
Fisher et al 91 Male Mouse
5600 ppm TCE Closed Chamber
~i	1	1	1	1	r
0 0 0.5 1.0 1.5 2 0 2.5 3 0
t(hr)
Fisher et al 93 Female Mouse
2000 mg/kg TCE Oral gavage (oil)
Fisher et al. 93 Female Mouse
2000 mgflcg TCE Oral gavage (oi)
Fisher et al 93 Female Mouse
487 mg/kg TCE Oral gavage (oil)
_j o
1 00 5.00 20 00
t(hr)
Fisher et al 93 Female Mouse
487 mgfcg TCE Oral gavage (oil)
~l	1	1	1	1	1	1	r
0.05 0.20 0.50 2.00 5.00
Fisher et al 93 Female Mouse
973 mg/kg TCE Oral gavage (oil)
i	1—i	1	1	r
0.5 2.0 5.0 20.0
t (hr)
Fisher et al 93 Female Mouse
973 mg/kg TCE Oral gavage (oil)
—i	1	1	1	r
0.1 0.2 0.5 1.0 2.0
t(hr)
Fisher et aJ 93 Male Mouse
2000 mg/kg TCE Oral gavage (oil)
h—i	1—i—r
0.5 2.0 5.0 20.0
t (hr)
Fisher et al 93 Male Mouse
2000 mgfkg TCE Oral gavage (oil)
Fisher et al 93 Male Mouse
487 mg/kg TCE Oral gavage (ofl)
I I I I	1
0.20 0.50 2 00 5 00
~l—I—I—I—I—T
0.05 0.20 1 00
i—i—i—r
5 00 20 00
i—i	1—i—r
2.0 5.0 20.0
Fisher et al 93 Male Mouse
487 mg/kg TCE Oral gavage (oil)
Frsher et al 93 Male Mouse
973 mg/kg TCE Oral gavage (oil)
Fisher et al 93 Male Mouse
973 mg/kg TCE Oral gavage (oil)
0.05 0.20 1.00 5.00 20.00
020 050
2 00 5.00
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-l55 DRAFT—DO NOT CITE OR QUOTE
-------
1	Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
2	model predictions (red line: using the posterior mean of the subject-specific
3	parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
4	50, 75, and 97.5% population-based predictions) (continued).
5
This document is a draft for review purposes only and does not constitute Agency policy.
A-156 DRAFT—DO NOT CITE OR QUOTE
-------
Green et al 86 Male Mouse
10 mg/kg TCE Orel gavage (oil)
Green et al 85 Male Mouse
10 mg/kg TCE Oral gavage (oil)
Green et al 85 Male Mouse
10 mg/kg TCE Oral gavage (oil)
I—
30
—I—
40
—i	r~
50 60
Khr)
Green et al 85 Male Mouse
1000 mg/kg TCE Oral gavage (oiO
Green et al. 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
t(hr)
Green et al 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
Green et al 85 Male Mouse
2000 mg/kg TCE Oral gavage (oil)
Green et al. 85 Male Mouse
2000 mg/kg TCE Oral gavage (oil)
Green et al 85 Male Mouse
2000 mg/kg TCE Oral gavage (oil)
Green et aJ 85 Male Mouse
500 mg/kg TCE Oral gavage (oil)
Green et al 85 Male Mouse
500 mg/kg TCE Oral gavage (oil)
Green et al 85 Male Mouse
500 mg/kg TCE Oral gavage (oil)
lu o
Prout et al 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
Prout et al. 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
Prout et al 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-157 DRAFT—DO NOT CITE OR QUOTE
-------
Prout et al 85 Male Mouse
1000 mg/kg TCE Oral gavage (oil)
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al. 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
i	1	1	1	1	r
2.0 2.5 3.0 3.5 4.0 4.5
t (hr)
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
Is-

n	1	1	r
2.0 2.5 3.0 3.5 4.0 45
t (hr)
Greenberg et al. 99 Male Mouse
100 ppm TCE 4 hr inhalation
00-
E
h- CM-
O -

~i	1	r
2.0 2.5 3.0 3.5 4.0 4.5
t (hr)
Greenberg et al. 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al. 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
100 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
600 ppm TCE 4 hr inhalation
Greenberg et al 99 Male Mouse
0 ppm TCE 4 hr inhalation
N3
—I—
25
3.0 3.5
t (hr)
—I—
4 0
—r
4 5
Greenberg et al. 99 Male Mouse
600 ppm TCE 4 hr inhalation
t (hr)
Greenberg et al 99 Male Mouse
600 ppm TCE 4 hr inhalation
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-l58 DRAFT—DO NOT CITE OR QUOTE
-------
Greenberg et al. 99 Male Mouse	Greenberg et al. 99 Male Mouse
600 ppm TCE 4 hr inhalation	600 ppm TCE 4 hr inhalation
Greenberg et al. 99 Male Mouse	Greenberg et al 99 Male Mouse
600 ppm TCE 4 hr inhalation	600 ppm TCE 4 hr inhalation
Larson et al. 92a Male Mouse	Larson et al. 92a Male Mouse
197 mg/kg TCE Oral gavage (aq)	197 mg/kg TCE Oral gavage (aq)
Larson et al. 92a Male Mouse	Larson et al 92a Male Mouse
2000 mg/kg TCE Oral gavage (aq)	2000 mg/kg TCE Oral gavage (aq)
Larson et al. 92a Male Mouse	Larson et al. 92a Male Mouse
592 mg/kg TCE Oral gavage (aq)	592 mg/kg TCE Oral gavage (aq)
Greenberg et al 99 Male Mouse
600 ppm TCE 4 hr inhalation
	1	1	1	r
5 10 20	50
«(hr)
Greenberg et al. 99 Male Mouse
600 ppm TCE 4 hr inhalation
Larson et al. 92a Male Mouse
197 mg/kg TCE Oral gavage (aq)
~~i	1	1	1	r~
0.3 0.4 0.5 0.6 0.7
«(hr)
Larson et al. 92a Male Mouse
2000 mg/kg TCE Oral gavage (aq)
t (hr)
Larson et al. 92a Male Mouse
592 mg/kg TCE Oral gavage (aq)
~i	1	1	1	r~
0.5 1.0 2.0 5.0 10.0
t (hr)
~i	1	1	1	1	1	r
0.5 1.0 2.0 5.0 20.0 50.0
t (hr)
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-159 DRAFT—DO NOT CITE OR QUOTE
-------
Larson et al. 92b Male Mouse
20 mg/kg TCA Oral gavage (aq)
Larson et al. 92b Male Mouse
100 mg/kg TCA Oral gavage (aq)
Merdink et al. 98 Male Mouse
100 mg/kg TCE iv
0.5 1.0 2.0 5.0 10.0
t(hr)
50 10.0
Merdink et al 98 Male Mouse
100 mg/kg TCE iv
Templin et al. 93 Male Mouse
500 mg/kg TCE Oral gavage (aq)
Templin et al. 93 Male Mouse
500 mg/kg TCE Oral gavage (aq)
i	T	1
0.5 1.0 2.0
Templin et al. 93 Male Mouse
500 mg/kg TCE Oral gavage (aq)
This document is a draft for review purposes only and does not constitute Agency policy.
A-160 DRAFT—DO NOT CITE OR QUOTE
-------
1	Figure A-31. Comparison of mouse calibration data (boxes) and PBPK
2	model predictions (red line: using the posterior mean of the subject-specific
3	parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
4	50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-161 DRAFT—DO NOT CITE OR QUOTE
-------
A.4.2.2. Rat Data and Model Predictions
Bernauer et al. 96 Male Rat
40 ppm TCE 6 hr inhalation
Bernauer et al. 96 Male Rat
40 ppm TCE 6 hr inhalation
Bernauer et al. 96 Male Rat
40 ppm TCE 6 hr inhalation
-------
Fisher et al. 89 Female Rat
300 ppm TCE Closed Chamber
Fisher et al. 89 Female Rat
5100 ppm TCE Closed Chamber
~i	1	1	1	T
0.2 0.4 0.6 0.8 1.0 1.2 1.4
Fisher et al. 91 Female Rat
600 ppm TCE 4 hr inhalation
Fisher et al. 91 Female Rat
600 ppm TCE 4 hr inhalation
Fisher et al. 91 Female Rat
600 ppm TCE 4 hr inhalation
5 10
t (hr)
Fisher et al. 91 Male Rat
505 ppm TCE 4 hr inhalation
¦§> + -
E i?
—r~
5
t (hr)
—I—
10
~~r~
20
Fisher et al. 91 Male Rat
505 ppm TCE 4 hr inhalation
t (hr)
Green et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
t (hr)
Green et al. 85 Male Rat
10 mg/kg TCAiv
~l	1	1	1	
15 20 25 30
t (hr)
Green et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
t (hr)
Green et al. 85 Male Rat
75 mg/kg TCA Oral gavage (aq)
t(hr)
Green et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil) bile cannulated
t (hr)
Green et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil) bile cannulated
t (hr)
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
t (hr)
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
t (hr)
t (hr)
t (hr)
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-163 DRAFT—DO NOT CITE OR QUOTE
-------
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
t (hr)
Hissink et al. 02 Male Rat
10 mg/kg TCE iv
1	I	T
2.0	5.0 10.0
t (hr)
Hissink et al. 02 Male Rat
75 mg/kg TCE iv
~~r~
20
~~r~
50
—I—
100
t (hr)
Hissink et al. 02 Male Rat
100 mg/kg TCE Oral gavage (oil)
t (hr)
Hissink et al. 02 Male Rat
100 mg/kg TCE Oral gavage (oil)
20	50 100
t (hr)
Hissink et al. 02 Male Rat
1000 mg/kg TCE Oral gavage (oil)
t (hr)
Kaneko et al. 94 Male Rat
50 ppm TCE 6 hr inhalation
Hissink et al. 02 Male Rat
10 mg/kg TCE iv
~~r~
10
~~r~
20
~~r~
50
t (hr)
Hissink et al. 02 Male Rat
75 mg/kg TCE iv
Hissink et al. 02 Male Rat
1000 mg/kg TCE Oral gavage (oil)
I
20
I
50
t (hr)
Kaneko et al. 94 Male Rat
50 ppm TCE 6 hr inhalation
Kaneko et al. 94 Male Rat
50 ppm TCE 6 hr inhalation
Kaneko et al. 94 Male Rat
100 ppm TCE 6 hr inhalation
Kaneko et al. 94 Male Rat
100 ppm TCE 6 hr inhalation
E, -
<= P_
-r~
10
i—
20
~~r~
30
I—
40
t (hr)
—I—
10
—
20
~n—
30
-r~
40
t (hr)
t (hr)
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-164 DRAFT—DO NOT CITE OR QUOTE
-------
Keys et al. 03 Male Rat
50 ppm TCE 2 hr inhalation
Figure A-32. (Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25
50 75, and 97.5% population-based predictions) (continued).
Kaneko et al. 94 Male Rat
100 ppm TCE 6 hr inhalation
~i	1	1	T
10 20 30 40
t (hr)
Kaneko et al. 94 Male Rat
500 ppm TCE 6 hr inhalation
Keys et al. 03 Male Rat
8 mg/kg TCE ia
Keys et al. 03 Male Rat
8 mg/kg TCE ia
~l	1	1	1	T"
0.1 0.2 0.5 1.0 2.0
t (hr)
Keys et al. 03 Male Rat
50 ppm TCE 2 hr inhalation
Kaneko et al. 94 Male Rat
1000 ppm TCE 6 hr inhalation
Keys et al. 03 Male Rat
8 mg/kg TCE ia
—i	1	1	1	
10 20 30 40
t (hr)
Keys et al. 03 Male Rat
8 mg/kg TCE ia
o_
E m
Kaneko et al. 94 Male Rat
1000 ppm TCE 6 hr inhalation
~i	1	1	T
10 20 30 40
t (hr)
Kaneko et al. 94 Male Rat
500 ppm TCE 6 hr inhalation
Keys et al. 03 Male Rat
8 mg/kg TCE ia
.E m
Ui °"
o °_
—i i r~
0.2 0.5 1.0
t (hr)
Keys et al. 03 Male Rat
8 mg/kg TCE ia
Kaneko et al. 94 Male Rat
1000 ppm TCE 6 hr inhalation
Kaneko et al. 94 Male Rat
500 ppm TCE 6 hr inhalation
i	i	i	i	r
6	7	8	9 10
t Chr)
This document is a draft for review purposes only and does not constitute Agency policy.
A-165 DRAFT—DO NOT CITE OR QUOTE
-------
Keys et al. 03 Male Rat
50 ppm TCE 2 hr inhalation
~l	1	1	T"
0.2 0.5 1.0 2.0
t (hr)
Keys et al. 03 Male Rat
50 ppm TCE 2 hr inhalation
ni	1	1	1	r
0.2 0.5 1.0 2.0 5.0
t (hr)
Keys et al. 03 Male Rat
500 ppm TCE 2 hr inhalation
1	1	1	1	
0.2 0.5 1.0 2.0
t (hr)
Keys et al. 03 Male Rat
500 ppm TCE 2 hr inhalation
I	1	1	1	r
0.2 0.5 1.0 2.0 5.0
t (hr)
Keys et al. 03 Male Rat
8 mg/kg TCE Oral gavage (aq)
-------
Keys et al. 03 Male Rat
8 mg/kg TCE Oral gavage (aq)
Kimmerle et al. 73 Male Rat
49 ppm TCE 4 hr inhalation
Kimmerle et al. 73 Male Rat
49 ppm TCE 4 hr inhalation
—I
30
~r~
35
—I—
40
t (hr)
—I—
45
I—
50
Kimmerle et al. 73 Male Rat
54 ppm TCE 4 hr inhalation
Kimmerle et al. 73 Male Rat
49 ppm TCE 4 hr inhalation
Kimmerle et al. 73 Male Rat
175 ppm TCE 4 hr inhalation
40
t (hr)
—I—
45
I
50
Kimmerle et al. 73 Male Rat
175 ppm TCE 4 hr inhalation
> 4.4 4.6 4.8
t (hr)
Kimmerle et al. 73 Male Rat
175 ppm TCE 4 hr inhalation
~r~
30
I
40
I	
50
t (hr)
Kimmerle et al. 73 Male Rat
330 ppm TCE 4 hr inhalation
® io
111 o -
o o
"S>
_£_ o
0) o~
"a
E r
~i	1	1	r
9 10 11 12
t (hr)
Kimmerle et al. 73 Male Rat
330 ppm TCE 4 hr inhalation
~i	1	1	T
9 10 11 12
t (hr)
Kimmerle et al. 73 Male Rat
3000 ppm TCE 4 hr inhalation
-1—
4.2
4.4 4.6
t (hr)
—I—
4.8
~r~
40
I	
50
t (hr)
—I—
60
I
70
Kimmerle et al. 73 Male Rat
330 ppm TCE 4 hr inhalation
~~r~
30
I
40
I	
50
t (hr)
—I—
60
Kimmerle et al. 73 Male Rat
3000 ppm TCE 4 hr inhalation
Larson et al. 92A Male Rat
20 mg/kg TCA Oral gavage (aq)
Larson et al. 92A Male Rat
100 mg/kg TCA Oral gavage (aq)
~i	1	1	1	r
0.5 1.0 2.0 5.0 10.0
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-167 DRAFT—DO NOT CITE OR QUOTE
-------
Larson et al. 92B Male Rat
197 mg/kg TCE Oral gavage (aq)
XI o
•— (N —
< -
	1—
5
t (hr)
—I—
10
Larson et al. 92B Male Rat
3000 mg/kg TCE Oral gavage (aq)
t (hr)
Larson et al. 92B Male Rat
592 mg/kg TCE Oral gavage (aq)
Larson et al. 92B Male Rat
197 mg/kg TCE Oral gavage (aq)
Larson et al. 92B Male Rat
197 mg/kg TCE Oral gavage (aq)
t (hr)
Larson et al. 92B Male Rat
3000 mg/kg TCE Oral gavage (aq)
	1	
5
t (hr)
I
10
Larson et al. 92B Male Rat
592 mg/kg TCE Oral gavage (aq)
t (hr)
Larson et al. 92B Male Rat
3000 mg/kg TCE Oral gavage (aq)
t (hr)
Larson et al. 92B Male Rat
592 mg/kg TCE Oral gavage (aq)
I
10
I
50
t (hr)
Lee et al. 00A Male Rat
16 mg/kg TCE iv
t (hr)
Lee et al. 00A Male Rat
16 mg/kg TCE iv
t(hr)
Lee et al. 00A Male Rat
16 mg/kg TCE pv
t (hr)
Lee et al. 00A Male Rat
16 mg/kg TCE pv
0.04 0.06
t (hr)
Merdink et al. 99 Male Rat
100 mg/kg TCOH ivf_aq
t (hr)
Prout et al. 85 Male Rat
10 mg/kg TCE Oral gavage (oil)
*3)
	1	
0.06
	1—
0.08
-r~
40
t (hr)
t (hr)
	r~
50
t (hr)
—I—
60
~~r~
70
Figure A-32 Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-168 DRAFT—DO NOT CITE OR QUOTE
-------
Prout et al. 85 Male Rat
10 mg/kg TCE Oral gavage (oil)
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
~r~
30
I	
35
t (hr)
—I—
40
Prout et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
I—
40
r~
50
t (hr)
—I—
60
I
70
Prout et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
~~r~
25
I
30
35 40
t (hr)
—I—
45
Prout et al. 85 Male Rat
10 mg/kg TCE Oral gavage (oil)
~~r~
30
40 50 60
t (hr)
Prout et al. 85 Male Rat
10 mg/kg TCE Oral gavage (oil)
~~r~
25
~r~
30
I	
35
t (hr)
—I—
40
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
a> o
q-
O o
~~r~
40
	r~
50
t (hr)
—I—
60
~~r~
70
Prout et al. 85 Male Rat
1000 mg/kg TCE Oral gavage (oil)
~r~
30
I
40
50
t (hr)
Prout et al. 85 Male Rat
2000 mg/kg TCE Oral gavage (oil)
I
30
40 50
t (hr)
—I—
60
r~
70
Prout et al. 85 Male Rat
2000 mg/kg TCE Oral gavage (oil)
	r~
50
t (hr)
—I—
60
~r~
70
Prout et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
~~r~
30
I—
40
	r~
50
t (hr)
—I—
60
Prout et al. 85 Male Rat
500 mg/kg TCE Oral gavage (oil)
I
30
I—
40
	r~
50
t (hr)
—I—
60
	r~
50
t (hr)
—I—
60
~~r~
70
Simmons et al. 02 Male Rat
2000 ppm TCE 1 hr inhalation
~~r~
40
	r~
50
t (hr)
—I—
60
~~r~
70
Simmons et al. 02 Male Rat
2000 ppm TCE 1 hr inhalation
t (hr)
t (hr)
Figure A-32 Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-169 DRAFT—DO NOT CITE OR QUOTE
-------
Simmons et al. 02 Male Rat	Simmons et al. 02 Male Rat	Simmons et al. 02 Male Rat
2000 ppm TCE 1 hr inhalation	4000 ppm TCE 1 hr inhalation	4000 ppm TCE 1 hr inhalation
I	I	1	r
0.5 1.0 1.5 2.0
t (hr)
Simmons et al. 02 Male Rat
4000 ppm TCE 1 hr inhalation
t (hr)
Simmons et al. 02 Male Rat
1000 ppm TCE Closed Chamber
t (hr)
Simmons et al. 02 Male Rat
200 ppm TCE 1 hr inhalation
Simmons et al. 02 Male Rat
200 ppm TCE 1 hr inhalation
Simmons et al. 02 Male Rat
100 ppm TCE Closed Chamber
E o
Simmons et al. 02 Male Rat
500 ppm TCE Closed Chamber
>3> +-
E «
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCE id
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCE id
~i	1	1	1	T~
0.5 1.0 2.0 5.0 10.0
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv
o
U")
CN
O
in
o
0.5
1.0
1.5
2.0
t (hr)
Simmons et al. 02 Male Rat
200 ppm TCE 1 hr inhalation
o
OO
CD
(N
O
0.5
1.0
1.5
2.0
t (hr)
Simmons et al. 02 Male Rat
3000 ppm TCE Closed Chamber
o
o
o
CN
o
o
o
o
CN
0
1
2
3
4
5
6
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCE id
o
o
in
o
in
o
m
o
o
0.5
1.0
2.0
5.0
t (hr)
Figure A-32 Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-170 DRAFT—DO NOT CITE OR QUOTE
-------
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv
0.5 1.0 2.0
5.0 10.0 20.0
t (hr)
Stenner et al. 97 Male Rat
5 mg/kg TCOH iv bile cannulated
Stenner et al. 97 Male Rat
5 mg/kg TCOH iv bile cannulated
Stenner et al. 97 Male Rat
20 mg/kg TCOH iv bile cannulated
t (hr)
Stenner et al. 97 Male Rat
20 mg/kg TCOH iv bile cannulated
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv bile cannulated
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv bile cannulated
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv bile cannulated
t (hr)
Stenner et al. 97 Male Rat
100 mg/kg TCOH iv bile cannulated
t(hr)
Templin et al. 95 Male Rat
100 mg/kg TCE Oral gavage (aq)
I I I
0.5 1.0 2.0
i—i	r
20.0 50.0
t (hr)
Templin et al. 95 Male Rat
100 mg/kg TCE Oral gavage (aq)
t (hr)
Templin et al. 95 Male Rat
100 mg/kg TCE Oral gavage (aq)
t (hr)
Yu et al. 00 Male Rat
1 mg/kg TCA iv
"B)
E
~r
0.5
~r
2.0
~r
5.0
t (hr)
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-171 DRAFT—DO NOT CITE OR QUOTE
-------
Yu et al. 00 Male Rat
1 mg/kg TCA iv
Yu et al. 00 Male Rat
1 mg/kg TCA iv
Yu et al. 00 Male Rat
1 mg/kg TCA iv
~i—i	1—i—i	1—i—T
0.1	0.5 2.0 5.0 20.0
t (hr)
Yu et al. 00 Male Rat
10 mg/kg TCA iv
Yu et al. 00 Male Rat
10 mg/kg TCA iv
Yu et al. 00 Male Rat
10 mg/kg TCA iv
2.0 5.0
Yu et al. 00 Male Rat
10 mg/kg TCA iv
Yu et al. 00 Male Rat
50 mg/kg TCA iv
Yu et al. 00 Male Rat
50 mg/kg TCA iv
~r~
10
i
15
t (hr)
—I—
20
Yu et al. 00 Male Rat
50 mg/kg TCA iv
I	1	1	1	1	1
0.5 2.0 5.0 20.0
2.0 5.0
Yu et al. 00 Male Rat
50 mg/kg TCA iv
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model
predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-172 DRAFT—DO NOT CITE OR QUOTE
-------
Lee et al. 96 Male Rat	Lee et al. 96 Male Rat	Lee et al. 96 Male Rat
0.71 mg/kg TCE ia	8 mg/kg TCE ia	16mg/kgTCEia
t(hr)
Lee et al. 96 Male Rat
0.71 mg/kg TCE iv
1	1	r
0.20 0.50
t (hr)
Lee et al. 96 Male Rat
16 mg/kg TCE iv
~S)0
E?_
0.10 0.50 2.C
t (hr)
Lee et al. 96 Male Rat
8 mg/kg TCE pv
-1—i	1—i—i	T
0.10 0.50 2.00 10.00
t (hr)
Lee et al. 96 Male Rat
8 mg/kg TC E iv
"i—i—i	1—i—r~
0.10 0.50 2.00
t (hr)
Lee et al. 96 Male Rat
2 mg/kg TCE iv
i—i—i	i—i—r~
0.10 0.50 2.00
t (hr)
Lee et al. 96 Male Rat
64 mg/kg TCE iv
0.05	0.50	5.00
t (hr)
Lee et al. 96 Male Rat
16 mg/kg TCE pv
10.00
i—i—i	1—i—i	r
0.10 0.50 2.00 10.00
t (hr)
Lee et al. 96 Male Rat
0.71 mg/kg TCE pv
-1	1	1	1—i	1	r
0.05 0.20 0.50 2.00 5.00
t (hr)
Lee et al. 96 Male Rat
64 mg/kg TCE pv
0.50 2.00 10.00
0.10 0.50 2.00 10.00
t (hr)
Lee et al. 96 Male Rat
8 mg/kg TCE Oral gavage (aq)
Lee et al. 96 Male Rat
16 mg/kg TCE Oral gavage (aq)
Lee et al. 96 Male Rat
64 mg/kg TCE Oral gavage (aq)
10.00
10.00
This document is a draft for review purposes only and does not constitute Agency policy.
A-173 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al. 91 Male Rat
529 ppm TCE 4 hr inhalation
Jakobson et al. 86 Male Rat
500 ppm TCE 6 hr inhalation
Lee et al. 00B Male Rat
2 mg/kg TCE Oral gavage (aq)
—r~
0.5
t (hr)
Lee et al. 00B Male Rat
3 mg/kg TCE Oral gavage (aq)
1.0
t (hr)
—I—
2.0
0.20 0.50
2.00 5.00
Lee et al. 00B Male Rat
5 mg/kg TCE Oral gavage (aq)
Lee et al. 00B Male Rat
48 mg/kg TCE Oral gavage (aq)
"i—r
0.20 1.00
t (hr)
Lee et al. 00B Male Rat
144 mg/kg TCE Oral gavage (aq)
t (hr)
Lee et al. 00B Male Rat
432 mg/kg TCE Oral gavage (aq)
t (hr)
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
E
\ r
-0.02 0.10
0.50
t (hr)
I
2.00
i r
10.00
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
50 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
LU CN
Figure A-33. Comparison of rat evaluation data (boxes) and PBPK model
predictions (+ with error bars: single data points or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions).
This document is a draft for review purposes only and does not constitute Agency policy.
A-174 DRAFT—DO NOT CITE OR QUOTE
-------
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
w CM
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
Bruckner et al. XX Male Rat
500 ppm TCE 2 hr inhalation
DSouza et al. 85 Male Rat
5 mg/kg TCE iv
0.1 0.2
0.5 1.0 2.0
DSouza et al. 85 Male Rat
10 mg/kg TCE iv
DSouza et al. 85 Male Rat
25 mg/kg TCE iv
DSouza et al. 85 Male Rat
10 mg/kg TCE Oral gavage (aq)
LU CD
i—i—i	r
0.50 2.00 10.00
Andersen et al. 87 Male Rat
100 ppm TCE Closed Chamber
Andersen et al. 87 Male Rat
450 ppm TCE Closed Chamber
0.5 1.0 1.5
t (hr)
Andersen et al. 87 Male Rat
2000 ppm TCE Closed Chamber
t (hr)
Andersen et al. 87 Male Rat
5000 ppm TCE Closed Chamber
Andersen et al. 87 Male Rat
1000 ppm TCE Closed Chamber
~l	
2
t (hr)
t (hr)
t (hr)
Figure A-33. Comparison of rat evaluation data (boxes) and PBPK model
predictions (+ with error bars: single data points or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-175 DRAFT—DO NOT CITE OR QUOTE
-------
to
s§i
lo"
•	O -
CM.
c
3 -
*	°_
¦o csi
I :
Fisher et at. 1998 Human *1  _
£ o
< "I
O
1	 " 	1	1	1	1	1	1—
50 100 0.5 1.0 2.0 5.0 20.0 50.0
t (hr)
Figure A-33. Comparison of rat evaluation data (boxes) and PBPK model
predictions (+ with error bars: single data points or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-176 DRAFT—DG NOT CITE OR QUOTE
-------
A.4.2.3. Human Data and Model Predictions
This document is a draft for review purposes only and does not constitute Agency policy.
A-177 DRAFT—DO NOT CITE OR QUOTE
-------
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
Fisher et al.1998 Human #3 (sex=Male)
55.2 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
10.0
20.0
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
	1	
20.0
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
"9)
£ si
Fisher et al.1998 Human #3 (sex=Male)
55.2 ppm TCE 4 hr inhalation
I I	I I r
1 2	5 10 20
t (hr)
Fisher et al.1998 Human #3 (sex=Male)
55.2 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
i g-
o °
o
O)
.!
3 —
C 0_
T3 CN
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #3 (sex=Male)
101.5 ppm TCE 4 hr inhalation
*
E
~i	1	1	1	
1.0 2.0 5.0 10.0 20.0
t (hr)
Fisher et al.1998 Human #3 (sex=Male)
55.2 ppm TCE 4 hr inhalation
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions).
This document is a draft for review purposes only and does not constitute Agency policy.
A-178 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
E, LO-
TS d
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
^ S
O 9-
O m
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
5 °-
O o
O >o_
Fisher et al.1998 Human #4 (sex=Male)
53.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
"O o
a) o
qj + —
~i	1	1	1	r
5.0 20.0 50.0
t (hr)
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #4 (sex=Male)
97.8 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-179 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #5 (sex=Male)
105.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #6 (sex=Male
102.6 ppm TCE 4 hr inhalation
CD H
Fisher et al.1998 Human #8 (sex=Male)
102.6 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #6 (sex=Male)
102.6 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #6 (sex=Male)
102.6 ppm TCE 4 hr inhalation
n	1	1	T
0.5 1.0 2.0 5.0
~i	1	r
20.0 50.0
t (hr)
Fisher et al.1998 Human #6 (sex=Male)
102.6 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #6 (sex=Male)
102.6 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #6 (sex=Male)
102.6 ppm TCE 4 hr inhalation
~~T~
10
I
20
I
50
t (hr)
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TC E 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TC E 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TCE 4 hr inhalation
D U">
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-180 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #7 (sex=Male)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #8 (sex=Male)
101.1 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-181 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #9 (sex=Male)
103.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
~ m
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
E «
0.5 1.0 2.0 5.0 20.0 50.0
t (hr)
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
(j m
t (hr)
Fisher et al.1998 Human #10 (sex=Female)
55.1 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
~r~
10
t (hr)
—I—
10
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
	1	
20
t (hr)
—I—
50
—
100
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #10 (sex=Female)
101.4 ppm TCE 4 hr inhalation
XI q.
	1—
2.0
t (hr)
—I—
5.0
O CN
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-182 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
o
E q.
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
i	1—r
20.0 50.0
	1	1	1	r
5	10	15	20
t (hr)
Fisher et al.1998 Human #11 (sex=Female)
97.7 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #11 (sex=Female)
53 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-183 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #12 (sex=Female)
102.5 ppm TCE 4 hr inhalation
I
10
t (hr)
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
20
t (hr)
—I—
50
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
O iO
5 0.5
1.0 2.0 5.0
I
20.0
t (hr)
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
t(hr)
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
~~r~
10
t (hr)
LU rsl
Fisher et al.1998 Human #13 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
<
o pj-
h- o
20
t (hr)
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
I
10
I
12
~~r
14
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-184 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
I—
1.0
5.0
t (hr)
	1	
20.0
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #14 (sex=Female)
102 ppm TCE 4 hr inhalation
O m
I
10
20
t (hr)
—I—
50
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
~~r~
10
—i—
20
t (hr)
—I—
50
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
t(hr)
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
1.0
2.0
t (hr)
—I—
5.0
—I—
10.0
r
3
t (hr)
Fisher et al.1998 Human #15 (sex=Female)
101 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
<
o ¦<-.
o
-r~
10
—i—
20
t (hr)
—I—
50
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-185 DRAFT—DO NOT CITE OR QUOTE
-------
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
!?:
E,  _
I.5 1.0 2.0	5.0
t (hr)
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
Paykoc et al.1945 Human #18 (sex=unknown)
32.9 mg/kg TCA 1 hr IV
I.5 1.0 2.0 5.0 10.0 20.0
t (hr)
Fisher et al.1998 Human #16 (sex=Female)
103.3 ppm TCE 4 hr inhalation
~i	1	1—i—i	1—T
1.0 2.0 5.0 20.0 50.0
t (hr)
o-
E ^
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
t (hr)
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
"I	1	1	1—
0.5 1.0 2.0	5.0
t (hr)
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
o o
Fisher et al.1998 Human #17 (sex=Female)
102 ppm TCE 4 hr inhalation
"I	1	1	1	T"
1 2	5 10 20
t (hr)
Paykoc et al.1945 Human #18 (sex=unknown)
32.9 mg/kg TCA 1 hr IV
	1	1	1	1	r
5 10 20 50 100
t (hr)
Paykoc et al.1945 Human #18 (sex= unknown)
32.9 mg/kg TCA 1 hr IV
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-186 DRAFT—DO NOT CITE OR QUOTE
-------
Paykoc et al.1945 Human #19 (sex=unknown)
53.06 mg/kg TCA 1 hr IV
Paykoc et al.1945 Human #19 (sex=unknown)
53.06 mg/kg TCA 1 hr IV
Paykoc et al.1945 Human #19 (sex= unknown)
53.06 mg/kg TCA 1 hr IV
i	1	1	1	r~
40 60 80 100 120
t (hr)
Paykoc et al.1945 Human #20 (sex=unknown)
24.8 mg/kg TCA 1 hr IV
Paykoc et al.1945 Human #20 (sex=unknown)
24.8 mg/kg TCA 1 hr IV
50	100 150 200
t (hr)
Kimmerle and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
3 and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #21 (sex=Female)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #22 (sex=Female)
44 ppm TCE 4 hr inhalation
3 and Eben 1973 Human #22 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #22 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #22 (sex= Female)
44 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-187 DRAFT—DO NOT CITE OR QUOTE
-------
Kimmerle and Eben 1973 Human #22 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #22 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #23 (sex= Female)
44 ppm TCE 4 hr inhalation
3 and Eben 1973 Human #23 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #23 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #23 (sex= Female)
44 ppm TCE 4 hr inhalation
di o
Kimmerle and Eben 1973 Human #23 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #23 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #24 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #24 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #24 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #24 (sex=Female)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-188' DRAFT—DO NOT CITE OR QUOTE
-------
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #25 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
$
	1	1	1	
150	200	250
t (hr)
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
o-
o
o -
	1	1	1—
150	200	250
t (hr)
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #27 (sex=Male)
40 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #26 (sex=Male)
40 ppm TCE 4 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-189 DRAFT—DO NOT CITE OR QUOTE
-------
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #28 (sex=Male)
44 ppm TCE 4 hr inhalation
Kimmerle and Eben 1973 Human #29 (sex=unknown) ^ Kimmerle and Eben 1973 Human #29 (sex=unknown) Kimmerle and Eben 1973 Human #29 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
i i i i i i i r
20 40 60 80 100 140	<
t (hr)	I
Kimmerle and Eben 1973 Human #29 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
n	i	1	i	1	r
250 300 350 400 450 500 550
t (hr)
Kimmerle and Eben 1973 Human #29 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
~r~
50
—r~
100
150
t (hr)
—I—
200
Kimmerle and Eben 1973 Human #29 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
~~r~
10
20
t (hr)
—I—
50
~~r~
60
~l	1—
80 100
t (hr)
~i	1	1	1	1	r
250 300 350 400 450 500 550
t (hr)
Kimmerle and Eben 1973 Human #30 (sex=unknown) ~ Kimmerle and Eben 1973 Human #30 (sex=unknown) Kimmerle and Eben 1973 Human #30 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
150
t (hr)
~i	1	1	1	1	1	r
250 300 350 400 450 500 550
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-190 DRAFT—DO NOT CITE OR QUOTE
-------
Kimmerle and Eben 1973 Human #30 (sex=unknown) Kimmerle and Eben 1973 Human #30 (sex=unknown) Kimmerle and Eben 1973 Human #30 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days	48 ppm TCE 4 hr inhalation for 5 days	48 ppm TCE 4 hr inhalation for 5 days
~~r~
50
t (hr)
150
t (hr)
Kimmerle and Eben 1973 Human #31 (sex=unknown) ~ Kimmerle and Eben 1973 Human #31 (sex=unknown) Kimmerle and Eben 1973 Human #31 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
48 ppm TCE 4 hr inhalation for 5 days
Kimmerle and Eben 1973 Human #31 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
~i	1	1	1	1	T
250 300 350 400 450 500 550
Kimmerle and Eben 1973 Human #31 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
Kimmerle and Eben 1973 Human #31 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
I
10
20
t (hr)
—I—
50
O) lO	
Kimmerle and Eben 1973 Human #32 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
t (hr)
Kimmerle and Eben 1973 Human #32 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
t(hr)
Kimmerle and Eben 1973 Human #32 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
I—
50
150
t (hr)
—r~
250
~~r~
10
~~r~
20
Kimmerle and Eben 1973 Human #32 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
t (hr)
Kimmerle and Eben 1973 Human #32 (sex=unknown)
48 ppm TCE 4 hr inhalation for 5 days
t (hr)
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
I—
50
150
t (hr)
—r~
250
I I I I I I
250 300 350 400 450 500 550
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-191 DRAFT—DO NOT CITE OR QUOTE
-------
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
20
t (hr)
—I—
50
<
o 10.
I- o
~r
10
~T~
20
~r
50
t (hr)
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #33 (sex=Male)
65 ppm TCE 4 hr inhalation
i	1	1	1	1	1	1	r
20 30 40 50 60 70 80 90
t (hr)
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
I
20
~~r~
30
~1~
40
~~r~
50
t (hr)
—I—
60
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
~~r~
10
~~r~
20
~~r~
30
~~I—
40
t (hr)
—I—
50
~~r~
60
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
Monster et al.1976 Human #33 (sex=Male)
140 ppm TCE 4 hr inhalation
o m
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
I—
20
~~r~
30
—
40
nr~
50
t (hr)
—I—
60
-r~
70
O -
o o
i- o-
-r~
20
~i	1	1	r~
30 40 50 60
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-192 DRAFT—DO NOT CITE OR QUOTE
-------
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
o _
o o
-Q m -
~i	r
) 50 100 150 200
t (hr)
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
I
10
I
20
I
50
t (hr)
Monster et al.1976 Human #34 (sex=Male)
68 ppm TCE 4 hr inhalation
E <=!
^ O"
£ in
~r~
10
t (hr)
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
3.5 4.0 4.5
t (hr)
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
~~r~
20
~~r~
30
40 50
t (hr)
—I—
70
l
80
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
—I—
5.0
2.5 3.0 3.5 4.0 4.5
t (hr)
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
I
10
I
50
t(hr)
Monster et al.1976 Human #34 (sex=Male)
138 ppm TCE 4 hr inhalation
o (N
t (hr)
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
-r~
10
t (hr)
	r~
20
t (hr)
—I—
50
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-193 DRAFT—DO NOT CITE OR QUOTE
-------
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #35 (sex=Male)
70 ppm TCE 4 hr inhalation
~i	1	1	1	T
3.0 3.5 4.0 4.5 5.0
t (hr)
Monster et al.1976 Human #35 (sex=Male)
142 ppm TCE 4 hr inhalation
t (hr)
Monster et al.1976 Human #35 (sex=Male)
142 ppm TCE 4 hr inhalation
n LO
Monster et al.1976 Human #35 (sex=Male)
142 ppm TCE 4 hr inhalation

~r~
20
~~r~
40
—I—
60
t (hr)
—I—
80
Monster et al.1976 Human #35 (sex=Male)
142 ppm TCE 4 hr inhalation
> o
.E O-
~~r~
10
	r~
20
t (hr)
Monster et al.1976 Human #35 (sex=Male)
142 ppm TCE 4 hr inhalation
o o
o i -
2 
-------
Monster et al.1976 Human #36 (sex=Male)
76 ppm TCE 4 hr inhalation
Monster et al.1976 Human #36 (sex=Male)
76 ppm TCE 4 hr inhalation
Monster et al.1976 Human #36 (sex=Male)
76 ppm TCE 4 hr inhalation
~i	1	T
3.0 3.5 4.0
t (hr)
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
E -
w o
9> r-i-
I
20
I
40
t (hr)
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
I
60
t (hr)
I
80
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
O m
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
	r~
20
t (hr)
—I—
50
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
Monster et al.1976 Human #36 (sex=Male)
140 ppm TCE 4 hr inhalation
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
t(hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
~~r~
20
~~r~
40
—I	
60
t (hr)
—I—
80
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
E o_
—r~
10
t (hr)
—I—
20
o _
o b
I5
~~r~
40
I—
60
~~r~
80
—T"
120
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
I I I I I
2.0 5.0 20.0
100.0
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-195 DRAFT—DO NOT CITE OR QUOTE
-------
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
•§• o
dj LD —
.E O

~I
20
~~r~
40
~~r~
60
100 120 140
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
I I I
20 50 100
III III II
0.5 2.0 5.0 20.0 50.0
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
i	1	r
80 100 120
Chiu et al. 2007 Human #37 (sex=Male)
1 ppm TCE 6 hr inhalation
5 10 20
t (hr)
—I	T"
50 100
I
20
I
50
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
t(hr)
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
o m
t (hr)
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
E .
-r~
10
-r~
20
O
i	1	r
20 50 100
t (hr)
t (hr)
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-196 DRAFT—DO NOT CITE OR QUOTE
-------
t (hr)
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #38 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
i o
O 9
o m_
-------
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
—i	1	1	1	r~
2 5 10 20 50
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
n—i—i	1—i—i	1—T
0.5 2.0 5.0 20.0 100.0
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
I I I I I I I \
0.5 2.0 5.0 20.0 100.0
t (hr)
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
	1	1	1	1	1—
40 60 80 100 120
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
—i	1	1	1	r
2 5 10 20 50
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
—i	1	1	1	1	T
2 5 10 20 50 100
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #39 (sex=Male)
1 ppm TCE 6 hr inhalation
n	i	i	i
10 20	50 100
t (hr)
Chiu et al. 2007 Human #40 (sex=Male)
1 ppm TCE 6 hr inhalation
E o
5 10 15 20 25 30 35 40
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
"B> rS~
E o _
1
2
5 10 20
50 100
t (hr)
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-198 DRAFT—DO NOT CITE OR QUOTE
-------
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
E
20
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
O o
20
100
t (hr)
t
i
~i	1	—i	1	
20 25 30 35 40 45
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
n—r
100.0
120
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
i ®-r
O
P in-!
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
"i	1	1	1	T
5 10 15 20 25
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
~~I l l l l l l r
0.5 2.0 5.0 20.0 100.0
t (hr)
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
—i	1	1	1	i	T~
2 5 10 20 50 100
t (hr)
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
"i	1	1	1	1	r
1 2 5 10 20 50
t (hr)
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #41 (sex=Male)
1 ppm TCE 6 hr inhalation
o
o
o.
o
o
o.
o
o
d
50 100
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-199 DRAFT—DO NOT CITE OR QUOTE
-------
—I	1	1	1	1—
10 20 30 40 50
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
80 ppm TCE 6 hr inhalation
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
Bernauer et al.1996 Human #43 (sex=Male)
40 ppm TCE 6 hr inhalation
Bernauer et al.1996 Human #43 (sex=Male)
80 ppm TCE 6 hr inhalation
Bernauer et al.1996 Human #43 (sex=Male)
160 ppm TCE 6 hr inhalation
Muller et al.1974 Human #44 (sex=Male)
10 mg/kg TCOH oral
§ 8"
U CN-
1	1	
100.0
	1	1	1	1	1	
10 20 30 40 50
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
160 ppm TCE 6 hr inhalation
—i	1	1	1	T~
10 20 30 40 50
t (hr)
Muller et al.1974 Human #44 (sex=Male)
2.646 mg/kg TCA oral
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
40 ppm TCE 6 hr inhalation
	1	1	1	1	1	
10 20 30 40 50
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
160 ppm TCE 6 hr inhalation
~i	1	1	1	r~
10 20 30 40 50
t (hr)
Muller et al.1974 Human #44 (sex=Male)
2.646 mg/kg TCA oral
Chiu et al. 2007 Human #42 (sex=Male)
1 ppm TCE 6 hr inhalation
"i	1	1	1	1	T
1 2 5 10 20 50
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
40 ppm TCE 6 hr inhalation
	1	1	1	1	1	
10 20 30 40 50
t (hr)
Bernauer et al.1996 Human #43 (sex=Male)
80 ppm TCE 6 hr inhalation
Figure A-34. Comparison of human calibration data (boxes) and PBPK
model predictions (red line: using the posterior mean of the subject-specific
parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-200 DRAFT—DO NOT CITE OR QUOTE
-------
Muller et al.1974 Human #44 (sex=Male)	Muller et al.1974 Human #44 (sex=Male)	Muller et al.1974 Human #44 (sex=Male)
10 mg/kg TCOH oral	10 mg/kg TCOH oral	10 mg/kg TCOH oral
0.5
2.0 5.0
t (hr)
20.0 50.0
140
t (hr)
< o
0.5
2.0 5.0
20.0
100.0
1	Figure A-34. Comparison of human calibration data (boxes) and PBPK
2	model predictions (red line: using the posterior mean of the subject-specific
3	parameters; + with error bars: single data points; or shaded regions: 2.5, 25,
4	50, 75, and 97.5% population-based predictions) (continued).
5
This document is a draft for review purposes only and does not constitute Agency policy.
A-201 DRAFT—DO NOT CITE OR QUOTE
-------
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
I
50
~r~
60
70
t (hr)
—I—
80
r~
90
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
F. c
~r~
60
70
t (hr)
—I—
80
~~r~
90
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #1 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
o
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
E
w o _
JS c\i
I
50
70
t (hr)
I
80
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #2 (sex=unknown)
200 ppm TCE 5 hr inhalation
p o_
° (N
<
I
50
I
70
t (hr)
—I—
80
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
—r
100
This document is a draft for review purposes only and does not constitute Agency policy.
A-202 DRAFT—DO NOT CITE OR QUOTE
-------
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #3 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #4 (sex=unknown)
200 ppm TCE 5 hr inhalation
O cn
Bartonicek 1962 Human #4 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #4 (sex=unknown)
200 ppm TCE 5 hr inhalation
I
50
I
70
t (hr)
—I—
80
—r~
100
Bartonicek 1962 Human #4 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #4 (sex=unknown)
200 ppm TCE 5 hr inhalation
t(hr)
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
t (hr)
t (hr)
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions).
This document is a draft for review purposes only and does not constitute Agency policy.
A-203 DRAFT—DO NOT CITE OR QUOTE
-------
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #5 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
I
50
I
70
t (hr)
—I—
80
—r~
100
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #6 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
t (hr)
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-204 DRAFT—DO NOT CITE OR QUOTE
-------
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #7 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #8 (sex=unknown)
200 ppm TCE 5 hr inhalation
I
50
I—
60
70
t (hr)
—I—
80
—I
90
Bartonicek 1962 Human #8 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #8 (sex=unknown)
200 ppm TCE 5 hr inhalation
Bartonicek 1962 Human #8 (sex=unknown)
200 ppm TCE 5 hr inhalation
~i	1	1	1	1	T
50 60 70 80 90 100
Bartonicek 1962 Human #8 (sex=unknown)
200 ppm TCE 5 hr inhalation
Lapare et al.1995 Human #9 (sex=unknown)
multiple ppm TCE inhalation
Lapare et al.1995 Human #9 (sex=unknown)
multiple ppm TCE inhalation
3 o
c °H
- CO
500
t (hr)
—I—
600
Lapare et al.1995 Human #9 (sex=unknown)
multiple ppm TCE inhalation
t (hr)
Lapare et al.1995 Human #9 (sex=unknown)
multiple ppm TCE inhalation
1 (hr)
Lapare et al.1995 Human #10 (sex=unknown)
multiple ppm TCE inhalation
t (hr)
Lapare et al.1995 Human #10 (sex=unknown)
multiple ppm TCE inhalation
t (hr)
Lapare et al.1995 Human #10 (sex=unknown)
multiple ppm TCE inhalation
t (hr)
Lapare et al.1995 Human #10 (sex=unknown)
multiple ppm TCE inhalation
Figure A-35 Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-205 DRAFT—DO NOT CITE OR QUOTE
-------
Lapare et al.1995 Human #11 (sex=unknown)
multiple ppm TCE inhalation
Lapare et al.1995 Human #11 (sex=unknown)
multiple ppm TCE inhalation
Lapare et al.1995 Human #11 (sex=unknown)
multiple ppm TCE inhalation
100 150
t(hr)
Lapare et al.1995 Human #11 (sex=unknown)
multiple ppm TCE inhalation
Lapare et al.1995 Human #12 (sex=unknown)
multiple ppm TCE inhalation
850 900 950 1000
t (hr)
Lapare et al.1995 Human #12 (sex=unknown)
multiple ppm TCE inhalation
~Q) °
E
~l	1	
850	900	950	10C
t (hr)
Lapare et al.1995 Human #12 (sex=unknown)
multiple ppm TCE inhalation
Lapare et al.1995 Human #12 (sex=unknown)
multiple ppm TCE inhalation
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
I—
850
—I—
900
—I—
950
—r o —i	1	1	1	1	1	r
1000 O 30 35 40 45 50 55 60
t (hr)
t (hr)
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
t(hr)
Bloemen et al.2001 Human #13 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
I—
1.0
I—
1.4
I—
1.6
t (hr)
Bloemen et al.2001 Human #13 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
~i	1	1	1	1	r
30 35 40 45 50 55 60
t (hr)
Bloemen et al.2001 Human #13 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
t (hr)
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
—I—
2.5
3.0 3.5
t (hr)
—I—
4.0
O (N
t (hr)
t (hr)
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-206 DRAFT—DO NOT CITE OR QUOTE
-------
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
Bloemen et al.2001 Human #13 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
i	1	T
30 35 40 45 50 55 60
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #14 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
t (hr)
Bloemen et al.2001 Human #14 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
3 o_



excreted in urine i
0 50
TCA exc
0	10
1	I



I I I I I I I
30 35 40 45 50 55 60
t (hr)
0
O
o
i-
Bloemen et al.2001 Human #14 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
o in
o (n
t (hr)
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
t (hr)
Bloemen et al.2001 Human #14 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
t(hr)
Bloemen et al.2001 Human #15 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #15 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #15 (sex=Male)
100 ppm TCE 15 min inhalation 4 times
Bloemen et al.2001 Human #15 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
i	r
1.8 2.0
<
o
i- o-
-r~
30
-r~
40
—
60
t (hr)
t (hr)
t (hr)
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-207 DRAFT—DO NOT CITE OR QUOTE
-------
Bloemen et al.2001 Human #15 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
Bloemen et al.2001 Human #15 (sex=Male)
50 ppm TCE 15 min inhalation 8 times
Bloemen et al.2001 Human #15 (sex=Male)
100 ppm TCE 15 min inhalation 8 times
i o
O u"> —
O 
-------
Fernandez etal.1977 Human #18 (sex=Male)
160 ppm TCE 8 hr inhalation
E"-
D- O
Q. T	
i- O
"O T~-
~~r~
20
I	
30
t (hr)
—I—
50
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
1	1	1	1	1	1	1	
0 50 100 150 200 250 300
t (hr)
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
~r~
10
—i—
20
t (hr)
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
	r~
10
t (hr)
—I—
20
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
10
t (hr)
—I—
20
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
t (hr)
Monster et al.1979 Human #19 (sex=Male)
70 ppm TCE 4 hr inhalation for 5 days
t (hr)
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
t(hr)
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
	r~
50
t (hr)
—I—
60
~r~
70
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
t (hr)
Muller et al.1974 Human #20 (sex=Male)
100 ppm TCE 6 hr inhalation
t (hr)
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation
I—
30
	r~
50
t (hr)
—I—
60
I
70
I—
10
	1—
15
t (hr)
—I—
20
T
25
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-209 DRAFT—DO NOT CITE OR QUOTE
-------
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation
Muller et al.1975 Human #21 (sex=Male)
100 ppm TCE 6 hr inhalation

5	10
t (hr)
—I—
20
Muller et al.1975 Human #22 (sex=Male)
50 ppm TCE 6 hr inhalation for 5 days
t (hr)
Muller et al.1975 Human #22 (sex=Male)
50 ppm TCE 6 hr inhalation for 5 days
200
t (hr)
Stewart et al.1970 Human #23 (sex=unknown)
200 ppm TCE 7 hr inhalatino for 5 days
~~r~
10
i
15
~r~
20
t (hr)
Muller et al.1975 Human #22 (sex=Male)
50 ppm TCE 6 hr inhalation for 5 days
Stewart et al.1970 Human #23 (sex=unknown)
200 ppm TCE 7 hr inhalatino for 5 days
i	T
100 120
t (hr)
Muller et al.1975 Human #22 (sex=Male)
50 ppm TCE 6 hr inhalation for 5 days
~~r~
40
~r~
60
~~I	1	1—
80 100 120
t (hr)
Stewart et al.1970 Human #23 (sex=unknown)
200 ppm TCE 7 hr inhalatino for 5 days
x o-
O g
P s
I—
50
150
t (hr)
—r~
250
Treibig et al.1976 Human #24 (sex=Male)
136 ppm TCE 6 hr inhalation
-r~
10
—i—
15
t (hr)
—I—
20
O (N _
I- O
t (hr)
Treibig et al.1976 Human #24 (sex=Male)
136 ppm TCE 6 hr inhalation
t (hr)
Treibig et al.1976 Human #24 (sex=Male)
136 ppm TCE 6 hr inhalation
t (hr)
t (hr)
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
This document is a draft for review purposes only and does not constitute Agency policy.
A-210 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
Treibig et al.1976 Human #24 (sex=Male)
136 ppm TCE 6 hr inhalation
Sato et al.1977 Human #25 (sex=Male)
100 ppm TCE 4 hr inhalation
t (hr)
Sato et al.1977 Human #25 (sex=Male)
100 ppm TCE 4 hr inhalation
t (hr)
Sato et al.1977 Human #25 (sex=Male)
100 ppm TCE 4 hr inhalation
=3,c
E r
I—
10
t (hr)
Fernandez et al.1977 Human #26 (sex=Male)
54 ppm TCE 8 hr inhalation
t (hr)
Fernandez et al.1977 Human #26 (sex=Male)
54 ppm TCE 8 hr inhalation
E +-
q. a>
o. ID
~~r~
20
I	
30
t (hr)
I
40
~~r~
50
~~r~
10
~~r~
20
~~r~
50
t (hr)
Fernandez et al.1977 Human #27 (sex=Male)
97 ppm TCE 8 hr inhalation
Fernandez et al.1977 Human #27 (sex=Male)
97 ppm TCE 8 hr inhalation
~~r~
20
30
t (hr)
—I—
40
~r~
50
~~r~
20
50
t (hr)
100
Sato et al.1977 Human #25 (sex=Male)
100 ppm TCE 4 hr inhalation
~~T"
10
t (hr)
O in
Fernandez et al.1977 Human #26 (sex=Male)
54 ppm TCE 8 hr inhalation
~~r~
10
i—
20
~~r~
50
t (hr)
Fernandez et al.1977 Human #27 (sex=Male)
97 ppm TCE 8 hr inhalation
t (hr)
200
Figure A-35. Comparison of human evaluation data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions) (continued).
A.4. EVALUATION OF RECENTLY PUBLISHED TOXICOKINETIC DATA
5	Several in vivo toxicokinetic studies were published or became available during internal
6	EPA review and Interagency Consultation, and were not evaluated as part of the originally
7	planned analyses. Preliminary analyses of these data are summarized here. The general
8	approach is the same as that used for the evaluation data in the primary analysis—population
9	predictions from the PBPK model are compared visually with the toxicokinetic data.
This document is a draft for review purposes only and does not constitute Agency policy.
A-211 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
A.4.3. Trichloroethylene (TCE) Metabolite Toxicokinetics in Mice: Kim et al. (2009)
Kim et al. (2009) measured TCA, DCA, DCVG, and DCVC in blood of male B6C3F1
mice following a single gavage dose of 2,140 mg/kg. Of these data, only TCA and DCVG blood
concentrations are predicted by the updated PBPK model, so only those data are compared with
PBPK model predictions (prior values for the distribution volume and elimination rate constant
of DCVG were used, as there were no calibration data informing those parameters). These data
were within the interquartile region of the PBPK model population predictions, as shown in
Figure A-36.
Figure A-36. Comparison of Kim et al. (2009) mouse data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions).
Kim et al. (2009) Male Mouse
2140 mg/kg/d TCE corn oil gavage (single dose)
Kim et al. (2009) Male Mouse
2140 mg/kg/d TCE corn oil gavage (single dose)
o r\j
< LD
E _
E
"O
2 o
O .
>
o
Q oo
i	r
10.0 20.0
-------
o
o
CD
^r
o
o
o
^r
O
13
u
CM
O
o
o
0	2 5 4 10 6 15 8	20 12
t(hr)
1	Figure A-37. Comparison of best-fitting (out of 50,000 posterior samples)
2	PBPK model prediction and Kim et al. (2009) TCA blood concentration data
3	for mice gavaged with 2,140 mg/kg TCE.
4
This document is a draft for review purposes only and does not constitute Agency policy.
A-213 DRAFT—DO NOT CITE OR QUOTE
-------
10
10
10
10w
oral exposure (mg/kg/d continuous)
Figure A-38. Comparison of best-fitting (out of 50,000 posterior samples)
PBPK model prediction and Kim et al. (2009) DCVG blood concentration
data for mice gavaged with 2,140 mg/kg TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
A-214 DRAFT—DO NOT CITE OR QUOTE
-------
1	Figure A-39. PBPK model predictions for the fraction of intake undergoing
2	GSH conjugation in mice continuously exposed orally to TCE. Lines and
3	error bars represent the median and 95th percentile confidence interval for the
4	posterior predictions, respectively (also reported in Section 3.5.7.2.1). Filled
5	circles represent the predictions from the sample (out of 50,000 total posterior
6	samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG
7	blood concentration data for mice gavaged with 2,140 mg/kg TCE.
8
This document is a draft for review purposes only and does not constitute Agency policy.
A-215 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
10 1	1	101	102	103
Inhalation exposure (ppm continuous)
Figure A-40. PBPK model predictions for the fraction of intake undergoing
GSH conjugation in mice continuously exposed via inhalation to TCE. Lines
and error bars represent the median and 95th percentile confidence interval for the
posterior predictions, respectively (also reported in Section 3.5.7.2.1). Filled
circles represent the predictions from the sample (out of 50,000 total posterior
samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG
blood concentration data for mice gavaged with 2,140 mg/kg TCE.
compared to the full posterior distribution (see Figures A-15 and A-16). The predictions for this
"best fitting" parameter set was similar (within threefold) of the median of the full posterior
distribution (see Figures A-39 and A-40). While a formal assessment of the impact of these new
data (i.e., including its uncertainty and variability) would require a rerunning of the Bayesian
analysis, it appears that the median estimates for the mouse GSH conjugation dose metric used in
the dose-response assessment (see Chapter 5) are reasonably consistent with the Kim et al.
(2009) data.
An additional note of interest from the Kim et al. (2009) data is the interstudy variability
in TCA kinetics. In particular, the TCA blood concentrations reported by Kim et al. (2009) are
twofold lower than those reported by Abbas and Fisher (1997) in the same sex and strain of
This document is a draft for review purposes only and does not constitute Agency policy.
A-216 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
mouse, with a very similar corn oil gavage dose of 2,000 mg/kg (as compared to 2,140 mg/kg
used in S. Kim et al., 2009).
A.4.4. Trichloroethylene (TCE) Toxicokinetics in Rats: Liu et al. (2009)
Liu et al. (2009) measured TCE in blood of male rats after treatment with TCE by i.v.
injection (0.1, 1.0, or 2.5 mg/kg) or aqueous gavage (0.0001, 0.001, 0.01, 0.1, 1, 2.5, 5, or
10 mg/kg). Almost all of the data from gavage exposures were within the interquartile region of
the PBPK model population predictions, with all of it within the 95% confidence interval, as
shown in Figure A-41 For i.v. exposures, the data at 1 and 2.5 mg/kg were well simulated, but
the time-course data at 0.1 mg/kg were substantially different in shape from that predicted by the
PBPK model, with a lower initial concentration and longer half-life. The slower elimination rat
at 0.1 mg/kg was noted by the study authors through use of noncompartamental analysis. There
is no clear explanation for this discrepancy, particularly since the gavage data at this and even
lower doses were well predicted by the PBPK model.
A.4.5. Trichloroacetic Acid (TCA) Toxicokinetics in Mice and Rats: Mahle et al. (2001)
and Green (2003a, 2003b)
Three technical reports (Green 2003a, 2003b; Mahle et al., 2001) described by Sweeney
et al. (2009) contained data on TCA toxicokinetics in mice and rats exposed to TCA in drinking
water. These technical reports were provided to EPA by the Sweeney et al. (2009) authors.
A.4.5.1. Analysis Using Evans et al. (2009) and Chiu et al. (2009) Physiologically Based
Pharmacokinetic (PBPK) Model
TCA blood and liver concentrations were reported by Mahle et al. (2001) for male
B6C3F1 mice and male Fischer 344 rats exposed to 0.1 g/L to 2 g/L TCA in drinking water for 3
or 14 days (12-270 mg/kg-day in mice and 7-150 mg/kg-day in rats). For mice, these data were
all within the 95% confidence interval of PBPK model population predictions, with about half of
these data within the interquartile region. For rats, all these data, except those for the 3-day
exposure at 0.1 g/L, were within the 95% confidence interval of the PBPK model predictions. In
addition, the median rat predictions were consistently higher than the data, although this could be
explained by interstudy (strain, lot, etc.) variability.
This document is a draft for review purposes only and does not constitute Agency policy.
A-217 DRAFT—DO NOT CITE OR QUOTE
-------
1	TCA blood concentrations were reported by Green (2003a) for male and female B6C3F1
2	mice exposed to 0.5 g/L to 2.5 g/L TCA in drinking water for 5 days (130-600 mg/kg-day in
3	males and 160-750 mg/kg-day in females). Notably, these animals consumed around twice as
4	much water per day as compared to the mice reported by Mahle et al. (2001), and therefore
5	received comparatively higher doses of TCA for the same TCE concentration in drinking water.
This document is a draft for review purposes only and does not constitute Agency policy.
A-218 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
0
1
Liu et al. (2009) Male Rat
0.1 mg/kg TCE iv 0.1 mg/kg
Liu et al. (2009) Male Rat
1 mg/kg TCE iv 1 mg/kg
Liu et al. (2009) Male Rat
2.5 mg/kg TCE iv 2.5 mg/kg
Liu et al. (2009) Male Rat
0.001 mg/kg TCE aqueous gavage 0.001 mg/kg
Liu et al. (2009) Male Rat
0.01 mg/kg TCE aqueous gavage 0.01 mg/kg
Liu et al. (2009) Male Rat
0.1 mg/kg TCE aqueous gavage 0.1 mg/kg
"i	1	T
0.05 0.10 0.20
—I—I—I	1—I—I	T~
0.05 0.20 0.50 2.00 5.00
t(hr)
Liu et al. (2009) Male Rat
1 mg/kg TCE aqueous gavage 1 mg/kg
Liu et al. (2009) Male Rat
2.5 mg/kg TCE aqueous gavage 2.5 mg/kg
Liu et al. (2009) Male Rat
5 mg/kg TCE aqueous gavage 5 mg/kg
I I
2.00 5.00
0.20 0.50
t(hr)
Liu et al. (2009) Male Rat
10 mg/kg TCE aqueous gavage 10 mg/kg
i—i	1—i—i	r
0.20 0.50 2.00 5.00
t (hr)
~l—I—I	1—I—I	T
0.05 0.20 0.50 2.00 5.00
t (hr)
Figure A-41. Comparison of Liu et al. (2009) rat data (boxes) and PBPK
model predictions (+ with error bars: single data points or shaded regions:
2.5, 25, 50, 75, and 97.5% population-based predictions).
In male mice, the data at the lower two doses (130 and 250 mg/kg-day) were within the
interquartile region of the PBPK model predictions. The data for male mice at the highest dose
(600 mg/kg-day) were below the interquartile region, but within the 95% confidence interval of
the PBPK model predictions. In females, the data at the lower two doses (160 and
360 mg/kg-day) were mostly below the interquartile region, but within the 95% confidence
This document is a draft for review purposes only and does not constitute Agency policy.
A-219 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
interval of the PBPK model predictions, while about half the data at the highest dose were just
below the 95% confidence interval.
TCA blood, plasma, and liver concentrations were reported by Green (2003b) for male
PPARa-null mice, male 129/sv mice (the background strain of the PPARa-null mice), and male
and female B6C3F1 mice, exposed to 1.0 g/L or 2.5 g/L TCA in drinking water for 5 days (male
B6C3F1 only) to 14 days.2 In male PPARa-null mice, plasma and blood concentrations were
within the interquartile region of the PBPK model predictions, while liver concentrations were
below the interquartile region but within the 95% confidence interval. In male 129/sv mice, the
plasma concentrations were within the interquartile region of the PBPK model predictions, while
blood and liver concentrations were below the interquartile region but within the 95% confidence
interval. In male B6C3F1 mice, all data were within the 95% confidence intervals of the PBPK
model predictions, with about half within the interquartile region, and the rest above (plasma
concentrations at the lower dose) or below (liver concentrations at all but the lowest dose at
5 days). In female B6C3F1 mice, plasma concentrations were below the interquartile region but
within the 95% confidence region, while liver and blood concentrations were at or below the
lower 95% confidence bound.
Overall, the predictions of the TCA submodel of the updated TCE PBPK model appear
consistent with these data on the toxicokinetics of TCA after drinking water exposure in male
rats and male mice. In female mice, the reported concentrations tends to be at the low end of or
lower than those predicted by the PBPK model. Importantly, the data used for calibrating the
mouse PBPK model parameters were predominantly in males, with only Fisher et al. (1991) and
Fisher and Allen (1993) reporting TCA plasma levels in female mice after TCE exposure. In
addition, median PBPK model predictions at higher doses (>300 mg/kg-day), even in males,
tended to be higher than the concentrations reported. While TCA kinetics after TCE exposure
includes predicted internal production at these higher levels, previously published data on TCA
kinetics alone only included doses up to 100 mg/kg, and only in males. Therefore, these results
suggest that the median predictions of the TCA sub-model of the updated TCE PBPK model are
somewhat less accurate for female mice and for higher doses of TCA (>300 mg/kg-day) in mice,
though the 95% confidence intervals still cover the majority of the reported data. Finally, the
ratio of blood to liver concentrations of -1.4 reported in the mouse experiments in Mahle et al.
(2001) were significantly different from the ratios of-2.3 reported by Green (2003b), a
difference for which there is no clear explanation given the similar experimental designs and
common use the
2Sweeney et al. (2009) reported that blood concentrations in Green (2003b) were incorrect due to an arithmetic error
owing to a change in chemical analytic methodology, and should have been multiplied by 2. This correction was
included in the present analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
A-220 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
B6C3F1 mouse strain. Because median PBPK model predictions for the blood to liver
concentration ratio for these studies are -1.3, they are more consistent with the Mahle et al.
(2001) data than with the Green (2003b) data.
A.4.5 .2. Summary of Results From Chiu of Bayesian Updating of Evans et al. (2009) and
Chiu et al. (2009) Model Using Trichloroacetic Acid (TCA) Drinking Water
Data
Sweeney et al. (2009) also suggested that the available data, in conjunction with
deterministic modeling using the TCA portion of the Hack et al. (2006) TCE PBPK model,
supported a hypothesis that the bioavailability of TCA in drinking water in mice is substantially
less than 100%. Classically, oral bioavailability is assessed by comparing blood concentration
profiles from oral and i.v. dosing experiments, because blood concentration data from oral
dosing alone cannot distinguish fractional uptake from metabolism. Schultz et al. (1999) made
this comparison in rats at a single dose of 82 mg/kg, and reported an empirical bioavailability of
116%, consistent with complete absorption. A priori, there would not seem to be a strong reason
to suspect that oral absorption in mice would be significantly different from that in rats. As
discussed above in the evaluation of Hack et al. (2006) model, available data strongly support
clearance of TCA in addition to urinary excretion, based on the finding of less than
100%) recovery in urine after i.v. dosing. In addition, as the current TCE PBPK model assumes
100%) absorption for orally-administered TCA, and the PBPK model predictions are consistent
with these data, it is likely that the limited bioavailability determined by Sweeney et al. (2009)
was confounded by this additional clearance pathway unaccounted for by Hack et al. (2006).
Therefore, Chiu conducted a Bayesian reanalysis of the TCE mouse PBPK model, the results of
which are summarized here.
In brief, the TCA submodel from Evans et al. (2009) and Chiu et al. (2009) is augmented
by the addition of a fractional absorption parameter for drinking water exposures and parameters
reestimated by adding the newly available TCA drinking water kinetic studies in mice. Being
nocturnal animals, rodents do not have a steady pattern of drinking water consumption
throughout the day. It has been suggested that a 90/10%>-split between dark-cycle (night
time)/light-cycle (day time) drinking water consumption is a reasonable approximation (Yuan,
1995), and that pattern is assumed here. Most analyses assume something similar (e.g., Sweeney
et al., 2009, assumed 100%> consumption during the dark cycle).
However, TCA kinetics from drinking water exposures also depends on the relationship
between the times of the light/dark cycle and the times of specimen collection—i.e., at what time
This document is a draft for review purposes only and does not constitute Agency policy.
A-221 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
during the cycle did exposure begin (when is "t = 0")? These data are not specified in any of the
available technical reports cited by Sweeney et al. (2009). Therefore, in the present analysis,
three different assumptions that represent a range of possibilities were made, and the results of
each were carried through the analysis. These patterns are shown in Figure A-42 and designated
low-12/high-12 (LH), low-6/high-12/low-6 (LHL), and high-12/low-12 (HL). In the first, it is
assumed that the start of exposure coincided exactly with the start of the light cycle; in the
second, it is assumed that the start of exposure was exactly in the middle of the light cycle; and
in the last case, it is assumed that the start of exposure was exactly at the end of the light cycle.
A priori, one of the first two patterns (LH and LHL) would appear to be most likely, but the last
pattern (HL) was included for completeness. Sweeney et al. (2009) assumed drinking water
intake was most similar to the LH pattern.
Low-12/High-12 (LH)
12 18 24 30 36
time since beginning of exposure
Low-6/High-12/Low-6 (LHL)
12 18 24 30 36
time since beginning of exposure
This document is a draft for review purposes only and does not constitute Agency policy.
A-222 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
2
1.8
1.6
1.4
• 1.2
i
1
08
0.6
0.4
0.2
0
High-12/Low-12 (HL)
12 18 24 30 36
time since beginning of exposure
Figure A-42. Assumed drinking water patterns as a function of time since
beginning of exposure. The upper left panel (LH) assumes that t = 0 is at the
beginning of the "light" part of the "light/dark" cycle (light is dashed grey line at
the bottom, dark is thick black line at the bottom). The upper right panel (LHL)
assumes that t = 0 is in the middle of the "light" part of the cycle. The lower left
panel (HL) assumes that t = 0 is at the end of the "light" part of the cycle.
As was done by Evans et al. (2009) and Chiu et al. (2009), the PBPK parameter
estimation is performed in a hierarchical Bayesian population statistical framework, with
calculations performed using MCMC, using posteriors from the earlier analysis as priors for the
reanalysis. A total of six different model runs were made using the "harmonized" PBPK model,
as shown in Table A-18, using different assumptions for fractional absorption and for drinking
water intake patterns. Comparisons between different modeling assumptions (i.e., fixing or
estimating fractional absorption; assumed drinking water patterns) were made using the deviance
information criterion (DIC) (Spiegelhalter et al., 2002). The DIC is a Bayesian analogue to the
Akaike information criterion (AIC) and is used in a similar manner, with smaller values
indicating better model fits. As with the AIC, "small" differences in DIC (e.g., less than 5, as
suggested by the WinBUGS "DIC page" [http://www.mrc-bsu.cam.ac.uk/bugs/winbugs/
dicpage.shtml]) are not likely to be important, but much lower values suggest substantially better
fitting models. Results of these comparison are also shown in Table A-18. Adding the fractional
absorption parameter decreases the DIC by about 100 units, which strongly supports inclusion of
the parameter. In addition, in both cases of fixed and fitted fractional absorption, the lowest DIC
was for the LHL drinking water intake pattern, with the second lowest DIC for the LH pattern,
with a difference of 33 units in DIC. Given that these model runs are highly favored relative to
the others, the rest of this summary reports the results for the "LHL.fitted" run (see Chiu, In
Press, for additional details).
This document is a draft for review purposes only and does not constitute Agency policy.
A-223 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table A-18. Summary characteristics of model runs
2
Run
designation
Drinking water pattern
Fractional absorption
Convergence
DIC
Fixed
Fitted
LH. fixed
Low-12/high-12
a/

R< 1.04
895
LHL.fixed
Low-6/high-12/low-6
a/

R< 1.09
877
HL. fixed
High-12/low-12
V

R< 1.05
897
LH.fitted
Low-12/high-12

a/
R< 1.05
764
LHL.fitted
Low-6/high-12/low-6

V
R< 1.11
731
HL.fitted
High-12/low-12

V
R< 1.12
781
3	Posterior model fits for the LHL.fitted runs are shown in Figures A-43 and A-44, using a
4	representative sample from the converged MCMC chain. A dose-dependent fractional
5	absorption can account for the less-than-proportional increase in TCA blood concentrations
6	between the middle and high dose groups observed in Mahle et al. (2001) (see Figure A-43) and
7	among all the dose groups observed in Green (2003a, 2003b) (see Figure A-44).
Mahle et al. (2001) m B6C3F1 (3 d) [LHL.fitted]
Mahle etal. (2001) m B6C3F1 (14 d) [LHL.fitted]
t (hours)
t (hours)
This document is a draft for review purposes only and does not constitute Agency policy.
A-224 DRAFT—DO NOT CITE OR QUOTE
-------
Mahleetal. (2001) m B6C3F1 (14 d) [LHL.fitted]
Mahle et al. (2001) m B6C3F1 (3 d) [LHL.fitted]
—i
D)
E

c
o
<
o
I-
300 310 320 330 340 350 360
_i
O)
CD
>
c
o
<
o
I-
40
50
60
70
80
90
t (hours)	t (hours)
1
2	Figure A-43. PBPK model predictions for TCA in blood and liver of male
3	B6C3Fi mice from Mahle et al. (2001). Three- and 14-day exposures to 0.08
4	(data: open circles, predictions: solid line), 0.8 (data: open triangle, predictions:
5	dashed line), and 2 g/L TCA in drinking water (data: crosses, predictions: dotted
6	line). Predictions use a representative parameter sample from the converged
7	MCMC chain for the LHL drinking water intake pattern.
8
9
This document is a draft for review purposes only and does not constitute Agency policy.
A-225 DRAFT—DO NOT CITE OR QUOTE
-------
Green (2003a) m B6C3F1 (5 d) [LHL.fitted]
Green (2003b) m B6C3F1 (5 d) [LHL.fitted]

=d o -
O) T-

	 I11" 111 1"
,01 l
u) poo|q Uj \/01
0 20 40 60 80 100 120
t (hours)
Green (2003b) m B6C3F1 (14 d) [LHL.fitted]
~i i i i i i i r
85 90 95 100 105 110 115 120
t (hours)
Green (2003b) m B6C3F1 (5 d) [LHL.fitted]
t (hours)
t (hours)
Green (2003b) m B6C3F1 (14 d) [LHL.fitted]
n	1	1	1	1	1	1	r~
300 305 310 315 320 325 330 335
t (hours)
Figure A-44. PBPK model predictions for TCA in blood and liver of male
B6C3F1 mice from Green (2003a, 2003b). Green (2003a): 5-day drinking water
exposures to 0.5 (data: open circle; predictions: solid line), 1 (data: open triangle;
predictions: dashed line), and 2.5 g/L TCA (data: crosses; predictions: dotted
lines). Green (2003b): 5- and 14-day drinking water exposures to 1 (data: open
circle; predictions: solid line) and 2.5 g/L TCA (data: open triangle; predictions:
dashed line). Predictions use a representative parameter sample from the
converged MCMC chain for the LHL drinking water intake pattern.
This document is a draft for review purposes only and does not constitute Agency policy.
A-226 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
As was done by Sweeney et al. (2009), fractional absorption is separately estimated for
each drinking water dose group, and the results are fit to a parametric model, shown in
Figure A-45. Several features of the data and analysis are worth noting. First, there is a general
trend for decrease in fractional absorption with increasing concentration, evident even within
studies. Second, there appears to be substantial interstudy and intrastudy variability in the
apparent fractional absorption. This is particularly evident across strains in Green (2003b)—the
PPARa-null and 129/sv mice appear to have substantially higher fractional absorption than the
B6C3Fi mice, even though in all strains there appeared to be a decreasing trend with increasing
TCA concentration. Third, the fractional absorption estimates increase as the "start of exposure"
is assumed to be later and later in the "light" cycle. Fourth, the estimated fractional absorption at
low concentrations is fairly high, at more than 80%. Finally, the estimates for fractional
absorption from the current analysis are 3-4 times greater than those reported by Sweeney et al.
(2009). Because hepatic clearance was not included in the previous Hack et al. (2006) version of
the TCE model used by Sweeney et al. (2009), and this could partially explain why they found a
very low fractional absorption to be necessary to provide a fit to the observed data from drinking
water exposures.
In sum, comparing model results with complete- and less-than-complete-fractional
absorption, it is evident (e.g., through the much lower DIC) that including a
concentration-dependent fractional absorption substantially improves model fits. Thus, these
data are consistent with reduced bioavailability from drinking water, particularly at higher TCA
drinking water concentrations. However, the estimates of fractional absorption are three- to
fourfold higher than those estimated by Sweeney et al. (2009). In addition, there appeared to be
substantial inter- and intrastudy variability, with the fractional absorption for some mouse strains
estimated to be nearly complete even at the higher TCA drinking water concentrations. Thus, on
the whole, adding a fractional absorption parameter substantially improves the PBPK model
predictions, though the degree of absorption is greater than that reported by Sweeney et al.
(2009) and appears to be variable between studies and mouse strains. Data are lacking as to a
mechanistic basis for reduced absorption of TCA at higher doses. Biliary excretion is a
possibility, though data from rats suggest that the degree of biliary excretion of TCA is rather
modest (Stenner et al., 1997). It is also possible that the nonlinearity in TCA kinetics reflects a
difference in clearance processes, such as saturation of renal reabsorption, which would lead to
increased urinary clearance and reduced internal dose. This could be tested experimentally by
simultaneously measuring blood and urinary kinetics of TCA at different doses. However, this
would not explain differences between drinking water and gavage dosing.
This document is a draft for review purposes only and does not constitute Agency policy.
A-227 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
O Mahleetal. (2001) B6C3F1
~ Green (2003a, 2003b)
B6C3F1
A Green (2003a, 2003b)
sv129 (PPARalpha-null and
wild)
^"MichaelisMenten Fit
o Sweeneyetal. (2009)
estimates for Mahle et al.
(2001) B6C3F1
~ Sweeneyetal. (2009)
estimates for Green
(2003a,2003b) B6C3F1
Michaelis-Menten Fit to
Sweeneyetal. (2009)
Figure A-45. Distribution of fractional absorption fit to each TCA drinking
water kinetic study group in mice, using LHL drinking water intake
patterns. Fits are to a Michaelis-Menten function for "effective" concentration
Ceff = Cmax x C/(Cv2 + C), so that the fractional absorption Fabs = Ceff/C
= Cmax/(Ci 2 + C). Sweeney et al. (2009) estimates of Fabs, along with a
Michaelis-Menten fit, are included for comparison. The ratio Cmax/Cvi gives the
fractional uptake at low concentrations.
The degree of interexperimental variability raises the question of whether the apparent
fractional absorption may be due, in part, to experimental factors, such as analytical errors due to
incomplete/inadequate procedures to prevent TCA degradation or experimental losses in
estimating drinking water consumption rates. With respect to TCA degradation, Mahle et al.
(2001) appeared to be specifically aware of the issue and froze biological samples prior to
analysis in order to address it. However, lacking any external validation, the extent to which this
was completely successful is unclear. On the other hand, Green (2003a, 2003b) did not appear to
have any particular procedure designed to address TCA degradation. Thus, the extent and
impact of TCA degradation is not clear, though it may be a plausible explanation for the degree
of variability observed across data sets. With respect to drinking water consumption,
experimental variance is notable with respect to reported drinking water consumption rates, with
Low-6/High-12/Low-6 Drinking Water Intake
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0-
0.1
0
0
1
2
3
TCA Concentration (g/L)
This document is a draft for review purposes only and does not constitute Agency policy.
A-228 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Green (2003a) > Green (2003b) > Mahle et al. (2001) > other TCA drinking water studies. One
may hypothesize that the actual drinking water consumption rates are roughly equal, with
differences in reported values reflecting experimental losses. However, in this case, reported
drinking water consumption would inversely correlate with fractional absorption, and no such
correlation is evident. In addition, this does not explain the consistent dose-related trends within
a study or data set, even if the slope of the trend varies between experiments.
Overall, then, it may be more accurate to characterize the fractional absorption as an
empirical parameter reflecting unaccounted-for biological processes as well as experimental
variation.
A.5. UPDATED PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK)
MODEL CODE
The following pages contain the updated PBPK model code for the MCSim software
(version 5.0.0). Additional details on baseline parameter derivations are included as inline
documentation. Example simulation files containing prior distributions and experimental
calibration data are available electronically:
•	Mouse: Appendix.linked.files\TCE. 1.2.3.3.Mouse.pop.example.in
•	Rat: Appendix.linked.files\TCE. 1.2.3.3.Rat.pop.example.in
•	Human: Appendix.linked.files\TCE. 1.2.3.3.Human.pop.example.in.
This document is a draft for review purposes only and does not constitute Agency policy.
A-229 DRAFT—DO NOT CITE OR QUOTE
-------
#	TCE.risk.1.2.3.3.pop.model — Updated TCE Risk Assessment Model
#
#### HISTORY OF HACK ET AL. (2006) MODEL
#	Model code to correspond to the block diagram version of the model
#	Edited by Deborah Keys to incorporate Lapare et al. 1995 data
#	Last edited: August 6, 2004
^ # Translated into MCSim from acslXtreme CSL file by Eric Hack, started 31Aug2004
S"* # Removed nonessential differential equations (i.e., AUCCBld) for MCMC runs.
^	# Changed QRap and QSlw calculations and added QTot to scale fractional flows
O	# back to 1 after sampling.
#	Finished translating and verifying results on 15Sep2004.
2	# Changed QSlw calculation and removed QTot 21Sep2004.
S	# Removed diffusion-limited fat uptake 24Sep2004.
#### HISTORY OF U.S. EPA (2009) MODEL (CHIU ET AL., 2009)
>1
^	# Extensively revised by U.S. EPA June 2007-June 2008
#	- Fixed hepatic plasma flow for TCA-submodel to include
S"4
5^	#	portal vein (i.e., QGutLivPlas -- originally was just
#	QLivPlas, which was only hepatic artery).
^	#	- Clearer coding and in-line documentation
•>5	#	- Single model for 3 species
^	#	- Revised physiological parameters, with discussion of
^	#	uncertainty and variability,
.	#	- In vitro data used for default metabolism parameters,
s^, ^3
Si	#	with discussion of uncertainty and variability
1 K?
^3	#	- added TCE blood compartment
\ O
>j	#	- added TCE kidney compartment, with GSH metabolism
O fis
>!	#	- added DCVG compartment
O
a
#	- removed DCA compartment
C5	#	- added IA and PV dosing (for rats)
5^ #	- Version 1.1 -- fixed urinary parameter scaling
^ #	— fixed VBod in kUrnTCOG (should be VBodTCOH)
#	- Version 1.1.1 -- changed some truncation limits (in commments only)
^ #	- Version 1.2 --
O	#	-- removed TB compartment as currently coded
Ci	#	-- added respiratory oxidative metabolism:
81
2; ^
o ^
- added additional outputs available from in vivo data
^2	#	3 states: AInhResp, AResp, AExhResp
-- removed clearance from respiratory metabolism
-	Version 1.2.1 -- changed oral dosing to be similar to IV
-	Version 1.2.2 -- fixed default lung metabolism (additional
scaling by lung/liver weight ratio)
oS	#	- Version 1.2.3 — fixed FracKidDCVC scaling
3 Ci	#	_ Version 1.2.3.1 -- added output CDCVG_ND (no new dynamics)
O ^	# for non-detects of DCVG in blood
HH ^3
O	#	- Version 1.2.3.2 -- Exact version of non-detects likelihood
ffl.	#	- Version 1.2.3.3 -- Error variances changed to "Ve_xxx"
•	#	NOTE -- lines with comment " (vrisk) " are used only for
#	calculating dose metrics, and are commented out
—#	when doing MCMC runs.
o
H #
w
State Variable Specifications
States = {
##— TCE uptake
AStom,
ADuod,
AExc,
AO,
InhDose, #
##— TCE in the body
#	Amount of TCE in stomach
#	oral gavage absorption -- mice and rats only
#(vrisk) excreted in feces from gavage (currently 0)
#(vrisk) total absorbed
Amount inhaled
ARap,
# Amount
in
rapidly perfused tissues
ASlw,
# Amount
in
slowly perfused tissues
AFat,
# Amount
in
fat
AGut,
# Amount
in
gut
ALiv,
# Amount
in
liver
AKid,
# Amount
in
Kidney -- previously in Rap tissue
ABld,
# Amount
in
Blood -- previously in Rap tissue
AInhResp, #
AResp,
AExhResp, #
—	TCA in the body
AOTCA,
AStomTCA, #
APlasTCA, #
ABodTCA, #
ALivTCA, #
—	TCA metabolized
AUrnTCA, #
AUrnTCA sat
Amount in respiratory lumen during inhalation
# Amount in respiratory tissue
Amount in respiratory lumen during exhalation
#(vrisk)
Amount of TCA in stomach
Amount of TCA in plasma #comment out for
Amount of TCA in lumped body compartment
Amount of TCA in liver
Cumulative Amount of TCA excreted in urine
#	Amount of TCA excreted that during times that had
#	saturated measurements (for lower bounds)
AUrnTCA_collect,# Cumulative Amount of TCA excreted in urine during
#	collection times (for intermittent collection)
TCOH in body
AOTCOH,
AStomTCOH,
ABodTCOH,
ALivTCOH,
TCOG in body
ABodTCOG,
ALivTCOG,
ABileTCOG,
ARecircTCOG,
TCOG excreted
AUrnTCOG,
AUrnTCOG sat,
AUrnTCOG_collect,:
f(vrisk)
f Amount of TCOH in stomach
f Amount of TCOH in lumped body compartment
t Amount of TCOH in liver
f Amount of TCOG in lumped body compartment
t Amount of TCOG in liver
t Amount of TCOG in bile (incl. gut)
f(vrisk)
Amount of TCOG excreted in urine
#	Amount of TCOG excreted that during times that had
#	saturated measurements (for lower bounds)
Cumulative Amount of TCA excreted in urine during
#	collection times (for intermittent collection)
—	DCVG in body
ADCVGIn,
ADCVGmol,
AMetDCVG,
—	DCVC in body
ADCVCIn,
ADCVC,
t(vrisk)
i Amount of DCVG in body in mmoles
t(vrisk)
f (vrisk)
# Amount of DCVC in body
-------
s?
>!
§•
rs
s;
>1
£
S^. ^
K) ^
U>
o
o
>!
>!
o
a
a
a
§•
>!
5
o
rs
o
a
>i

H
I
o
o
2! . _
O^
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
ABioactDCVC,
NAcDCVC excreted
AUrnNDCVC,
Other states for TCE
ACh,
AExh,
AExhExp,
Metabolism
AMetLivl,
AMetLiv2,
AMetLng,
AMetKid,
AMetTCOHTCA,
AMetTCOHGluc,
AMetTCOHOther,
AMetTCA, #(vrisk)
Other Dose metrics
*(vrisk)
Amount of NAcDCVC excreted
#	Amount in closed chamber -- mice and rats only
#	Amount exhaled
Amount exhaled during expos [to calc. retention]
$(vrisk)
*(vrisk)
$(vrisk)
*(vrisk)
Amount metabolized by P450 in liver
Amount metabolized by GSH conjugation in liver
Amount metabolized in the lung
#(vrisk) Amount of TCOH metabolized to TCA
#(vrisk) Amount of TCOH glucuronidated
#(vrisk)
Amount of TCA metabolized
* (vrisk)
$(vrisk)
*(vrisk)
*(vrisk)
*(vrisk)
AUCCBld,
AUCCLiv,
AUCCKid,
AUCCRap,
AUCCTCOH,
AUCCBodTCOH,
AUCTotCTCOH,
AUCPlasTCAFree,
AUCPlasTCA,
AUCLivTCA,
AUCCDCVG #(vrisk)
*(vrisk)
*(vrisk)
*(vrisk)
$(vrisk)
$(vrisk)
Input Variable Specifications
Inputs = {
##— TCE dosing
Cone,
IVDose,
PDose,
Drink,
IADose,
PVDose,
##— TCA dosing
IVDoseTCA,
PODoseTCA,
##— TCOH dosing
IVDoseTCOH,
PODoseTCOH,
##— Potentially time-
QPmeas,
TCAUrnSat,
TCOGUrnSat,
UrnMissing
#	Inhalation exposure conc. (ppm)
#	IV dose (mg/kg)
#	Oral gavage dose (mg/kg)
#	Drinking water dose (mg/kg/day)
#	Inter-arterial
#	Portal Vein
#	IV dose (mg/kg) of TCA
#	Oral dose (mg/kg) of TCA
#	IV dose (mg/kg) of TCOH
#	Oral dose (mg/kg) of TCOH
•varying parameters
#	Measured value of Alveolar ventilation QP
#	Flag for saturated TCA urine
#	Flag for saturated TCOG urine
#	Flag for missing urine collection times
#*•*•*	Output Variable Specifications	**"*¦
Outputs = {
#*** Outputs for mass balance check
MassBalTCE,
TotDose,
TotTissue,
MassBalTCOH,
TotTCOHIn,
TotTCOHDose,
TotTissueTCOH,
TotMetabTCOH,
MassBalTCA,
TotTCAIn,
TotTissueTCA,
MassBalTCOG,
TotTCOGIn,
TotTissueTCOG,
MassBalDCVG,
MassBalDCVC,
AUrnNDCVCequiv,
#•*••1-* Outputs that are potential dose metrics
TotMetab, #(vrisk) Total metabolism
TotMetabBW34, # (vrisk) Total metabolism/BW'v3/4
ATotMetLiv, #(vrisk) Total metabolism in liver
AMetLivlLiv, #(vrisk) Total oxidation in liver/liver volume
AMetLivOther, #(vrisk) Total "other" oxidation in liver
AMetLivOtherLiv, #(vrisk) Total "other" oxidation in liver/liver vol
AMetLngResp, #(vrisk) oxiation in lung/respiratory tissue volume
AMetGSH, #(vrisk) total GSH conjugation
AMetGSHBW34, # (vrisk) total GSH conjugation/BW'%3/4
ABioactDCVCKid,	#(vrisk) Amount of DCVC bioactivated/kidney volume
#	NEW
TotDoseBW34, # (vrisk) mg intake / BW3/4
AMetLivlBW34, # (vrisk) mg hepatic oxidative metabolism / 61^3/4
TotOxMetabBW34, # (vrisk) mg oxidative metabolism / BW-"3/4
TotTCAInBW, #(vrisk) TCA production / BW
AMetLngBW34, # (vrisk) oxiation in lung/BWr>3/4
ABioactDCVCBW34, # (vrisk) Amount of DCVC bioactivated/BW-3/4
AMetLivOtherBW34, #(vrisk) Total "other" oxidation in liver/BW^3/4
#*** Outputs for comparison to in vivo data
#	TCE
RetDose, # human - = (InhDose - AExhExp)
CAlv,	# needed for CAlvPPM
CAlvPPM, # human
CInhPPM, # mouse, rat
-------
s?
>!
§•
rs
s;
>1
£
S^. ^
K) ^
OJ g
to TO
>!
o
a
a
a
§•
>!
5
o
o
H
I
o
o
2! . -
O^
<3
H
1—1 s
H 2.
W|
O •
o
c
o
H
W
CInh,	# needed for CMixExh
CMixExh,	# rat - Mixed exhaled breath (mg/1)
CArt,	# rat, human - Arterial blood concentration
CVen,	# mouse, rat, human
CBldMix,	# rat - Concentration in mixed arterial+venous blood
# (used for cardiac puncture)
CFat,	# mouse, rat - Concentration in fat
CGut,	# rat
CRap,	# needed for unlumped tissues
CSlw,	# needed for unlumped tissues
CHrt,	# rat - Concentration in heart tissue [use CRap]
CKid,	# mouse, rat - Concentration in kidney
CLiv,	# mouse, rat - Concentration in liver
CLung,	# mouse, rat - Concentration in lung [use CRap]
CMus,	# rat - Concentration in muscle [use CSlw]
CSpl,	# rat - Concentration in spleen [use CRap]
CBrn,	# rat - Concentration in brain [use CRap]
zAExh,	# mouse
zAExhpost,	# rat - Amount exhaled post-exposure (mg)
# TCOH
CTCOH,	# mouse, rat, human - TCOH concentration in blood
CKidTCOH,	# mouse - TCOH concentration in kidney
CLivTCOH,	# mouse - TCOH concentration in liver
CLungTCOH,	# mouse - TCOH concentration in lung
# TCA
CPlasTCA,
CBldTCA,
CBodTCA,
CKidTCA,
CLivTCA,
CLungTCA,
zAUrnTCA,
zAUrnTCA_collect,
zAUrnTCA sat,
mouse, rat, human - TCA concentration in plasma
mouse, rat, human - TCA concentration in blood
needed for CKidTCA and CLungTCA
mouse - TCA concentration in kidney
mouse, rat - TCA concentration in liver
mouse - TCA concentration in lung
mouse, rat, human - Cumulative Urinary TCA
human - TCA measurements for intermittent collection
human - Saturated TCA measurements
# TCOG
zABileTCOG,	# rat - Amount of TCOG in bile
CTCOG, # needed for CTCOGTCOH
(mg)
CTCOGTCOH,	1
CKidTCOGTCOH,	1
CLivTCOGTCOH,	1
CLungTCOGTCOH,	1
AUrnTCOGTCOH,	1
AUrnTCOGTCOH_collect,
AUrnTCOGTCOH sat,	1
mouse - TCOG concentration in blood (in TCOH-equiv)
mouse - TCOG concentration in kidney (in TCOH-equiv)
mouse - TCOG concentration in liver (in TCOH-equiv)
mouse - TCOG concentration in lung (in TCOH-equiv)
mouse, rat, human - Cumulative Urinary TCOG (in TCOH-equiv)
#	human - TCOG (in TCOH-equiv) measurements for
#	intermittent collection
human - Saturated TCOG (in TCOH-equiv) measurements
# Other
CDCVGmol,	# concentration of DCVG (mmol/1)
CDCVGmolO,	# Dummy variable without likelihood (for plotting)#(vl.2.3.1)
CDCVG_ND, # Non-detect of DCVG (<0.05 pmol/ml= 5e-5 mmol/1 )#(vl.2.3.1)
# Output -In(likelihood)#(vl.2.3.1)
zAUrnNDCVC,	# rat, human - Cumulative urinary NAcDCVC
AUrnTCTotMole,	# rat, human - Cumulative urinary TCOH+TCA in mmoles
TotCTCOH, # mouse, human - TCOH+TCOG Concentration (in TCOH-equiv)
TotCTCOHcomp,	# ONLY FOR COMPARISON WITH HACK
ATCOG,	# ONLY FOR COMPARISON WITH HACK
QPsamp, # human - sampled value of alveolar ventilation rate
## PARAMETERS #(vrisk)
QCnow, # (vrisk) #Cardiac output (L/hr)
QP, # (vrisk) #Alveolar ventilation (L/hr)
QFatCtmp, # (vrisk) #Scaled fat blood flow
QGutCtmp, # (vrisk) #Scaled gut blood flow
QLivCtmp, # (vrisk) #Scaled liver blood flow
QSlwCtmp, # (vrisk) #Scaled slowly perfused blood flow
QRapCtmp, # (vrisk) #Scaled rapidly perfused blood flow
QKidCtmp, # (vrisk) #Scaled kidney blood flow
DResp, # (vrisk) #Respiratory lumen:tissue diffusive clearance rate
VFatCtmp, # (vrisk) #Fat fractional compartment volume
VGutCtmp, # (vrisk) #Gut fractional compartment volume
VLivCtmp, # (vrisk) #Liver fractional compartment volume
VRapCtmp, # (vrisk) #Rapidly perfused fractional compartment volume
VRespLumCtmp, # (vrisk) # Fractional volume of respiratory lumen
VRespEffCtmp, # (vrisk) #Effective fractional volume of respiratory tissue
VKidCtmp, # (vrisk) #Kidney fractional compartment volume
VBldCtmp, # (vrisk) #Blood fractional compartment volume
VSlwCtmp, # (vrisk) #Slowly perfused fractional compartment volume
VPlasCtmp, # (vrisk) #Plasma fractional compartment volume
VBodCtmp, # (vrisk) #TCA Body fractional compartment volume [not incl.
blood+liver]
VBodTCOHCtmp, # (vrisk) #TCOH/G Body fractional compartment volume [not incl.
liver]
PB, # (vrisk) #TCE Blood/air partition coefficient
PFat, # (vrisk) #TCE Fat/Blood partition coefficient
PGut, # (vrisk) #TCE Gut/Blood partition coefficient
PLiv, # (vrisk) #TCE Liver/Blood partition coefficient
PRap, # (vrisk) #TCE Rapidly perfused/Blood partition coefficient
PResp, # (vrisk) #TCE Respiratory tissue:air partition coefficient
PKid, # (vrisk) #TCE Kidney/Blood partition coefficient
PSlw, # (vrisk) #TCE Slowly perfused/Blood partition coefficient
TCAPlas, # (vrisk) #TCA blood/plasma concentration ratio
PBodTCA, # (vrisk) #Free TCA Body/blood plasma partition coefficient
PLivTCA, # (vrisk) #Free TCA Liver/blood plasma partition coefficient
kDissoc, # (vrisk) #Protein/TCA dissociation constant (umole/L)
BMax, # (vrisk) #Maximum binding concentration (umole/L)
PBodTCOH, # (vrisk) #TCOH body/blood partition coefficient
PLivTCOH, # (vrisk) #TCOH liver/body partition coefficient
PBodTCOG, # (vrisk) #TCOG body/blood partition coefficient
PLivTCOG, # (vrisk) #TCOG liver/body partition coefficient
VDCVG, # (vrisk) #DCVG effective volume of distribution
kAS, # (vrisk) #TCE Stomach absorption coefficient (/hr)
kTSD, # (vrisk) #TCE Stomach-duodenum transfer coefficient (/hr)
-------
kAD, # (vrisk) #TCE Duodenum absorption coefficient (/hr)
kTD, # (vrisk) #TCE Duodenum-feces transfer coefficient (/hr)
kASTCA, # (vrisk) #TCA Stomach absorption coefficient (/hr)
kASTCOH, # (vrisk) #TCOH Stomach absorption coefficient (/hr)
VMAX, # (vrisk) #VMAX for hepatic TCE oxidation (mg/hr)
KM, # (vrisk) #KM for hepatic TCE oxidation (mg/L)
^	FracOther, # (vrisk) #Fraction of hepatic TCE oxidation not to TCA+TCOH
U*	FracTCA, # (vrisk) #Fraction of hepatic TCE oxidation to TCA
^	VMAXDCVG, # (vrisk) #VMAX for hepatic TCE GSH conjugation (mg/hr)
O"4	KMDCVG, # (vrisk) #KM for hepatic TCE GSH conjugation (mg/L)
§	VMAXKidDCVG, # (vrisk) #VMAX for renal TCE GSH conjugation (mg/hr)
2	KMKidDCVG, # (vrisk) #KM for renal TCE GSH conjugation (mg/L)
S	FracKidDCVC, # (vrisk) #Fraction of renal TCE GSH conj. "directly" to DCVC
# (vrisk) #(i.e., via first pass)
Go
^	VMAXClara, # (vrisk) #VMAX for Tracheo-bronchial TCE oxidation (mg/hr)
KMClara, # (vrisk) #KM for Tracheo-bronchial TCE oxidation (mg/L)
S"4
5^	FracLungSys, # (vrisk) #Fraction of respiratory metabolism to systemic circ
^	VMAXTCOH, # (vrisk) #VMAX for hepatic TCOH->TCA (mg/hr)
O	KMTCOH, # (vrisk) #KM for hepatic TCOH->TCA (mg/L)
•>S	VMAXGluc, # (vrisk) #VMAX for hepatic TCOH->TCOG (mg/hr)
5	KMGluc, # (vrisk) #KM for hepatic TCOH->TCOG (mg/L)
kMetTCOH, # (vrisk) #Rate constant for hepatic TCOH->other (/hr)
^	kUrnTCA, # (vrisk) #Rate constant for TCA plasma->urine (/hr)
^ Si	kMetTCA, # (vrisk) #Rate constant for hepatic TCA->other (/hr)
I ""'S
^	kBile, # (vrisk) #Rate constant for TCOG liver->bile (/hr)
^ §	kEHR, # (vrisk) #Lumped rate constant for TCOG bile->TCOH liver (/hr)
Gn	kUrnTCOG, # (vrisk) #Rate constant for TCOG->urine (/hr)
^	kDCVG, # (vrisk) #Rate constant for hepatic DCVG->DCVC (/hr)
kNAT, # (vrisk) #Lumped rate constant for DCVC->Urinary NAcDCVC (/hr)
kKidBioact, # (vrisk) #Rate constant for DCVC bioactivation (/hr)

## Misc
O ^	RUrnTCA, #(vrisk)
^0 Grj
^	RUrnTCOGTCOH, #(vrisk)
O	RUrnNDCVC, #(vrisk)
CPlasTCAMole,
CPlasTCAFreeMole
Cb	RAO,
I §	CVenMole,
I >!
81
w ^ } ;
Cb	Global Constants
O ^
M s
H 2.
W.	# Molecular Weights
o
c
o
H
ffl
MWTCE =
131.39;
# TCE
MWDCA =
129.0;
# DCA
MWDCVC =
216.1;
# DCVC
MWTCA =
163.5;
# TCA
MWChlor =
147.5;
# Chloral
MWTCOH =
14 9.5;
# TCOH
MWTCOHGluc =
325.53;
# TCOH-Gluc
MWNADCVC = 25 8.8;
# N Acetyl DCVC
# Stoichiometry
StochChlorTCE	=
StochTCATCE	=
StochTCATCOH	=
StochTCOHTCE	=
StochGlucTCOH	=
StochTCOHGluc	=
StochTCEGluc	=
StochDCVCTCE	=
StochN	=
StochDCATCE	=
MWChlor / MWTCE;
MWTCA / MWTCE;
MWTCA / MWTCOH;
MWTCOH / MWTCE;
MWTCOHGluc / MWTCOH;
MWTCOH / MWTCOHGluc;
MWTCE / MWTCOHGluc;
MWDCVC / MWTCE;
MWNADCVC / MWDCVC;
MWDCA / MWTCE;
#*"*"*	Global Model Parameters	***
#	These are the actual model parameters used in "dynamics."
#	Values that are assigned in the "initialize" section,
#	are all set to 1 to avoid confusion.
# Flows

QC
= 1
QPsamp
= 1
VPR
= 1
QFatCtmp
= 1
QGutCtmp
= 1
QLivCtmp
= 1
QSlwCtmp
= 1
DResptmp
= 1
[scaled tc
QP]
QKidCtmp
= 1,
FracPlas
= 1,

****
# Volumes
VFat
VGut
VLiv	=
VRap	=
VRespLum =
VRespEfftmp
VRespEff =
Cardiac output (L/hr)
Alveolar ventilation (L/hr)
Alveolar ventilation-perfusion ratio
Scaled fat blood flow
Scaled gut blood flow
Scaled liver blood flow
Scaled slowly perfused blood flow
Respiratory lumen:tissue diffusive clearance rate (L/hr)
Scaled kidney blood flow
Fraction of blood that is plasma (1-hematocrit)
Fat compartment volume (L)
Gut compartment volume (L)
Liver compartment volume (L)
Rapidly perfused compartment volume (L)
Volume of respiratory lumen (L air)
1;	#(vrisk) volume for respiratory tissue (L)
Effective volume for respiratory tissue (L air) = V(tissue)
Resp:Air partition coefficient
VKid
VBld
VSlw
VPlas
VBod
VBodTCOH
Kidney compartment volume (L)
Blood compartment volume (L)
Slowly perfused compartment volume (L)
Plasma compartment volume [fraction of blood] (L)
TCA Body compartment volume [not incl. blood+liver] (L)
TCOH/G Body compartment volume [not incl. liver] (L)
Distribution/partitioning
B	=1;	# TCE Blood/air partition coefficient
-------
s?
>!
§•
rs
s;
>1
£
PFat
PGut
PLiv
PRap
PResp
PKid
PSlw
TCAPlas
PBodTCA
PLivTCA
kDissoc
BMax
PBodTCOH
PLivTCOH
PBodTCOG
PLivTCOG
VDCVG
TCE Fat/Blood partition coefficient
TCE Gut/Blood partition coefficient
TCE Liver/Blood partition coefficient
TCE Rapidly perfused/Blood partition coefficient
TCE Respiratory tissue:air partition coefficient
TCE Kidney/Blood partition coefficient
TCE Slowly perfused/Blood partition coefficient
TCA blood/plasma concentration ratio
Free TCA Body/blood plasma partition coefficient
Free TCA Liver/blood plasma partition coefficient
Protein/TCA dissociation constant (umole/L)
Protein concentration (UNITS?)
TCOH body/blood partition coefficient
TCOH liver/body partition coefficient
TCOG body/blood partition coefficient
TCOG liver/body partition coefficient
DCVG effective volume of distribution
to ^
£ §
>!
O
a
ca
a
>!
o
0
p
H
1
o
o
z;. _
a
Cs
o
a
>j
S"
(3s
H
1—1 s
H 2.
O •
o
c
o
H
# Oral
absorption


kTSD
= 1.4;
# TCE
Stomach-duodenum transfer coefficient (/hr)
kAS
= 1.4;
# TCE
Stomach absorption coefficient (/hr)
kTD
= 0.1;
# TCE
Duodenum-feces transfer coefficient (/hr)
kAD
= 0.75;
# TCE
Duodenum absorption coefficient (/hr)
kASTCA
= 0.75;
# TCA
Stomach absorption coefficient (/hr)
kASTCOH
= 0.75;
# TCOH Stomach absorption coefficient (/hr)
TCE Metabolism
MAX	= 1
1
1
1
1
1
KM
FracOther
FracTCA
VMAXDCVG
KMDCVG
VMAXKidDCVG
KMKidDCVG = 1
VMAXClara = 1
KMClara = 1
FracLungSys
enters systemic circulation
VMAX for hepatic TCE oxidation (mg/hr)
KM for hepatic TCE oxidation (mg/L)
Fraction of hepatic TCE oxidation not to TCA+TCOH
Fraction of hepatic TCE oxidation to TCA
VMAX for hepatic TCE GSH conjugation (mg/hr)
KM for hepatic TCE GSH conjugation (mg/L)
1;	# VMAX for renal TCE GSH conjugation (mg/hr)
KM for renal TCE GSH conjugation (mg/L)
VMAX for Tracheo-bronchial TCE oxidation (mg/hr)
KM for Tracheo-bronchial TCE oxidation (mg/L)
but in units of air concentration
1;	# Fraction of respiratory oxidative metabolism that
# TCOH metabolism
VMAXTCOH = 1;
KMTCOH = 1;
VMAXGluc = 1 ;
KMGluc = 1 ;
kMetTCOH = 1;
VMAX for hepatic TCOH->TCA (mg/hr)
KM for hepatic TCOH->TCA (mg/L)
VMAX for hepatic TCOH->TCOG (mg/hr)
KM for hepatic TCOH->TCOG (mg/L)
Rate constant for hepatic TCOH->other (/hr)
# TCA metabolism/clearance
kUrnTCA = 1;	# Rate constant for TCA plasma->urine (/hr)
kMetTCA = 1;	# Rate constant for hepatic TCA->other (/hr)
kBile	= 1;	# Rate constant for TCOG liver->bile (/hr)
kEHR	= 1;	# Lumped rate constant for TCOG bile->TCOH liver (/hr)
kUrnTCOG = 1;	# Rate constant for TCOG->urine (/hr)
#	DCVG metabolism
kDCVG	= 1;	# Rate constant for hepatic DCVG->DCVC (/hr)
FracKidDCVC	=1;	# Fraction of renal TCE GSH conj. "directly" to DCVC
(i.e., via first pass)
#	DCVC metabolism/clearance
kNAT	= 1;	# Lumped rate constant for DCVC->Urinary NAcDCVC (/hr)
kKidBioact	= 1;	# Rate constant for DCVC bioactivation (/hr)
#	Closed chamber and other exposure parameters
Rodents
= 1;
# Number of rodents in closed chamber data
VCh
= 1;
# Chamber volume for closed chamber data
kLoss
= 1;
# Rate constant for closed chamber air loss
CC
= 0.0;
# Initial chamber concentration (ppm)
TChng
= 0.003;
# IV infusion duration (hour)

## Flag
for species
sex -- these are global parameters
BW
=0.0; #
Species-specific defaults during initialization
BW75
= 0.0;
#(vrisk) Variable for BW^3/4
Male
= 1.0;
# 1 = male, 0 = female
Species
= 1.0;
# 1 = human, 2 = rat, 3 = mouse
#***	Potentially measured covariates (constants)	**"*¦
BWmeas =0.0; # Body weight
VFatCmeas = 0.0; # Fractional volume fat
PBmeas = 0.0; # Measured blood-air partition coefficient
Hematocritmeas = 0.0; # Measured hematocrit -- used for FracPlas = 1 - HCt
CDCVGmolLD = 5e-5; # Detection limit of CDCVGmol#(vl.2.3.1)
Global Sampling Parameters
These parameters are potentially sampled/calibrated in the MCMC or MC
analyses. The default values here are used if no sampled value is given.
M_ indicates population mean parameters used only in MC sampling
V_ indicates a population variance parameter used in MC and MCMC sampling
# Flow
Rates





InQCC
= 0.0
# Scaled
by
BW^0.75 and species-specific central
InVPRC
= 0.0
# Scaled
to
species-specific
central
estimates
# Fractional Blood Flows to Tissues (fraction of
cardiac
output)
QFatC
= 1.0
# Scaled
to
species-specific
central
estimates
QGutC
= 1.0
# Scaled
to
species-specific
central
estimates
QLivC
= 1.0
# Scaled
to
species-specific
central
estimates
QSlwC
= 1.0
# Scaled
to
species-specific
central
estimates
-------
QKidC	= 1.0; # Scaled to species-specific central estimates
FracPlasC = 1.0; # Scaled to species-specific central estimates
InDRespC =0.0; # Scaled to alveolar ventilation rate in dynamics
S?
>!
cs
s;
>1
£
K) ^
s s
L/l
>!
o
a
ca
a
§•
>!
o
0
p
H
1
o
o
2! . -
<3
Cs
o
a
>j
S"
(3s
H
1—1 s
H 2.
W|
O •
o
c
o
H
W
# Fractional
Tissue
Volumes
(fraction of BW)


VFatC
=
1.0;
# Scaled
to
species-specific
central
estimates
VGutC
=
1.0;
# Scaled
to
species-specific
central
estimates
VLivC
=
1.0;
# Scaled
to
species-specific
central
estimates
VRapC
=
1.0;
# Scaled
to
species-specific
central
estimates
VRespLumC
=
1.0;
# Scaled
to
species-specific
central
estimates
VRespEffC
=
1.0;
# Scaled
to
species-specific
central
estimates
VKidC
=
1.0;
# Scaled
to
species-specific
central
estimates
VBldC
=
1.0;
# Scaled
to
species-specific
central
estimate
# Partition
Coefficients for
TCE


InPBC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPFatC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPGutC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPLivC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPRapC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPRespC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPKidC
=
0.0;
# Scaled
to
species-specific
central
estimates
InPSlwC
=
0.0;
# Scaled
to
species-specific
central
estimates
Partition Coefficients for TCA
InPRBCPlasTCAC
InPBodTCAC
InPLivTCAC
0.0
0.0
0.0
Scaled to species-specific central estimates
Scaled to species-specific central estimates
Scaled to species-specific central estimates
#	Plasma Binding for TCA
InkDissocC	= 0.0; # Scaled to species-specific central estimates
InBMaxkDC = 0.0; # Scaled to species-specific central estimates
#	Partition Coefficients for TCOH and TCOG
InPBodTCOHC
InPLivTCOHC
InPBodTCOGC
InPLivTCOGC
InPeffDCVG
Scaled to species-specific	central	estimates
Scaled to species-specific	central	estimates
Scaled to species-specific	central	estimates
Scaled to species-specific	central	estimates
Scaled to species-specific	central	estimates
# Oral Absorpt
ion r
InkTSD
0.
336;
InkAS =
0.
336;
InkTD
-2
.303;
InkAD
-0
.288;
InkASTCA =
-0
.288;
InkASTCOH =
-0
.288;
# TCE Metabolism
InVMAXC =0.0
InKMC	=0.0
InCIC	=0.0
Scaled by liver weight and species-specific central estimates
Scaled to species-specific central estimates
Scaled to species-specific central estimates
InFracOtherC
InFracTCAC
InVMAXDCVGC
estimates
InClDCVGC = 0.0;
InKMDCVGC = 0.0;
InVMAXKidDCVGC
estimates
InClKidDCVGC
InKMKidDCVGC
InVMAXLungLivC
InKMClara = 0.0;
0.0
0.0
0.0
Ratio of DCA to non-DCA
Ratio of TCA to TCOH
Scaled by liver weight and species-specific central
Scaled to species-specific central estimates
Scaled to species-specific central estimates
0.0; # Scaled by kidney weight and species-specific central
0.0; # Scaled to species-specific central estimates
0.0; # Scaled to species-specific central estimates
0.0; # Ratio of lung VMAX to liver VMAX,
# Scaled to species-specific central estimates
now in units of air concentration
# Clearance in lung
InFracLungSysC
oxidation
ratio of systemic to local clearance of lung
# TCOH Metabolism
InVMAXTCOHC
InClTCOHC = 0.0;
InKMTCOH = 0.0;
InVMAXGlucC
InClGlucC = 0.0;
InKMGluc = 0.0;
InkMetTCOHC
0.0; # Scaled by BW^0.75
Scaled by BW^0.75
0.0; # Scaled by BW"0.75
Scaled by BW^0.75
Scaled by BW^-0.25
# TCA Metabolism/clearance
InkUrnTCAC	= 0.0;
central estimates
InkMetTCAC	= 0.0;
Scaled by (plasma volume)^-1 and species-specific
Scaled by BW^-0.25
TCOG excretion and reabsorption
InkBileC = 0.0;
InkEHRC = 0.0;
InkUrnTCOGC
central estimates
Scaled by BW^-0.25
Scaled by BW-0.25
0.0; # Scaled by (blood volume)^-l and species-specific
# DCVG metabolism
InFracKidDCVCC
InkDCVGC = 0.0;
0.0; # Ratio of "directly" to DCVC to systemic DCVG
Scaled by BW^-0.25
# DCVC metabolism
InkNATC = 0.0;
InkKidBioactC
Scaled by BW^-0.25
0.0; # Scaled by BW^-0.25
# Closed chamber parameters
NRodents =1;	#
VChC	=1;	#
InkLossC =0;	#
Population means
-------
s?
>!
§•
rs
s;
These are given truncated normal or uniform distributions, depending on
what prior information is available. Note that these distributions
reflect uncertainty in the population mean, not inter-individual
variability. Normal distributions are truncated at 2, 3, or 4 SD.
For fractional volumes and flows, 2xSD
For plasma fraction, 3xSD
For cardiac output and ventilation-perfusion ratio, 4xSD
For all others, 3xSD
For uniform distributions, range of le2 to le8 fold, centered on
central estimate.
£
to ^
LtJ
On
0
P
fe
H
1
o
o
2! . _
a
<3
o
>!
>!
O
a
a
a
§•
TO
>!
a
o
cs
o
a
S"
(3s
H
1—1 s
H 2.
O •
o
c
o
H
W
M InQCC =1.0

M InVPRC =1.0

M QFatC =1.0

M QGutC =1.0

M QLivC =1.0

M QSlwC =1.0

M QKidC =1.0

M FracPlasC
= 1
M InDRespC = 1.0;
M VFatC =1.0

M VGutC =1.0

M VLivC =1.0

M VRapC =1.0

M VRespLumC = 1
0;
M VRespEffC = 1
0;
M VKidC =1.0

M VBldC =1.0

M InPBC =1.0

M InPFatC =1.0

M InPGutC =1.0

M InPLivC =1.0

M InPRapC =1.0

M InPRespC
= 1
M InPKidC =1.0

M InPSlwC =1.0

M InPRBCPlasTCAC = 1
M InPBodTCAC
= 1
M InPLivTCAC
= 1
M InkDissocC
= 1
M InBMaxkDC
= 1
M InPBodTCOHC
= 1
M InPLivTCOHC
= 1
M InPBodTCOGC
= 1
M InPLivTCOGC
= 1
M InPeffDCVG
= 1
M InkTSD =1.0

M InkAS =1.0

M InkTD =1.0

M InkAD =1.0

M InkASTCA
= 1
M InkASTCOH
= 1
M
InVMAXC = 1.0;


M
InKMC = 1.0;


M
InCIC = 1.0;


M
InFracOtherC
=
1.0
M
InFracTCAC
=
1.0
M
InVMAXDCVGC
=
1.0
M
InClDCVGC
=
1.0
M
InKMDCVGC
=
1.0
M
InVMAXKidDCVGC
=
1.0
M
InClKidDCVGC
=
1.0
M
InKMKidDCVGC
=
1.0
M
InVMAXLungLivC
=
1.0
M
InKMClara
=
1.0
M
InFracLungSysC
=
1.0
M
InVMAXTCOHC
=
1.0
M
InClTCOHC
=
1.0
M
InKMTCOH
=
1.0
M
InVMAXGlucC
=
1.0
M
InClGlucC
=
1.0
M
InKMGluc
=
1.0
M
InkMetTCOHC
=
1.0
M
InkUrnTCAC
=
1.0
M
InkMetTCAC
=
1.0
M
InkBileC
=
1.0
M
InkEHRC = 1.0;


M
InkUrnTCOGC
=
1.0
M
InFracKidDCVCC
=
1.0
M
InkDCVGC
=
1.0
M
InkNATC = 1.0;


M
InkKidBioactC
=
1.0
#	Population Variances
#
# These are given InvGamma(alpha,beta) distributions. The parameterization
#	for alpha and beta is given by:
#	alpha = (n-1)/2
#	beta = s"2*(n-1)/2
#	where n = number of data points, and s^ is the sample variance
#	Sum (x_ir>2 )/n - ^2.
#	Generally, for parameters for which there is no direct data, assume a
#	value of n = 5 (alpha = 2). For a sample variance s^2, this give
#	an expected value for the standard deviation  = 0.9*s,
#	a median [2.5%,97.5%] of l.l*s [ 0.6*s,2.9*s] .
#
V_lnQCC	= 1.0;
V_lnVPRC	= 1.0;
V_QFatC	= 1.0;
V_QGutC	= 1.0;
V_QLivC	= 1.0;
V_QSlwC	= 1.0;
V_QKidC	= 1.0;
-------
s?
>!
§•
rs
s;
>1
£
S^. ^
K) ^
LtJ
^1
o
o
>!
>!
O
a
a
a
§•
>!
5
o
rs
o
a
>i

H
I
o
o
z;. _
o^
H J
1—1 s
H 2.
W|
O •
o
c
o
H
W
V_FracPlasC
V_lnDRespC = 1.0;
V_VFatC
V_VGutC
V_VLivC
V_VRapC
V_VRespLumC
V_VRespEffC
V_VKidC
V_VBldC
V_lnPBC
V_lnPFatC =
V_lnPGutC =
V_lnPLivC =
V_lnPRapC =
V_lnPRespC
V_lnPKidC =
V_lnPSlwC =
V_lnPRBCPlasTCAC
V_lnPBodTCAC
V_lnPLivTCAC
V_lnkDissocC
V_lnBMaxkDC
V_lnPBodTCOHC
V_lnPLivTCOHC
V_lnPBodTCOGC
V_lnPLivTCOGC
V InPeffDCVG
V_lnkTSD = 1
V_lnkAS = 1
V_lnkTD = 1
V_lnkAD = 1
V_lnkASTCA
V_lnkASTCOH
V_lnVMAXC = 1
V_lnKMC = 1
V_lnClC = 1
V_lnFracOtherC
V_lnFracTCAC
V_lnVMAXDCVGC
V_lnClDCVGC
V_lnKMDCVGC
V_lnVMAXKidDCVGC
V_lnClKidDCVGC
V_lnKMKidDCVGC
V_lnVMAXLungLivC
V_lnKMClara
V_lnFracLungSysC
V_lnVMAXTCOHC
V_lnClTCOHC
V_lnKMTCOH
V_lnVMAXGlucC
V InClGlucC
1.0;
1.0;
V_lnKMGluc
V_lnkMetTCOHC
V_lnkUrnTCAC
V_lnkMetTCAC
V_lnkBileC
V_lnkEHRC = 1.
V_lnkUrnTCOGC
V_lnFracKidDCVCC
V_lnkDCVGC
V_lnkNATC = 1.0;
V InkKidBioactC
. 0;
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Measurement error variances for output
= 1;
Ve_RetDose
Ve_CAlv = 1 ;
Ve_CAlvPPM
Ve_CInhPPM
Ve_CInh = 1;
Ve_CMixExh
Ve_CArt = 1;
Ve_CVen = 1;
Ve CBldMix
1;
1;
Ve_CFat
Ve_CGut
Ve_CRap
Ve_CSlw
Ve_CHrt
Ve_CKid
Ve_CLiv
Ve_CLung
Ve_CMus
Ve_CSpl
Ve_CBrn
Ve_zAExh
Ve_zAExhpost
Ve_CTC0H = 1;
Ve_CKidTCOH
Ve_CLivTCOH
Ve_CLungTCOH
Ve_CPlasTCA
Ve_CBldTCA
Ve_CBodTCA
Ve_CKidTCA
Ve_CLivTCA
Ve_CLungTCA
Ve zAUrnTCA
-------
Ve_zAUrnTCA_collect = 1;
Ve_zAUrnTCA_sat	= 1;
S?
>!
rs
s;
>1
£
Ve_zABileTCOG
Ve_CTCOG = 1;
Ve_CTCOGTCOH
Ve_CKidTCOGTCOH
Ve_CLivTCOGTCOH
Ve_CLungTCOGTCOH
Ve AUrnTCOGTCOH
Ve AUrnTCOGTCOH collect
Ve AUrnTCOGTCOH sat
Ve_CDCVGmol
Ve_zAUrnNDCVC
Ve_AUrnTCTotMole
Ve_TotCTCOH
Ve_QPsamp = 1;
to ^
Defaults for input parameters
00
o
o
>!
>!
o
a
a
a
§•
>!
o
rs
o
a
>1

H
I
o
o
2! . _
H J
1—1 s
H 2.
O •
o
c
o
H
W
#— TCE dosing
Cone = 0.0;
IVDose = 0.0;
PDose = 0.0;
Drink = 0.0;
IADose = 0.0;
PVDose = 0.0;
#— TCA dosing
IVDoseTCA = 0.
PODoseTCA = 0.
#— TCOH dosing
IVDoseTCOH
PODoseTCOH
#	Inhalation exposure conc. (ppm)
#	IV dose (mg/kg)
#	Oral gavage dose (mg/kg)
#	Drinking water dose (mg/kg/day)
#	Intraarterial dose (mg/kg)
#	Portal vein dose (mg/kg)
IV dose (mg/kg) of TCA
Oral dose (mg/kg) of TCA
0.0;# IV dose (mg/kg) of TCOH
0.0;# Oral dose (mg/kg) of TCOH
Potentially time-varying parameters
QPmeas =0.0;	# Measured value of Alveolar ventilation QP
TCAUrnSat = 0.0;# Flag for saturated TCA urine
TCOGUrnSat = 0.0;# Flag for saturated TCOG urine
UrnMissing = 0.0;# Flag for missing urine collection times
Initialize
Parameter Initialization and Scaling
Model Parameters	(used in dynamics):
QC	Cardiac output (L/hr)
VPR	Ventilation-perfusion ratio
QPsamp	Alveolar ventilation (L/hr)
#	QFatCtmp	Scaled fat blood flow
#	QGutCtmp	Scaled gut blood flow
#	QLivCtmp	Scaled liver blood flow
#	QSlwCtmp	Scaled slowly perfused blood flow
#	DResptmp	Respiratory lumen:tissue diffusive clearance rate
#	QKidCtmp	Scaled kidney blood flow
#	FracPlas	Fraction of blood that is plasma (1-hematocrit)
#	VFat	Fat compartment volume (L)
#	VGut	Gut compartment volume (L)
#	VLiv	Liver compartment volume (L)
#	VRap	Rapidly perfused compartment volume (L)
#	VRespLum	Volume of respiratory lumen (L air)
#	VRespEff	Effective volume of respiratory tissue (L air)
#	VKid	Kidney compartment volume (L)
#	VBld	Blood compartment volume (L)
#	VSlw	Slowly perfused compartment volume (L)
#	VPlas	Plasma compartment volume [fraction of blood] (L)
#	VBod	TCA Body compartment volume [not incl. blood+liver]
(L)
#	VBodTCOH	TCOH/G Body compartment volume [not incl. liver] (L)
#	PB	TCE Blood/air partition coefficient
#	PFat	TCE Fat/Blood partition coefficient
#	PGut	TCE Gut/Blood partition coefficient
#	PLiv	TCE Liver/Blood partition coefficient
#	PRap	TCE Rapidly perfused/Blood partition coefficient
#	PResp	TCE Respiratory tissue:air partition coefficient
#	PKid	TCE Kidney/Blood partition coefficient
#	PSlw	TCE Slowly perfused/Blood partition coefficient
#	TCAPlas	TCA blood/plasma concentration ratio
#	PBodTCA	Free TCA Body/blood plasma partition coefficient
#	PLivTCA	Free TCA Liver/blood plasma partition coefficient
#	kDissoc	Protein/TCA dissociation constant (umole/L)
#	BMax	Maximum binding concentration (umole/L)
#	PBodTCOH	TCOH body/blood partition coefficient
#	PLivTCOH	TCOH liver/body partition coefficient
#	PBodTCOG	TCOG body/blood partition coefficient
#	PLivTCOG	TCOG liver/body partition coefficient
#	kAS	TCE Stomach absorption coefficient (/hr)
#	kTSD	TCE Stomach-duodenum transfer coefficient (/hr)
#	kAD	TCE Duodenum absorption coefficient (/hr)
#	kTD	TCE Duodenum-feces transfer coefficient (/hr)
#	kASTCA	TCA Stomach absorption coefficient (/hr)
#	kASTCOH	TCOH Stomach absorption coefficient (/hr)
#	VMAX	VMAX for hepatic TCE oxidation (mg/hr)
#	KM	KM for hepatic TCE oxidation (mg/L)
#	FracOther	Fraction of hepatic TCE oxidation not to TCA+TCOH
#	FracTCA	Fraction of hepatic TCE oxidation to TCA
#	VMAXDCVG	VMAX for hepatic TCE GSH conjugation (mg/hr)
#	KMDCVG	KM for hepatic TCE GSH conjugation (mg/L)
#	VMAXKidDCVG	VMAX for renal TCE GSH conjugation (mg/hr)
#	KMKidDCVG	KM for renal TCE GSH conjugation (mg/L)
#	VMAXClara	VMAX for Tracheo-bronchial TCE oxidation (mg/hr)
#	KMClara	KM for Tracheo-bronchial TCE oxidation (mg/L)
-------
FracLungSys	Fraction of respiratory metabolism to systemic circ.
VMAXTCOH	VMAX for hepatic TCOH->TCA (mg/hr)
KMTCOH	KM for hepatic TCOH->TCA (mg/L)
VMAXGluc	VMAX for hepatic TCOH->TCOG (mg/hr)
KMGluc	KM for hepatic TCOH->TCOG (mg/L)
kMetTCOH	Rate constant for hepatic TCOH->other (/hr)
kUrnTCA	Rate constant for TCA plasma->urine (/hr)
U* # kMetTCA	Rate constant for hepatic TCA->other (/hr)
^ # kBile	Rate constant for TCOG liver->bile (/hr)
O # kEHR	Lumped rate constant for TCOG bile->TCOH liver (/hr)
§ # kUrnTCOG	Rate constant for TCOG->urine (/hr)
2 # kDCVG	Rate constant for hepatic DCVG->DCVC (/hr)
S # FracKidDCVC	Fraction of renal TCE GSH conj. "directly" to DCVC


(i.e., via first pass)
VDCVG	DCVG effective volume of distribution
#	kNAT	Lumped rate constant for DCVC->Urinary NAcDCVC (/hr)
S"4
5^ #	kKidBioact	Rate constant for DCVC bioactivation (/hr)
#	Rodents	Number of rodents m closed chamber data
2 #	VCh	Chamber volume for closed chamber data
*
#	kLoss	Rate constant for closed chamber air loss
^ # Parameters used	(not assigned here)
BW	Body weight in kg
#	Species	1 = human (default), 2 = rat, 3 = mou
SJ	#	Male	0 = female, 1 (default) = male
1
K) "5	#	CC	Closed chamber initial concentration
>!	# Sampling/scaling parameters (assigned or sampled)
InQCC
InVPRC
O
a
#	InDRespC
£ #	QFatC
#	QGutC
Si, #	QLivC
g §	#	QSlwC
£ ^	#	QKidC
^	#	FracPlasC
U C5	#	VFatC
•5	#	VGutC
VLivC
VRapC
VRespLumC
VRespEffC
>1
81
2; ^
W "5	#	VKidC
Cb	#	VBldC
O ^	#	InPBC
HH ^3
O	#	InPFatC
tn	#	InPGutC
Q •	#	InPLivC
#	InPRapC
O
C
o
H
tn
InPSlwC
InPRespC
InPKidC
InPRBCPlasTCAC
InPBodTCAC
#	InPLivTCAC
#	InkDissocC
#	InBMaxkDC
#	InPBodTCOHC
#	InPLivTCOHC
#	InPBodTCOGC
#	InPLivTCOGC
#	InPeffDCVG
#	InkTSD
#	InkAS
#	InkTD
#	InkAD
#	InkASTCA
#	InkASTCOH
#	InVMAXC
#	InKMC
#	InCIC
#	InFracOtherC
#	InFracTCAC
#	InVMAXDCVGC
#	InClDCVGC
#	InKMDCVGC
#	InVMAXKidDCVGC
#	InClKidDCVGC
#	InKMKidDCVGC
#	InVMAXLungLivC
#	InKMClara
#	InFracLungSysC
#	InVMAXTCOHC
#	InClTCOHC
#	InKMTCOH
#	InVMAXGlucC
#	InClGlucC
#	InKMGluc
#	InkMetTCOHC
#	InkUrnTCAC
#	InkMetTCAC
#	InkBileC
#	InkEHRC
#	InkUrnTCOGC
#	InFracKidDCVCC
#	InkDCVGC
#	InkNATC
#	InkKidBioactC
#	NRodents
#	VChC
#	InkLossC
#	Input parameters
#	none
#	Notes:
#	use measured value of > 0, otherwise use 0.03 for mouse,
#	0.3 for rat, 60 for female human, 70 for male human
-------
BW = (BWmeas > 0.0 ? BWmeas : (Species ==3 ? 0.03 : (Species ==2 ? 0.3
(Male == 0 ? 60.0 : 70.0) )));
5
BW75 = pow(BW, 0.75);
BW25 = pow(BW, 0.25);
Cardiac Output and alveolar ventilation (L/hr)
U*	QC = exp(lnQCC) * BW75 *	# Mouse, Rat, Human (default)
^	(Species == 3 ? 11.6 : (Species == 2 ? 13.3 : 16.0 ));
O"4	# Mouse: CO=13.98 +/- 2.85 ml/min, BW=3G g (Brown et al. 1997, Tab. 22)
§	#	Uncertainty CV is 0.20
2	# Rat: CO=110.4 ml/min + /- 15.6, BW=396 g (Brown et al. 1997, Tab. 22,
S	#	p 441). Uncertainty CV is 0.14.
#	Human: Average of Male CO=6.5 1/min, BW=73 kg
^	#	and female CO= 5.9 1/min, BW=60 kg (ICRP #89, sitting at rest)
#	From Price et al. 2003, estimates of human perfusion rate were
S"4
5^	#	4.7-6.5 for females and 5.5-7.1 1/min for males (note
#	portal blood was double-counted, and subtracted off here)
^	# Thus for uncertainty use CV of 0.2, truncated at 4xCV
-*S	#	Variability from Price et al. (2003) had CV of 0.14-0.20,
^	#	so use 0.2 as central estimate
g	VPR = exp(InVPRC)*
^	(Species == 3 ? 2.5 : (Species == 2 ? 1.9 : 0.96 ));
SJ	# Mouse: QP/BW=116.5 ml/min/100 g (Brown et al. 1997, Tab. 31), VPR=2.5
1	^
K}	#	Assume uncertainty CV of 0.2 similar to QC, truncated at 4xCV
rv O
>!	#	Consistent with range of QP m Tab. 31
^	# Rat: QP/BW=52.9 ml/min/100 g (Brown et al. 1997, Tab. 31), VPR=1.9
2	#	Assume uncertainty CV of 0.3 similar to QC, truncated at 4xCV
rS
#	Used larger CV because Tab. 31 shows a very large range of QP
^	# Human: Average of Male VE=9 1/min, resp. rate=12 /min,
5^	#	dead space=0.15 1 (QP=7.2 1/min), and Female
^	#	VE=6.5 1/min, resp. rate=14 /min, dead space=0.12 1
2	#	(QP=4 . 8 1/min), VPR =0.96
^0 >!
Assume uncertainty CV of 0.2 similar to QC, truncated at 4xCV
Respiratory diffusion flow rate
Will be scaled by QP in dynamics
Use log-uniform distribution from le-5 to 10
DResptmp = exp(InDRespC);
#	Consistent with range of QP in Tab. 31
Cb	QPsamp = QC*VPR;
I S
I >1
81
z; ^
3 Cb	# Fractional Flows scaled to the appropriate species
#	Fat = Adipose only
O	# Gut = GI tract + pancreas + spleen (all drain to portal vein)
ffl.	# Liv = Liver, hepatic artery
•	# Slw = Muscle + Skin
#	Kid = Kidney
Rap = Rapidly perfused (rest of organs, plus bone marrow, lymph, etc.),
derived by difference in dynamics
Mouse and rat data from Brown et al. (1997). Human data from
O
c
o
H
#	ICRP-89 (2002), and is sex-specific.
)) ;
)) ;
)) ;
QFatCtmp = QFatC*
(Species == 3 ? 0.07 : (Species == 2 ? 0.07 : (Male == 0 ? 0.085 : 0.05)
QGutCtmp = QGutC*
(Species == 3 ? 0.141 : (Species == 2 ? 0.153 : (Male == 0 ? 0.21 : 0.19)
QLivCtmp = QLivC*
(Species == 3 ? 0.02 : (Species == 2 ? 0.021 : 0.065 ));
QSlwCtmp = QSlwC*
(Species == 3 ? 0.217 : (Species == 2 ? 0.336 : (Male == 0 ? 0.17 : 0.22)
QKidCtmp = QKidC*
(Species == 3 ? 0.091 : (Species == 2 ? 0.141 : (Male == 0 ?
0.17 : 0.19) ));
# Plasma Flows to Tissues (L/hr)
## Mice and rats from Hejtmancik et al. 2002,
##	control F344 rats and B6C3F1 mice at 19 weeks of age
## However, there appear to be significant strain differences in rodents, so
##	assume uncertainty CV=0.2 and variability CV=0.2.
## Human central estimate from ICRP. Well measured in humans, from Price et al.
##	human SD in hematocrit was 0.029 in females, 0.027 in males,
##	corresponding to FracPlas CV of 0.047 in females and
##	0.04 8 in males. Use rounded CV = 0.05 for both uncertainty and
variability
## Use measured 1-hematocrit if available
## Truncate distributions at 3xCV to encompass clinical "normal range"
FracPlas = (Hematocritmeas > 0.0 ? (1-Hematocritmeas) : (FracPlasC *
(Species == 3 ? 0.52 : (Species == 2 ? 0.53 : (Male == 0 ? 0.615 :
0.567)))));
#	Tissue Volumes (L)
#	Fat = Adipose only
#	Gut = GI tract (not contents) + pancreas + spleen (all drain to portal vein)
#	Liv = Liver
#	Rap = Brain + Heart + (Lungs-TB) + Bone marrow + "Rest of the body"
#	VResp = Tracheobroncial region (trachea+broncial basal+
#	broncial secretory+bronchiolar)
#	Kid = Kidney
#	Bid = Blood
#	Slw = Muscle + Skin, derived by difference
#	residual (assumed unperfused) = (Bone-Marrow)+GI contents+other
#
#	Mouse and rat data from Brown et al. (1997) . Human data from
#	ICRP-89 (2002), and is sex-specific.
VFat = BW * (VFatCmeas > 0.0 ? VFatCmeas : (VFatC * (Species ==3 ? 0.07 :
(Species == 2 ? 0.07 : (Male == 0 ? 0.317 : 0.199) ))));
VGut = VGutC * BW *
(Species == 3 ? 0.049 : (Species == 2 ? 0.032 : (Male == 0 ? 0.022 :
0.020) ));
VLiv = VLivC * BW *
-------
(Species == 3 ? 0.055 : (Species == 2 ? 0.034 : (Male == 0 ? 0.023 :
0.025) ));
VRap = VRapC * BW *
(Species == 3 ? 0.100 : (Species == 2 ? 0.088 : (Male == 0 ? 0.093 :
0.088) ));
VRespLum = VRespLumC * BW *
^	(Species == 3 ? (0.00014/0.03) : (Species == 2 ? (0.0014/0.3) : (0.167/70)
S"* )); # Lumenal volumes from Styrene model (Sarangapani et al. 2002)
^	VRespEfftmp = VRespEffC * BW *
O"4	(Species == 3 ? 0.0007 : (Species == 2 ? 0.0005 : 0.00018 ));
^	# Respiratory tract volume is TB region
2	# will be multiplied by partition coef. below
S	VKid = VKidC * BW *
£«.	(Species == 3 ? 0.017 : (Species == 2 ? 0.007 : (Male == 0 ? 0.0046 :
0.0043) ));
VBld = VBldC * BW *
^	(Species == 3 ? 0.049 : (Species == 2 ? 0.074 : (Male == 0 ? 0.068 :
^ 0.077) ));
^	VSlw = (Species == 3 ? 0.8897 : (Species == 2 ? 0.8995 : (Male == 0 ?
>2 0.85778 : 0.856) ) ) * BW
ri	- VFat - VGut - VLiv - VRap - VRespEfftmp - VKid - VBld;
^	# Slowly perfused:
^ # Baseline mouse: 0.8 8 97-0.04 9-0.017-0.0 0 07-0.1-0.055-0.04 9-0.07= 0.549
S # Baseline rat: 0.8995 -0.074-0.0 07-0.0 0 05-0.0 8 8-0.034-0.032-0.07= 0.594
to	# Baseline human F: 0.8577 8-0.0 68-0.0 04 6-0.0 0 018-0.0 93-0.023-0.022-0.317= 0.33
§ # Baseline human M: 0.856-0.077-0.0043-0.00018-0.088-0.025-0.02-0.199=0.4425
l—' TO
>!
O
a
VBod = VFat + VGut + VRap + VRespEfftmp + VKid + VSlw; # For TCA
^	VBodTCOH = VBod + VBld;	# for TCOH and TCOG — body without liver
S
a.
,, ^ # Partition coefficients
^ ^	PB = (PBmeas > 0.0 ? PBmeas : (exp(lnPBC) * (Species == 3 ? 15. : (Species
^ ^ 2 ? 22. : 9.5 )))); # Blood-air
^	# Mice: pooling Abbas and Fisher 1997, Fisher et al. 1991
Ci	#	each a single measurement, with overall CV = 0.07.
I §	#	Given small number of measurements, and variability
¦ ^	#	in rat, use CV of 0.25 for uncertainty and variability.
#	Rats: pooling Sato et al. 1977, Gargas et al. 1989,
#	Barton et al. 1995, Simmons et al. 2002, Koizumi 1989,
#	Fisher et al. 1989. Fisher et al. measurement substantially
VPlas = FracPlas * VBld;
81
2!
c>
#	smaller than others (15 vs. 21-26). Recent article
3 Ci	#	by Rodriguez et al. 2007 shows significant change with
^	#	age (13.1 at PND10, 17.5 at adult, 21.8 at aged), also seems
O	#	to favor lower values than previously reported. Therefore
#	use CV = 0.25 for uncertainty and variability,
pooling Sato and Nakajima 1979, Sato et al. 1977,
#	Gargas et al. 1989, Fiserova-Bergerova et al. 1984,
Fisher et al. 1998, Koizumi 1989
O	I
a
o	*
H
ffl	#	uncertainty CV
Overall variability CV = 0.185. Consistent with
within study inter-individual variability CV = 0.07-0.22.
Study-to-study, sex-specific means range 8.1-11, so
0.2.
PFat = exp(lnPFatC) *	# Fat/blood
(Species == 3 ? 36. : (Species == 2 ? 27. : 67. ));
# Mice: Abbas and Fisher 1997. Single measurement. Use
#	rat uncertainty of CV = 0.3.
#	Rats: Pooling Barton et al. 1995, Sato et al. 1977,
#	Fisher et al. 1989. Recent article by Rodriguez et al.
#	(2007) shows higher value of 36., so assume uncertainty
#	CV of 0.3.
#	Humans: Pooling Fiserova-Bergerova et al. 1984, Fisher et al. 1998,
#	Sato et al. 1977. Variability in Fat:Air has CV = 0.07.
#	For uncertainty, dominated by PB uncertainty CV = 0.2
#	For variability, add CVs in guadrature for
#	sgrt(0.07^2+0.185"2)=0.2 0
PGut = exp(lnPGutC) *	# Gut/blood
(Species ==3 ? 1.9 : (Species ==2 ? 1.4 : 2.6 ));
#	Mice: Geometric mean of liver, kidney
#	Rats: Geometric mean of liver, kidney
#	Humans: Geometric mean of liver, kidney
#	Uncertainty of CV = 0.4 due to tissue extrapolation
PLiv = exp(lnPLivC) *	# Liver/blood
(Species ==3 ? 1.7 : (Species ==2 ? 1.5 : 4.1 ));
#	Mice: Fisher et al. 1991, single datum, so assumed uncert CV = 0.4
#	Rats: Pooling Barton et al. 1995, Sato et al. 1977,
#	Fisher et al. 1989, with little variation (range 1.3-1.7).
#	Recent article by Rodriguez et al.reports 1.34. Use
#	uncertainty CV = 0.15.
#	Humans: Pooling Fiserova-Bergerova et al. 1984, Fisher et al. 1998
#	almost 2-fold difference in Liver:Air values, so uncertainty
#	CV = 0.4
PRap = exp(lnPRapC) *	# Rapidly perfused/blood
(Species == 3 ? 1.9 : (Species == 2 ? 1.3 : 2.6 ));
#	Mice: Similar to liver, kidney. Uncertainty CV = 0.4 due to
#	tissue extrapolation
#	Rats: Use brain values Sato et al. 1977. Recent article by
#	Rodriguez et al. (2007) reports 0.99 for brain. Uncertainty
#	CV of 0.4 due to tissue extrapolation.
#	Humans: Use brain from Fiserova-Bergerova et al. 1984
#	Uncertainty of CV = 0.4 due to tissue extrapolation
PResp = exp(InPRespC) *	# Resp/blood =
(Species ==3 ? 2.6 : (Species ==2 ? 1.0 : 1.3 ));
#	Mice: Abbas and Fisher 1997, single datum, so assumed uncert CV = 0.4
#	Rats: Sato et al. 1977, single datum, so assumed uncert CV = 0.4
#	Humans: Pooling Fiserova-Bergerova et al. 1984, Fisher et al. 1998
#	> 2-fold difference in lung:air values, so uncertainty
#	CV = 0.4
VRespEff = VRespEfftmp * PResp * PB; # Effective air volume
PKid = exp(lnPKidC) *	# Slowly perfused/blood
(Species == 3 ? 2.1 : (Species == 2 ? 1.3 : 1.6 ));
#	Mice: Abbas and Fisher 1997, single datum, so assumed uncert CV = 0.4
#	Rats: Pooling Barton et al. 1995, Sato et al. 1977. Recent article
#	by Rodriguez et al. (2007) reports 1.01, so use uncertainty
#	CV of 0.3. Pooled variability CV = 0.39.
#	Humans: Pooling Fiserova-Bergerova et al. 1984, Fisher et al. 1998
-------
#	For uncertainty, dominated by PB uncertainty CV = 0.2
#	Variability in kidney:air CV = 0.23, so add to PB variability
#	in quadrature	sqrt(0.23^2+0.185^2)=0.30
PSlw = exp(lnPSlwC) * # Slowly perfused/blood
(Species == 3 ? 2.4 : (Species == 2 ? 0.58 : 2.1 ));
#	Mice: Muscle - Abbas and Fisher 1997, single datum, so assumed
#	uncert CV = 0.4
#	Rats: Pooling Barton et al. 1995, Sato et al. 1977,
#	Fisher et al. 1989. Recent article by Rodriguez et al. (2007)
#	reported 0.72, so use uncertainty CV of 0.25. Variability
#	in Muscle:air and muscle:blood - CV = 0.3
#	Humans: Pooling Fiserova-Bergerova et al. 1984, Fisher et al. 1998
#	Range of values 1.4~2.4, so uncertainty CV = 0.3
#	Variability in muscle:air CV = 0.3, so add to PB variability
#	in quadrature sqrt(0.3^2+0.185^2)=0.35
TCA partitioning
TCAPlas = FracPlas + (1 - FracPlas) * 0.5 * exp(InPRBCPlasTCAC);
#	Blood/Plasma concentration ratio. Note dependence
#	on fraction of blood that is plasma. Here
#	exp(InPRBCPlasTCA) = partition coefficient
#	C(blood minus plasma)/C(plasma)
#	Default of 0.5, corresponding to Blood/Plasma
#	concentration ratio of 0.76 in
#	rats (Schultz et al 1999)
#	For rats, Normal uncertainty with GSD = 1.4
#	For mice and humans, diffuse prior uncertainty of
#	100-fold up/down
PBodTCA = TCAPlas * exp(InPBodTCAC) *
(Species == 3 ? 0.88 : (Species == 2 ? 0.88 : 0.52 ));
#	Note -- these were done at 10~20 microg/ml (Abbas and Fisher 1997),
#	which is 1.635-3.27 mmol/ml (1.635-3.27 x 10^6 microM).
#	At this high concentration, plasma binding should be
#	saturated -- e.g., plasma albumin concentration was
#	measured to be P=190-239 microM in mouse, rat, and human
#	plasma by Lumpkin et al. 2003, or > 6800 molecules of
#	TCA per molecule of albumin. So the measured partition
#	coefficients should reflect free blood-tissue partitioning.
#	Used muscle values, multiplied by blood:plasma ratio to get
#	Body:Plasma partition coefficient
#	Rats = mice from Abbas and Fisher 1997
#	Humans from Fisher et al. 1998
#	Uncertainty in mice, humans GSD = 1.4
#	For rats, GSD = 2.0, based on difference between mice
#	and humans.
PLivTCA = TCAPlas * exp(InPLivTCAC) *
(Species == 3 ? 1.18 : (Species == 2 ? 1.18 : 0.66 ));
#	Multiplied by blood:plasma ratio to get Liver:Plasma
#	Rats = mice from Abbas and Fisher 1997
#	Humans from Fisher et al. 1998
#	Uncertainty in mice, humans GSD = 1.4
#	For rats, GSD = 2.0, based on difference between mice
#	and humans.
#	Binding Parameters for TCA
#	GM of Lumpkin et al. 2003; Schultz et al. 1999;
#	Templin et al. 1993, 1995; Yu et al. 2000
#	Protein/TCA dissociation constant (umole/L)
#	note - GSD = 3.29, 1.84, and 1.062 for mouse, rat, human
kDissoc = exp(InkDissocC) *
(Species == 3 ? 107. : (Species == 2 ? 275. : 182. ));
#	BMax = NSites * Protein concentration. Sampled parameter is
#	BMax/kD (determines binding at low concentrations)
#	note - GSD = 1.64, 1.60, 1.20 for mouse, rat, human
BMax = kDissoc * exp(InBMaxkDC) *
(Species == 3 ? 0.88 : (Species == 2 ? 1.22 : 4.62 ));
#	TCOH partitioning
#	Data from Abbas and Fisher 1997 (mouse) and Fisher et al.
#	1998 (human). For rat, used mouse values.
#	Uncertainty in mice, humans GSD = 1.4
#	For rats, GSD = 2.0, based on difference between mice
#	and humans.
PBodTCOH = exp(InPBodTCOHC) *
(Species == 3 ? 1.11 : (Species == 2 ? 1.11 : 0.91 ));
PLivTCOH = exp(InPLivTCOHC) *
(Species == 3 ? 1.3 : (Species == 2 ? 1.3 : 0.59 ));
#	TCOG partitioning
#	Use TCOH as a proxy, but uncertainty much greater
#	(e.g., use uniform prior, 100-fold up/down)
PBodTCOG = exp(InPBodTCOGC) *
(Species == 3 ? 1.11 : (Species == 2 ? 1.11 : 0.91 ));
PLivTCOG = exp(InPLivTCOGC) *
(Species == 3 ? 1.3 : (Species == 2 ? 1.3 : 0.59 ));
#	DCVG distribution volume
#	exp(InPeffDCVG) is the effective partition coefficient for
#	the "body" (non-blood) compartment
#	Diffuse prior distribution: loguniform le-3 to le3
VDCVG = VBld +	# blood plus body (with "effective" PC)
exp(InPeffDCVG) * (VBod + VLiv);
#	Absorption Rate Constants (/hr)
#	All priors are diffuse (log)uniform distributions
#	transfer from stomach centered on 1.4/hr, range up or down 100-fold,
#	based on human stomach half-time of 0.5 hr.
kTSD = exp(InkTSD);
#	stomach absorption centered on 1.4/hr, range up or down 1000-fold
kAS = exp(lnkAS);
#	assume no fecal excretion -- 100% absorption
kTD = 0.0 * exp(InkTD);
#	intestinal absorption centered on 0.75/hr, range up or down
#	1000-fold, based on human transit time of small intestine
#	of 4 hr (951 throughput in 4 hr)
-------
kAD = exp(lnkAD);
kASTCA = exp(InkASTCA);
kASTCOH = exp(InkASTCOH);
#	TCE Oxidative Metabolism Constants
#	For rodents, in vitro microsomal data define priors (pooled).
^ # For human, combined in vitro microsomoal+hepatocellular individual data
S"* #	define priors.
^	# All data from Elfarra et al. 1998; Lipscomb et al. 1997, 1998a,b
O"4 # For VMAX, scaling from in vitro data were (Barter et al. 2007):
§	#	32 mg microsomal protein/g liver
2	#	99 x le6 hepatocytes/g liver
S	#	Here, human data assumed representative of mouse and rats.
#	For KM, two different scaling methods were used for microsomes:
^	#	Assume microsomal concentration = liver concentration, and
#	use central estimate of liver:blood PC (see above)
S"4
5^	#	Use measured microsome:air partition coefficient (1.78) and
#	central estimate of blood:air PC (see above)
^	# For human KM from hepatocytes, used measured human hepatocyte:air
#	partition coefficient (21.62, Lipscomb et al. 1998), and
^ # central estimate of blood:air PC.
^	#	Note that to that the hepatocyte:air PC is similar to that
#	found in liver homogenates (human: 29.4+/-5.1 from Fiserova-
SJ	#	Bergerova et al. 1984, and 54 for Fisher et al. 1998; rat:
I ""'S
^	#	27.2+/-3.4 from Gargas et al. 1989, 62.7 from Koisumi 1989,
§	#	43.6 from Sato et al. 1977; mouse: 23.2 from Fisher et al. 1991).
>!	# For humans, sampled parameters are VMAX and C1C (VMAX/KM), due to
2	#	improved convergence. VMAX is kept as a parameter because it
#	appears less uncertain (i.e., more consistent across microsomal
C5	#	and hepatocyte data).
S
a.
^	# Central estimate of VMAX is 342, 76.2, and 32.3 (micromol/min/
#	kg liver) for mouse, rat, human. Converting to /hr by
^ ? #
5
81
2! ^
O ^
(60 min/hr * 0.1314 mg/micromol) gives
#	2700, 600, and 255 mg/hr/kg liver
^ Cb	# Observed variability of about 2-fold GSD. Assume 2-fold GSD for
I §	#	both uncertainty and variability
' &	VMAX = VLiv*exp(InVMAXC)*
(Species == 3 ? 2700. : (Species == 2 ? 600. : 255.));
For mouse and rat central estimates for KM are 0.068~1.088 and
#	0.039-0.679 mmol/1 in blood, depending on the scaling
3 fv,	#	method used. Taking the geometric mean, and converting
#	to mg/1 by 131.4 mg/mmol gives 36. and 21. mg/1 in blood.
O # For human, central estimate
ffl.	#	for CI are 0.306~3.95 1/min/kg liver. Taking the geometric
•	#	mean and converting to /hr gives a central estimate of
^	#	66. 1/hr/kg.
#	KM is then derived from KM = VMAX/(Cl*Vliv) (central estimate
O
c
o
H
W	#	2- to 4- fold in humans (depending on scaling).
of
Note uncertainty due to scaling is about 4-fold.
Variability is about 3-fold in mice, 1.3-fold in rats, and
KM = (Species == 3 ? 36.*exp(InKMC) : (Species == 2 ? 21.*exp(InKMC)
VMAX/(VLiv* 66.*exp(InCIC))));
Oxidative metabolism splits
#	Fractional split of TCE to DCA
#	exp(InFracOtherC) = ratio of DCA to non-DCA
#	Diffuse prior distribution: loguniform le-4 to le2
FracOther = exp(InFracOtherC)/(1+exp(InFracOtherC));
#	Fractional split of TCE to TCA
#	exp(InFracTCAC) = ratio of TCA to TCOH
#	TCA/TCOH = 0.1 from Lipscomb et al. 1998 using fresh hepatocytes,
#	but TCA/TCOH ~ 1 from Bronley-DeLancey et al 2006
#	GM = 0.32, GSD =3.2
FracTCA = 0.32*exp(InFracTCAC)*(1-FracOther)/(1+0.32*exp(InFracTCAC));
TCE GSH Metabolism Constants
Human in vitro data from Lash et al. 1999, define human priors.
VMAX (nmol/min/	KM (mM)
g tissue)
CLeff (ml/min/
g tissue)
Human liver cytosol:
Human liver cytosol+
microsomes
Human hepatocytes*
Human kidney cytosol:
[high affinity pathway only] [total]
-423	0.0055-0.023
-211
[total]
12-30**
81
[total]
0.012-0.039***
0.0164-0.0263
21.2-87.0
[total]
0.2-0.5***
3.08-4.93
* estimated visually from Fig 1, Lash et al. 1999
** Fig 1A, data from 50-500 ppm headspace at 60 min
and Fig IB, data at 100-5000 ppm in headspace for 120 min
*** Fig IB, 30-100 ppm headspace, converted to blood concentration
using blood:air PC of 9.5
**** Fig 1A, data at 50 ppm headspace at 120 min and Fig IB, data at
25 and 50 ppm headspace at 120 min.
Overall, human liver hepatocytes are probably most like the
intact liver (e.g., accounting for the competition between
GSH conjugation and oxidation). So central estimates based
on those: CLeff - 0.32 ml/min/g tissue, KM - 0.022 mM in blood.
CLeff converted to 19 1/hr/kg; KM converted to 2.9 mg/1 in blood
However, uncertainty in CLeff is large (values in cytosol
-100-fold larger). Moreover, Green et al. 1997 reported
DCVG formation in cytosol that was -30,000-fold smaller
than Lash et al. (1998) in cytosol, which would be a VMAX
-300-fold smaller than Lash et al. (1998) in hepatocytes.
Uncertainty in KM appears smaller (-4-fold)
CLC: GM = 19., GSD = 100; KM: GM = 2.9., GSD = 4.
In addition, at a single concentration, the variability
in human liver cytosol samples had a GSD=1.3.
For the human kidney, the kidney cytosol values are used, with the same
uncertainty as for the liver. Note that the DCVG formation rates
in rat kidney cortical cells and rat cytosol are guite similar
(see below).
CLC: GM = 230., GSD = 100; KM: GM = 2.7., GSD = 4.
-------
s?
>!
§•
rs
s;
>1
£
S^. ^
TO
>!
O
a
a
a
§•
>!
5
o
rs
o
a
>i
o
&
H
I
o
o
z;. _
o^
a

H
1—1 s
H 2.
W|
O •
o
c
o
H
W
Rat and mouse in vitro data from Lash et al. 1995,1998 define rat and mouse
priors. However, rats and mice are only assayed at 1 and 2	mM
providing only a bound on VMAX and very little data on KM.
Rate at 2 mM	Equivalent	CLeff
blood conc.	at 2 mM
(nmol/min/	(mM)	(ml/min/
g tissue)	g tissue)
Rat
hepatocytes:
liver cytosol:
kidney cells:
kidney cytosol:
liver cytosol:
kidney cytosol:
4.4-16
8.0-12
0.79-1.1 2.2
0.53-0.75 1.1-2.
36-40
6.2-9.3
2.0
1.7-2.0
1.1-2.0
0.91-2.0
0.0022-0.0079
0.0040-0.0072
0.00036-0.00049
0.00027-0.00068
0.018-0.036
0.0031-0.0102
In most cases, rates were increased over the same sex/species at 1 mM,
indicating VMAX has not yet been reached. The values between cells
and cytosol are more much consistent that in the human data.
These data therefore put a lower bound on VMAX and a lower bound
on CLC. To account for in vitro-in vivo uncertainty, the lower
bound of the prior distribution is set 100-fold below the central
estimate of the measurements here. In addition, Green et al.
(1997) found values 100-fold smaller than Lash et al. 1995, 1998.
Therefore diffuse prior distributions set to le-2~le4.
Rat liver: Bound on VMAX of 4.4-16, with GM of 8.4. Converting to
mg/hr/kg tissue (* 131.4 ng/nmol * 60 min/hr * le3 g/kg / le6 mg/ng)
gives a central estimate of 66. mg/hr/kg tissue. Bound on CL of
0.0022-0.0079, with GM of 0.0042. Converting to 1/hr/kg tissue
(* 60 min/hr) gives 0.25 1/hr/kg tissue.
Rat kidney: Bound on VMAX of 0.53-1.1, with GM of 0.76. Converting
to mg/hr/kg tissue gives a central estimate of 6.0 mg/hr/kg.
Bound on CL of 0.00027-0.00068, with GM of 0.00043. Converting
to 1/hr/kg tissue gives 0.026 1/hr/kg tissue.
Mouse liver: Bound on VMAX of 36-40, with GM of 38. Converting
to mg/hr/kg tissue gives a central estimate of 300. mg/hr/kg.
Bound on CL of 0.018-0.036, with GM of 0.025. Converting
to 1/hr/kg tissue gives 1.53 1/hr/kg tissue.
Mouse kidney: Bound on VMAX of 6.2-9.3, with GM of 7.6. Converting
to mg/hr/kg tissue gives a central estimate of 60. mg/hr/kg.
Bound on CL of 0.0031-0.0102, with GM of 0.0056. Converting
to 1/hr/kg tissue gives 0.34 1/hr/kg tissue.
VMAXDCVG = VLiv*(Species == 3 ? (300.*exp(InVMAXDCVGC)) : (Species == 2
exp(InVMAXDCVGC)) : (2.9*19.*exp(InClDCVGC+lnKMDCVGC))));
KMDCVG = (Species == 3 ? (VMAXDCVG/(VLiv*1.53*exp(InClDCVGC))) : (Species
0.25*exp(InClDCVGC))) : 2.9* exp(InKMDCVGC)));
VKid*(Species == 3 ? (60.*exp(InVMAXKidDCVGC)) : (Species
(2.7*230.*exp(InClKidDCVGC+lnKMKidDCVGC))));
3 ? (VMAXKidDCVG/(VKid*0.34 *exp(InClKidDCVGC))) :
exp(InClKidDCVGC))) :
(66.
2 ? (VMAXDCVG/(VL
VMAXKidDCVG
2 ? (6.0*exp(InVMAXKidDCVGC))
KMKidDCVG = (Species
(Species == 2 ? (VMAXKidDCVG/(VKid*0.02
2.7
kexp(InKMKidDCVGC)));
TCE Metabolism Constants for Chloral Kinetics in Lung (mg/hr)
#	Scaled to liver VMAX using data from Green et al. (1997)
#	in microsomal preparations (nmol/min/mg protein) at -1 mM.
#	For humans, used detection limit of 0.03
#	Additional scaling by lung/liver weight ratio
#	from Brown et al. Table 21 (mouse and rat) or
#	ICRP Pub 89 Table 2.8 (Human female and male)
#	Uncertainty - 3-fold truncated at 3 GSD
VMAXClara = exp(InVMAXLungLivC) * VMAX *
(Species == 3 ? (1. 03/1. 87*0 . 7/5 . 5) : (Species == 2 ?
(0.08/0.82*0.5/3.4) : (0.03/0.33*(Male == 0 ? (0.42/1.4) : (0.5/1.8)))));
KMClara = exp(InKMClara);
#	Fraction of Respiratory Metabolism that goes to system circulation
#	(translocated to the liver)
FracLungSys = exp(InFracLungSysC)/(1 + exp(InFracLungSysC));
#	TCOH Metabolism Constants (mg/hr)
#	No in vitro data. So use diffuse priors of
#	le-4 to le4 mg/hr/kg^O.75 for VMAX
#	(4e-5 to 4000 mg/hr for rat),
#	le-4 to le4 mg/1 for KM,
#	and le-5 to le3 l/hr/kg^0.75 for CI
#	(2e-4 to 2.4e4 1/hr for human)
VMAXTCOH = BW75*
(Species == 3 ? (exp(InVMAXTCOHC)) : (Species == 2 ?
(exp(InVMAXTCOHC)) : (exp(InClTCOHC+lnKMTCOH))));
KMTCOH = exp(InKMTCOH);
VMAXGluc = BW75*
(Species == 3 ? (exp(InVMAXGlucC)) : (Species == 2 ?
(exp(InVMAXGlucC)) : (exp(InClGlucC+lnKMGluc))));
KMGluc = exp(InKMGluc);
#	No in vitro data. So use diffuse priors of
#	le-5 to le3 kg^0.25/hr (3.5e-6/hr to 3.5e2/hr for human)
kMetTCOH = exp(InkMetTCOHC) / BW25;
#	TCA kinetic parameters
#	Central estimate based on GFR clearance per unit body weight
#	10.0, 8.7, 1.8 ml/min/kg for mouse, rat, human
#	(= 0.6, 0.522, 0.108 1/hr/kg) from Lin 1995.
#	= CL_GFR / BW (BW=0.02 for mouse, 0.265 for rat, 70 for human)
#	kUrn = CL_GFR / VPlas
#	Diffuse prior with uncertainty of up,down 100-fold
kUrnTCA = exp(InkUrnTCAC) * BW / VPlas *
(Species == 3 ? 0.6 : (Species == 2 ? 0.522 : 0.108));
#	No in vitro data. So use diffuse priors of
#	le-4 to le2 /hr/kg^O.25 (0.3/hr to 35/hr for human)
kMetTCA = exp(InkMetTCAC) / BW25;
#	TCOG kinetic parameters
#	No in vitro data. So use diffuse priors of
#	le-4 to le2 /hr/kg^O.25 (0.3/hr to 35/hr for human)
kBile = exp(InkBileC) / BW25;
kEHR = exp(InkEHRC) / BW25;
#	Central estimate based on GFR clearance per unit body weight
-------
#	10.0, 8.7, 1.8 ml/min/kg for mouse, rat, human
#	(= 0.6, 0.522, 0.108 1/hr/kg) from Lin 1995.
#	= CL_GFR / BW (BW=0.02 for mouse, 0.265 for rat, 70 for human)
#	kUrn = CL_GFR / VBld
#	Diffuse prior with Uncertainty of up,down 1000-fold
kUrnTCOG = exp(InkUrnTCOGC) * BW / (VBodTCOH * PBodTCOG) *
^	(Species == 3 ? 0.6 : (Species == 2 ? 0.522 : 0.108));
^	# DCVG Kinetics (/hr)
O	# Fraction of renal TCE GSH con], "directly" to DCVC via "first pass"
^	# exp(InFracOtherCC) = ratio of direct/non-direct
2	# Diffuse prior distribution: loguniform le-3 to le3
S	# FIXED in vl.2.3
#	In ".in" files, set to 1, so that all kidney GSH conjugation
^	# is assumed to directly produce DCVC (model lacks identifiability
#	otherwise).
5^	FracKidDCVC = exp(InFracKidDCVCC)/(1 + exp(InFracKidDCVCC));
#	No m vitro data. So use diffuse priors of
^	#	le-4 to le2 /hr/kg"0.25 (0.3/hr to 35/hr for human)
•>S	kDCVG = exp (InkDCVGC) / BW25;
DCVC Kinetics in Kidney (/hr)
# No in vitro data. So use diffuse priors of
SJ	#	le-4 to le2 /hr/kg^O.25 (0.3/hr to 35/hr for human)
I ""'S
(\J ^3	kNAT = exp (InkNATC) / BW25;
§	kKidBioact = exp(InkKidBioactC) / BW25;
L/i
>!
2	# CC data initialization
S
Rodents = (CC > 0 ? NRodents : 0.0); # Closed chamber simulation
VCh = (CC > 0 ? VChC - (Rodents * BW) : 1.0);
5^	# Calculate net chamber volume
^	kLoss = (CC > 0 ? exp(InkLossC) : 0.0);
0	£
>	a
r"	O	State Variable Initialization and Scaling
1	^	# NOTE: All State Variables are automatically set to 0 initially,
I >1
81
2! ^
o
l-J s
^ ,Ci	###################### End of Initialization
HH ^3
O	Dynamics{
w|"
#***	Dynamic physiological parameter scaling
O
c
o
H
unless re-initialized here
ACh = (CC * VCh * MWTCE) / 24450.0; # Initial amount in chamber
State Variables with dynamics:
none
Input Variables:
ffl	#	QPmeas
#	Other State Variables and Global Parameters:
#	QC
#	VPR
#	DResptmp
#	QPsamp
#	QFatCtmp
#	QGutCtmp
#	QLivCtmp
#	QSlwCtmp
#	QKidCtmp
#	FracPlas
#	Temporary variables used:
#	none
#	Temporary variables assigned:
#	QP
#	DResp
#	QCnow
#	QFat
#	QGut
#	QLiv
#	QSlw
#	QKid
#	QGutLiv
#	QRap
#	QCPlas
#	QBodPlas
#	QGutLivPlas
#	Notes:
#	QP uses QPmeas if value is > 0, otherwise uses sampled value
QP = (QPmeas > 0 ? QPmeas : QPsamp);
DResp = DResptmp * QP;
#	QCnow uses QPmeas/VPR if QPmeas > 0, otherwise uses sampled value
QCnow = (QPmeas > 0 ? QPmeas/VPR : QC);
These done here in dynamics in case QCnow changes
Blood Flows to Tissues (L/hr)
QFat = (QFatCtmp) * QCnow; #
QGut = (QGutCtmp) * QCnow; #
QLiv = (QLivCtmp) * QCnow; #
QSlw = (QSlwCtmp) * QCnow; #
QKid = (QKidCtmp) * QCnow; #
QGutLiv = QGut + QLiv; #
QRap = QCnow - QFat - QGut - QLiv - QSlw - QKid;
QRapCtmp = QRap/QCnow; #(vrisk)
QBod = QCnow - QGutLiv;
# Plasma Flows to Tissues (L/hr)
QCPlas = FracPlas * QCnow; #
QBodPlas = FracPlas * QBod; #
-------
QGutLivPlas = FracPlas * QGutLiv; #
Exposure and Absorption calculations
O
c
o
H
W
State Variables with dynamics:
AS torn
U* #	ADuod
^ #	AStomTCA
#	AStomTCOH
^ #	Input Variables:
2 #	IVDose
S #	PDose
Drink
Cone
#	IVDoseTCA
S"4
^ #	PODoseTCA
^ #	IVDoseTCOH
O #	PODoseTCOH
*
"-S #	Other State Variables and Global Parameters:
^ #	ACh
g #	CC
^ #	VCh
^ SJ #	MWTCE
IS) ^ #	BW
£ § #	TChng
G\ ri
&a #	kAS
kTSD
kAD
S	#
a
^	#
£	#	kTD
#	kASTCA
S^,	# kASTCOH
#	Temporary variables used:
&rj
r s	#
lIj O	# Temporary variables assigned:
^ Cb	#	klV - rate into CVen
I §	#	klA - rate into CArt
¦ ^	#	kPV - rate into portal vein
kStom - rate into stomach
kDrink - incorporated into RAO
RAO - rate into gut (oral absorption - both gavage and drinking wat
81
2!
c>
#	CInh - inhalation exposure concentration
l-J 5
~ Cb	#	klVTCA - rate into blood
O	#	kStomTCA - rate into stomach
HH ^3
^ O	#	kPOTCA - rate into liver (oral absorption)
fb' #	klVTCOH - rate into blood
•	#	kStomTCOH - rate into stomach
#	kPOTCOH - rate into liver (oral absorption)
Notes:
For oral dosing, using "Spikes" for instantaneous inputs
Inhalation Concentration (mg/L)
CInh uses Cone when open chamber (CC=0) and
ACh/VCh when closed chamber COO.
## TCE DOSING
IV route
klV = (IVDose *
klA = (IADose *
kPV = (PVDose *
kStom = (PDose
BW) / TChng;# IV infusion rate (mg/hr)
# (IVDose constant for duration TChng)
BW) / TChng;	# IA infusion rate (mg/hr)
BW) / TChng;	# PV infusion rate (mg/hr)
* BW) / TChng;# PO dose rate (into stomach) (mg/hr)
## Oral route
# Amount of TCE in stomach -- for oral dosing only (mg)
dt(AStom) = kStom - AStom * (kAS + kTSD);
#	Amount of TCE in duodenum -- for oral dosing only (mg)
dt(ADuod) = (kTSD * AStom) - (kAD + kTD) * ADuod;
#	Rate of absorption from drinking water
kDrink = (Drink * BW) / 24.0; #Ingestion rate via drinking water (mg/hr)
#	Total rate of absorption including gavage and drinking water
RAO = kDrink + (kAS * AStom) + (kAD * ADuod);
## Inhalation route
CInh = (CC > 0 ? ACh/VCh : Conc*MWTCE/24450.0); # in mg/1
#### TCA Dosing
klVTCA = (IVDoseTCA * BW) / TChng; # TCA IV infusion rate (mg/hr)
kStomTCA = (PODoseTCA * BW) / TChng; # TCA PO dose rate into stomach
dt(AStomTCA) = kStomTCA - AStomTCA * kASTCA;
kPOTCA = AStomTCA * kASTCA; # TCA oral absorption rate (mg/hr)
#### TCOH Dosing
klVTCOH = (IVDoseTCOH * BW) / TChng;#TCOH IV infusion rate (mg/hr)
kStomTCOH = (PODoseTCOH * BW) / TChng; # TCOH PO dose rate into stomach
dt(AStomTCOH) = kStomTCOH - AStomTCOH * kASTCOH;
kPOTCOH = AStomTCOH * kASTCOH;# TCOH oral absorption rate (mg/hr)
TCE Model
State Variables with dynamics
ARap,	#
ASlw,	#
AFat,	#
AGut,	#
ALiv,	#
AInhResp,
AResp,
AExhResp,
AKid,	#
ABld,	#
ACh,	#
Input Variables:
none
Other State Variables and Global Parameters
Amount in rapidly perfused tissues
Amount in slowly perfused tissues
Amount in fat
Amount in gut
Amount in liver
Amount in Kidney
Amount in Blood -
- currently in Rap tissue
currently in Rap tissue
Amount of TCE in closed chamber
-------
#	VRap
#	PRap
#	VSlw
#	PSlw
#	VFat
#	PFat
^ #	VGut
S- #	PGut
>!
o
c
o
H
W
VLiv
O	#	PLiv
^	#	VRespLum
2	#	VRespEff
S	#	FracLungSys
VKid
PKid
#	VBld
S"4
^	#	VMAXClara
^	#	KMClara
^	#	PB
•>S	#	Rodents
^	#	VCh
^	#	kLoss
#	VMAX
Si	#	KM
1 K?
(\J ^3	#	VMAXDCVG
§	#	KMDCVG
&5	#	VMAXKidDCVG
2	#	KMKidDCVG
#	Temporary variables used:
£	#	QM
^	#	QFat
.	#	QGutLiv
O TO	*	QSlw
^ S	#	QRap
L O	#	QKid
lj n	#	kiv
~ o
^	#	QCnow
CInh
QP
RAO
Temporary variables assigned:
81
z; ^
O 'to #	QM
l-J S
^ £	#	CRap
O J?	#	CSlw
HH ^3
O	#	CFat
W	#	CGut
Q •	#	CLiv
^	#	CInhResp
CResp
CExhResp
ExhFactor
CMixExh
CKid
#	CVRap
#	CVSlw
#	CVFat
#	CVGut
#	CVLiv
#	CVTB
#	CVKid
#	CVen
#	RAMetLng
#	CArt_tmp
#	CArt
#	CAlv
#	RAMetLivl
#	RAMetLiv2
#	RAMetKid
#	Notes:
#
#* * * *Blood (venous
# Tissue Concentrations (mg/L)
CRap
= ARap/VRap
CSlw
= ASlw/VSlw
CFat
= AFat/VFat
CGut
= AGut/VGut
CLiv
= ALiv/VLiv
CKid
= AKid/VKid
Venous Concent
rations
(mg/L
CVRap
= CRap

PRap
CVSlw
= CSlw

PSlw
CVFat
= CFat

PFat
CVGut
= CGut

PGut
CVLiv
= CLiv

PLiv
CVKid
= CKid

PKid
#	Concentration of TCE in mixed venous blood (mg/L)
CVen = ABld/VBld;
#	Dynamics for blood
dt(ABld) = (QFat*CVFat + QGutLiv* CVLiv + QSlw*CVSlw +
QRap*CVRap + QKid*CVKid + klV) - CVen * QCnow;
#****Gas exchange and Respiratory Metabolism*4"*'* **•*•* * * * * * * * * * * * * ***
#
QM = QP/0.7; # Minute-volume
CInhResp = AInhResp/VRespLum;
CResp = AResp/VRespEff;
CExhResp = AExhResp/VRespLum;
dt(AInhResp) = (QM*CInh + DResp*(CResp-CInhResp) - QM*CInhResp);
RAMetLng = VMAXClara * CResp/(KMClara + CResp);
dt(AResp) = (DResp*(CInhResp + CExhResp - 2*CResp) - RAMetLng);
CArt_tmp = (QCnow*CVen + QP*CInhResp)/(QCnow + (QP/PB));
dt(AExhResp) = (QM*(CInhResp-CExhResp) + QP*(CArt_tmp/PB-CInhResp)
DResp*(CResp-CExhResp));
CMixExh = (CExhResp > 0 ? CExhResp : le-15); # mixed exhaled breath
-------
#	Concentration in alveolar air (mg/L)
#	Correction factor for exhaled air to account for
#	absorption/desorption/metabolism in respiratory tissue
#	= 1 if DResp = 0
ExhFactor_den = (QP * CArt_tmp / PB + (QM-QP)*CInhResp);
^	ExhFactor = (ExhFactor_den > 0) ? (
U*	QM * CMixExh / ExhFactor_den) : 1;
^	# End-exhaled breath (corrected for absorption/
O	#	desorption/metabolism in respiratory tissue)
§	CAlv = CArt_tmp / PB * ExhFactor;
2	# Concentration in arterial blood entering circulation (mg/L)
S	CArt = CArt_tmp + klA/QCnow; # add inter-arterial dose
5^	t^^Other dynamics for inhalation/exhalation * * * * * * * * * * * * * * * * * * * * * * * * * 4
#	Dynamics for amount of TCE in closed chamber
5^ dt(ACh) = (Rodents * (QM * CMixExh - QM * ACh/VCh)) - (kLoss * ACh);
O	# * * * * Non-metabolizing tissues ******************************************
"-S	# Amount of TCE in rapidly perfused tissues (mg)
^	dt(ARap) = QRap * (CArt - CVRap);
^	# Amount of TCE in slowly perfused tissues
^ dt(ASlw) = QSlw * (CArt - CVSlw);
Si	# Amount of TCE in fat tissue (mg)
1 K?
KJ ^3	dt(AFat) = QFat*(CArt - CVFat);
rv O
>!	# Amount of TCE m gut compartment (mg)
00
dt(AGut) = (QGut * (CArt - CVGut)) + RAO;
O
a
^	# Rate of TCE oxidation by P450 to TCA, TCOH, and other (DCA) in liver (mg/h
RAMetLivl = (VMAX * CVLiv) / (KM + CVLiv);
,, ^ # Rate of TCE metabolized to DCVG in liver (mg)
2	RAMetLiv2 = (VMAXDCVG * CVLiv) / (KMDCVG + CVLiv);
pO >!
5
Liver
Dynamics for amount of TCE in liver (mg)
O	dt(ALiv) = (QLiv * (CArt - CVLiv)) + (QGut * (CVGut - CVLiv))
Cb	- RAMetLivl - RAMetLiv2 + kPV; # added PV dose
I ^
I >1
81
Rate of TCE metabolized to DCVG in kidney (mg) #
RAMetKid = (VMAXKidDCVG * CVKid) / (KMKidDCVG + CVKid);
Amount of TCE in kidney compartment (mg)
2!
O ^
H
HH ^3
O	#***	TCOH Sub-model
W <3'
O '
P
o
c
o
H
W
OCK)
"5	dt(AKid) = (QKid * (CArt - CVKid)) - RAMetKid;
H s
<3 		
State Variables with dynamics:
ABodTCOH
ALivTCOH
Input Variables:
none
Other State Variables and Global Parameters:
ABileTCOG
#	kEHR
#	VBodTCOH
#	PBodTCOH
#	VLiv
#	PLivTCOH
#	VMAXTCOH
#	KMTCOH
#	VMAXGluc
#	KMGluc
#	kMetTCOH - hepatic metabolism of TCOH (e.g., to DCA)
#	FracOther
#	FracTCA
#	StochTCOHTCE
#	StochTCOHGluc
#	FracLungSys
#	Temporary variables used:
#	QBod
#	QGutLiv
#	QCnow
#	kPOTCOH
#	RAMetLivl
#	RAMetLng
#	Temporary variables assigned:
#	CVBodTCOH
#	CVLivTCOH
#	CTCOH
#	RAMetTCOHTCA
#	RAMetTCOHGluc
#	RAMetTCOH
#	RARecircTCOG
#	Notes:
^* * * * B]_ood (venous=arterial) * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
#	Venous Concentrations (mg/L)
CVBodTCOH = ABodTCOH / VBodTCOH / PBodTCOH;
CVLivTCOH = ALivTCOH / VLiv / PLivTCOH;
CTCOH = (QBod * CVBodTCOH + QGutLiv * CVLivTCOH + klVTCOH)/QCnow;
£***•*• Body ********************************************************************
#	Amount of TCOH in body
dt(ABodTCOH) = QBod * (CTCOH - CVBodTCOH);
#	Rate of oxidation of TCOH to TCA (mg/hr)
RAMetTCOHTCA = (VMAXTCOH * CVLivTCOH) / (KMTCOH + CVLivTCOH);
#	Amount of glucuronidation to TCOG (mg/hr)
RAMetTCOHGluc = (VMAXGluc * CVLivTCOH) / (KMGluc + CVLivTCOH);
#	Amount of TCOH metabolized to other (e.g., DCA)
RAMetTCOH = kMetTCOH * ALivTCOH;
#	Amount of TCOH-Gluc recirculated (mg)
RARecircTCOG = kEHR * ABileTCOG;
#	Amount of TCOH in liver (mg)
-------
dt(ALivTCOH) = kPOTCOH + QGutLiv * (CTCOH - CVLivTCOH)
- RAMetTCOH - RAMetTCOHTCA - RAMetTCOHGluc
+ ((1.0 - FracOther - FracTCA) * StochTCOHTCE
(RAMetLivl + FracLungSys^RAMetLng))
+ (StochTCOHGluc * RARecircTCOG);
TCA Sub-model
^ #
O	# State Variables with dynamics:
§	#	APlasTCA
2	#	ABodTCA
S	#	ALivTCA
AUrnTCA
AUrnTCA sat
#	AUrnTCA collect
S"4	_
5^	# Input Variables:
^ #	TCAUrnSat
^ #	UrnMissing
#	Other State Variables and Global Parameters:
^ #	VPlas
^ #	MWTCA
#	kDissoc
^ SJ #	BMax
1 K?
^3 #	kMetTCA — hepatic metabolism of TCA (e.g., to DCA)
§ #	VBod
#	PBodTCA
2 #	PLivTCA
#	kUrnTCA
G #	FracTCA
#	StochTCATCE
#	StochTCATCOH
#	FracLungSys
pO >!
r,, O #	klVTCA
J Cl #	kPOTCA
~ o
Temporary variables used:
^	#	QBodPlas
QGutLivPlas
QCPlas
RAMetLivl
RAMetTCOHTCA
81
z; ^
c>
s-o'	#	RAMetLng
' ^	# Temporary variables assigned:
OS	#	CPlasTCA
HH ^3
O	#	CPLasTCAMole
W K'	#	a, b, c
O •
o
c
o
H
W
CPlasTCAFreeMole
CPlasTCAFree
APlasTCAFree
CPlasTCABnd
CBodTCAFree
CLivTCAFree
CBodTCA
#	CLivTCA
#	CVBodTCA
#	CVLivTCA
#	RUrnTCA
#	RAMetTCA
#	Notes:
#* * * * Plasma
#	Concentration of TCA in plasma (umoles/L)
CPlasTCA = (APlasTCACl.0e-15 ? 1.0e-15 : APlasTCA/VPlas);
#	Concentration of free TCA in plasma in (umoles/L)
CPlasTCAMole = (CPlasTCA / MWTCA) * 1000.0;
a = kDissoc+BMax-CPlasTCAMole;
b = 4.0*kDissoc*CPlasTCAMole;
c = (b < O.Ol^a^a ? b/2.0/a : sqrt(a*a+b)-a);
CPlasTCAFreeMole = 0.5*0;
#	Concentration of free TCA in plasma (mg/L)
CPlasTCAFree = (CPlasTCAFreeMole * MWTCA) / 1000.0;
APlasTCAFree = CPlasTCAFree * VPlas;
#	Concentration of bound TCA in plasma (mg/L)
CPlasTCABnd = (CPlasTCACCPlasTCAFree ? 0 : CPlasTCA-CPlasTCAFree)
#	Concentration in body and liver
CBodTCA = (ABodTCACO ? 0 : ABodTCA/VBod);
CLivTCA = (ALivTCACl.0e-15 ? 1.0e-15 : ALivTCA/VLiv);
#	Total concentration in venous plasma (free+bound)
CVBodTCAFree = (CBodTCA / PBodTCA);	# free in equilibrium
CVBodTCA = CPlasTCABnd + CVBodTCAFree;
CVLivTCAFree = (CLivTCA / PLivTCA);
CVLivTCA = CPlasTCABnd + CVLivTCAFree; # free in equilibrium
#	Rate of urinary excretion of TCA
RUrnTCA = kUrnTCA * APlasTCAFree;
#	Dynamics for amount of total (free+bound) TCA in plasma (mg)
dt(APlasTCA) = klVTCA + (QBodPlas+CVBodTCA) + (QGutLivPlas+CVLivTCA)
- (QCPlas * CPlasTCA) - RUrnTCA;
# Dynamics for amount of TCA in the body (mg)
dt(ABodTCA) = QBodPlas * (CPlasTCAFree - CVBodTCAFree);
Rate of metabolism of TCA
RAMetTCA = kMetTCA * ALivTCA;
Dynamics for amount of TCA in the liver (mg)
dt(ALivTCA) = kPOTCA + QGutLivPlas*(CPlasTCAFree - CVLivTCAFree)
- RAMetTCA + (FracTCA * StochTCATCE *
(RAMetLivl + FracLungSys*RAMetLng))
+ (StochTCATCOH * RAMetTCOHTCA);
Urine ********************************************************-*
Dynamics for amount of TCA in urine (mg)
dt(AUrnTCA) = RUrnTCA;
dt(AUrnTCA_sat) = TCAUrnSat*(1-UrnMissing)* RUrnTCA;
# Saturated, but not missing collection times
-------
dt(AUrnTCA_collect) = (1-TCAUrnSat)*(1-UrnMissing)* RUrnTCA;
# Not saturated and not missing collection times
TCOG Sub-model

State Variables with dynamics:
jj1 #	ABodTCOG
^	#	ALivTCOG
#	ABileTCOG
§	#	AUrnTCOG
2	#	AUrnTCOG sat
S	#	AUrnTCOG collect
O
a
2; ^
O ^
Input Variables:
TCOGUrnSat
#	UrnMissing
S"4
5^	# Other State Variables and Global Parameters:
^	#	VBodTCOH
2^	#	VLiv
•>S	#	PBodTCOG
ri	#	PLivTCOG
g	#	kUrnTCOG
^	#	kBile
^ Si	#	StochGlucTCOH
1
JsJ	# Temporary variables used:
§	#	QBod
O "5
&0	#	QGutLiv
QCnow
RAMetTCOHGluc
CVLivTCOG
CTCOG
G	#	RARecircTCOG
5^	# Temporary variables assigned:
.	.	#	CVBodTCOG
0	^	#
r,, O	#	RUrnTCOG
Ci	#	RBileTCOG
1	§	# Notes:
I ^	„******************************
u s
Sj	#**** Blood (venous=arterial) *
^	# Venous Concentrations (mg/L)
CVBodTCOG = ABodTCOG / VBodTCOH / PBodTCOG;
"5	CVLivTCOG = ALivTCOG / VLiv / PLivTCOG;
fv,	CTCOG = (QBod * CVBodTCOG + QGutLiv * CVLivTCOG)/QCnow;
O ^	#4-4-4-4- Body 4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-4-
HH ^3
^ O	# Amount of TCOG m body
fTj C}'	RUrnTCOG = kUrnTCOG * ABodTCOG;
dt(ABodTCOG) = QBod * (CTCOG - CVBodTCOG) - RUrnTCOG;
RUrnTCOGTCOH = RUrnTCOG* StochTCOHGluc; #(vrisk)
Amount of TCOG in liver (mg)
RBileTCOG = kBile * ALivTCOG;
dt(ALivTCOG) = QGutLiv * (CTCOG - CVLivTCOG)
P
O
c
o
H
+ (StochGlucTCOH * RAMetTCOHGluc)
RBileTCOG;
#	Amount of TCOH-Gluc excreted into bile (mg)
dt(ABileTCOG) = RBileTCOG - RARecircTCOG;
#+++4r urine *******************************************************************
#	Amount of TCOH-Gluc excreted in urine (mg)
dt(AUrnTCOG) = RUrnTCOG;
dt(AUrnTCOG_sat) = TCOGUrnSat*(1-UrnMissing)* RUrnTCOG;
#	Saturated, but not missing collection times
dt(AUrnTCOG_collect) = (1-TCOGUrnSat)*(1-UrnMissing)*RUrnTCOG;
#	Not saturated and not missing collection times
#***	DCVG Sub-model	***
#	State Variables with dynamics:
#	ADCVGmol
#	Input Variables:
#	none
#	Other State Variables and Global Parameters:
#	kDCVG
#	FracKidDCVC	# Fraction of kidney DCVG going to DCVC in first pass
#	VDCVG
#	Temporary variables used:
#	RAMetLiv2
#	RAMetKid
#	Temporary variables assigned:
#	RAMetDCVGmol
#	CDCVGmol
#	Notes:
#	Assume negligible GGT activity in liver as compared to kidney,
#	supported by in vitro data on GGT (even accounting for 5x
#	greater liver mass relative to kidney mass), as well as lack
#	of DCVC detected in blood.
#	"FracKidDCVC" Needed to account for "first pass" in
#	kidney (TCE->DCVG->DCVC without systemic circulation of DCVG).
#	Rate of metabolism of DCVG to DCVC
RAMetDCVGmol = kDCVG * ADCVGmol;
#	Dynamics for DCVG in blood
dt(ADCVGmol) = (RAMetLiv2 + RAMetKid*(1-FracKidDCVC)) / MWTCE
- RAMetDCVGmol;
#	Concentration of DCVG in blood (in mmoles/1)
CDCVGmol = ADCVGmol / VDCVG;
#***	DCVC Sub-model	***
#	State Variables with dynamics:
#	ADCVC
#	AUrnNDCVC
#	Input Variables:
-------
#	none
#	Other State Variables and Global Parameters:
#	MWDCVC
#	FracKidDCVC
#	StochDCVCTCE
#	kNAT
.	#	kKidBioact
#	StochN
^	# Temporary variables used:
O"4 #	RAMetDCVGmol
§	#	RAMetKid
2 # Temporary variables assigned:
S	#	RAUrnDCVC
. # Notes:
>1
^	#	Cannot detect DCVC in blood, so assume all is locally generated
^ #	and excreted or bioactivated in kidney.
#	Amount of DCVC in kidney (mg)
O	dt(ADCVC) = RAMetDCVGmol * MWDCVC
>2	+ RAMetKid * FracKidDCVC * StochDCVCTCE
ri	- ((kNAT + kKidBioact) * ADCVC);
^	# Rate of NAcDCVC excretion into urine (mg)
.	RAUrnDCVC = kNAT * ADCVC;
^3
Si	# Dynamics for amount of N Acetyl DCVC excreted (mg)
1 K?
to ^3	dt(AUrnNDCVC) = StochN * RAUrnDCVC;
^ §	RUrnNDCVC = StochN * RAUrnDCVC; #(vrisk)
>!
O
5L
^	^4-4-4-jr jvjas5 Balance for TCE * * * * * * * *
5^	# Total intake from inhalation (mg)
^ RInhDose = QM * CInh;
2 ^ dt(InhDose) = RInhDose;
pO >!
5
Total Mass Balance
Amount of TCE absorbed by non-inhalation routes (mg)
O	dt(AO) = RAO + klV + klA + kPV; #(vrisk)
^ Cb #	Total dose
I §	TotDose = InhDose + AO; #(vrisk)
#	Total in tissues
q	TotTissue = #(vrisk)
fiS	ARap + ASlw + AFat + AGut + ALiv + AKid + ABld + #(vrisk)
AInhResp + AResp + AExhResp; #(vrisk)
oS	# Total metabolized
2!
Cb	dt(AMetLng) = RAMetLng; #(vrisk)
O ^	dt(AMetLivl) = RAMetLivl; #(vrisk)
HH ^3
^ O	dt(AMetLiv2) = RAMetLiv2; #(vrisk)
fb'	dt(AMetKid) = RAMetKid; #(vrisk)
•	ATotMetLiv = AMetLivl + AMetLiv2; #(vrisk)
TotMetab = AMetLng + ATotMetLiv + AMetKid; #(vrisk)
AMetLivOther = AMetLivl * FracOther; #(vrisk)
AMetGSH = AMetLiv2 + AMetKid; #(vrisk)
# Amount of TCE excreted in feces (mg)
RAExc = kTD * ADuod; #(vrisk)
O
c
o
H
dt(AExc) = RAExc; #(vrisk)
#	Amount exhaled (mg)
RAExh = QM * CMixExh;
dt(AExh) = RAExh;
#	Mass balance
TCEDiff = TotDose - TotTissue - TotMetab; #(vrisk)
MassBalTCE = TCEDiff - AExc - AExh; #(vrisk)
Mass Balance for TCOH * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
#	Total production/intake of TCOH
dt(ARecircTCOG) = RARecircTCOG; #(vrisk)
dt(AOTCOH) = kPOTCOH + klVTCOH; #(vrisk)
TotTCOHIn = AOTCOH + ((1.0 - FracOther - FracTCA) * #(vrisk)
StochTCOHTCE * (AMetLivl + FracLungSys^AMetLng)) + #(vrisk)
(StochTCOHGluc * ARecircTCOG); #(vrisk)
TotTCOHDose = AOTCOH + ((1.0 - FracOther - FracTCA) * #(vrisk)
StochTCOHTCE * (AMetLivl + FracLungSys^AMetLng)); #(vrisk)
#	Total in tissues
TotTissueTCOH = ABodTCOH + ALivTCOH; #(vrisk)
#	Total metabolism of TCOH
dt(AMetTCOHTCA) = RAMetTCOHTCA; #(vrisk)
dt(AMetTCOHGluc) = RAMetTCOHGluc; #(vrisk)
dt(AMetTCOHOther) = RAMetTCOH; #(vrisk)
TotMetabTCOH = AMetTCOHTCA + AMetTCOHGluc + AMetTCOHOther; #(vrisk)
#	Mass balance
MassBalTCOH = TotTCOHIn - TotTissueTCOH - TotMetabTCOH; #(vrisk)
#**** Mass Balance for TCA ****************************************************
#	Total production/intake of TCA
dt(AOTCA) = kPOTCA + klVTCA; #(vrisk)
TotTCAIn = AOTCA + (FracTCA*StochTCATCE*(AMetLivl + #(vrisk)
FracLungSys*AMetLng)) + (StochTCATCOH*AMetTCOHTCA); #(vrisk)
#	Total in tissues
TotTissueTCA = APlasTCA + ABodTCA + ALivTCA; #(vrisk)
#	Total metabolism of TCA
dt(AMetTCA) = RAMetTCA; #(vrisk)
#	Mass balance
TCADiff = TotTCAIn - TotTissueTCA - AMetTCA; #(vrisk)
MassBalTCA = TCADiff - AUrnTCA; #(vrisk)
Mass Balance for TCOG ***************************************************
#	Total production of TCOG
TotTCOGIn = StochGlucTCOH * AMetTCOHGluc; #(vrisk)
#	Total in tissues
TotTissueTCOG = ABodTCOG + ALivTCOG + ABileTCOG; #(vrisk)
#	Mass balance
MassBalTCOG = TotTCOGIn - TotTissueTCOG - #(vrisk)
ARecircTCOG - AUrnTCOG; #(vrisk)
#	*•*•*•*• Mass Balance for DCVG ***************************************************
#	Total production of DCVG
dt(ADCVGIn) = (RAMetLiv2 + RAMetKid*(1-FracKidDCVC)) / MWTCE; #(vrisk)
#	Metabolism of DCVG
dt(AMetDCVG) = RAMetDCVGmol; #(vrisk)
-------
# Mass balance
MassBalDCVG = ADCVGIn - ADCVGmol - AMetDCVG; #(vrisk)
#**** Mass Balance for DCVC *****************************************
# Total production of DCVC
dt(ADCVCIn) = RAMetDCVGmol * MWDCVC #(vrisk)
+ RAMetKid * FracKidDCVC * StochDCVCTCE;#(vrisk)
U* # Bioactivation of DCVC
^	dt(ABioactDCVC) = (kKidBioact * ADCVC);#(vrisk)
O	# Mass balance
§	AUrnNDCVCequiv = AUrnNDCVC/StochN;
5
MassBalDCVC = ADCVCIn - ADCVC - ABioactDCVC - AUrnNDCVCequiv;#(vrisk)
>1
^	#***	Dynamic Outputs
^	# Amount exhaled during	exposure (mg)
dt(AExhExp) = (CInh	> 0 ? RAExh : 0);
>8
ri	#***	Dose Metrics
^	^ -± -k -k -k j\ucs in mg-hr/L unless otherwise noted
Si	#AUC of TCE in arterial blood
1 K?
^3	dt(AUCCBld) = CArt; # (vrisk)
[-0 §	#AUC of TCE in liver
N> "5
Gn	dt(AUCCLiv) = CLiv; # (vrisk)
#AUC of TCE in kidney
O
a
dt(AUCCKid) = CKid; #(vrisk)
^	#AUC of TCE in rapidly perfused
dt(AUCCRap) = CRap; #(vrisk)
#AUC of TCOH in blood
2	dt (AUCCTCOH) = CTCOH; # (vrisk)
pO Grj
#AUC of TCOH in body
^ O	dt(AUCCBodTCOH) = ABodTCOH / VBodTCOH; #(vrisk)
^ Cb	#AUC of free TCA in the plasma (mg/L * hr)
I §	dt(AUCPlasTCAFree) = CPlasTCAFree; #(vrisk)
#AUC of total TCA in plasma (mg/L * hr)
^ g"	dt(AUCPlasTCA) = CPlasTCA; #(vrisk)
#AUC of TCA in liver (mg/L * hr)
dt(AUCLivTCA) = CLivTCA; #(vrisk)
2; ^
r^i ^
s-o'	#AUC of total TCOH (free + gluc) in TCOH-equiv in blood (mg/L * hr)
^	dt(AUCTotCTCOH) = CTCOH + CTCOGTCOH; #(vrisk)
O	#AUC of DCVG in blood (mmol/L * hr) — NOTE moles, not mg
HH ^3
O	dt (AUCCDCVG) = CDCVGmol; # (vrisk)
w| >-¦
Q •	################ End of Dynamics
P
o
c
o
H
W
CalcOutputs{
ft**** static outputs for comparison to data
# TCE
RetDose = ((InhDose-AExhExp) > 0 ? (InhDose - AExhExp) : le-15);
CAlvPPM = (CAlv < 1.0e-15 ? 1.0e-15 : CAlv * (24450.0 / MWTCE));
CInhPPM = (ACh< 1.0e-15 ? 1.0e-15 : ACh/VCh*24450.0/MWTCE);
# CInhPPM Only used for CC inhalation
CArt = (CArt
<
1.0e-15 ?
1.0e-15
CArt)
CVen = (CVen
<
1.0e-15 ?
1.0e-15
CVen)
CBldMix = (CArt+CVen)/2;


CFat = (CFat
<
1.0e-15 ?
1.0e-15
CFat)
CGut = (CGut
<
1.0e-15 ?
1.0e-15
CGut)
CRap = (CRap
<
1.0e-15 ?
1.0e-15
CRap);
CSlw = (CSlw
<
1.0e-15 ?
1.0e-15
CSlw)
CHrt = CRap;




CKid = (CKid
<
1.0e-15 ?
1.0e-15
CKid)
CLiv = (CLiv
<
1.0e-15 ?
1.0e-15
CLiv);
CLung = CRap




CMus = (CSlw
<
1.0e-15 ?
1.0e-15
CSlw);
CSpl = CRap;




CBrn = CRap;




zAExh = (AExh <
: 1.0e-15
? 1.0e-15
: AExh)
zAExhpost =
((AExh - AExhExp) < 1
0e-15 ?
1.0e-15 : AExh - AExhExp);
CTCOH = (CTCOH
CBodTCOH =
CKidTCOH =
CLivTCOH =
CLungTCOH
(ABodTCOH <
CBodTCOH;
(ALivTCOH <
= CBodTCOH;
: CTCOH);
.0e-15
ABodTCOH/VBodTCOH);
ALivTCOH/VLiv);
CPlasTCA = (CPlasTCA < 1.0e-15 ? 1.0e-l
CBldTCA = CPlasTCA* TCAPlas;
CBodTCA = (CBodTCA < 1.0e-15 ? 1.0e-15
CLivTCA = (CLivTCA < 1.0e-15 ? 1.0e-15
CKidTCA = CBodTCA;
CLungTCA = CBodTCA;
zAUrnTCA = (AUrnTCA < 1.0e-15 ? 1.0e-15
zAUrnTCA_sat = (AUrnTCA_sat < 1.0e-15 ?
zAUrnTCA_collect = (AUrnTCA_collect < 1
AUrnTCA_collect);
# TCOG
zABileTCOG = (ABileTCOG < 1.0e-15 ? 1.0e-15
# Concentrations are in TCOH-equivalents
.0e-15 : CTCOG);
CPlasTCA);
CBodTCA);
CLivTCA);
: AUrnTCA);
1.0e-15 : AUrnTCA_sat);
0e-15 ? 1.0e-15 :
ABileTCOG);
CTCOG = (CTCOG < 1.0e-15 ? 1
CTCOGTCOH = (CTCOG < 1.0e-15 ? 1.0e-15
CBodTCOGTCOH = (ABodTCOG < 1.0e-15 ? 1.
StochTCOHGluc*ABodTCOG/VBodTCOH);
CKidTCOGTCOH = CBodTCOGTCOH;
CLivTCOGTCOH = (ALivTCOG < 1.0e-15 ? 1.
StochTCOHGluc+ALivTCOG/VLiv);
CLungTCOGTCOH = CBodTCOGTCOH;
AUrnTCOGTCOH = (AUrnTCOG < 1.0e-15 ? 1.
AUrnTCOGTCOH_sat = (AUrnTCOG_sat <1.0*
StochTCOHGluc*AUrnTCOG_sat);
AUrnTCOGTCOH_collect = (AUrnTCOG_collect < 1.0e-15
StochTCOHGluc*AUrnTCOG_collect);
: StochTCOHGluc*CTCOG);
0e-15 :
StochTCOHGluc*AUrnTCOG);
0e-15 :
-------
# Other
CDCVGmol = (CDCVGmol < 1.0e-15 ? 1.0e-15 : CDCVGmol);
CDCVGmolO = CDCVGmol; #(vl.2.3.2)
CDCVG_NDtmp = CDFNormal(3*(1-CDCVGmol/CDCVGmolLD));
# Assuming LD = 3*sigma blank, Normally distributed
CDCVG_ND = ( CDCVG_NDtmp < 1.0 ? ( CDCVG_NDtmp >= le-100 ? -
^ log(CDCVG_NDtmp) : -log(le-100)) : le-100 );
g-	#(vl.2.3.2)
^	zAUrnNDCVC =(AUrnNDCVC < 1.0e-15 ? 1.0e-15 : AUrnNDCVC);
O"4	AUrnTCTotMole = zAUrnTCA / MWTCA + AUrnTCOGTCOH / MWTCQH;
§	TotCTCOH = CTCOH + CTCOGTCOH;
2	TotCTCOHcomp = CTCOH + CTCOG; # ONLY FOR COMPARISON WITH HACK
S	ATCOG = ABodTCOG + ALivTCOG; # ONLY FOR COMPARISON WITH HACK
Misc
CVenMole = CVen / MWTCE;
CPlasTCAMole = (CPlasTCAMole < 1.0e-15 ? 1.0e-15 : CPlasTCAMol
S"4
^	CPlasTCAFreeMole = (CPlasTCAFreeMole < 1.0e-15 ? 1.0e-15 :
CPlasTCAFreeMole);
>8
#* * * * Additional Dose Metrics * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
s #
CS"
k?	TotTCAInBW = TotTCAIn/BW;#(vrisk)
£0 ^	# Scaled by BW^3/4
^ §	TotMetabBW34 = TotMetab/BW75;#(vrisk)
AMetGSHBW34 = AMetGSH/BW75;#(vrisk)
^	TotDoseBW34 = TotDose/BW75;#(vrisk)
AMetLivlBW34 = AMetLivl/BW75;#(vrisk)
^	TotOxMetabBW34 = (AMetLng+AMetLivl)/BW75;#(vrisk)
AMetLngBW34 = AMetLng/BW75; #(vrisk)
ABioactDCVCBW34 = ABioactDCVC/BW75;#(vrisk)
2 ^	AMetLivOtherBW34 = AMetLivOther/BW75; #(vrisk)
pO >!
O	# Scaled by tissue volume
^ Cb	AMetLivlLiv = AMetLivl/VLiv; #(vrisk)
I §	AMetLivOtherLiv = AMetLivOther/VLiv; #(vrisk)
lLj ^	AMetLngResp = AMetLng/VRespEfftmp; #(vrisk)
gf"	ABioactDCVCKid = ABioactDCVC/VKid;#(vrisk)
Fractional Volumes
2; ^
Cb	VFatCtmp = VFat/BW; # (vrisk)
O	VGutCtmp = VGut/BW; #(vrisk)
HH ^3
^ O	VLivCtmp = VLiv/BW; #(vrisk)
ffl.	VRapCtmp = VRap/BW; # (vrisk)
•	VRespLumCtmp = VRespLum/BW; # (vrisk)
VRespEffCtmp = VRespEfftmp/BW; #(vrisk)
VKidCtmp = VKid/BW; #(vrisk)
VBldCtmp = VBld/BW; #(vrisk)
VSlwCtmp = VSlw/BW; #(vrisk)
VPlasCtmp = VPlas/BW; #(vrisk)
O
c
o
H
VBodCtmp = VBod/BW; # (vrisk)
VBodTCOHCtmp = VBodTCOH/BW; #(vrisk)
-------
A.6. REFERENCES
1
This document is a draft for review purposes only and does not constitute Agency policy.
A-254 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
APPENDIX B
Systematic Review of Epidemiologic Studies
on Cancer and Trichloroethylene (TCE)
Exposure
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-cclv DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.l. INTRODUCTION
The epidemiologic evidence on trichloroethylene (TCE) is large with over 50 studies
identified as of June 2009 and includes occupational cohort studies, case-control studies, both
nested within a cohort (nested case-control study) or population based, and geographic based
studies. The analysis of epidemiologic studies on cancer and TCE serves to document essential
design features, exposure assessment approaches, statistical analyses, and potential sources of
confounding and bias. These studies are described below and reviewed according to criteria to
assess (1) their ability to inform weight of evidence evaluation for TCE exposure and a cancer
hazard and (2) their utility for examination using meta-analysis approaches. A secondary goal of
the qualitative review is to provide transparency on study strengths and weaknesses, providing
background for inclusion or exclusion of individual studies for quantitative treatment using meta-
analysis approaches. Individual study qualities are discussed according to specific criteria in
Section B.2.1 to B.2.8., and rationale for studies examined using meta-analysis approaches, the
systematic review, contained in Section B.2.9. Appendix C contains a full discussion of the
meta-analysis, its analytical methodology, including sensitivity analyses, and findings. This
analysis supports discussion of site-specific cancer observations in Chapter 4 where a
presentation may be found of study findings with assessment and discussion of observations
according to a study's weight of evidence and potential for alternative explanations, including
bias and confounding.
B.2. METHODOLOGIC REVIEW OF EPIDEMIOLOGIC STUDIES ON CANCER
AND TRICHLOROETHYLENE
Epidemiologic studies considered in this analysis assess the relationship between TCE
exposure and cancer, and are identified using several sources and their utility for characterizing
hazard and quantitative treatment is based on recommendations in National Research Council
(2006). A thorough search of the literature was carried out through December 2010 without
restriction on year of publication or language using the following approaches: a search of the
bibliographic database PubMed (http://www.ncbi.nlm.nih.gov/ pubmed/). TOXNET
(http://toxnet.nlm.nih.gov/) and EMBASE (http://www.embase.com/) using the terms
"trichloroethylene cancer epidemiology" and ancillary terms, "degreasers," "aircraft, aerospace
or aircraft maintenance workers," "metal workers," and "electronic workers," "trichloroethylene
and cohort," or, "trichloroethylene and case-control;" bibliographies of reviews of the TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-l DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
epidemiologic literature such as those of the Institute of Medicine (IOM, 2003), NRC (2006,
2009), and Scott and Chiu (2006) and, review of bibliographies of individual studies for relevant
studies not identified in the previous two approaches. The search strategy identified studies that
were either published or available on-line (in press). NRC (2006) noted "a full review of the
literature should identify all published studies in which there was a possibility that
trichloroethylene was investigated, even though results per se may not have been reported."
Additional steps of U.S. Environmental Protection Agency (U.S. EPA) staff to identify
studies not published in the literature included contacting primary investigators for case-control
studies of liver, kidney and lymphoma and occupation, asking for information on analyses
examining trichloroethylene uniquely and a review of Agency for Toxic Substances and Disease
Registry (ATSDR) or state health department community health surveys or statistics reviews for
information on TCE exposure and cancer incidence or mortality.
The breadth of the available epidemiologic database on trichloroethylene and cancer is
wide compared to that available for other chemicals assessed by U.S. EPA. However, few
studies were designed with the sole, or primary, objective of this report—to characterize the
magnitude of underlying association, if such exists, between TCE and cancer. Yet, many studies
in the body of evidence can provide information for identifying cancer hazard and dose-response
inferences. The weight a study contributes to the overall evidence on TCE and cancer depends
on a number of characteristics regarding the design, exposure assessment, and analysis
approaches. Epidemiologic studies were most informative for analysis if they approached ideals
described below, as evaluated using objective criteria for identifying a cancer hazard.
Seventy-five studies potentially relevant to health assessment of TCE exposure and
cancer and identified from the above comprehensive search are presented in Tables B-l, B-2, and
B-3. The studies vary widely in their approaches to study design, exposure assessment, and
statistical analysis; for these reasons, studies vary in their usefulness for identifying cancer
hazard. Studies are reviewed according to a set of a priori guidelines of their utility for assessing
TCE exposure and cancer according to the below criteria. Studies approaching criteria ideals
contribute greater weight in the weight of evidence analysis than studies with significant
deficiencies. These criteria are not meant to be used to "accept" or "reject" a particular study for
identifying cancer hazard. Rather, they are to be used as measurement tools for evaluating a
study's ability to identify TCE exposure and cancer outcomes. Studies suitable for meta-analysis
treatment are selected according to specific criteria identified in B.2.9.4. Individual study
descriptions and abstract sheets according to these criteria are found in Section B.3. Appendix C
describes meta-analysis methods and findings.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-2 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
>3
o
o
s
Co
I?
£u 2
Oq ^
53
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Aircraft and aerospace workers
Radican et al.
(2008), Blair
et al. (1998)
Civilian aircraft-maintenance
workers with at least 1 yr in
1952-1956 at Hill Air Force Base,
UT. Vital status (VS) to 1990
(Blair etal., 1998) or 2000 (Radican
et al., 2008); cancer incidence
1973-1990 (Blairetal., 1998).
14,457 (7,204 ever exposed to TCE).
Incidence (Blair et al., 1998) and
mortality rates (Blair et al., 1998;
Radican et al., 2008) of nonchemical
exposed subjects.
Most subjects (n = 10,718) with potential exposure to 1 to 25
solvents. Cumulative TCE assigned to individual subjects using
JEM. Exposure-response patterns assessed using cumulative
exposure, continuous or intermittent exposures, and peak exposure.
TCE replaced in 1968 with 1,1,1-trichloroethane and was
discontinued in 1978 in vapor degreasing activities. Median TCE
exposures were about 10 ppm for rag and bucket; 100-200 ppm for
vapor degreasing. Poisson regression analyses controlled for age,
calendar time, sex (Blair et al., 1998) or Cox proportional hazard
model for age and race.
Krishnadasan
et al. (2007)
Nested case-control study within a
cohort of 7,618 workers employed
for between 1950 and 1992, or who
had started employment before
1980 at Boeing/Rockwell/
Rocketdyne (Santa Susana Field
Laboratory, [the UCLA cohort of
(Morgensternetal., 1997)]. Cancer
incidence 1988-1999.
326 prostate cancer cases, 1,805
controls.
Response rate:
Cases, 69%; Controls, 60%.
JEM for TCE, hydrazine, PAHs, benzene, mineral oil constructed
from company records, walk-through, or interviews. Lifestyle
factors obtained from living subjects through mail and telephone
surveys. Conditional logistic regression controlled for cohort, age at
diagnosis, physical activity, SES and other occupational exposure
(benzene, PAHs, mineral oil, hydrazine).
TO
*
s
8
TO
>;
0
s
1
a.
to
o
VO
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Zhao et al.
Aerospace workers with >2 yrs of
6,044 (2,689 with high cumulative
JEM for TCE, hydrazine, PAHs, mineral oil, and benzene. IH
(2005); Ritz
employment at Rockwell/
exposure to TCE). Mortality rates of
ranked each job title ranked for presumptive TCE exposure as high
et al. (1999a)
Rocketdyne (now Boeing) and who
subjects in lowest TCE exposure
(3), medium (2), low (1), or no (0) exposure for 3 time periods

worked at Santa Susana Field
category.
(1951-1969, 1970-1979, 1980-1989). Cumulative TCE score: low

Laboratory, Ventura, CA, from
5,049 (2,227 with high cumulative
(up to 3), medium (over 3 up to 12), high (over 12) assigned to

1950-1993 (the UCLA cohort of
exposure to TCE). Incidence rates of
individual subjects using JEM. Cox proportional hazard, controlled

(Morgensternetal., 1997)]).
subjects in lowest TCE exposure
for time, since 1st employment, SES, age at diagnosis and

Cancer mortality as of December
category.
hydrazine.

31,2001. Cancer incidence 1988-



2000 for subjects alive as of 1988.


-------

§•
to
s
a
to
s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
o
o
s
Co
I?
Oq ^
53
Co
S5
<§>
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
to
*
s
8
TO
0
s
1
a.
to
o
VO
Boice et al.
(2006b)
Aerospace workers with >6 months
employment at Rockwell/
Rocketdyne (Santa Susana Field
Laboratory and nearby facilities)
from 1948-1999 (IEI cohort, IEI
[2005]). VS to 1999.
41,351, 1,642 male hourly test stand
mechanics (1,111 with potential TCE
exposure).
Mortality rates of United States
population and California
population. Internal referent groups
including male hourly
nonadministrative Rocketdyne
workers; male hourly,
nonadministrative SSFL workers;
and test stand mechanics with no
potential exposure to TCE.
Potential TCE exposure assigned to test stands workers only whose
tasks included the cleaning or flushing of rocket engines (engine
flush) (n = 639) or for general utility cleaning (n = 472); potential
for exposure to large quantities of TCE was much greater during
engine flush than when TCE used as a utility solvent. JEM for TCE
and hydrazine without semiquantitative intensity estimates.
Exposure to other solvents not evaluated due to low potential for
confounding (few exposed, low exposure intensity, or not
carcinogenic). Exposure metrics included employment duration,
employment decade, years worked with potential TCE exposure, and
years worked with potential TCE exposure via engine cleaning,
weighted by number of tests. Lifetable (SMR); Cox proportional
hazard controlling for birth year, hire year, and hydrazine exposure.
Boice et al.
(1999)
Aircraft-manufacturing workers
with at least 1 yr >1960 at
Lockheed Martin (Burbank, CA).
VS to 1996.
77,965 (2,267 with potential routine
TCE exposures and 3,016 with
routine or intermittent TCE
exposure).
Mortality rates of United States
population (routine TCE exposed
subjects) and non-exposed internal
referents (routine and intermittent
TCE exposed subjects).
12% with potential routine mixed solvent exposure and 30% with
route or intermittent solvent exposure. JEM for potential TCE
exposure on (1) routine basis or (2) intermittent or routine basis
without semiquantitative intensity estimate. Exposure-response
patterns assessed by any exposure or duration of exposure and
internal control group. Vapor degreasing with TCE before 1966 and
PCE, afterwards. Lifetable analyses (SMR); Poisson regression
analysis adjusting for birth date, starting employment date, finishing
employment date, sex and race.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Morgan et al.
(1998)
Aerospace workers with >6 months
1950-1985 at Hughes (Tucson,
AZ). VS to 1993.
20,508 (4,733 with TCE exposures).
Mortality rates of United States
population for overall TCE exposure;
mortality rates of all-other cohort
subjects (internal referents) for
exposure-response analyses.
TCE exposure intensity assigned using JEM. Exposure-response
patterns assessed using cumulative exposure (low versus high) and
job with highest TCE exposure rating (peak, medium/high exposure
versus no/low exposure). "High exposure" job classification defined
as >50 ppm. Vapor degreasing with TCE 1952-1977, but limited IH
data <1975. Limited IH data before 1975 and medium/ low rankings
likely misclassified given temporal changes in exposure intensity not
fully considered (NRC, 2006).
Costa et al.
(1989)
Aircraft manufacturing workers
employed 1954-1981at plant in
Italy. VS to 1981.
8,626 subjects
Mortality rates of the Italian
population.
No exposure assessment to TCE and job titles grouped into one of
four categories: blue- and white-collar workers, technical staff, and
administrative clerks. Lifetable (SMR).
Garabrant et
al. (1988)
Aircraft manufacturing workers >4
yrs employment and who had
worked at least 1 d at San Diego,
CA, plant 1958-1982. VS to 1982.
14,067
Mortality rates of United States
population.
TCE exposure assessment for 70 of 14,067 subjects; 14 cases of
esophageal cancer and 56 matched controls. For these 70 subjects,
company work records identified 37% with job title with potential
TCE exposure without quantitative estimates. Lifetable (SMR).
Cohorts Identified From Biological Monitoring (U-TCA)
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
VO
-------

§•
to
s
a
to
s
Co
a,
>r
>S
>S
to'
*
S
s
TO
>1
0
s
1
a.
to
o
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Hansen et al.
(2001)
Workers biological monitored using
U-TCA and air-TCE, 1947-1989.
Cancer incidence from 1964-1996.
803 total
Cancer incidence rates of the Danish
population.
712 with U-TCA, 89 with air-TCE measurement records, 2 with
records of both types. U-TCA from 1947-1989; air TCE
measurements from 1974. Historic median exposures estimated
from the U-TCA concentrations were: 9 ppmfor 1947 to 1964,
5 ppm for 1965 to 1973, 4 ppm for 1974 to 1979, and 0.7 ppm for
1980 to 1989. Air TCE measurements from 1974 onward were
19 ppm (mean) and 5 ppm (median). Overall, median TCE
exposure to cohort as extrapolated from air TCE and U-TCA
measurements was 4 ppm (arithmetic mean, 12 ppm). Exposure
metrics: year 1st employed, employment duration, mean exposure,
cumulative exposure. Exposure metrics: employment duration,
average TCE intensity, cumulative TCE, period 1st employment.
Lifetable analysis (SIR).
Anttila et al.
(1995)
Workers biological monitored using
U-TCA, 1965-1982. VS
1965-1991 and cancer incidence
1967-1992.
3,974 total (3,089 with U-TCA
measurements]).
Mortality and cancer incidence rates
of the Finnish population.
Median U-TCA, 63 |imol/L for females and 48 |imol/L for males;
mean U-TCA was 100 |imol/L. Average 2.5 U-TCA measurements
per individual. Using the Ikeda et al. (1972) relationship for TCE
exposure to U-TCA, TCE exposures were roughly 4 ppm (median)
and 6 ppm (mean). Exposure metrics: years since 1st measurement.
Lifetable analysis (SMR, SIR).
Axelson et al.
(1994)
Workers biological monitored using
U-TCA, 1955-1975. VS to 1986
and cancer incidence 1958-1987.
1,4,21 males
Mortality and cancer incidence rates
of Swedish male population.
Biological monitoring for U-TCA from 1955 and 1975. Roughly %
of cohort had U-TCA concentrations equivalent to <20 ppm TCE.
Exposure metrics: duration exposure, mean U-TCA. Lifetable
analysis (SMR, SIR).
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
S
s
TO
>1
0
s
1
a.
to
o
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Other Cohorts
Clapp and
Hoffman
(2008)
Deaths between 1969-2001 among
employees >5 yrs employment
duration at an IBM facility
(Endicott, NY).
360 deaths
Proportion of deaths among New
York residents during 1979 to 1998.
No exposure assessment to TCE. PMR analysis.
Sung et al.
(2007; 2008)
Female workers 1st employed
1973-1997 at an electronics (RCA)
manufacturing factory (Taoyuan,
Taiwan). Cancer incidence 1979-
2001 (Sung etal., 2007).
Childhood leukemia 1979-2001
among first born of female subjects
in (Sung et al., 2007; 2008)
63,982 females and 40,647 females
with 1st live born offspring.
Cancer incidence rates of Taiwan
population (Sung et al., 2007).
Childhood leukemia incidence rates
of first born live births of Taiwan
population (Sung et al.,
http://birenheide.com/sra/2011 AM/pr
ogram/singlesession.php3?sessid=M
3-J2007).
No exposure assessment. Chlorinated solvents including TCE and
PCE found in soil and groundwater at factory site. Company records
indicated TCE not used 1975-1991 and PCE 1975-1991 and PCE
after 1981. No information for other time periods. Exposure-
response using employment duration. Lifetable analysis (SMR, SIR)
(Chang et al., 2003; Chang et al., 2005; Sung et al., 2007) orPoisson
regression adjusting for maternal age, education, sex, and birth year
(Sung et al., 2008).
Chang et al.
(2003; 2005)
Male and female workers employed
1978-1997 at electronics factory as
studied by Sung et al. (2007). VS
from 1985-1997 and cancer
incidence 1979-1997.
86,868 total
Incidence (Chang et al., 2005) or
mortality (Chang et al., 2003) rates
Taiwan population.
ATSDR
(2004)
Workers 1952-1980 at the View-
Master factory (Beaverton, OR).
616 deaths 1989-2001
Proportion of deaths between
1989-2001 in Oregon population.
No exposure information on individual subjects. TCE and other
VOCs detected in well water at the time of the plant closure in 1998
were TCE, 1,220-1,670 |ig/L: 1,1-DCE, up to 33 |ig/L: and, PCE up
to 56 |ig/L. PMR analysis.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Raaschou-
Nielsen et al.
(2003)
Blue-collar workers employed
>1968 at 347 Danish TCE-using
companies. Cancer incidence
through 1997.
40,049 total (14,360 with presumably
higher level exposure to TCE).
Cancer incidence rates of the Danish
population.
Employers had documented TCE usage but no information on
individual subjects. Blue-collar versus white-collar workers and
companies with <200 workers were variables identified as increasing
the likelihood for TCE exposure. Subjects from iron and metal,
electronics, painting, printing, chemical, and dry cleaning industries.
Median exposures to trichloroethylene were 40-60 ppm for the years
before 1970, 10-20 ppm for 1970 to 1979, and approximately 4 ppm
for 1980 to 1989. Exposure metrics: employment duration, year 1st
employed, and # employees in company. Lifetable (SIR).
Ritz (1999a)
Male uranium-processing plant
workers >3 months employment
1951-1972 at DOE facility
(Fernald, OH). VS 1951-1989,
cancer.
3,814 white males monitored for
radiation (2,971 with potential TCE
exposure).
Mortality rates of the United States
population; Non-TCE exposed
internal controls for TCE exposure-
response analyses.
JEM for TCE, cutting fluids, kerosene, and radiation generated by
employees and industrial hygienists. Subjects assigned potential
TCE according to intensity: light (2,792 subjects), moderate
(179 subjects), heavy (no subjects). Lifetable (SMR) and
conditional logistic regression adjusted for pay status, date first hire,
radiation.
Henschler et
al. (1995)
Male workers > 1 yr 1956-1975 at
cardboard factory (Arnsberg region,
Germany). VS to 1992.
169 exposed; 190 unexposed
Mortality rates from German
Democratic Republic (broad
categories) or renal cell carcinoma
incidence rates from Danish
population, German Democratic, or
non-TCE exposed subjects.
Walk-through surveys and employee interviews used to identify
work areas with TCE exposure. TCE exposure assigned to renal
cancer cases using workman's compensation files. Lifetable (SMR,
SIR) or Mantel-Haenszel.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Greenland et
al. (1994)
Cancer deaths, 1969-1984, among
pensioned workers employed
<1984 at GE transformer
manufacturing plant (Pittsfield,
MA), and who had job history
record; controls were noncancer
deaths among pensioned workers.
512 cases, 1,202 controls.
Response rate:
Cases, 69%;
Controls, 60%.
Industrial hygienist assessment from interviews and position
descriptions. TCE (no/any exposure) assigned to individual subjects
using JEM. Logistic regression.
Sinks et al.
(1992)
Workers employed 1957-1980 at a
paperboard container
manufacturing and printing plant
(Newnan, GA). VS to 1988.
Kidney and bladder cancer
incidence through 1990.
2,050 total
Mortality rates of the United States
population, bladder and kidney
cancer incidence rates from the
Atlanta-SEER registry for the years
1973-1977.
No exposure assessment to TCE; analyses of all plant employees
including white- and blue-collar employees. Assignment of work
department in case-control study based upon work history; Material
Safety Data Sheets identified chemical usage by department.
Lifetable (SMR, SIR) or conditional logistic regression adjusted for
hire date and age at hire, and using 5- and 10-year lagged
employment duration.
Blair et al.
(1989)
Workers employed 1942- 1970 in
U.S. Coast. VS to 1980.
3,781 males of whom 1,767 were
marine inspectors (48%).
Mortality rates of the United States
population. Mortality rates of marine
inspectors also compared to that of
noninspectors.
No exposure assessment to TCE. Marine inspectors worked in
confined spaces and had exposure potential to multiple chemicals.
TCE was identified as one of 10 potential chemical exposures.
Lifetable (SMR) and directly adjusted relative risks.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
S
s
TO
>1
0
s
1
a.
to
o
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE exposure
(continued)
Reference
Description
Study group (N)
Comparison group (N)
Exposure assessment and other information
Shannon et
al. (1988)
Workers employed >6 months at
GE lamp manufacturing plant,
1960-1975. Cancer incidence from
1964-1982.
1,870 males and females, 249 (13%)
in coiling and wire-drawing area.
Cancer incidence rates from Ontario
Cancer Registry.
No exposure assessment to TCE. Workers in coiling and wire
drawing (CWD) had potential exposure to many chemicals including
metals and solvents. A 1955-dated engineering instruction sheet
identified trichloroethylene used as degreasing solvent in CWD.
Lifetable (SMR).
Shindell and
Ulrich (1985)
Workers employed >3 months at a
TCE manufacturing plant 1957-
1983. VS to 1983.
2,646 males and females
Mortality rates of the United States
population.
No exposure assessment to TCE; job titles categorized as either
white- or blue-collar. Lifetable analysis (SMR).
Wilcosky et
al. (1984)
Respiratory, stomach, prostate,
lymphosarcoma, and lymphatic
leukemia cancer deaths 1964-1972
among 6,678 active and retired
production workers at a rubber
plant (Akron, OH); controls were a
20% age-stratified random sample
of the cohort.
183 cases (101 respiratory,
33 prostate, 30 stomach, 9
lymphosarcoma and 10 lymphatic
leukemia cancer deaths).
JEM without quantitative intensity estimates for 20 exposures
including TCE. Exposure metric: ever held job with potential TCE
exposure.
DCE = dichloroethylene, DOE = U.S. Department of Energy, IEI = International Epidemiology Institute, JEM = job-exposure matrix, NRC = National Research
Council, PCE = perchloroethylene, PMR = proportionate mortality ratio, SIR = standardized incidence ratio, SMR = standardized mortality ratio, SSFL = Santa
Susanna Field Laboratory, U-TCA = urinary trichloroacetic acid, UCLA = University of California, Los Angeles, VOCs = volatile organic compounds, VS =
vital status.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-2. Case-control epidemiologic studies examining cancer and TCE exposure
Reference
Population
Study group (N)
Comparison group (N)
Response rates
Exposure assessment and other information
Bladder
Pesch et al.
(2000a)
Histologically confirmed
urothelial cancer (bladder, ureter,
renal pelvis) cases from German
hospitals (5 regions) in
1991-1995; controls randomly
selected from residency registries
matched on region, sex, and age.
1,035 cases
4,298 controls
Cases, 84%; Controls, 71%
Occupational history using job title or self-reported exposure. JEM and
JTEM to assign exposure potential to metals and solvents (chlorinated
solvents, TCE, PCE). Lifetime exposure to TCE exposure examined as 30th,
60th, and 90th percentiles (medium, high, and substantial) of exposed control
exposure index. Duration used to examine occupational title and job task
duties and defined as 30th, 60th, and 90th percentiles (medium, long, and
very long) of exposed control durations.
Logistic regression with covariates for age, study center, and smoking.
Siemiatycki
(1994);
Siemiatycki
(1991)
Male bladder cancer cases, age
35-75 yrs, diagnosed in 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and random digit
dialing (RDD).
484 cases
533 population controls;
740 other cancer controls
Cases, 78%; Controls, 72%
JEM to assign 294 exposures including TCE on semiquantitative scales
categorized as any or substantial exposure. Other exposure metrics included
exposure duration in occupation or job title.
Logistic regression adjusted for age, ethnic origin, socioeconomic status,
smoking, coffee consumption, and respondent status [occupation or job title]
or Mantel-Haenszel stratified on age, income, index for cigarette smoking,
coffee consumption, and respondent status (TCE).
Brain
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
VO
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
De Roos et al.
(2001a);
Olshan et al.
(1999)
Neuroblastoma cases in children
of <19 yrs selected from
Children's Cancer Group and
Pediatric Oncology Group with
diagnosis in 1992-1994;
population controls (RDD)
matched to control on birth date.
504 cases
504 controls
Cases, 73%; Controls, 74%
Telephone interview with parent using questionnaire to assess parental
occupation and self-reported exposure history and judgment-based attribution
of exposure to chemical classes (halogenated solvents) and specific solvents
(TCE). Exposure metric was any potential exposure.
Logistic regression with covariate for child's age and material race, age, and
education.
Heineman et
al. (1994)
White, male cases, age >30 yrs,
identified from death certificates
in 1978-1981; controls identified
from death certificates and
matched for age, year of death and
study area.
300 cases
386 controls
Cases, 74%; Controls, 63%
In-person interview with next-of-kin; questionnaire assessing lifetime
occupational history using job title and JEM of Gomez et al. (1994).
Cumulative exposure metric (low, medium or and high) based on weighted
probability and duration.
Logistic regression with covariates for age and study area.
Table B-2. Case-control epidemiologic studies examining cancer and TCE exposure (continued)
Reference
Population
Study group (N)
Comparison group (N)
Response rates
Exposure assessment and other information
Colon and Rectum
Goldberg et al.
(2001);
Siemiatycki
(1991)
Male colon cancer cases, 35-75
yrs, from 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and random digit
dialing (RDD).
497 cases
533 population controls and
740 cancer controls
Cases, 82%; Controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales);
potential TCE exposure defined as any or substantial exposure.
Logistic regression adjusted for age, ethnic origin, birthplace, education,
income, parent's occupation, smoking, alcohol consumption, tea
consumption, respondent status, heating source socioeconomic status,
smoking, coffee consumption, and respondent status [occupation, some
chemical agents] or Mantel-Haenszel stratified on age, income, index for
cigarette smoking, coffee consumption, and respondent status [TCE],
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Dumas et al.
(2000);
Simeiatycki
(1991)
Male rectal cancer cases, age
35-75 yrs, diagnosed in 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
292 cases
533 population controls and
740 other cancer controls
Cases, 78%; Controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales);
potential TCE exposure defined as any or substantial exposure.
Logistic regression adjusted for age, education, respondent status, cigarette
smoking, beer consumption and body mass index [TCE] or Mantel-Haenszel
stratified on age, income, index for cigarette smoking, coffee consumption,
ethnic origin, and beer consumption [TCE],
Fredriksson et
al. (1989)
Colon cancer cases aged 30-75
yrs identified through the Swedish
Cancer Registry among patients
diagnosed in 1980-1983;
population-based controls were
frequency-matched on age and sex
and were randomly selected from
a population register.
329 cases
658 controls
Not available
Mailed questionnaire assessing occupational history with telephone interview
follow-up. Self-reported exposure to TCE defined as any exposure.
Mantel-Haenszel stratified on age, sex, and physical activity.
Esophagus
Parent et al.
(2000a),
Siemiatycki
(1991)
Male esophageal cancer cases,
35-75 yrs, diagnosed in 19 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
292 cases
533 population controls;
740 subjects with other
cancers
Cases, 78%; controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales);
potential TCE exposure defined as any or substantial exposure.
Logistic regression adjusted for age, education, respondent status, cigarette
smoking, beer consumption and body mass index [solvents] or Mantel-
Haenszel stratified on age, income, index for cigarette smoking, coffee
consumption, ethnic origin, and beer consumption [TCE],
Lymphoma
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Purdue et al.
(2011); Gold
et al.
Cases aged 20-74 with
histologically-confirmed NHL(B-
cell diffuse and follicular, T-cell,
lymphoreticular) without HIV in
1998-2000 and identified from
four SEER areas (Los Angelos
County and Detrioit metropolitan
area, random sample;
SeattlePuget Sound and Iowa, all
consecutive cases); population
controls aged 20-74 with no
previous diagnosis of HIV
infection or NHL,identified
through (1) if >65 yrs of age,
RDD, or (2) if >65 years,
identified from Medicare
eligibility files and stratified on
geographic area, age, and race,
1,321 cases
1,057 controls
Cases , 76% ; Controls,
78%
In-person interview using questionnaire or computer-assisted personal
interviewquestionnaire specific for jobs held for >1 yr since the age of 16
years, hobbies, and medical and family history. For occupational history, 32
job- or industry-specific interview modules asked for detailed information on
individual jobs and focused on solvents exposure, including TCE, assessment
by expert industrial hygienist blinded to case and control status by levels of
probability, frequency, and intensity. Exposure metric of overall exposure,
average weekly exposure, years exposed, average exposure intensity, and
cumulative exposure.
Logistic regression adjusted for sex, age, race, education and SEER site.
Gold et al.
Cases aged 35-74 with
histologically-confirmed multiple
myeloma in 2000-2002 and
identified from Seer areas
(Detrioit, Seattle-Puget Sound);
population controls
181 cases
481 controls
Cases, 71%; Controls, 52%
In-person interview using computer-assisted personal interview questionnaire
for jobs held >1 year since 1941 (cases) or 1946 (controls) and since age 18
years. For occupational history, 20 occupations, job- or industry-specific
interview modules asked for detailed information on individual jobs held at
least 2 years and focused on solvents exposure, including TCE, assessment
by expert industrial hygienist blinded to case and control status by levels of
probability, duration and cumulative exposure.
Logistic regression adjusted for sex, age, race, education and SEER site.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Cocco et al.
Cases aged >17 years with
2,348 cases
In-person interviews using same structured questionnaire translated to the
(2010)
lymphoma (B-cell, T-cell, CLL,
2,462 controls
local language for information on sociodemographic factors, lifestyle, health

multiple myeloma, Hodgkin) in
Cases, 88%; controls, 81%
history and all full-time job held >1 year. Assessment by industrial

1998-2004 and residents of
hospital and 52%
hygienists in each participating center to 43 agents, including TCE, by

referral areas from 7 European
population
confidence, exposure intesntiy, and exposure frequency. Exposure metric of

countries (Czech Republic,

overall TCE exposure and cumulative TCE exposure for subjects assessed

Finland, France, Germany,

with high degree of confidence (defined as low, medium, and high).

Ireland, Italy, and Spain); hospital

Logistic regression adjusted for age, gender, education and study center.

(4 participating countries) or



population controls (all others);



controls from (1) Germany and



Italy selected by RDD from



general population and matched



(individually in German and



group-based in Italy) to cases by



sex, age and residence area, and,



(2) for all other countries, matched



hospital controls with diagnoses



other than cancer, infectious



diseases and immundeficient



diseases.


German
NHL and Hodgkin's disease cases
710 cases
n-person interview using questionnaire assessing personal characteristics,
centers:
aged 18-80 yrs identified through
710 controls
ifestyle, medical history, UV light exposure, and occupational history of all
Seidler et al.
all hospitals and ambulatory
Cases, 87%; Controls, 44%
obs held for >1 yr. Exposure of a prior interest were assessed using job task-
(2007); Mester
physicians in six regions of

ipecific supplementary questionnaires. JEM used to assign cumulative
et al. (2006);
Germany between 1998 and 2003;

luantitative TCE exposure metric, categorized according to the distribution
Becker et al.
population controls were

imong the control persons (50th and 90th percentile of the exposed controls).
(2004)
identified from population

Conditional logistic regression adjusted for age, sex, region, smoking and

registers and matched on age, sex,

ilcohol consumption.

and region.


-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Wang etal.
(2009)
Cases among females aged 21 and
84 yrs with NHL in 1996-2000
and identified from Connecticut
Cancer Registry; population-based
female controls (1) if <65 yrs of
age, having Connecticut address
stratified by 5-yr age groups
identified from random digit
dialing or (2) >65 yrs of age, by
random selection from Centers for
Medicare and Medicaid Service
files.
601 cases
717 controls
Cases, 72%; Controls, 69%
(<65 yrs), 47% (>65 yrs)
In-person interview with using questionnaire assessment specific jobs held
for >1 yr. Intensity and probability of exposure to broad category of organic
solvents and to individual solvents, including TCE, estimated using JEM
(Dosemeci et al., 1999; Gomez et al., 1994) and assigned blinded. Exposure
metric of any exposure, exposure intensity (low, medium/high), and exposure
probability (low, medium/high).
Logistic regression adjusted for age, family history of hematopoietic cancer,
alcohol consumption and race.
Costantini et
al. (2008);
Miligi et al.
(2006)
Cases aged 20-74 with NHL,
including CLL, all forms of
leukemia, or multiple myeloma
(MM) in 1991-1993 and
identified through surveys of
hospital and pathology
departments in study areas and in
specialized hematology centers in
8 areas in Italy; population-based
controls stratified by 5-yr age
groups and by sex selected
through random sampling of
demographic or of National Health
Service files.
1,428 NHL + CLL, 586
Leukemia,
263, MM
1,278 controls (leukemia
analysis)
1,100 controls (MM
analysis)
Cases, 83%; Controls, 73%
In-person interview primarily at interviewee's home (not blinded) using
questionnaire assessing specific jobs, extra occupational exposure to solvents
and pesticides, residential history, and medical history. Occupational
exposure assessed by job-specific or industry-specific questionnaires. JEM
used to assign TCE exposure and assessed using intensity (2 categories) and
exposure duration (2 categories). All NHL diagnoses and 20% sample of all
cases confirmed by panel of 3 pathologists.
Logistic regression with covariates for sex, age, region, and education.
Logistic regression for specific NHL included an additional covariate for
smoking.




-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Persson and
Fredriksson
(1999);
Combined
analysis of
NHL cases in
Persson et al.
(1993);
Persson et al.
(1989)
Histologically confirmed cases of
B-cell NHL, age 20-79 yrs,
identified in two hospitals in
Sweden: Oreboro in 1964-1986
(Persson et al., 1989) and in
Linkoping between 1975-1984
(Persson et al., 1993); controls
were identified from previous
studies and were randomly
selected from population registers.
199 NHL cases,
479 controls
Cases, 96% (Oreboro),
90% (Linkoping);
controls, not reported
Mailed questionnaire to assess self reported occupational exposures to TCE
and other solvents.
Mantel-Haenszel chi-square.
Nordstrom et
al. (1998)
Histologically-confirmed cases in
males of hairy-cell leukemia
reported to Swedish Cancer
Registry in 1987-1992 (includes
one case latter identified with an
incorrect diagnosis date);
population-based controls
identified from the National
Population Registry and matched
(1:4 ratio) to cases for age and
county.
111 cases
400 controls
Cases, 91%; Controls, 83%
Mailed questionnaire to assess self reported working history, specific
exposure, and leisure time activities.
Univariate analysis for chemical-specific exposures (any TCE exposure).
Fritschi and
Siemiatycki
(1996a);
Siemiatycki
(1991)
Male NHL cases, age 35-75 yrs,
diagnosed in 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
215 cases
533 population controls
(Group 1) and
1,900 subjects with other
cancers (Group 2)
Cases, 83%; Controls, 71%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales).
Exposure metric defined as any or substantial exposure.
Logistic regression adjusted for age, proxy status, income, and ethnicity
[solvents] or Mantel-Haenszel stratified by age, body mass index, and
cigarette smoking [TCE],
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Hardell et al.
(1994; 1981)
Histologically-confirmed cases of
NHL in males, age 25-85 yrs,
admitted to Swedish (Umea)
hospital between 1974-1978;
living controls (1:2 ratio) from the
National Population Register,
matched to living cases on sex,
age, and place of residence;
deceased controls from the
National Registry for Causes of
Death, matched (1:2 ratio) to dead
cases on sex, age, place of
residence, and year of death.
105 cases
335 controls
Response rate not available
Self-administered questionnaire assessing self-reported solvent exposure;
phone follow-up with subject, if necessary.
Mantel-Haenszel chi-square.
Persson et al.
(1993);
Persson et al.
(1989)
Histologically confirmed cases of
Hodgkin's disease, age 20-80 yrs,
identified in two hospitals in
Sweden: Oreboro in 1964-1986
(Persson et al., 1989) and in
Linkoping between 1975-1984
(Persson et al., 1993); controls
randomly selected from
population registers.
54 cases (1989 study);
31 cases (1993 study)
275 controls (1989 study);
204 controls (1993 study)
Response rate not available
Mailed questionnaire to assess self reported occupational exposures to TCE
and other solvents.
Logistic regression with adjustment for age and other exposure; unadjusted
Mantel-Haenszel chi-square.
Childhood Leukemia
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Shu et al.
(2004; 1999)
Childhood leukemia cases, <15
yrs, diagnosed between 1989 and
1993 by a Children's Cancer
Group member or affiliated
institute; population controls
(random digit dialing), matched
for age, race, and telephone area
code and exchange.
1,842 cases
1,986 controls
Cases, 92%; controls, 77%
Telephone interview with mother, and whenever available, fathers using
questionnaire to assess occupation using job-industry title and self-reported
exposure history. Questionnaire included questions specific for solvent,
degreaser or cleaning agent exposures.
Logistic regression with adjustment for maternal or paternal education, race,
and family income. Analyses of paternal exposure also included age and sex
of the index child.
Costas et al.
(2002);
MDPH
(1997a)
Childhood leukemia (<19 yrs age)
diagnosed in 1969-1989 and who
were resident of Woburn. MA;
controls randomly selected from
Woburn public School records,
matched for age.
19 cases
37 controls
Cases, 91%; Controls, not
available
Questionnaire administered to parents separately assessing demographic and
lifestyle characteristics, medical history information, environmental and
occupational exposure and use of public drinking water in the home.
Hydraulic mixing model used to infer delivery of TCE and other solvents
water to residence.
Logistic regression with composite covariate, a weighted variable of
individual covariates.
McKinney et
al. (1991)
Incident childhood leukemia and
non-Hodgkin's lymphoma cases,
1974-1988, ages not identified,
from three geographical areas in
England; controls randomly
selected from children of residents
in the three areas and matched for
sex and birth health district.
109 cases
206 controls
Cases, 72%; Controls, 77%
In-person interview with questionnaire with mother to assess maternal
occupational exposure history, and with father and mother, as surrogate, to
assess paternal occupational exposure history. No information provided in
paper whether interviewer was blinded as to case and control status.
Matched pair design using logistic regression for univariate and multivariate
analysis.
Lowengart et
al. (1987)
Childhood leukemia cases aged
<10 yrs and identified from the
Los Angeles (CA) Cancer
Surveillance Program in
1980-1984; controls selected from
RDD or from friends of cases and
matched on age, sex, and race.
123 cases
123 controls
Cases, 79%; Controls,
not available
Telephone interview with questionnaire to assess parental occupational and
self-reported exposure history.
Matched (discordant) pair analysis.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Melanoma
Fritschi and
Siemiatycki
(1996b);
Siemiatycki
(1991)
Male melanoma cases, age 35-75
yrs, diagnosed in 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
103 cases
533 population controls and
533 other cancer controls
Cases, 78%; Controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales);
potential TCE exposure defined as any or substantial exposure.
Logistic regression adjusted for age, education, and ethic origin [TCE] or
Mantel-Haenszel stratified on age, income, index for cigarette smoking, and
ethnic origin [TCE],
Prostate
Aronson et al.
(1996);
Siemiatycki
(1991)
Male prostate cancer cases, age
35-75 yrs, diagnosed in 16 large
Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
449 cases
533 population controls
(Group 1) and
other cancer cases from
same study (Group 2)
Cases, 81%; Controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales).
Logistic regression adjusted for age, ethnic origin, socioeconomic status,
Quetlet, and respondent status [occupation] or Mantel-Haenszel stratified on
age, income, index for cigarette smoking, ethnic origin, and respondent status
[TCE],
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Renal Cell
Moore et al.
(2010)
Cases aged 20-74 year from four
European countries (Czech
Republic, Poland, Russia,
Romania) with histologically -
confirmed kidney cancer in
1999-2003; hospital controls with
diagnoses unrelated to smoking or
genitourinary disorders in 1998-
2003 and frequencymatched by
sex, age and study center.
1,097 cases (825 reanl cell
carcinomas)
1,184 controls
Cases, 90-99%; Controls,
90.3-96%
In-person interview using questionnaire for information on lifestyle habits,
smoking, antopometric measures, personal and family medical history and
occupational history. Specialized job-specific questionnaire for specific jobs
or industries of interest focused on solvents exposure, including.TCE, with
exposure assignmnet by expert blinded to case and control status by
frequency, intensity and confidence of TCE exposure. Exposure metric of
overall exposure, duration (total hours, years) and cumulative exposure.
Logistic regression adjusted for sex, age, and study center. BMI,
hypertension, smoking, and residence location also included in initial models
but did not alter odds ratios by >10%.
Charbotel et
al. (2006;
2009)
Cases from Arve Valley region in
France identified from local
urologists files and from area
teaching hospitals; age- and sex-
matched controls chosen from file
of same urologist as who treated
case or recruited among the
patients of the case's general
practitioner.
87 cases
316 controls
Cases, 74%; controls, 78%
Telephone interview with case or control, or, if deceased, with next-of-kin
(22% cases, 2% controls). Questionnaire assessing occupational history,
particularly, employment in the screw cutting jobs, and medical history.
Semiquantitative TCE exposure assigned to subjects using a task/TCE-
Exposure Matrix designed using information obtained from questionnaires
and routine atmospheric monitoring of work shops or biological monitoring
(U-TCA) of workers carried out since the 1960s. Cumulative exposure,
cumulative exposure with peaks, and TWA.
Conditional logistic regression with covariates for tobacco smoking and body
mass index.
Briining et al.
(2003)
Histologically-confirmed cases
1992-2000 from German
hospitals (Arnsberg); hospital
controls (urology department)
serving area, and local geriatric
department, for older controls,
matched by sex and age.
134 cases
401 controls
Cases, 83%; Controls, not
available
In-person interviews with case or next-of-kin; questionnaire assessing
occupational history using job title. Exposure metrics included longest job
held, JEM of Pannett et al. (1985) to assign cumulative exposure to TCE and
PCE, and exposure duration.
Logistic regression with covariates for age, sex, and smoking.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Pesch et al.
(2000b)
Histologically-confirmed cases
from German hospitals (5 regions)
in 1991-1995; controls randomly
selected from residency registries
matched on region, sex, and age.
935 cases
4,298 controls
Cases, 88%; Controls, 71%
In-person interview with case or next-of-kin; questionnaire assessing
occupational history using job title (JEM approach), self-reported exposure,
or job task (JTEM approach) to assign TCE and other exposures.
Logistic regression with covariates for age, study center, and smoking.
Parent et al.
(2000a);
Siemiatycki
(1991)
Male renal cell carcinoma cases,
age 35-75 yrs, diagnosed in 16
large Montreal-area hospitals in
1979-1985 and histologically
confirmed; controls identified
concurrently at 18 other cancer
sites; age-matched, population-
based controls identified from
electoral lists and RDD.
142 cases
533 population controls
(Group 1) and
other cancer controls
(excluding lung and bladder
cancers) (Group 2)
Cases, 82%; Controls, 71%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (about 300 exposures on semiquantitative scales);
TCE defined as any or substantial exposure.
Mantel-Haenszel stratified by age, body mass index, and cigarette smoking
[TCE] or logistic regression adjusted for respondent status, age, smoking,
and body mass index [occupation, job title].
Dosemeci et
al. (1999)
Histologically-confirmed cases,
1988-1990, white males and
females, 20-85 yrs, from
Minnesota Cancer Registry;
controls stratified for age and sex
using RDD, 21-64 yrs, or from
HCFA records, 64-85 yrs.
438 cases
687 controls
Cases, 87%; Controls, 86%
In-person interviews with case or next-of-kin; questionnaire assessing
occupational history of TCE using job title and JEM of Gomez et al. (1994).
Exposure metric was any TCE exposure.
Logistic regression with covariates for age, smoking, hypertension, and body
mass index.
Vamvakas et
al. (1998)
Cases who underwent
nephrectomy in 1987-1992 in a
hospital in Arnsberg region of
Germany; controls selected
accident wards from nearby
hospital in 1992.
58 cases
84 controls
Cases, 83%; Controls, 75%
In-person interview with case or next-of-kin; questionnaire assessing
occupational history using job title or self-reported exposure to assign TCE
and PCE exposure.
Logistic regression with covariates for age, smoking, body mass index,
hypertension, and diuretic intake.
Multiple or Other Sites
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
Lee et al.
(2003)
Liver, lung, stomach, colorectal
cancer deaths in males and
females between 1966-1997 from
two villages in Taiwan; controls
were cardiovascular and cerebral-
vascular disease deaths from same
underlying area as cases.
53 liver,
39 stomach,
26 colorectal,
41 lung cancer cases
286 controls
Response rate not reported
Residence as recorded on death certificate.
Mantel-Haenszel stratified by age, sex, and time period.
Kernan et al.
(1999)
Pancreatic deaths, 1984-1993, in
24 states; non-cancer death and
non-pancreatic disease death
controls, frequency matched to
cases by age, gender, race and
state.
63,097 pancreatic cancer
cases
252,386 non-cancer
population controls
Response rate not reported
Usual occupation and industry on death certificate coded to standardized
occupation codes and industry codes for 1980 U. S. census. Potential
exposure to 11 chlorinated hydrocarbons, including TCE, assessed using job-
exposure matrix of Gomez et al. (1994).
Logistic regression adjusted for age, marital status, gender, race, and
metropolitan and residential status.
Siemiatycki
(1991)
Male cancer cases, 1979-1985,
35-75 yrs, diagnosed in
16 Montreal-area hospitals,
histologically confirmed; cancer
controls identified concurrently;
age-matched, population-based
controls identified from electoral
lists and RDD.
857 lung and
117 pancreatic cancer cases
533 population controls
(Group 1) and other cancer
cases from same study
(Group 2)
Cases, 79% (lung), 71%
(pancreas); Controls, 72%
In-person interviews (direct or proxy) with segments on work histories (job
titles and self-reported exposures); analyzed and coded by a team of chemists
and industrial hygienists (294 exposures on semiquantitative scales); TCE
defined as any or substantial exposure.
Mantel-Haenszel stratified on age, income, index for cigarette smoking,
ethnic origin, and respondent status (lung cancer) and age, income, index for
cigarette smoking, and respondent status (pancreatic cancer).
HCFA = Health Care Financing Administration, JEM = job-exposure matrix, JTEM = job-task-exposure matrix, NCI = National Cancer Institute,
PCE = perchloroethylene, RDD = random digit dialing, U-TCA = urinary trichloroacetic acid, UV = ultra-violet.
-------

§•
to
s
a
to
s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-3. Geographic-based studies assessing cancer and TCE exposure
Reference
Description
Analysis approach
Exposure assessment
o
o
s
Co
I?
Oq ^
53
Broome County, NY Studies
Co
a
a,
>r
<§>
>S
TO
*
s
s
TO
0
s
1
a.
to
o
ATSDR
(2006c, 2008)
Total, 22 site-specific, and
childhood cancer incidence
from 1980-2001 among
residents in 2 areas in
Endicott, NY.
SIR among all subjects (ATSDR,
2006c) or among white subjects
only (ATSDR, 2008) with expected
numbers of cancers derived using
age-specific cancer incidence rates
for New York State, excluding New
York City. Limited assessment of
smoking and occupation using
medical and other records in lung
and kidney cancer subjects
(ATSDR, 2008).
Two study areas, Eastern and Western study areas, identified based on
potential for soil vapor intrusion exposures as defined by the extent of
likely soil vapor contamination. Contour lines of modeled VOC soil vapor
contamination levels based on exposure model using GIS mapping and soil
vapor sampling results taken in 2003. The study areas were defined by
2000 Census block boundaries to conform to model predicted areas of soil
vapor contamination. TCE was the most commonly found contaminant in
indoor air in Eastern study area at levels ranging from 0.18 to 140 |ig/m3 ,
with tetrachloroethylene, cis-l,2-dichloroethene, 1,1,1-trichloroethane, 1,1-
dichloroethylene, 1,1-dichloroethane, andFreon 113 detected at lower
levels. PCE was most common contaminant in indoor air in Western study
area with other VOCs detected at lower levels.
Maricopa County, AZ Studies
vo
Aickin et al.
(2004; 1992)
Cancer deaths, including
leukemia, 1966-1986, and
childhood (<19 yrs old)
leukemia incident cases
(1965-1986), Maricopa
County, AZ.
Standardized mortality RR from
Poisson regression modeling.
Childhood leukemia incidence data
evaluated using Bayes methods and
Poisson regression modeling.
Location of residency in Maricopa County, AZ, at the time of death as
surrogate for exposure. Some analyses examined residency in West Central
Phoenix and cancer. Exposure information is limited to TCE concentration
in two drinking water wells in 1982.
Pima County, AZ Studies
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
s
s
TO
>;
0
s
1
a.
VO
to
o
ADHS (1990,
1995)
Cancer incidence in
children (<19 yrs old) and
testicular cancer in
1970-1986 and
1987-1991, Pima County,
AZ.
Standardized incidence RR from
Poisson regression modeling using
method of Aickin et al. (1992).
Analysis compares incidence in
Tucson Airport Area to rate for rest
of Pima County.
Location of residency in Pima, County, AZ, at the time of diagnosis or
death as surrogate for exposure. Exposure information is limited to
monitoring since 1981 and includes VOCs in soil gas samples (TCE, PCE,
1,1-dichloroethylene, 1,1,1-trichloroacetic acid); PCBs in soil samples, and
TCE in municipal water supply wells.
Other
Coyle et al.
(2005)
Incident breast cancer
cases among men and
women, 1995-2000,
reported to Texas Cancer
Registry
Correlation study using rank order
statistics of mean average annual
breast cancer rate among women
and men and atmospheric release of
12 hazardous air pollutants.
Reporting to EPA Toxic Release Inventory the number of pounds released
for 12 hazardous air pollutants, (carbon tetrachloride, formaldehyde,
methylene chloride, styrene, tetrachloroethylene, trichloroethylene, arsenic,
cadmium, chromium, cobalt, copper, and nickel).
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-3. Geographic-based studies assessing cancer and TCE exposure (continued)
Reference
Description
Analysis approach
Exposure assessment
Morgan and
Cassady (2002)
Incident cancer cases,
1988-1989, among resi-
dents of 13 census tracts in
Redlands area, San
Bernardino County, CA.
SIR for all cancer sites and 16 site-
specific cancers; expected numbers
using incidence rates of site-specific
cancer of a four-county region
between 1988-1992.
TCE and perchlorate detected in some county wells; no information on
location of wells to residents, distribution of contaminated water, or TCE
exposure potential to individual residents in studied census tracts.
Vartiainen et al.
(1993)
Total cancer and site-
specific cancer cases
(NHLsites and liver) from
1953-1991 in two Finnish
municipalities.
SIR with expected number of
cancers and site-specific cancers
derived from incidence of the
Finnish population.
Monitoring data from 1992 indicated presence of TCE, tetrachloroethylene
and 1,1,1,-trichloroethane in drinking water supplies in largest towns in
municipalities. Residence in town used to infer exposure to TCE.
Cohn et al.
(1994a);
Fagliano et al.
(1990)
Incident leukemia and
NHL cases, 1979-1987,
from 75 municipalities and
identified from the New
Jersey State Cancer
Registry. Histological
type classified using WHO
scheme and the
classification of NIH
Working Formulation
Group for grading NHL.
Logistic regression modeling
adjusted for age.
Monitoring data from 1984-1985 on TCE, THM, and VOCs concentrations
in public water supplies, and historical monitoring data conducted in
1978-1984.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-3. Geographic-based studies assessing cancer and TCE exposure (continued)
Reference
Description
Analysis approach
Exposure assessment
Mallin (1990)
Incident bladder cancer
cases and deaths,
1978-1985, among
residents of 9 NW Illinois
counties.
SIR and SMR by county of
residence and zip code; expected
numbers of bladder cancers using
age-race-sex specific incidence
rates from SEER or bladder cancer
mortality rates of the United States
population from 1978-1985.
Exposure data are lacking for the study population with the exception of
noting one of two zip code areas with observed elevated bladder cancer
rates also had groundwater supplies contaminated with TCE, PCE and other
solvents.
Isacson et al.
(1985)
Incident bladder, breast,
prostate, colon, lung and
rectal cancer cases repor-
ted to Iowa cancer registry
between 1969-1981.
Age-adjusted site-specific cancer
incidence in Iowa towns with
populations of 1,000-10,000 and
who were serviced by a public
drinking water supply.
Monitoring data of drinking water at treatment plant in each Iowa
municipality with populations of 1,000-10,000 used to infer TCE and other
volatile organic compound concentrations in finished drinking water
supplies.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
s
s
TO
>;
0
s
1
a.
to
o
GIS = geographic information system, NW = Northwestern, PCE = perchloroethylene, RR = rate ratio, SEER = Surveillance, Epidemiology, and End Results,
SIR = standardized incidence ratio, SMR = standardized mortality ratio, VOCs = volatile organic compounds, WHO = World Health Organization.
vo
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Category A: Study Design
•	Clear articulation of study objectives or hypothesis. The ideal is a clearly stated
hypothesis or study objectives and the study is designed to achieve the identified
objectives.
•	Selection and characterization in cohort studies of exposure and control groups and of
cases and controls (case-control studies) is adequate. The ideal is for selection of cohort
and referents from the same underlying population and differences between these groups
are due to TCE exposure or level of TCE exposure and not to physiological, health status,
or lifestyle factors. Controls or referents are assumed to lack or to have background
exposure to TCE. These factors may lead to a downward bias including one of which is
known as "healthy worker bias," often introduced in analyses when mortality or
incidence rates from a large population such as the U.S. population are used to derive
expected numbers of events. The ideal in case-control studies is cases and controls are
derived from the same population and are representative of all cases and controls in that
population. Any differences between controls and cases are due to exposure to TCE
itself and not to confounding factors related to both TCE exposure and disease.
Additionally, the ideal is for controls to be free of any disease related to TCE exposure.
In this latter case, potential bias is toward the null hypothesis.
Category B: Endpoint Measured
•	Levels of health outcome assessed. Three levels of health outcomes are considered in
assessing the human health risks associated with exposure to TCE: biomarkers of effects
and susceptibility, morbidity, and mortality. Both morbidity as enumerated by incidence
and mortality as identified from death certificates are useful indicators in risk assessment
for hazard identification. The ideal is for accurate and predictive indicator of disease.
Incidence rates are generally considered to provide an accurate indication of disease in a
population and cancer incidence is generally enumerated with a high degree of accuracy
in cancer registries. Death certifications are readily available and have complete national
coverage but diagnostic accuracy is reduced and can vary by specific diagnosis.
Furthermore, diagnostic inaccuracies can contribute to death certificates as a poor
surrogate for disease incidence. Incidence, when obtained from population-based cancer
registries, is preferred for identifying cancer hazards.
•	Changes in diagnostic coding systems for lymphoma, particularly non-Hodgkin's
lymphoma. Classification of lymphomas today is based on morphologic,
immunophenotypic, genotypic, and clinical features and is based upon the World Health
Organization (WHO) classification, introduced in 2001, and incorporation of WHO
terminology into International Classification of Disease (ICD)-0-3. ICD Versions 7 and
earlier had rubrics for general types of lymphatic and hematopoietic cancer, but no
categories for distinguishing specific types of cancers, such as acute leukemia.
Epidemiologic studies based on causes of deaths as coded using these older ICD
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
classifications typically grouped together lymphatic neoplasms instead of examining
individual types of cancer or specific cell types. Before the use of immunophenotyping,
these grouping of ambiguous diseases such as non-Hodgkin's lymphoma and Hodgkin's
lymphoma may be have misclassified. Lymphatic tumors coding, starting in 1994 with
the introduction of the Revised European-American Lymphoma classification, the basis
of the current WHO classification, was more similar to that presently used.
Misclassification of specific types of cancer, if unrelated to exposure, would have
attenuated estimate of relative risk and reduced statistical power to detect associations.
When the outcome was mortality, rather than incidence, misclassification would be
greater because of the errors in the coding of underlying causes of death on death
certificates (IOM, 2003). Older studies that combined all lymphatic and hematopoietic
neoplasms must be interpreted with care.
Category C: TCE-Exposure Criteria
•	Adequate characterization of exposure. The ideal is for TCE exposure potential known
for each subject and quantitative assessment (job-exposure-matrix approach) of TCE
exposure assessment for each subject as a function of job title, year exposed, duration,
and intensity. Consideration of job task as additional information supplementing job title
strengthens assessment increases specificity of TCE assignment. The assessment
approach is accurate for assigning TCE intensity (TCE concentration or a time-weighted
average) to individual study subjects and estimates of TCE intensity are validated using
monitoring data from the time period. The objective for cohort and case-controls studies
is to differentiate TCE exposed subjects from subjects with little or no TCE exposure. A
variety of dose metrics may be used to quantify or classify exposures for an
epidemiologic study. They include precise summaries of quantitative exposure,
concentrations of biomarkers, cumulative exposure, and simple qualitative assessments of
whether exposure occurred (yes or no). Each method has implicit assumptions and
potential problems that may lead to misclassification. Exposure assessment approaches
in which it was unclear that the study population was actually exposed to TCE are
considered inferior since there may be a lower likelihood or degree of exposure to study
subjects compared to approaches which assign known TCE exposure potential to each
subject.
Category D: Follow-up (Cohort)
•	Loss to follow-up. The ideal is complete follow-up of all subjects; however, this is not
achievable in practice, but it seems reasonable to expect loss to follow-up not to exceed
10%. The bias from loss to follow-up is indeterminate. Random loss may have less
effect than if subjects who are not followed have some significant characteristics in
common.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
•	Follow-up period allows full latency period for over 50% of the cohort. The ideal to
follow all study subjects until death. Short of the ideal, a sufficient follow-up period to
allow for cancer induction period or latency over 15 or 20 years is desired for a large
percentage of cohort subjects.
Category E: Interview Type (Case-control)
•	Interview approach. The ideal interviewing technique is face-to-face by trained
interviewers with more than 90% of interviews with cases and control subjects conduced
face-to-face. The effect on the quality of information from other types of data collection
is unclear, but telephone interviews and mail-in questionnaires probably increase the rate
of misclassification of subject information. The bias is toward the null hypothesis if the
proportion of interview by type is the same for case and control, and of indeterminate
direction otherwise.
•	Blinded interviewer. The ideal is for the interviewer to be unaware whether the subject is
among the cases or controls and the subject to be unaware of the purpose and intended
use of the information collected. Although desirable for case-control studies, blinding is
usually not possible to fully accomplish because subject responses during the interview
provide clues as to subject status. In face-to-face and telephone interviews, potential bias
may arise from the interviewer expects regarding the relationship between exposure and
cancer incidence. The potential for bias from face-to-face interviews is probably less
than with mail-in interviews. Some studies have assigned exposure status in a blinded
manner using a job-exposure matrix and information collected in the unblinded interview.
The potential for bias in this situation is probably less with this approach than for
nonblinded assignment of exposure status.
Category F: Proxy Respondents
•	Proxy respondents. The ideal is for data to be supplied by the subject because the subject
generally would be expected to be the most reliable source; less than 10% of either total
cases or total controls for case-control studies. A subject may be either deceased or too
ill to participate, however, making the use of proxy responses unavoidable if those
subjects are to be included in the study. The direction and magnitude of bias from use of
proxies is unclear, and may be inconsistent across studies.
Category G: Sample Size
•	The ideal is for the sample size is large enough to provide sufficient statistical power to
ensure that any elevation of effect in the exposure group, if present, would be found, and
to ensure that the confidence bounds placed on relative risk estimates can be
well-characterized.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Category H: Analysis Issues
•	Control for potentially confounding factors of importance in analysis. The ideal in cohort
studies is to derive expected numbers of cases based on age-sex- and time-specific cancer
rates in the referent population and in case-control studies by matching on age and sex in
the design and then adjusting for age in the analysis of data. Age and sex are likely
correlated with exposure and are also risk factors for cancer development. Similarly,
other factors such as cigarette smoking and alcohol consumption are risk factors for
several site-specific cancers reported as associative with TCE exposure. To be a
confounder of TCE, exposure to the other factor must be correlated, and the association
of the factor with the site-specific cancer must be causal. The expected effect from
controlling for confounders is to move the estimated relative risk estimate closer to the
true value.
•	Statistical methods are appropriate. The ideal is that conclusions are drawn from the
application of statistical methods that are appropriate to the problem and accurately
interpreted.
•	Evaluation of exposure-response. The ideal is an examination of a linear
exposure-response as assessed with a quantitative exposure metric such as cumulative
exposure. Some studies, absent quantitative exposure metrics, examine exposure
response relationships using a semiquantitative exposure metric or by duration of
exposure. A positive dose-response relationship is usually more convincing of an
association as causal than a simple excess of disease using TCE dose metric. However, a
number of reasons have been identified for a lack of linear exposure-response finding and
the failure to find such a relationship means little from an etiological viewpoint and does
not minimize an observed association with overall TCE exposure.
•	Documentation of results. The ideal is for analysis observations to be completely and
clearly documented and discussed in the published paper, or provided in supplementary
materials accompanying publication.
B.2.1. Study Designs and Characteristics
The epidemiologic designs investigating TCE exposure and cancer include cohort studies
of occupationally exposure populations, population case-control studies, and geographic studies
of residents in communities with TCE in water supplies or ambient air. Analytical
epidemiologic studies, which include case-control and cohort designs, are generally relied on for
identifying a causal association between human exposure and adverse health effects (U.S. EPA,
2005c) due to their clear ability to show exposure precedes disease occurrence. In contrast,
ecologic studies such as health surveys of cancer incidence or mortality in a community during a
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
specified time period, i.e., geographic-based studies identified in Appendix B, Table B-3,
provide correlations between rates of cancer and exposure measured at the geographic level.
An epidemiologic study's ability to inform a question on TCE and cancer depends on
clear articulation of study objective or hypothesis and adequate selection of exposed and control
group in cohort studies and cases and controls in case-control studies are important. As the body
of evidence on trichloroethylene has grown over the past 20 years, so has the number of studies
with clearly articulated hypothesis. All Nordic cohort studies (Anttila et al., 1995; Axelson et al.,
1994; Hansen et al., 2001; Raaschou-Nielsen et al., 2003) are designed to examine cancer and
TCE, albeit some with limited statistical power, as are recent cohort studies of United States
occupationally exposed populations (Boice et al., 1999; Boice et al., 2006b; Radican et al., 2008;
Ritz, 1999a; Zhao et al., 2005). Exposure assessment approaches in these studies distinguished
subjects with varying potentials for TCE exposure, and in some cases, assigned a
semiquantitative TCE exposure surrogate to individual study subjects. Three case-control
studies nested in cohorts, furthermore, examined TCE exposure and site-specific cancer, albeit a
subject's potential and overall prevalence of TCE exposure greatly varied between these studies
(Greenland et al., 1994; Krishnadasan et al., 2007; Wilcosky et al., 1984). Typically, studies of
all workers at a plant or manufacturing facility (2004; Blair et al., 1989; Chang et al., 2003;
Chang et al., 2005; Clapp and Hoffman, 2008; Costa et al., 1989; Garabrant et al., 1988;
Shannon et al., 1988; Shindell and Ulrich, 1985; Sinks et al., 1992; 2007; 2008) are not designed
to evaluate cancer and TCE specifically, given their inability to identify varying TCE exposure
potential for individual study subjects; rather, such studies evaluate the health status of the entire
population working at that facility. Bias associated with exposure misclassification is greater in
these studies, and for this and other reasons more fully discussed below, they are of limited
utility for informing evaluations on TCE exposure and cancer.
Recent case-control studies with hypotheses specific for TCE exposure include the
kidney cancer case-control studies of Vamvakas et al. (1998), Briining et al. (2003), and
Charbotel et al. (2006; 2009). More common, population-based case-control studies assess
occupational exposure to organic solvents, using a job-exposure matrix approach for exposure
assessment to examine organic solvent categories, i.e., aliphatic hydrocarbons, or specific
solvents such as TCE. The case-control studies of Costas et al. (2002) and Lee et al. (2003) were
also designed to examine possible association with contaminated drinking water containing
trichloroethylene and other solvents detected at lower concentrations. The hypothesis of
Siemiatycki (1991) and ancillary publications (Dumas et al., 2000; Fritschi and Siemiatycki,
1996a; Goldberg et al., 2001; Parent et al., 2000a; Siemiatycki et al., 1994) explored possible
association between 20 site-specific cancers and occupational title or chemical exposures,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
including TCE exposure, using a contemporary exposure assessment approach for more focused
research investigation.
Cases and control selection in most population-based case-control studies of TCE
exposure are considered a random sample and representative of the source population
(Siemiatycki, 1991) [and related publications (Aronson et al., 1996; Briining et al., 2003;
Charbotel et al., 2006; Charbotel et al., 2009; Cocco et al., 2010; Costas et al., 2002; De Roos et
al., 2001a; Dosemeci et al., 1999; Dumas et al., 2000; Fritschi and Siemiatycki, 1996a, b; Gold et
al., In Press; Goldberg et al., 2001; Hardell et al., 1994; Heineman et al., 1994; Kernan et al.,
1999; Lee et al., 2003; Lowengart et al., 1987; McKinney et al., 1991; Miligi et al., 2006; Moore
et al., 2010; Nordstrom et al., 1998; Parent et al., 2000a; Persson and Fredrikson, 1999; Pesch et
al., 2000a; Pesch et al., 2000b; Seidler et al., 2007; Shu et al., 2004; Siemiatycki et al., 1994)].
Case and control selection in Vamvakas et al. (1998), a study conducted in the Arnsberg area of
Germany, is subject to criticism regarding possible selection bias resulting from differences in
selection criteria, cases worked in small industries and controls from a wider universe of
industries; differences in age, controls being younger than cases with possible lower exposure
potentials; and temporal difference in case and control selection, controls selected only during
the last year of the study period with possible lower exposure potential if exposure has decreased
over period of the study (NRC, 2006). The potential for selection bias in Briining et al. (2003),
another study in the same area as Vamvakas et al. (1998) but of later period of observation, was
likely reduced compared to Vamvakas et al. (1998) due to the broader region of southern
Germany from which cases were identified and interviewing cases and controls during the same
time. One case-control study nested in a cohort (Greenland et al., 1994) included subjects whose
deaths were reported to and known by the employer, e.g., occurred among vested or pensioned
employees or among currently employees. A 10- to 15-year employment period was required for
subjects in this study to receive a pension; deaths among employees who left employment before
this time were not known to the employer and not included the study. Survivor bias, a selection
bias, may be introduced by excluding nonpensioned workers or those who leave employment
before becoming vested in a company's retirement plan is more likely than in a study of all
employees with complete follow-up. The use of pensioned deaths as controls, as was done in
this study, would reduce potential bias if both cases and control had the same likelihood of
becoming pensioned. That is, the probability for becoming a pensioned worker is similar for all
deaths and unrelated to the likelihood of exposure or magnitude of exposure and disease. No
information was available in (Greenland et al., 1994) to evaluate this assumption.
Geographic-based and ecological studies of TCE contaminated water supplies typically
focus on estimating cancer or other disease rates in geographically circumscribed populations
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
who are geospatially located with a source containing TCE, e.g., a hazardous waste site, well
water, or air. These studies are often less informative for studying cancer due to their inability to
estimate incidence rate ratios, essential for causal inferences, inferior exposure assessment
approach, and to possible selection biases. Ecological studies also are subject to bias known as
"ecological fallacy" since variables of exposure and outcome measured on an aggregate level do
not represent association at the individual level. Consideration of this bias is important for
diseases with more than one risk factor, such as the site-specific cancers evaluated in this
assessment.
B.2.2. Outcomes Assessed in Trichloroethylene (TCE) Epidemiologic Studies
The epidemiologic studies consider at least three levels of health outcomes in their
examinations of human health risks associated with exposure to trichloroethylene: biomarkers of
effects and susceptibility, morbidity, and mortality (NRC, 2006). Few susceptibility biomarkers
have been examined and these are not specific to trichloroethylene (NRC, 2006). By far, the
bulk of the literature on cancer and trichloroethylene exposure is of cancer morbidity (ADHS,
1990, 1995; Aickin, 2004; Anttila et al., 1995; ATSDR, 2006c; Axelson et al., 1994; Briining et
al., 2003; Charbotel et al., 2006; Charbotel et al., 2009; Cocco et al., 2010; Cohn et al., 1994a;
Costas et al., 2002; Coyle et al., 2005; De Roos et al., 2001a; Dosemeci et al., 1999; Dumas et
al., 2000; Fredriksson et al., 1989; Gold et al., In Press; Hansen et al., 2001; Hardell et al., 1994;
Isacson et al., 1985; Lowengart et al., 1987; McKinney et al., 1991; Miligi et al., 2006; Moore et
al., 2010; Morgan and Cassady, 2002; Nordstrom et al., 1998; Persson and Fredrikson, 1999;
Persson et al., 1993; Pesch et al., 2000a; Pesch et al., 2000b; Purdue et al., 2011; Raaschou-
Nielsen et al., 2003; Seidler et al., 2007; Shannon et al., 1988; Shu et al., 2004; Siemiatycki,
1991; Sung et al., 2008; Vamvakas et al., 1998; Vartiainen et al., 1993; Wang et al., 2009),
mortality (Aickin et al., 1992; ATSDR, 2004; Blair et al., 1989; Boice et al., 1999; Boice et al.,
2006b; Clapp and Hoffman, 2008; Costa et al., 1989; Garabrant et al., 1988; Greenland et al.,
1994; Heineman et al., 1994; Kernan et al., 1999; Lee et al., 2003; Morgan et al., 1998; Radican
et al., 2008; Ritz, 1999a; Shindell and Ulrich, 1985; Wilcosky et al., 1984), or both (Blair et al.,
1998; Chang et al., 2003; Chang et al., 2005; Henschler et al., 1995; Sinks et al., 1992; Sung et
al., 2007; Zhao et al., 2005).
Mortality is readily identified from death certificates; however, diagnostic accuracy from
death certificates varies by the specific diagnosis (Brenner and Gefeller, 1993). Incident cancer
cases are enumerated more accurately by tumor registries and by hospital pathology records and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-35 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
cases identified from these sources are considered to have less bias resulting from disease
misclassification than cause or underlying cause of death as noted on death certificates. Studies
of incidence are preferred, particularly for examining association with site-specific cancers
having high 5-year survival rates or which may be misclassified on death certificate.
Misclassification of the cause of death as noted on death certificates attenuates statistical power
through errors of outcome identification. This nondifferential misclassification of outcome in
cohort studies will lead to attenuation of rate ratios, although the magnitude of is difficult to
predict (NRC, 2006). Cancer registries are used for cases diagnosed in more recent time periods
and cohorts whose entrance dates are 30 or 40 years may miss many incident cancers and
reduced statistical power as a consequence. Two studies examine both cancer incidence and
mortality (Blair et al., 1998; Zhao et al., 2005). The lapse of 20 or more years in Blair et al.
(1998) and 38 years in Zhao et al. (2005) between date of cohort identification and cancer
incidence ascertainment suggests these studies are missing cases and limits incidence
examinations.
B.2.3. Disease Classifications Adopted in Trichloroethylene (TCE) Epidemiologic Studies
Disease coding and changes over time are important in epidemiologic evaluations,
particularly in evaluation of heterogeneity or consistency of observations from a body of
evidence. The ICD, published by WHO, is used to code underlying and contributing cause of
death on death certificates and is updated periodically, adding to diagnostic inconsistency for
cross-study comparisons (NRC, 2006). Tumor registries use the International Classification of
Diseases-Oncology (ICD-O) for coding the site and the histology of neoplasms, principally
obtained from a pathology report.
The epidemiologic studies of TCE exposure have used a number of different
classification systems (Scott and Chiu, 2006). A number of studies classified neoplasms
according to ICD-0 (Chang et al., 2005; Costas et al., 2002; Gold et al., In Press; Moore et al.,
2010; Purdue et al., 2011; Siemiatycki, 1991) or to ICD-9 (Kernan et al., 1999; Nordstrom et al.,
1998; Ritz, 1999a; Zhao et al., 2005). Other ICD revisions used in recent studies include ICDA-
8 (Blair et al., 1989; Blair et al., 1998; Greenland et al., 1994), ICD-7 (Anttila et al., 1995;
Axelson et al., 1994; Hansen et al., 2001; Raaschou-Nielsen et al., 2003), or several ICD
revisions, whichever was in effect at the date of death (Boice et al., 1999; Garabrant et al., 1988;
Morgan et al., 1998, 2000b; Radican et al., 2008). In this latter case, changes in disease
classification over revisions are not harmonized or recoded to a common classification; and,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
diagnostic inconsistencies and disease misclassification errors leads to a greater likelihood for
bias in these studies. Greatest weight is placed on studies where all cases or deaths are classified
using current classification systems. However, association in studies adopting older revisions,
ICD 7 (Anttila et al., 1995; Axelson et al., 1994; Hansen et al., 2001; Raaschou-Nielsen et al.,
2003), for example, is noteworthy given the narrow consideration of lymphoid neoplasms
compared to contemporary classification systems. Consistency examinations of the overall body
of evidence using meta-analysis methods and examination of heterogeneity will need to consider
study differences in coding in interpreting findings.
A major shift in thinking occurred around 1995 with the Revised European-American
Lymphoma (REAL) classification of grouping diseases of the blood and lymphatic tissues along
their cell lines compared to previous approaches to group lymphomas by a cell's physical
characteristics. It was increasing recognized that some NHLs and corresponding lymphoid
leukemias were different phases (solid and circulating) of the same disease entity (Morton et al.,
2007). Many concepts of contemporary knowledge of lymphomas are incorporated in the WHO
Classification of Neoplastic Diseases of the Hematopoietic and Lymphoid Tissues, an
international consensus scheme for classifying leukemia and lymphoma now in use and the
predecessor to REAL (Jaffe et al., 2001). Both the ICD-O, 3rd edition, and ICD-10 have adopted
the WHO classification framework.
The only study coding NHLs using the WHO classification is (Cocco et al., 2010). Other
NHL studies have adopted older lymphoma classification systems, either the National Cancer
Institute's (NCI) Working Formulation (Costantini et al., 2008; Miligi et al., 2006) or other
systems coding lymphomas according to NCI's Working Formulation, i.e., International
Classification of Disease-Oncology, 2nd Edition (Gold et al., In Press; Purdue et al., 2011; Wang
et al., 2009), that divided lymphomas into low-grade, intermediate-grade and high grade, with
subgroups based on cell type and presentation, or Rappaport (Hardell et al., 1994; 1981), with
groupings based on microscopic morphology (Lymphoma Information Network, 2008). Both
Purdue et al. (2011) and Gold et al. do provide equivalent ICD-O-3 morphology codes
(http://www.seer.cancer.gov/tools/conversion/ICD02-3manual.pdf, accessed April 6, 2011,).
Lowengart et al. (1987), Persson et al. (1989; 1993), McKinney et al. (1991) nor Persson and
Fredriksson (1999) provide information in their published articles on lymphomas classification
systems used in these studies.
Implications of classification changes are most significant for NHL. As noted by the
IOM (2003), in Revision 7 and earlier editions of the ICD, all lymphatic and hematopoietic
neoplasms were grouped together instead of treated as individual types of cancer (such as
Hodgkin's disease) or specific cell types (such as acute lymphocytic leukemia). One limitation
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
of this treatment was the amalgamation of these relatively rare cancers would increase the
apparent sample size but could also result in diluted estimates of effect if etiologic heterogeneity
of different lymphoma subtypes existed, i.e., different sites of cancer were not associated in
similar ways with the exposures of interest. Additionally, immunophenotyping was not
available, leading to decreased ability to distinguish ambiguous diseases, and diagnoses of these
cancers may have been misclassified; for example, NHL may have been grouped with other
lymphatic and hematopoietic cancers to increase statistical power or misclassified as Hodgkin's
disease, for example. Examination of distinct lymphoma subtypes is expected to reduce disease
misclassification bias. Five case-control studies on non-Hodgkin's lymphoma (NHL) include
analysis of lymphoma subtype and trichloroethylene exposure (Cocco et al., 2010; Costantini et
al., 2008; Gold et al., In Press; Miligi et al., 2006; Purdue et al., 2011).
A change in liver cancer coding occurred between ICDA-8 and ICD-9 and is important to
consider in examinations of liver cancer observations across the TCE studies. With ICD-9, liver
cancer "not specified as primary or secondary" was moved from the grouping of secondary
malignant neoplasms and added to the larger class of malignant liver neoplasms. Thus, a similar
grouping of liver cancer causes is necessary to cross-study comparisons. For example, an
examination of liver cancer, based on ICDA-8, would need to include codes for liver and
intrahepatic bile duct (code 155) and liver, not specified as primary or secondary (code 197.8),
but, for ICD-9, would include liver and intrahepatic bile duct (code 155) only. The effect of
adding "liver cancer, not specified as primary or secondary" to the larger liver and intrahepatic
bile duct category in ICD-9 was a 2-fold increase in the overall liver cancer mortality (Percy et
al., 1990).
B.2.4. Exposure Classification
Adequacy of exposure assessment approaches and their supporting data are a critical
determinant of a study's contribution in a weight-of-evidence evaluation (Checkoway et al.,
1989). Exposure assessment approaches in studies of TCE and cancer vary greatly. At one
extreme, studies assume subjects are exposed by residence in a defined geographic area (ADHS,
1990, 1995; Aickin, 2004; Aickin et al., 1992; AT SDR, 2006c, 2008; Cohn et al., 1994a; Coyle
et al., 2005; Isacson et al., 1985; Lee et al., 2003; Morgan and Cassady, 2002; Vartiainen et al.,
1993) or by employment in a plant or job title (ATSDR, 2004; Blair et al., 1989; Chang et al.,
2003; Chang et al., 2005; Clapp and Hoffman, 2008; Costa et al., 1989; Garabrant et al., 1988;
Shannon et al., 1988; Shindell and Ulrich, 1985; Sung et al., 2007; Sung et al., 2008). This is a
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-38 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
poor exposure surrogate given potential for TCE exposure can vary in these broad categories
depending on job function, year, use of personal protection, and, for residential exposure,
pollutant fate and transport, water system distribution characteristics, percent of time per day in
residence, presence of mitigation devices, drinking water consumption rates, and showering
times. Another example comprises measurement from a subset of workers with jobs where TCE
is routinely used to infer TCE exposure and TCE intensity to all subjects. In both examples,
exposure misclassification potential may be extensive and with a downward bias in risk
estimates.
At the other extreme and preferred given a reduced likelihood for misclassification bias,
quantitative exposure assessment based upon a subject's job history, job title, and monitoring
data are used to develop estimates of TCE intensity and cumulative exposure (quantitative
exposure metrics or measures) and is know as job-exposure matrix (JEM) approaches. Peak
exposure is also well characterized. Addition to JEM approaches of information on job tasks
(JTEM) associated with exposure such as that done by Pesch et al. (2000a; 2000b) is expected to
reduce potential exposure misclassification. In between these two extremes, semiquantitative
estimates of low, medium, and high TCE exposure are assigned to subjects. Twenty-one studies
assigned a quantitative or semiquantitive TCE surrogate metrics to individual subjects using a
JEM , job-task-exposure-matrix (JTEM), or expert knowledge: (Siemiatycki, 1991)[and related
publications, (Aronson et al., 1996; Dumas et al., 2000; Fritschi and Siemiatycki, 1996a, b;
Goldberg et al., 2001; Parent et al., 2000a; Siemiatycki et al., 1994)], Blair et al. (1998) and
follow-up by Radican et al. (2008), Morgan et al. (1998), Vamvakas et al. (1998), Kernan et al.
(1999), Ritz (1999a), Pesch et al. (2000a; 2000b), Briining et al. (2003), Zhao et al. (2005),
Miligi et al. (2006), Charbotel et al. (2006; 2009), Krishnadansen et al. (2007), Seidler et al.
(2007), Costantini et al. (2008), Wang et al. (2009), Cocco et al. (2010), Gold et al., Moore et al.
(2010) and Purdue et al. (2011).
Thirteen other studies assigned a qualitative TCE surrogate metric (ever exposed or never
exposed), less preferred to a semi-quantitative exposure surrogate given greater likelihood for
error associated exposure misclassification, using general job classification of job title by
reference to industrial hygiene records indicating a high probability of TCE use, individual
biomarkers, job exposure matrices, water distribution models, for cohort studies, or obtained
from subjects using questionnaire for case-control studies. The 13 studies were: Wilcosky et al.
(1984), Lowengart et al. (1987), McKinney et al. (1991), Greenland et al. (1994), Hardell et al.
(1994), Nordstrom et al. (1998), Shu et al. (1999), Boice et al. (1999; 2006b), Dosemeci et al.
(1999), Persson and Fredriksson (1999), Costas et al. (2002), Raaschou-Nielsen et al. (2003),.
Without quantitative measures, however, it is not possible to quantify exposure difference
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
between groupings nor is it possible to compare similarly named categories across studies.
Exposure misclassification for dichotomous exposure defined in these studies, if nondifferential,
would downward bias resulting risk estimates.
Zhao et al. (2005), Krishnadansen et al. (2007), and Boice et al. (2006b) are studies with
overlap in some subjects, but with different exposure assessment approaches, more fully
discussed in B.3.1.1., with implication on study ability to identify cancer hazard. While these
studies used job title to assign TCE exposure potential, Zhao et al. (2005) and Krishnadansen et
al. (2007) developed a semiquantitative estimate of TCE exposure potential, whereas, Boice et
al. (2006b) classified subjects as either "exposed" or "unexposed" using a qualitative surrogate.
These studies, furthermore, identify TCE exposure potentially differently for possibly similar job
titles. For example, jobs as instrument mechanics, inspectors, test stand engineers, and research
engineers are identified with medium potential exposure in Zhao et al. (2005) and Krishnadansen
et al. (2007); however, these job titles were considered in Boice et al. (2006b) as having
background exposure and were combined with unexposed subjects, the referent population in
Cox Proportional Hazard analyses.
Three Nordic cohorts have TCE exposure as indicated from biological markers, assigning
TCE exposure to subjects using either concentration of trichloroacetic acid (TCA) in urine or
TCE in blood (Anttila et al., 1995; Axelson et al., 1994; Hansen et al., 2001). The utility of a
biomarker depends on it selectivity and the exposure situation. Urinary TCA (U-TCA) is a
nonselective marker since other chlorinated solvents besides TCE are metabolized to TCA and
resultant urinary elimination. If only TCE is the only exposure, urinary TCE may be a useful
marker; however, in setting with mixed exposure, urinary TCA may serve as an integrated
exposure marker of several chlorinated solvents. The Nordic studies used the linear relationship
-3
found for average inhaled trichloroethylene versus U-TCA: trichloroethylene (mg/m ) = 1.96;
U-TCA (mg/L) = 0.7 for exposures lower than 375 mg/m3 (69.8 ppm) (Ikeda et al., 1972). This
relationship shows considerable variability among individuals, which reflects variation in urinary
output and activity of metabolic enzymes. Therefore, the estimated inhalation exposures are
only approximate for individuals but can provide reasonable estimates of group exposures.
-3
There is evidence of nonlinear formation of U-TCA above about 400 mg/m or 75 ppm of
trichloroethylene. The half-life of U-TCA is about 100 hours. Therefore, the U-TCA value
represents roughly the weekly average of exposure from all sources, including skin absorption.
The Ikeda et al. (1972) relationship can be used to convert urinary values into approximate
airborne concentration, which can lead to misclassification if tetrachloroethylene and
1,1,1-trichloroethane are also being used because they also produce U-TCA. In most cases, the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Ikeda et al. (1972) relationship provides a rough upper boundary of exposure to
tri chl oroethyl ene.
B.2.5. Follow-up in Trichloroethylene (TCE) Cohort Studies
Cohort studies are most informative if vital status is ascertained for all cohort subjects
and if the period of time for disease ascertainment is sufficient to allow for long latencies,
particularly for cancer detection and death, in the case of mortality studies. Inability to ascertain
vital status for all subjects, or, conversely, subjects who are loss-to-follow-up, can affect the
validity of observations and lead to biased results. Both power and rate ratios estimated in
cohort studies can be underestimated due to bias introduced if the follow-up period was not long
enough to account for latency (NRC, 2006). The probability of loss to follow-up may be related
to exposure, disease, or both. The multiple-stage process of cancer development occurs over
decades after first exposure and studies with full latent periods are considered to provide greater
weight to the evaluation compared to cohort studies with shortened follow-up period and lower
percentage of subjects whose vital status was known on the date follow-up ended. Vital status
ascertainment for over 90% of all cohort studies and long mean follow-up periods, say 15 years
of longer, characterized many occupational cohort studies on trichloroethylene and cancer
(Anttila et al., 1995; Blair et al., 1998; Costa et al., 1989; Garabrant et al., 1988) and the
follow-up study of Radican et al. (Boice et al., 1999; Boice et al., 2006b; Morgan et al., 1998;
Raaschou-Nielsen et al., 2003; 2008; Ritz, 1999a; Zhao et al., 2005). Information is lacking in
two biomarker studies (Axelson et al., 1994; Hansen et al., 2001), additionally, to estimate the
mean follow-up period for TCE-exposed subjects; although, Hansen et al. (2001) state "some
workers were followed for as long as 50 years after their exposure, which allowed the detection
of cancers with long latency periods." Other studies of trichloroethylene and cancer did not
identify a latent period, information for calculating a latent period, or contained other
deficiencies in follow-up criteria (Blair et al., 1989; Chang et al., 2005; Costa et al., 1989;
Henschler et al., 1995; Shannon et al., 1988; Sinks et al., 1992; Sung et al., 2007; Wilcosky et al.,
1984). Proportionate mortality ratio studies, based only on deaths and which lack information on
person-year structure as cohort studies, by definition, do not contain information on cancer latent
periods or follow-up (ATSDR, 2004; Clapp and Hoffman, 2008).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.2.6. Interview Approaches in Case-Control Studies of Cancer and Trichloroethylene
(TCE) Exposure
Interview approaches and the percentage of subjects with information obtained from
proxy or next-of-kin respondents need consideration in interpreting population and hospital-
based case-control studies in light of possible biases. Biases resulting from proxy respondent or
from low participation related to mailed questionnaires are not relevant to cohort or geographic
studies since information is obtained from local, national, or corporate records. Both face-to-
face and telephone interviews are common and valid approaches used in population or
hospital-based case-control studies. Important to each is the use of a structured questionnaires
combined with intensive training as ways to minimize a high potential for biases often associated
with mailed questionnaires (Blatter et al., 1997; Schlesselman and Stolley, 1982). Studies with
information limited to job title, type of business and dates of employment and aided with
computer or job-exposure-matrix approaches are preferred to studies of job title only; the added
approaches can reduce exposure misclassification bias and improve disease risk estimates
(Stewart et al., 1996). Moreover, interview with respondents other than the individual case or
control, through proxy or next-of-kin respondents, may also introduce bias in case-control
studies. Proxy respondents are used when cases or control are either too sick to respond or if
deceased. This bias would dampen observed associations if proxy respondents did not fully
provide accurate information. Boyle et al. (1992), for example, in their study of several site-
specific cancers and occupational exposures found low sensitivity, or correct reporting, for
occupational exposure to solvents among proxy respondents. The weight of evidence analysis on
trichloroethylene and cancer, for this reason, places greatest weight on observations from studies
which obtain information on personal, medical, and occupational histories from each case and
control with lesser weight is placed on studies where 10 percent or more of interviews are with
proxy respondents.
Many of the more recent case-control studies include face-to-face (Briining et al., 2003;
Cocco et al., 2010; Costas et al., 2002; Dosemeci et al., 1999; Gold et al., In Press; McKinney et
al., 1991; Miligi et al., 2006; Moore et al., 2010; Pesch et al., 2000a; Pesch et al., 2000b; Purdue
et al., 2011; Seidler et al., 2007; Siemiatycki, 1991; Vamvakas et al., 1998; Wang et al., 2009) or
telephone (Charbotel et al., 2006; Charbotel et al., 2009; Lowengart et al., 1987; Shu et al., 2004;
Shu et al., 1999) interviews. Few of these studies included interviewers who were blinded or did
not know the identity of who is a case and who is a control. Although desirable for case-control
studies, blinding is usually not possible to fully accomplish because subject resposonses during
the interview provide clues as to subject status. For this reason, the lack of blinded interviewers
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-42 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
is not considered a serious limitation. More importantly, most studies assigned exposure to cases
and controls in a blinded manner
Information obtained from mailed questionnaire predominantly characterized older
Nordic studies (Fredriksson et al., 1989; Hardell et al., 1994; Hardell et al., 1981; Nordstrom et
al., 1998; Persson et al., 1989; Persson and Fredrikson, 1999; Persson et al., 1993). One case-
control study did not ascertain information from a questionnaire or through interviews, instead
using occupation coded on death certificates to infer TCE exposure potential (Kernan et al.,
1999). In all studies except Costas et al. (2002) and Kernan et al. (1999), assignment of potential
TCE exposure to cases and controls, to different degrees depending on each study, is based on
self-reported information on job title, and in some cases, to specific chemicals.
More common to the case-control studies on trichloroethylene and cancer was possible
bias related to a higher percentage of proxy interviews. Seven studies (Dosemeci et al., 1999;
Gold et al., In Press; Moore et al., 2010; Pesch et al., 2000a; Pesch et al., 2000b; Purdue et al.,
2011; Wang et al., 2009) excluded subjects with proxy interviews and the percentage of proxy
interview among subjects in one other study is less than 10 percent (Nordstrom et al., 1998).
Charbotel et al. (2006; 2009) furthermore presents analyses for data they considered as better
quality, including higher confidence exposure information and excluding proxy respondents, in
addition to analyses using both living and proxy respondents. A consideration of proxy
interviews in studies of childhood cancers which include an examination of paternal occupational
exposure is needed given a greater likelihood for bias if fathers are not directly interviewed and
the father's occupational information is provided only by the child's mother. A good practice is
for statistical analyses examining paternal occupational exposure to include only cases and
controls with direct information provided by the fathers, such as De Roos et al. (2001a), the only
childhood cancer study (neuroblastoma) to exclude the use of proxy information.
B.2.7. Sample Size and Approximate Statistical Power
Cancer is generally considered a rare disease compared to more common health outcomes
such as cardiovascular disease. Of all site-specific cancers, endocrine cancers of the breast
prostate and lung cancer are most common, with age-adjusted incidence rates of 126.0 per
100,000 women (breast), 163 per 100,000 men (prostate), and 63.9 per 100,000 men and women
(lung) (NCI, 2008). Several site-specific cancers including kidney cancer, liver cancer, and NHL
that are of interest to trichloroethylene are rarer and consideration of study size and the influence
on statistical power are factors forjudging a study's validity and assessment of a study's
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
contribution to the overall weight-of-evidence for identifying a hazard. For example, the age-
adjusted incidence rates of NHL, liver and intrahepatic bile duct cancer, and kidney and renal
pelvis cancer in the United States population are 19.5 per 100,000, 6.4 per 100,000, and 13.2 per
100,000; rates vary by sex and race. Age-adjusted mortality rates for these cancers are lower:
7.3 per 100,000 (NHL), 5.0 per 100,000 (liver and intrahepatic bile duct), 4.2 per 100,000
(kidney and renal pelvis). Rates of the childhood cancer, acute lymphocytic leukemia, are even
lower: 1.6 (incidence) and 0.5 (mortality) per 100,000 (NCI, 2008).
Only very large cohort or case-control studies would have a sufficient number of cases
and statistical power to estimate excess risks and exposure-response relationships (NRC, 2006).
Observations from studies with large numbers of TCE-exposed subjects, given consideration of
exposure conditions and other criteria discussed in this section, can provide useful information
on hazard and may provide quantitative information on possible upper bound trichloroethylene
cancer risks. Alternatively, studies of small numbers of subjects or cases and controls, typically,
studies with statistical power less than 80% to detect risk of a magnitude of 2 or less, are not
likely to provide useful evidence for or against the hypothesis that trichloroethylene is a human
carcinogen.
Studies with either a large number of TCE-exposed subjects or with large numbers of
total deaths, cancer deaths, or cancer cases among TCE-exposed subjects are the cohort studies
of Blair et al. (1998), Raaschou-Nielsen et al. (2003), and Zhao et al. (2005), and the case-control
studies of Pesch et al. (2000a), Shu et al. (2004; 1999) [paternal exposure assessment, only]),
Wang et al. (2009) and Cocco et al. (2010), with 50 or more TCE-exposed cases. The cohorts of
Boice et al. (1999; 2006b) and Morgan et al. (1998), like that of Blair et al. (1998), comprised
over 10,000 subjects both with and without potential TCE exposure; however, the number of
subjects and the percentage of the larger cohort identified with TCE exposure in these studies
was less than that in Blair et al. (1998); 23% of all subjects in Morgan et al. (1998), 3% in Boice
et al. (1999), 2% in Boice et al. (2006b) compared to 50% in Blair et al. (1998). Moreover,
although the cohorts of Garabrant et al. (1988), Chang et al. (2005) and Sung et al. (2007) are
also of population sizes greater than 10,000, these studies of employees at one manufacturing
facility lack assignment of potential TCE exposure to individual subjects and include subjects
with varying exposure potential, some of whom are likely with very low to no exposure potential
to TCE. Rate ratios estimated from cohorts that include unexposed subjects would be
underestimated although the magnitude of this bias can not be calculated given the absence in
individual studies of information on the percentage of subjects lacking potential TCE exposure.
Examination of the statistical power or ability to detect a rate ratio magnitude for site-
specific cancer in an epidemiologic study informs weight-of-evidence evaluations and provides
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
perspective on a study's validity and robustness of observations. Although statistical power
calculations are traditionally carried out during the design phase for sample size estimation,
examination of a study's statistical power post hoc is one of several tools to evaluate a study's
validity; however, such calculations must be interpreted in context of exposure conditions in the
study. Given the lower average exposure concentrations in the cohort studies and in population
case-control studies, an assumption of low relative risks is plausible. Approximate statistical
power to detect a relative risk of 2.0 with a = 0.05 was calculated for site-specific cancers in
cohort and geographic-based studies according to the methods of Beaumont and Breslow (1981),
as suggested by NRC (2006), and are found in Table B-4. Approximate statistical power was
calculated for kidney, NHL, and liver cancers as examples. Radican et al. (2008), the previously
follow-up of this cohort by Blair et al. (1998), and Raaschou-Nielsen et al. (2003) have over 80%
statistical power to detect relative risk of 2.0 for kidney and liver cancers and NHL and overall
TCE exposure. However, while these studies may appear sufficient for examining overall TCE
exposure and relative risks of 2.0, they have a greatly reduced ability to detect underlying risks
of this magnitude in analyses using rank-ordered exposure- or duration-response analyses. Other
studies with fewer TCE-exposed subjects and of similar or lower exposure conditions as Blair et
al. (1998) will decreased statistical power to detect most site-specific cancer risks of less than
2.0. Statistical power in Morgan et al. (2000a; 1998) and Boice et al. (1999) approaches that in
Blair et al. (1998) and Raaschou-Nielsen et al. (2003). As further identified in Table B-4,
Garabrant et al. (1988) and Morgan and Cassady (2002) each had over 80% statistical power to
detect relative risks of 2.0 for liver and kidney cancer and reflects the number of subjects in each
of these studies. However, underlying risk in both studies and other studies such as these which
lack characterization of TCE exposure to individual subjects is likely lower than 2.0 because of
inclusion of subjects with varying exposure potential, including low exposure potential. Case-
control studies such as Charbotel et al. (2006) and Briining et al. (2003) examine higher level
exposure to TCE than average exposure in the population case-control studies, and although
these two studies contain fewer subjects than population case-control studies such as Cocco et al.
(2010), a higher statistical power is expected related to the different and higher exposure
conditions and to the higher prevalence of exposure.
Overall, except for a few studies noted above, the body of evidence has limited statistical
power for evaluating low level cancer risk and trichloroethylene. For this reason, studies
reporting statistically significant association between trichloroethylene and site-specific cancer
are noteworthy if positive biases such as confounding are minimal.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.2.8. Statistical Analysis and Result Documentation
Appropriate analysis approaches characterize most cohort and case-control studies on
trichloroethylene cancer. Many studies clearly documented statistical analyses, evaluated
possible confounding factors, and included an examination of exposure-response. In
occupational cohort studies, potential confounding factors other than age, sex, race, and calendar
year are, generally, not evaluated. Expected numbers of outcomes (deaths or incident cancers)
were calculated using life table analysis and an external comparison group, national or regional
population mortality or incidence rates (Anttila et al., 1995; ATSDR, 2004; Axelson et al., 1994;
Blair et al., 1989; Blair et al., 1998; Boice et al., 1999; 2006b; Chang et al., 2003; Chang et al.,
2005; Costa et al., 1989; Garabrant et al., 1988; Henschler et al., 1995; Morgan et al., 1998;
Raaschou-Nielsen et al., 2003; Shannon et al., 1988; Shindell and Ulrich, 1985; Sinks et al.,
1992; Sung et al., 2007). Risk ratios are also presented in some cohort studies using proportional
hazard and logistic regression statistical methods using mortality or incidence rates of non-TCE
exposed cohort subjects as referent or internal controls (Blair et al., 1998; Boice et al., 1999;
Radican et al., 2008; Ritz, 1999a). Use of a non-TCE exposed referent group employed at the
same facility as exposed generally reduces downward bias or bias having potential associations
masked by a healthy worker work or other factors such as smoking that may be more similar
within an occupational cohort than between the cohort and the general population. However, the
advantage is minimized if subjects with lower TCE exposure potential are included in the
referent group as in Boice et al. (2006b). One referent group (the SSFL group) of Boice et al.
(2006b) included individuals with low TCE potential, a treatment different from the overlapping
study of Zhao et al. (2005) whose exposure assessment adopted a semi-quantitative approach,
grouping subjects identified with low TCE exposure potential separately from subjects with no
TCE exposure potential. A second referent group of all Rocketdyne workers in Boice et al.
(2006b) for whom TCE exposure potential was not examined may, also, have potential for
greater than background exposure since TCE use was widespread and rocket engine cleaning
occurred at other locations besides at test sites (Morgenstern, 1998). The inclusion of
nonexposed subjects in the low exposure group can obscure resultant associations due to
misclassification bias (Stewart et al., 1991).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-46 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
Exposure group
NHL
Kidney
Liver
Reference
Cohort studies—incidence
Aerospace workers (Rocketdyne)
Zhao et al. (2005)

Any exposure to TCE
Not reported
Not reported
Not reported


Low cumulative TCE score
Referent
Referent
Referent


Medium cumulative TCE score
97.0
43.8
Not reported


High TCE score
58.2
18.7
Not reported

All employees at electronics factory (Taiwan)
Chang et al. (2005)

Males
Not reported
Not reported
16.9


Females
Not reported
92. la
15.4

Danish blue-collar worker with TCE exposure
Raaschou-Nielsen et al. (2003)

Any exposure, all subjects
100.0
100.0
100.0


Employment duration, males





<1 yr
98.4
96.6
85.2


1-4.9 yrs
99.4
98.4
92.7


>5 yrs
97.7
97.0
93.1


Employment duration, females





<1 yr
40.3
30.1
27.3


1-4.9 yrs
48.4
37.1
34.1


>5 yrs
39.6
31.9
30.5

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE


to
s
a, 5
c> to
to s
Co
<3 Co'
s.	a
£
O	>r
S	vgs
£. >
s ?
^ 5
Oq 5.
TO
^ ^3
^3 5
S. ^
.TO Co
V; to
Co
0
s
1
a.
\o ^
o
to
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
Exposure group
NHL
Kidney
Liver
Reference
Biologically-monitored Danish workers
Hansen et al. (2001)

Any TCE exposure
37.9
47.9
35.7


Cumulative exposure (Ikeda)

Not reported
Not reported


<17 ppm-yr
17.9




>17 ppm-yr
20.3




Mean concentration (Ikeda)

Not reported
Not reported


<4 ppm
21.0




4+ ppm
23.6




Employment duration

Not reported
Not reported


<6.25 yr
18.3




>6.25
20.1



Aircraft maintenance workers from Hill Air Force Base
Blair et al. (1998)

TCE subcohort
Not reported
Not reported
Not reported


Males, cumulative exposure





0
Referent
Referent
Referent


<5 ppm-yr
79.5
67.8
58.2


5-25 ppm-yr
63.1
49.4
44.7


>25 ppm-yr
70.8
58.4
47.4

-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>s
to
to'
*
^S
«
8
TO
Co
0
S
1
a.
to
o
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
>3
o
o
s
Co
I?
Exposure group
NHL
Kidney
Liver
Reference

Females, cumulative exposure





0
Referent
Referent
Referent


<5 ppm-yr
28.2
0 cases
0 cases


5-25 ppm-yr
0 cases
0 cases
0 cases


>25 ppm-yr
34.1

0 cases

VO
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR =
(continued)
2
Exposure group
NHL
Kidney
Liver
Reference
Biologically-monitored Finnish workers
Anttila et al. (1995)

All subjects
53.8
70.4
56.5


Mean air-TCE (Ikeda extrapolation)





<6 ppm
36.8
Not reported
23.2


6+ ppm
25.6
Not reported
17.4

Cardboard manufacturing workers in Arnsberg, Germany
Henschler et al. (1995)

Exposed workers
Not reported
16.3
Not reported

Biologically-monitored Swedish workers
Axel son et al. (1994)

Any TCE exposure, males
43.5
59.6
40.1


Any TCE exposure, females
Not reported
Not reported
Not reported

Cardboard manufacturing workers, Atlanta area, GA
Sinks et al. (1992)

All subjects
Not reported
27.9
Not reported

Cohort studies—mortality
Aerospace workers (Rocketdyne)


Any TCE (utility/engine flush)
56.0
43.5
42.6
Boice et al. (2006b)

Any exposure to TCE
Not reported
Not reported
Not reported
Zhao et al. (2005)

Low cumulative TCE score
Referent
Referent
Referent


Medium cumulative TCE score
97.0
57.6
Not reported

-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>s
to
to'
*
^S
«
8
TO
Co
0
S
1
a.
to
o
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
>3
o
o
s
Co
I?
Exposure group
NHL
Kidney
Liver
Reference

High TCE score
55.4
26.4
Not reported

View-Master employees
AT SDR (2004)

Males
40.9
17.3
23.4

Females
74.1
24.1
0 deaths

VO
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
Exposure group
NHL
Kidney
Liver
Reference
All employees at electronics factory (Taiwan)
Chang et al. (2003)

Males
49.8
0 deaths
16.9


Females
79.0
37.5
15.4

United States uranium-processing workers (Fernald)
Ritz (1999a)

Any TCE exposure





Light TCE exposure, >2 yrs duration
91.6b
59.7°
10.1


Mod. TCE exposure, >2 yrs duration
20.9b
0 deaths0
0.08

Aerospace workers (Lockheed)
Boice et al. (1999)

Routine exposure
88.4
71.3
72.9


Duration of exposure, routine-intermittent





0 yrs
Referent
Referent
Referent


<1 yr
81.7
66.3
73.6


1-4 yrs
73.5
60.3
63.5


>5 yrs
78.5
63.8
67.3


p for trend




Aerospace workers (Hughes)
Morgan et al. (1998)

TCE subcohort
42.6, 79.6d
65.5
65.6


Low intensity (<50 ppm)
22.1
33.3
34.7

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
Exposure group
NHL
Kidney
Liver
Reference

High intensity (>50 ppm)
31.8
50.1
49.2

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
Exposure group
NHL
Kidney
Liver
Reference
Aircraft maintenance workers (Hill AFB, UT)
Blair et al. (1998)

TCE subcohort
92.7
81.5
87.9


Males, cumulative exposure


0





<5 ppm-yr
62.1
50.7
61.4


5-25 ppm-yr
43.1
37.1
44.7


>25 ppm-yr
54.8
44.9
52.8


Females, cumulative exposure


0





<5 ppm-yr
18.2
0 deaths
0 deaths


5-25 ppm-yr
0 deaths
8.4
0 deaths


>25 ppm-yr
22.0
11.5
19.1


TCE subcohort
99.9
94.4
99.7
Radican et al. (2008)

Males, cumulative exposure


0





<5 ppm-yr
83.0
43.8
59.4


5-25 ppm-yr
64.9
53.0
70.6


>25 ppm-yr
75.7
33.4
50.9

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR =
(continued)
2
Exposure group
NHL
Kidney
Liver
Reference

Females, cumulative exposure

0




<5 ppm-yr
38.9
0 deaths
25.9

5-25 ppm-yr
0 deaths
12.4
0 deaths

>25 ppm-yr
49.2
21.1
32.2

Cardboard manufacturing workers in Arnsberg, Germany
Henschler et al. (1995)

TCE exposed workers
19.6b
16.0
Not reported

Cardboard manufacturing workers, Atlanta area, GA
45.3b
17.3
Not reported
Sinks et al. (1992)
Coast Guard employees (US)
Blair et al. (1989)

Marine inspectors
31.8
31.8
38.6

Aircraft manufacturing plant employees (Italy)
Costa et al. (1989)

All subjects
94. lb
Not reported
63.1

Aircraft manufacturing plant employees (San Diego, CA)
Garabrant et al. (1988)

All subjects
95.1s, 74.2f
90.9
77.9

Geographic based studies
Residents in two study areas in Endicott, NY
90.8
41.7
31.8
AT SDR (2006b)
Residents of 13 census tracts in Redlands, CA
100
100.0
98.7
Morgan and Cassady (2002)
Finnish residents
Vartiainen et al. (1993)
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
is*
to
s
a
to
s
3"
a
a,
>r
<§>
>s
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect an RR = 2
(continued)
Exposure group
NHL
Kidney
Liver
Reference

Residents of Hausjarvi
98.8
Not reported
84.2

Residents of Huttula
98.7
Not reported
83.2

>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
aKidney cancer and other urinary organs, excluding bladder, as reported in Sung et al. (2008).
bAll cancers of hematopoietic and lymphatic tissues.
°Bladder and kidney cancer, as reported inNRC (2006).
dBased on number of observed cases of NHL reported in Mandel et al. (2006).
"Lymphosarcoma and reticulosarcoma.
fOther lymphatic and hematopoietic tissue neoplasms.

vo
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Cohort studies additionally evaluate a limited number of other factors associated with
employment which could be easily obtained from company and other records such as hire date,
time since first employment, socioeconomic status or pay status, and termination date (Boice et
al., 1999; 2006b; Greenland et al., 1994; Zhao et al., 2005), and three studies (Boice et al.,
2006b; Ritz, 1999a; Zhao et al., 2005) included a limited evaluation of smoking using
information collected by survey on smoking patterns from a subgroup of subjects. Neither
analysis of Morgan et al. (1998) nor Zhao et al. (2005) control for race, although Morgan et al.
(1998)	stated that "data concerning race were too sparse to use." The direction of any bias
introduced depends on proportion of nonwhites in the referent (internal) group compared to
TCE-exposed and on differences between racial groups in site-specific cancer incidence and
mortality rates. Blair et al. (1998), furthermore, presumed all subjects of unknown race were
white, an assumption with little associated error as shown later by Radican et al. (2008) whose
relative risk estimates were adjusted for race in follow-up analysis of this cohort.
The case-control studies on trichloroethylene are better able than cohort studies to
evaluate other possible confounders besides age and sex using logistic regression approaches
since such information can be obtained directly through interview and questionnaires. The case-
control studies of Hardell et al. (1994), Nordstrom et al. (1998) and Persson and Fredriksson
(1999)	lack evaluation of possible confounding factors other than age, sex and other
demographic information used to match control subjects to case subjects. Renal cell carcinoma
(RCC) case-control studies included evaluation of suggested risk factors for RCC such as
smoking (Briining et al., 2003; Charbotel et al., 2006; Pesch et al., 2000a; Siemiatycki, 1991;
Vamvakas et al., 1998), weight, or obesity (Charbotel et al., 2006; Dosemeci et al., 1999), and
diuretics (Dosemeci et al., 1999; Vamvakas et al., 1998). Moore et al. (2010) examined the
effect on renal cell carcinoma by smoking in univariate analyses and reported a change in their
odds ratio of less than 10% compared to that for TCE and renal cell carcinoma. They concluded
that smoking was not a confounder of the observed association with TCE. NHL and childhood
leukemia case-control studies included evaluation and control for possible confounding due to
smoking (Costas et al., 2002; Seidler et al., 2007; Siemiatycki, 1991), alcohol consumption
(Costas et al., 2002; Seidler et al., 2007), education (Costantini et al., 2008; Miligi et al., 2006),
although etiological factors for these cancers are not well identified other than a suggestion of a
role of immune function and some infectious agents in NHL (Alexander et al., 2007b). Smoking
was not controlled in other NHL case-control studies; however, neither smoking nor alcohol is a
strong risk factor for NHL (Besson et al., 2006; Morton et al., 2005).
Mineral oils such as cutting fluids or hydrazine common to some job titles with potential
TCE exposure as machinists, metal workers, and test stand mechanics are included as covariates
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
in statistical analyses of Zhao et al. (2005), Boice et al. (2006b) and Charbotel et al. (2006; 2009)
or evalauated as a single exposure for cases and controls in Moore et al. (Karami et al., 2011;
2010).	Two other kidney case-control studies of TCE exposure examined the effect of cutting oil
as a single occupational exposure on kidney cancer risk (Briining et al., 2003; Karami et al.,
2011).	In Briining et al. (2003), cutting oil exposure did not appear highly correlated with TCE
exposure as only 5 cases reported exposure to cutting oils compared to 25 cases reporting TCE
exposure. Karami et al. (Karami et al., 2011), who examined mineral oil or cutting fluid
exposure among cases and controls in Moore et al. (2010), reported an odds ratio of 0.8 (95% CI:
0.6, 1,1) and 1.1 (95% CI: 0.8, 1.4), for cutting oil mists or other mineral oil mists respectively,
and provides little evidence for confounding in Moore et al. (2010) by cutting or mineral oil
exposures. Moreover, cutting oils and mineral oils have not been associated with kidney cancer
in other cohort or case-control studies (Mirer, 2010; NIOSH, 1998). In all other studies,
exposure to cutting oils or to hydrazine did not greatly affect magnitude of risk estimates for
TCE exposure.
Geographical studies do not examine possible confounding factors other than sex, age
and calendar year. These studies are generally health surveys using publically-available records
such as death certificates and lack information on other risk factors such as smoking and
exposure to viruses, important to Lee et al. (2003), introduces uncertainties for informing
evaluations of trichloroethylene and cancer.
B.2.9. Systematic Review for Identifying Cancer Hazards and Trichloroethylene (TCE)
Exposure
The epidemiological studies on cancer and trichloroethylene are reviewed systematically
and transparently using criteria to identify studies for meta-analysis. Section B.3 contains a
description of and comment on 79 studies of varying qualities for identifying cancer hazard, a
question complementary but separate from that examined using meta-analysis. This section
identifies of the studies reviewed, studies in which there is a high likelihood of TCE exposure in
individual study subjects (e.g., based on job-exposure matrices, biomarker monitoring, or
industrial hygiene data indicating a high probability of TCE use) and were judged to have met
the inclusion criteria identified below. Lack of inclusion of an individual study in the meta-
analysis does not necessarily imply an inability to identify cancer hazard. Not all questions
associated with identifying a cancer hazard are addressed using meta-analyses and the 79 studies
with varying abilities approached, to sufficient degrees, the standards of epidemiologic design
and analysis, identified in the beginning of Section B.2.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-59 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
The NRC (2006) suggested U.S. EPA conduct a new meta-analysis of the epidemiologic
data on trichloroethylene to synthesize the epidemiologic data on TCE exposure. Meta-analysis
approaches are feasible for examining cancers of the liver, kidney, and NHL given most studies
presented risks for these sites in their published papers and these cancer sites are of interest given
observations in the animal studies. Examination of site-specific cancers other than kidney
cancer, liver cancer, and NHL, such as for childhood leukemia, bladder cancer, esophageal
cancer, or cervical cancer is more difficult and not recommended due to fewer available high-
quality studies. NRC (2006) specifically suggested EPA to:
1.	Document essential design features, exposure, and results from the epidemiologic
studies—Information on study design, exposure assessment approach, statistical
analysis, and other aspects important to interpreting observations in a weight of
evidence evaluation for individual studies is found in Section B.3. and
site-specific estimated relative risks or measures of association are presented in
Section 4;
2.	Analyze the epidemiologic studies to discriminate the amount of exposure
experience by the study population; exclude studies in meta-analysis based on
objective criteria (e.g., studies in which it was unclear that the study population
was exposed)—Appendix B.3. describes exposure assessment approach for
individual studies and inclusion criteria for identifying studies for meta-analysis
are identified below;
3.	Classify studies in terms of objective characteristics, such as on the basis of the
study's design characteristics or documentation of exposure —Section B.3.
groups studies by study design, analytical designs and geographic-based designs,
with discussion of factors important to study design, endpoint measured, exposure
assessment approach, study size, and statistical analysis methods including
adjustment for potential confounding exposures;
4.	Assess statistical power of each study—Table B.3 presents power calculations for
cohort studies;
5.	Combine case-control and cohort studies in the analysis, unless it introduces
substantial heterogeneity—Appendix C discusses the meta-analysis statistical
methods and findings;
6.	Testing of heterogeneity (e.g., fixed or random effect models)—Appendix C
discusses the meta-analysis statistical methods and findings;
7.	Perform a sensitivity analysis in which each study is excluded from the analysis to
determine whether any study significantly influences the finding—Appendix C
discusses the meta-analysis statistical methods and findings.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Studies selected for inclusion in the meta-analysis met the following criteria: (1) cohort
or case-control designs; (2) evaluation of incidence or mortality; (3) adequate selection in cohort
studies of exposure and control groups and of cases and controls in case-control studies; (4) TCE
exposure potential inferred to each subject and quantitative assessment of TCE exposure for each
subject by reference to industrial hygiene records indicating a high probability of TCE use,
individual biomarkersjob exposure matrices, water distribution models, or obtained from
subjects using questionnaire (case-control studies); (5) relative risk estimates for kidney cancer,
liver cancer, or NHL adjusted, at minimum, for possible confounding of age, sex, and race.
Table B-5 in Section B.2.9.4 identifies studies included in the meta-analysis and studies that did
not meet the inclusion criteria and the primary reasons for their deficiencies.
B.2.9.1. Cohort Studies
The cohort studies (Anttila et al., 1995; Axelson et al., 1994; Blair et al., 1989; Blair et
al., 1998; Boice et al., 1999; Boice et al., 2006b; Chang et al., 2003; Chang et al., 2005; Costa et
al., 1989; Garabrant et al., 1988; Greenland et al., 1994; Hansen et al., 2001; Henschler et al.,
1995; Krishnadasan et al., 2007; Morgan et al., 1998; Raaschou-Nielsen et al., 2003; Radican et
al., 2008; Ritz, 1999a; Shannon et al., 1988; Shindell and Ulrich, 1985; Sinks et al., 1992; Sung
et al., 2007; Sung et al., 2008; Wilcosky et al., 1984; Zhao et al., 2005), with data on the
incidence or morality of site-specific cancer in relation to trichloroethylene exposure range in
size (803 (Hansen et al., 2001) to 86,868 (Chang et al., 2003; Chang et al., 2005)), and were
conducted in Denmark, Sweden, Finland, Germany, Taiwan and the United States (see
Table B-l). Three case-control studies nested within cohorts (Greenland et al., 1994;
Krishnadasan et al., 2007; Wilcosky et al., 1984) are considered as cohort studies because the
summary risk estimate from a nested case-control study, the odds ratio, was estimated from
incidence density sampling and is considered an unbiased estimate of the hazard ratio, similar to
a relative risk estimate from a cohort study. Two studies of deaths within a cohort were included
in the group, but these studies lacked information on the person-year structure; i.e., both are
proportionate mortality ratio studies, and did not satisfy the meta-analysis inclusion criteria for
analytical study design (ATSDR, 2004; Clapp and Hoffman, 2008).
Cohort and nested case-control study designs are analytical epidemiologic studies and are
generally relied on for identifying a causal association between human exposure and adverse
health effects (Zhou et al., 2003). Some subjects in the Hansen et al. study are also included in a
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-61 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
study reported by Raaschou-Nielsen et al. (2003); however, any contribution from the former to
the latter are minimal given the large differences in cohort sizes of these studies (Hansen et al.,
2001; Raaschou-Nielsen et al., 2003). Similarly, some females in Chang et al. (2003; 2005), a
large cohort of 70,735 female and 16,133 male subjects, are included in Sung et al. (2007), a
cohort of 63,982 female electronic workers from the same factory who were followed an
additional 4-year period than subjects in Chang et al. (2003; 2005). Cancer observations for
female subjects in these studies are considered as equivalent since they are derived from
essentially the same population. Krishnadasan et al. (2007) is a nested case-control study of
prostate cancer with cases and controls drawn from subjects in a large cohort of aerospace
workers as subjects in Zhao et al. (2005), who did not report on prostate cancer, and met all the
inclusion criteria except that for reporting a relative risk estimate for cancer of the kidney, liver
or NHL.
Eleven of the cohort studies met all five inclusion criteria: the cohorts of Blair et al.
(1998) and its further follow-up by Radican et al. (2008), Morgan et al. (1998), Boice et al.
(1999; 2006b) and Zhao et al. (2005) of aerospace workers or aircraft mechanics; Axelson et al.
(1994), Anttila et al. (1995), Hansen et al. (2001), and Raaschou-Nielsen et al. (2003) of Nordic
workers in multiple industries with TCE exposure; and Greenland et al. (1994) of electrical
manufacturing workers. All eleven cohort studies adopted statistical methods, e.g., life table
analysis, Poisson regression analysis, or Cox Proportional Hazard analysis, that met
epidemiologic standards, and were able to control for age, race, sex, and calendar time trends in
cancer rates. Statistical analyses in Boice et al. (1999) adjusted for demographic variable such as
age, race, and sex, and, also, included date of first employment and terminating date of
employments, which may have decreased the statistical power of their analyses due to colinearity
between age, first and last employment dates. Statistical analyses in Zhao et al. (2005) and
Boice et al. (2006b) adjusted for potential effects by other occupational exposures on cancer and
both Raaschou-Nielsen et al. (2003) and Zhao et al. (2005) examined possible confounding by
smoking on TCE exposure and cancer risks using indirect approaches.
Of the eleven studies, two studies reported risk estimates for both site-specific cancer
incidence and mortality (Blair et al., 1998; Zhao et al., 2005) four studies reported risk estimates
for cancer incidence only (Anttila et al., 1995; Axelson et al., 1994; Hansen et al., 2001;
Krishnadasan et al., 2007; Raaschou-Nielsen et al., 2003) and four studies reported risk estimates
for mortality only (Boice et al., 1999; 2006b; Morgan et al., 1998; Radican et al., 2008).
Incidence ascertainment in two cohorts began 21 (Blair et al., 1998) and 38 years (Zhao et al.,
2005) after the inception of the cohort. Specifically, Zhao et al. (2005) note "results may not
accurately reflect the effects of carcinogenic exposure that resulted in nonfatal cancers before
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-62 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1988." Because of the issues concerning case ascertainment raised by this incomplete coverage,
incidence observations must be interpreted in light of possible bias reflecting incomplete
ascertainment of incident cases. Furthermore, use of an internal referent population, nonexposed
subjects drawn from the same or near-by facilities as exposed workers, in Blair et al. (1998) and
Radican et al. (2008) for overall TCE exposure, and in Blair et al. (1998), Morgan et al. (1998),
Boice et al. (1999), Zhao et al. (2005), Boice et al. (2006b), and Radican et al. (2008) for rank-
ordered TCE exposure is expected to reduce bias associated with the healthy worker effect.
Morgan et al. (1998) presents risk estimates for overall TCE exposure comparing mortality in
their TCE subcohort to that expected using mortality rate of the U.S. population in an
Environmental Health Strategies Final Report and sent to U.S. EPA by Paul Cammer, Ph.D., on
behalf of the Trichloroethylene Issues Group (EHS, 1997). The final report also contained risk
estimates from internal analyses of rank-order TCE exposure and published as Morgan et al.
(1998). Both internal cohort analyses of the rank-ordered exposure, presented in both the final
report of Environmental Health Strategies (1997) and Morgan et al. (1998), and overall TCE
exposure, available in the final report or upon request, are based on the same group of internal
referents, nonexposed TCE subjects employed at the same facility.
Subjects in these studies had a high likelihood or potential for TCE exposure, although
estimated average exposure intensity for overall TCE exposure in some cohorts was considered
as less than 10 or 20 ppm (time-weighted average). The exposure assessment techniques used in
these cohort studies included a detailed job-exposure matrix (Blair et al., 1998; Greenland et al.,
1994); its follow-up by Radican et al. (Boice et al., 1999; Boice et al., 2006b; Morgan et al.,
1998; 2008; Zhao et al., 2005); Radican et al. (2008), biomonitoring data (Anttila et al., 1995;
Axelson et al., 1994; Hansen et al., 2001), or use of industrial hygiene data on TCE exposure
patterns and factors that affect such exposure (Raaschou-Nielsen et al., 2003), with high
probability of TCE exposure potential to individual subjects. The job-exposure matrix in six
studies provided rank-ordered surrogate metrics for TCE exposure (Anttila et al., 1995; Axelson
et al., 1994; Blair et al., 1998; Hansen et al., 2001) and its follow-up by Radican et al. (2008;
Zhao et al., 2005), a strength compared to use of duration of employment as an exposure
surrogate, e.g., Boice et al. (1999; 2006b) or Raaschou-Nielsen et al. (2003), which is a poorer
exposure metric given subjects may have differing exposure intensity with similar exposure
duration (NRC, 2006). Rank-ordered TCE dose surrogates for low and medium exposure from
the job-exposure matrix of Morgan et al. (1998) are uncertain because of a lack on information
on frequency of exposure-related tasks and on temporal changes (NRC, 2006); only the high
category for TCE exposure is unambiguous. The nested case-control study of Greenland et al.
(1994) examined TCE as one of seven exposures and potential assigned to individual cases and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
controls using a job-exposure-matrix approach. However, the low exposure prevalence, missing
job history information for 34% of eligible subjects, and study of pensioned workers only were
other factors judged to lower this study's sensitivity for cancer hazard identification.
The remaining cohort studies (Blair et al., 1989; Chang et al., 2003; Chang et al., 2005;
Costa et al., 1989; Garabrant et al., 1988; Henschler et al., 1995; Ritz, 1999a; Shannon et al.,
1988; Shindell and Ulrich, 1985; Sinks et al., 1992; Wilcosky et al., 1984) Sung et al. (2007;
Sung et al., 2008) less satisfactorily meet inclusion criteria. These studies, while not meeting the
meta-analysis inclusion criteria, can inform the hazard analysis although their findings are
weighted less than for observations in the other studies, and observations may have alternative
causes. Reasons for study insufficiencies varied. Nine studies do not assign TCE exposure
potential to individual subjects (ATSDR, 2004; Chang et al., 2003; Chang et al., 2005; Clapp and
Hoffman, 2008; Costa et al., 1989; Garabrant et al., 1988; Shindell and Ulrich, 1985; Sinks et al.,
1992; Sung et al., 2007; Sung et al., 2008) all subjects are presumed as "exposed" because of
employment in the plant or facility although individual subjects would be expected to have
differing exposure potentials.
TCE exposure potential is ambiguous in both Wilcosky et al. (1984) and Ritz (1999a),
two studies of low potential, low intensity TCE exposure compared to studies using exposure
assessment approaches supported by information on job titles, tasks, and industrial hygiene
monitoring data. Furthermore, high correlation in Ritz (1999a) between TCE and other
exposures, particularly cutting fluids and radiation, may not have been sufficiently controlled in
statistical analyses. Ritz et al. (1999a), furthermore, did not report estimated relative risks for
kidney or NHL separately; rather, presenting relative risk estimates for kidney and bladder
cancer combined and for all hemato- and lymphopoietic cancers.
Two studies do not sufficiently define the underlying cohort or there is uncertainty in
cancer case or death ascertainment (Henschler et al., 1995; Shindell and Ulrich, 1985).
Furthermore, magnitude of observed risk in Henschler et al. (1995), ATSDR (2004), and Clapp
and Hoffman (2008) must be interpreted in a weight-of-evidence evaluation in light of possible
bias introduced through use of analysis of proportion of deaths (proportionate mortality ratio) in
ATSDR (2004) and Clapp and Hoffman (2008), or to inclusion of index kidney cancer cases in
Henschler et al. (1995).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-64 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.2.9.2. Case-Control Studies
Case-control studies on TCE exposure are of several site-specific cancers and include
bladder (Pesch et al., 2000a; Siemiatycki, 1991; Siemiatycki et al., 1994), brain (De Roos et al.,
2001a; Heineman et al., 1994) childhood lymphoma or leukemia (Costas et al., 2002; Lowengart
et al., 1987; McKinney et al., 1991; Shu et al., 2004; Shu et al., 1999); colon cancer (Goldberg et
al., 2001; Siemiatycki, 1991); esophageal cancer (Parent et al., 2000a; Siemiatycki, 1991); liver
cancer (Lee et al., 2003) lung (Siemiatycki, 1991), lymphoma (Hardell et al., 1994)[NHL,
Hodgkin lymphoma]; (Fritschi and Siemiatycki, 1996a; Nordstrom et al., 1998; Siemiatycki,
1991); [hairy cell leukemia]; (Persson and Fredrikson, 1999) [NHL]; (Miligi et al., 2006) [NHL
and chronic lymphocytic leukemia (CLL)]; (Seidler et al., 2007)[NHL, Hodgkin lymphoma];
(Costantini et al., 2008) [leukemia types, CLL included in Miligi et al. (2006); Wang et al.
(2009) [NHL]; (Cocco et al., 2010) [NHL, CLL, MM]; (Gold et al., In Press) [MM]; Purdue et
al. (2011) [NHL]; melanoma (Fritschi and Siemiatycki, 1996a; Siemiatycki, 1991); rectal cancer
(Dumas et al., 2000; Siemiatycki, 1991); renal cell carcinoma, a form of kidney cancer (Briining
et al., 2003; Charbotel et al., 2006; Charbotel et al., 2009; Dosemeci et al., 1999; Moore et al.,
2010; Parent et al., 2000a; Pesch et al., 2000b; Siemiatycki, 1991; Vamvakas et al., 1998);
pancreatic cancer (Siemiatycki, 1991); and prostate cancer (Aronson et al., 1996; Siemiatycki,
1991). No case-control studies of reproductive cancers (breast or cervix) and TCE exposure
were found in the peer-reviewed literature.
Several of the above publications are studies of cases and controls drawn from the same
underlying population with a common control series. Miligi et al. (2006) and Costantini et al.
(2008) presented observations from the Italian multicenter lymphoma population case-control
study; Miligi et al. (2006) on occupation or specific solvent exposures and NHL, and who also
included CLL and Hodgkin's lymphoma in the overall NHL category, and Costantini et al.
(2008) who examined leukemia subtypes, and included CLL as a separate disease outcome.
Seidler et al. (2007)analyzed independently the German subjects of the six European country,
multicenter lymphoma population case-control study (EPILYMPH study) of Cocco et al. (2010).
Each study adopted a different approach to calculate cumulative exposure and apparent
inconsistency in their conclusions may reflect the slightly different ranking of cases and controls
in each study (personnal communication from Pierluigi Cocco to Cheryl Siegel Scott). Gold et
al. and Purdue et al. (2011) presented observations from the NCI-SEER population case-control
studies and share a common control series; Purdue et al. (2011) of NHL in four SEER reporting
areas and Gold et al. of muliplte myeloma in two of the four SEER sites. Pesch et al. (2000a;
2000b) a multiple center population case- control study of urothelial cancers in Germany,
presented observations on TCE and bladder cancer, including cancer of the ureter and renal
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-65 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
pelvis, in Pesch et al. (2000a) and renal cell carcinoma in Pesch et al. (2000b). Siemiatycki
(1991), a case-control of occupational exposures and several site-specific cancers (bladder,
colon, esophagus, lung, rectum, pancreas, and prostate) and designed to generate hypotheses
about possible occupational carcinogens, presents risk estimates associated with TCE exposure
using Mantel-Haentszel methods. Subsequent publications examine either TCE exposure
(analyses of melanoma and colon cancers) or job title/occupation (all other cancer sites) using
logistic regression methods (Aronson et al., 1996; Dumas et al., 2000; Fritschi and Siemiatycki,
1996a, b; Goldberg et al., 2001; Parent et al., 2000a; Siemiatycki et al., 1994).
The population case-control studies with data on cancer incidence (Siemiatycki,
1991)[and related publications, (Aronson et al., 1996; Briining et al., 2003; Charbotel et al.,
2006; Charbotel et al., 2009; Cocco et al., 2010; Costantini et al., 2008; Costas et al., 2002; De
Roos et al., 2001a; Dosemeci et al., 1999; Dumas et al., 2000; Fritschi and Siemiatycki, 1996a, b;
Gold et al., In Press; Goldberg et al., 2001; Hardell et al., 1994; Kernan et al., 1999; Lowengart
et al., 1987; McKinney et al., 1991; Miligi et al., 2006; Moore et al., 2010; Nordstrom et al.,
1998; Parent et al., 2000a; Persson and Fredrikson, 1999; Pesch et al., 2000a; Pesch et al., 2000b;
Purdue et al., 2011; Seidler et al., 2007; Shu et al., 2004; Siemiatycki et al., 1994; Vamvakas et
al., 1998; Wang et al., 2009);(Briining et al., 2003; Charbotel et al., 2006; Charbotel et al., 2009;
Cocco et al., 2010; Costantini et al., 2008; Costas et al., 2002; De Roos et al., 2001; Dosemeci et
al., 1999; Gold et al., In Press; Goldberg et al., 2001; Hardell et al., 1994; Kernan et al., 1999;
Lowengart et al., 1987; McKinney et al., 1991; Miligi et al., 2006; Moore et al., 2010; Nordstrom
et al., 1998; Parent et al., 2000a; Persson and Fredrikson, 1999; Pesch et al., 2000a; Pesch et al.,
2000b; Purdue et al., 2011; Seidler et al., 2007; Shu et al., 2004; Siemiatycki et al., 1994;
Vamvakas et al., 1998; Wang et al., 2009) or mortality (Heineman et al., 1994; Lee et al., 2003)
in relation to trichloroethylene exposure range in size, from small studies with less than 100
cases and control (Costas et al., 2002) to multiple-center studies large-scale studies of over 2,000
cases and controls (Costantini et al., 2008; Miligi et al., 2006; Pesch et al., 2000a; Pesch et al.,
2000b; Shu et al., 2004; Shu et al., 1999), and were conducted in Sweden, Germany, Italy,
Taiwan, Canada and the United States (see Table B-2).
Fifteen of the case-control studies met the meta-analysis inclusion criteria identified in
Section B.2.9 (Briining et al., 2003; Charbotel et al., 2006; Charbotel et al., 2009; Cocco et al.,
2010; Dosemeci et al., 1999; Hardell et al., 1994; Miligi et al., 2006; Moore et al., 2010;
Nordstrom et al., 1998; Persson and Fredrikson, 1999; Pesch et al., 2000a; Purdue et al., 2011;
Seidler et al., 2007; Siemiatycki, 1991; Wang et al., 2009). They were of analytical study
design, cases and controls were considered to represent underlying populations and selected with
minimal potential for bias; exposure assessment approaches included assignment of TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-66 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
exposure potential to individual subjects using information obtained from face-to- face, mailed,
or telephone interviews; analyses methods were appropriate, well-documented, included
adjustment for potential confounding exposures, with relative risk estimates and associated
confidence intervals reported for kidney cancer, liver cancer or NHL. All fifteen studies
evaluated TCE exposure potential to individual cases and controls and a structured questionnaire
sought information on self-reported occupational history and specific exposures such as TCE.
Three studies assigned TCE exposure potential to cases and controls using self-reported
information (Hardell et al., 1994; Nordstrom et al., 1998) and two of these studies used judgment
to assign potential exposure intensity (Nordstrom et al., 1998; Persson and Fredrikson, 1999).
Persson and Fredriksson (1999) also assigned TCE exposure potential from both occupational
and leisure use, the only study to do so. The twelveother studies assigned TCE exposure
potential using self-reported job title and occupational history, a superior approach compared to
use of a job exposure matrix (JEM) supported by expert judgment and information on only self-
reported information given its expect greater specificity (Briining et al., 2003; Charbotel et al.,
2006; Charbotel et al., 2009; Cocco et al., 2010; Dosemeci et al., 1999; Miligi et al., 2006;
Moore et al., 2010; Pesch et al., 2000b; Purdue et al., 2011; Seidler et al., 2007; Siemiatycki,
1991; Wang et al., 2009). Pesch et al. (2000b) assigned TCE exposure potential using both job
exposure matrix and job-task exposure matrix (JTEM). The inclusion of task information is
considered superior to exposure assignment using only job title since it likely reduces potential
misclassification and, for this reason, relative risk estimates in Pesch et al. (2000b) for TCE from
a JTEM are preferred. All studies except Hardell et al. (1994) and Dosemeci et al. (1999)
developed a semiquantitative or quantitative TCE exposure surrogate.
These studies to varying degrees were considered as stronger studies for weight-of
evidence characterization of hazard. Both Briining et al. (2003) and Charbotel et al. (2006),
(2009) had a priori hypotheses for examining renal cell carcinoma and TCE exposure. Strengths
of both studies are in their examination of populations with potential for high exposure intensity
and in areas with high frequency of TCE usage and their assessment of TCE potential. An
important feature of the exposure assessment approach of Charbotel et al. (2006) is their use of a
large number of studies on biological monitoring of workers in the screw-cutting industry a
predominant industry with documented TCE exposures as support. The other studies were either
large multiple-center studies (Cocco et al., 2010; Miligi et al., 2006; Moore et al., 2010; Pesch et
al., 2000a; Pesch et al., 2000b; Purdue et al., 2011; Wang et al., 2009); or reporting from one
location of a larger international study (Dosemeci et al., 1999; Seidler et al., 2007). In contrast to
Briining et al. (2003) and Charbotel et al. (2006; 2009), two studies conducted in geographical
areas with widespread TCE usage and potential for exposure to higher intensity, a lower
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-67 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
exposure prevalence to TCE is found [any TCE exposure: 15% of cases (Dosemeci et al., 1999);
6% of cases (Miligi et al., 2006); 13% of cases (Seidler et al., 2007); 13% of cases (Wang et al.,
2009)] and most subjects identified as exposed to TCE probably had minimal contact [3% of
cases with moderate/high TCE exposure (Miligi et al., 2006); 1% of cases with high cumulative
TCE (Seidler et al., 2007); 2% of cases with high intensity, but of low probability TCE exposure
(Wang et al., 2009)]. This pattern of lower exposure prevalence and intensity is common to
community-based population case-control studies (Teschke et al., 2002).
Fifteen case-control studies did not meet specific inclusion criterion (Costantini et al.,
2008; Costas et al., 2002; Dumas et al., 2000; Fritschi and Siemiatycki, 1996b; Gold et al., In
Press; Goldberg et al., 2001; Kernan et al., 1999; Lee et al., 2003; Parent et al., 2000a; Pesch et
al., 2000a; Shu et al., 2004; Shu et al., 1999; Siemiatycki, 1991; Vamvakas et al., 1998).
Costantini et al. (2008) and Gold et al. examined multiple myeloma or leukemias, not included
in a older NHL classification schemes, although these neoplasms are now considered as
lymphoams under the World Health Organiziation Lymphoma Classification. Vamvakas et al.
(1998) has been subject of considerable controversy (Bloemen and Tomenson, 1995; Cherrie et
al., 2001; Green and Lash, 1999; Mandel, 2001; McLaughlin and Blot, 1997; Swaen, 1995) with
questions raised on potential for selection bias related to the study's controls. This study was
deficient in the criterion for adequacy of case and control selection. Briining et al. (2003), a
study from the same region as Vamvakas et al. (1998), is considered a stronger study for
identifying cancer hazard since it addresses many of the deficiencies of Vamvakas et al. (1998).
Lee et al. (2003) in their study of hepatocellular cancer assigns one level of exposure to all
subjects in a geographic area, and inherent measurement error and misclassification bias because
not all subjects are exposed uniformly. Additionally, statistical analyses in this study did not
control for hepatitis viral infection, a known risk factor for hepatocellular cancer and of high
prevalence in the study area, Ten of twelve studies reported relative risk estimates for site-
specific cancers other than kidney, liver, and NHL (Aronson et al., 1996; Costas et al., 2002;
Dumas et al., 2000; Fritschi and Siemiatycki, 1996b; Garabrant et al., 1988; Goldberg et al.,
2001; Kernan et al., 1999; Parent et al., 2000a; Pesch et al., 2000a; Shu et al., 2004; Shu et al.,
1999; Siemiatycki etal., 1994).
B.2.9.3. Geographic-Based Studies
The geographic-based studies (ADHS, 1990, 1995; Aickin, 2004; Aickin et al., 1992;
ATSDR, 2006c, 2008; Cohn et al., 1994a; Isacson et al., 1985; Mallin, 1990; Morgan and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-68 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Cassady, 2002; Vartiainen et al., 1993) with data on cancer incidence (all studies) are correlation
studies to examine cancer outcomes of residents living in communities with TCE and other
chemicals detected in groundwater wells or in municipal drinking water supplies. These eight
studies did not meet inclusion criteria and were deficient in a number of criteria.
All geographic-based studies are surveys of cancer rates for a defined time period among
residents in geographic areas with TCE contamination in groundwater or drinking water
supplies, or soil and are not of analytical designs such as cohort and case-control designs. A
major shortcoming in all studies is, also, their low level of detail to individual subjects for TCE
potential. The exposure surrogate is assigned to a community, town, or a geographically-defined
area such as a contiguous grouping of census tracts as an aggregate level, typically based on
limited number of water monitoring data from a recent time period and is a poor exposure
surrogate because potential for TCE exposure can vary in these broad categories depending on
job function, year, use of personal protection, and, for residential exposure, pollutant fate and
transport, water system distribution characteristics, percent of time per day in residence, presence
of mitigation devices, drinking water consumption rates, and showering times. Additionally,
ATSDR (2008), the only geographic-based study to examine other possible risk factors on
individual subjects, reported smoking patterns and occupational exposures may partly contribute
to the observed elevated rates of kidney and renal pelvis cancer and lung cancer in subjects living
in a community with contaminated groundwater and with TCE exposure potential from vapor
intrusion into residences.
B.2.9.4. Recommendation of Studies for Treatment Using Meta-Analysis Approaches
All studies are initially considered for inclusion in the meta-analysis; however, as
discussed through-out this section, some studies are better than others for inclusion in a
quantitative examination of cancer and trichloroethylene. Twenty-six of the studies included in
the meta-analysis (statistical methods and findings discussed in Appendix C) met the following
five inclusion criteria: (1) cohort or case-control designs; (2) evaluation of incidence or
mortality; (3) adequate selection in cohort studies of exposure and control groups and of cases
and controls in case-control studies; (4) TCE exposure potential inferred to each subject and
quantitative assessment of TCE exposure assessment for each subject by reference to industrial
hygiene records indicating a high probability of TCE use, individual biomarkers, job exposure
matrices, water distribution models, or obtained from subjects using questionnaire (case-control
studies); (5) relative risk estimates for kidney cancer, liver cancer, or NHL adjusted, at
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
minimum, for possible confounding of age, sex, and race. The twenty-six studies that met these
inclusion are: Siemiatycki (1991), Axelson et al. (1994), Greenland et al. (1994), Hardell et al.
(1994), Anttila et al. (1995), Blair et al. (1998), Morgan et al. (1998), Nordstrom et al. (1998),
Dosemeci et al. (1999), Boice et al. (1999; 2006b), Persson and Fredriksson (1999), Pesch et al.
(2000b), Hansen et al. (2001), Briining et al. (2003), Raaschou-Nielsen et al. (2003), Zhao et al.
(2005), Miligi et al. (2006), Charbotel et al. (2006; 2009), Seidler et al. (2007), Radican et al.
(2008), Wang et al. (2009), Cocco et al. (2010), Moore et al. (2010), and Purdue et al. (2011).
Table B-5 identifies studies included in the meta-analysis and studies that did not meet the
inclusion criteria and the primary reasons for their deficiencies.
There is some overlap between the cohorts of Zhao et al. (2005) and Boice et al. (2006b),
each cohort is identified from a population of workers, but these studies differ on cohort
definition, cohort identification dates, disease outcome examined, and exposure assessment
approach. Zhao et al. (2005) who adopted a semiquantitative approach for TCE exposure
assessment is preferred to Boice et al. (2006b), whose TCE subcohort included subjects with a
lower likelihood for TCE exposure and duration of exposure, a poor exposure metric given
subjects may have differing exposure intensity with similar exposure duration (NRC, 2006).
Additionally, a larger number of site-specific cancer deaths identified with potential TCE
exposure is observed by Zhao et al. (2005) compared to Boice et al. (2006b); e. g., 95 lung
cancer cases with medium or high TCE exposure (Zhao et al., 2005) and 51 lung cancer cases
with any TCE exposure (Boice et al., 2006b) (see further discussion in B.3.1.1.1.3.). Radican et
al. (2008) studied the same subjects as Blair et al. (1998), adding an additional 10 years of
follow-up and updating mortality. Observed site-specific cancer mortality risk estimates in
Radican et al. (2008) did not change appreciably and were consistent with those reported in Blair
et al. (1998) and is preferred. Blair et al. (1998) who also presented incidence relative risk
estimates is recommended for inclusion in sensitivity analyses. Cocco et al. (2010) is preferred to
Seidler et al. (2007) whose subjects are included in the larger multicenter population case-control
study.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3. INDIVIDUAL STUDY REVIEWS AND ABSTRACTS
B.3.1. Cohort Studies
B .3 .1.1. Studies of Aerospace Workers
Seven papers reported on cohort studies of aerospace or aircraft maintenance and
manufacturing workers in large facilities.
B.3.1.1.1. Studies of Santa Susanna Field Laboratory workers. Trichloroethylene exposure
to workers at Santa Susanna Field Laboratory (SSFL), an aerospace facility located nearby Los
Angeles, California, operated by Rocketdyne/Atomics International, formerly a division of
Boeing and currently owned by Pratt-Whitney, is subject of two research efforts: (1) the
University of California at Los Angeles (UCLA) study, overseen by the California Department
of Health Services and funded by the U.S. Department of Energy (DOE) (Morgenstern et al.,
1997, 1999; Ritz et al., 1999), with two publications on trichloroethylene exposure and cancer
incidence (Krishnadasan et al., 2007; Zhao et al., 2005a) and mortality (Zhao et al., 2005); and,
(2) the International Epidemiology Institute study (IEI), funded by Boeing after publication of
the initial UCLA reports, of all Rocketdyne employees which included a mortality analysis of
trichloroethylene exposure in a subcohort of SSFL test stand mechanics (Boice et al., 2006b). In
addition to chemical exposure, both groups examine radiation exposure and cancer among
Rocketdyne workers monitored for radiation (Boice et al., 2006a; Ritz et al., 2000).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-71 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1	Table B-5. Summary of rationale for study selection for meta-analysis
2
Decisio
n
Outcom
e
Studies
Primary reason(s)
Studies Recommended for Meta-analysis:

Siemiatycki (1991); Axelson et al.
(1994); Hardell (1994); Greenland
et al. (1994); Anttila et al. (1995);
Morgan et al. (1998); Nordstrom et
al. (1998); Boice et al. (1999;
2006b); Dosemeci et al., (1999);
Persson and Fredriksson, (1999);
Pesch et al. (2000a); Hansen et al.
(2001); Briining et al. (2003);
Raaschou-Nielsen et al. (2003);
Zhao et al. (2005); Miligi et al.
(2006); Seidler et al. (2007);
Charbotel et al., (2006; 2009);
Radican et al. (2008) [Blair et al.
(1998), incidence]; Wang et al.
(2009); Cocco et al. (2010); Moore
et al. (2010); Purdue et al. (2011)
Analytical study designs of cohort or case-control
approaches;
Evaluation of cancer incidence or cancer mortality;
Specifically identified TCE exposure potential to
individual study subjects by reference to industrial
hygiene records, individual biomarkers, job exposure
matrices, water distribution models, industrial
hygiene data indicating a high probability of TCE use
(cohort studies), or obtained information on TCE
exposure from subjects using questionnaire (case-
control studies);
Reported results for kidney cancer, liver cancer, or
NHL with relative risk estimates and corresponding
confidence intervals (or information to allow
calculation).
Studies Not Recommended for Meta-analysis:

AT SDR (2004); Clapp and
Hoffman, (2008); Cohn et al.
(1994a)
Weakness with respect to analytical study design (i.e.,
geographic-based, ecological or proportional
mortality ratio design)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-72 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE

Wilcosky et al. (1984); Isacson et
al. (1985); Shindell andUlrich
(1985); Garabrant et al. (1988);
Shannon et al.(1988); Blair et al.
(1989); Costa et al. (1989); ADHS
(1990, 1995); Mallin (1990); Aickin
et al. (1992); Sinks et al. (1992);
Vartiainen et al. (1993); Morgan
and Cassady (2002); Lee et al.
(2003); Aickin (2004); Chang et al.
(2003; 2005); Coyle et al. (2005);
AT SDR (2006b, 2008); Sung et al.
(2007; 2008)
TCE exposure potential not assigned to individual
subjects using job exposure matrix, individual
biomarkers, water distribution models, or industrial
hygiene data indicating a high probability of TCE use
(cohort studies)

Lowengart et al. (1987);
Fredriksson et al. (1989);
McKinney et al. (1991); Heineman
et al. (1994); Siemiatycki et al.
(1994); Aronson et al. (1996);
Fritchi and Siemiatycki (1996b);
Dumas et al. (2000); Kernan et
al.(1999); Shuetal. (2004; 1999);
Parent et al. (Parent et al., 2000a);
Pesch et al., (2000a); De Roos et al.
(2001a); Goldberg et al. (2001);
Costas et al. (2002); Krishnadasan
et al. (2007); Costantini et al.
(2008); Gold et al.
Cancer incidence or mortality reported for cancers
other than kidney, liver, or NHL

Ritz (1999a)
Subjects monitored for radiation exposure with
likelihood for potential confounding;
Cancer mortality and TCE exposure not reported for
kidney cancer and all hemato- and lymphopoietic
cancer reported as broad category

Henschler et al. (1995)
Incomplete identification of cohort and index kidney
cancer cases included in case series
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-73 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Vamvakas et al. (1998)
Control selection may not represent case series with
potential for selection bias
1
2
B.3.1.1.1.1. International epidemiology institute study of Rocketdyne workers.
3	B . 3 .1.1.1.1.1. Boice et al (2006b).
4	B.3.1.1.1.1.1.1. Author's abstract.
5
6	Objective: The objective of this study was to evaluate potential health risks
7	associated with testing rocket engines. Methods: A retrospective cohort mortality
8	study was conducted of 8372 Rocketdyne workers employed 1948 to 1999 at the
9	Santa Susana Field Laboratory (SSFL). Standardized mortality ratios (SMRs) and
10	95% confidence intervals (CIs) were calculated for all workers, including those
11	employed at specific test areas where particular fuels, solvents, and chemicals were
12	used. Dose-response trends were evaluated using Cox proportional hazards
13	models. Results: SMRs for all cancers were close to population expects among
14	SSFL workers overall (SMR = 0.89; CI = 0.82-0.96) and test stand mechanics in
15	particular (n = 1651; SMR= 1.00; CI = 0.86-1.1.6), including those likely
16	exposure to hydrazines (n = 315; SMR= 1.09; CI = 0.75-1.52) or trichloroethylene
17	(TCE) (n= 1111; SMR= 1.00; CI = 0.83-1.19). Nonsignificant associations were
18	seen between kidney cancer and TCE, lung cancer and hydrazines, and stomach
19	cancer and years worked as a test stand mechanic. No trends over exposure
20	categories were statistically significant. Conclusion: Work at the SSFL rocket
21	engine test facility or as a test stand mechanic was not associated with a significant
22	increase in cancer mortality overall or for any specific cancer.
23
24	B.3.1.1.1.1.1.2. Study description and comment. Boice et al. (2006b) examined all cause, all
25	cancer and site-specific mortality in a subcohort of 1,651 male and female test stand
26	mechanics who had been employed on or after 1949 to 1999, the end of follow-up, for at
27	least 6 months at SSFL. Subjects were identified from 41,345 male and female Rocketdyne
28	workers at SSFL (n = 8.372) and two nearby facilities (32,979). Of the 1,642 male test stand
29	mechanics, 9 females were excluded due to few numbers, personnel listing in company
30	phone directories were used to identify test stand assignments (and infer potential specific
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-74 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
chemical exposures) for 1,440 subjects, and of this group, 1,111 male test stand mechanics
were identified with potential trichloroethylene exposure either from the cleaning of rocket
engines between tests or from more generalized use as a utility degreasing solvent. Cause-
specific mortality is compared to several referents: (1) morality rates of the U.S.
population, (2) mortality rates of California residents, (3) hourly nonadministrative
workers at SSFL and two nearby facilities, and (4) 1,598 SSFL hourly workers; however,
the published paper does not clearly present details of all analyses. For example, the
referent population is not identified for the standardized mortality ratio (SMR) analysis of
the 1,111 male subjects with TCE potential exposure and analyses examining exposure
duration present point estimates and p-values from tests of linear trend, but not always
confidence intervals (e.g., Boice et al. {2006b) Table 7, table footnotes).
Exposure assessment to trichloroethylene is qualitative without attempt to characterize
exposure level as was done in the exposure assessment approach of Zhao et al. (2005) and
Krishnadsen et al. (2007). Test stand mechanics were nonadministrative hourly positions and
had the greatest potential for chemical exposures to TCE and hydrazine. Potential exposure to
chemicals also existed for other subjects associated with test stand work such as instrument
mechanics, inspectors, test stand engineers, and research engineers potential for chemical
exposure, although Boice et al. (2006b) considered their exposure potential lower compared to
that received by test stand mechanics and, thus, were not included in the cohort. Like that
encountered by UCLA researchers, work history information in the personnel file was not
specific to identify work location and test stand and Boice et al. (2006b) adopted ancillary
information, company phone directories, as an aid to identify subjects with greater potential for
TCE exposure. From these aids, investigators identified rocket stand assignment for 1,440 or
87% of the SSFL test stand mechanics. Bias is introduced through missing information on the
other 211 subjects or if phone directories were not available for the full period of the study. Test
stand mechanics, if exposed, had the likelihood for exposure to high TCE concentrations
associated with flushing or cleaning of rocket engines; 593 of the 1,111 subjects (53%) were
identified as having potential TCE exposure through rocket engine cleaning. The removal or
flushing of hydrocarbon deposits in fuel jackets and in liquid oxygen dome of large engines
entailed the use of 5 to 100 gallons of TCE, with TCE use starting around 1956 and ceased by
the late 1960's at all test stands except one which continued until 1994. No information was
provided on test stand and working conditions or the frequency of exposure-related tasks, and no
atmospheric monitoring data were available on TCE. A small number of these subjects (121)
also had potential exposure to hydrazines. The remaining 518 subjects in the TCE subcohort
were presumed exposed to TCE as a utility solvent. Information on use of TCE as a utility
solvent is lacking except that TCE as a utility solvent was discontinued in 1974 except at one test
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-75 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
stand where it was used until 1984. These subjects have a lower likelihood of exposure
compared to subjects with TCE exposure from cleaning rocket engines.
Several study design and analysis aspects limit this study for assessing risks associated
with trichloroethylene exposure. Overall, exposures were likely substantially misclassified and
their frequency likely low, particularly for subjects identified with TCE use as a utility solvent
who comprise roughly 50% of the TCE subcohort. Analyses examining number of years
employed at SSFL or worked as test stand mechanic as a surrogate for cumulative exposure has a
large potential for misclassification bias due to the lack of air monitoring data and inability to
account to temporal changes in TCE usage. Moreover, the exposure metric used in some dose-
response analyses is weighted by the number of workers without rationale provided and would
introduce bias if the workforce changed over the period covered by this study. Some information
suggests this was likely (1) the number of cohort subjects entering the cohort decreased over the
time period of this study, as much as a 20% decrease between 1960's and 1970s, and
(2) ancillary information (http://www.thewednesdavreport.com/twr/twr48v7.htm. accessed
March 11, 2008; DOE Closure Project, http://www.etec.energy.gov/Reading-
Room/DeSoto.html accessed March 11, 2008). Study investigators did not carry out exposure
assessment for referents and no information is provided on potential trichloroethylene exposure.
If referents had more than background exposure, likely for other hourly subjects with direct
association with test stand work but with a job title other than test stand mechanic, the bias
introduced leads to an underestimation of risk. TCE use at SSFL was widespread and rocket
engine cleaning occurred at other locations besides at test sites (Morgenstern et al., 1999),
locations from which the referent population arose.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-76 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Boice JD, Marano DE, Cohen SS, Mumma MT, Blott WJ, Brill AB, Fryzek JP, Henderson BE, McLaughlin JK. (2006b) .
Mortality among Rocketdyne workers who tested rocket engines, 1948-1999. J Occup Environ Med 48:1070-1092.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract "objective of this study was to evaluate potential health risks
associated with testing rocket engines."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
54,384 Rocketdyne workers of which 41,351 were employed on or after 1-1-1948
and for at least 6 mos at Santa Susana Field Laboratory or nearby facilities. Of the
41,351 subjects, 1,651 were identified as having a job title of test stand mechanic and
exposure assignments could be made for 1,440 of these subjects.
Site-specific mortality rates of U.S. population and of all-other Rocketdyne
employees. Potential TCE exposures of all other subjects (referents) not documented
but investigators assumed referents are unexposed to TCE.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality from 1948 to 12-31-1999.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Coding to ICD in use at time of death.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Qualitative exposure assessment, any TCE exposure. No quantitative information on
TCE intensity by job title or to individual subjects or referents.
Missing exposure potential to 12% of test stand mechanics; potential exposure
hydrazine and/or TCE assigned to 1,440 of 1,651 test stand mechanics. Of 1,440 test
stand mechanics, 1,111* identified with potential TCE exposure, 518 of the
1,111 identified as having presumed high intensity exposure from the cleaning of
rocket engines. The remaining 593 subjects with potential exposure to TCE through
use as "utility solvent," a job task with low likelihood or potential for TCE exposure.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
0.4%) for test stand mechanic cohort (1,651 subjects).
>50% cohort with full latency
35 years average follow-up; 88%> of 1,651 test stand mechanics >20 yr follow-up.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
TCE exposed subcohort—391 total deaths, 121 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
SMR analysis restricted to male hourly test stand mechanics using U.S. population
rates as referent—no adjustment of potential confounders other than age and
calendar-year.
Cox proportional hazard models examining TCE exposure adjusted for birth year,
year of hire and potential hydrazine exposure. Race was not included in Cox
proportional hazard analysis.
Statistical methods
SMR analysis and Cox proportional hazard.
Exposure-response analysis presented in
published paper
Duration of exposure (employment): 2-sided tests for linear trend.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Documentation of results
All analyses are not presented in published paper. Follow-up correspondence of C

Scott, U.S. EPA, to J. Boice, of 12-31-06 and 02-28-07 remain unanswered as of

November 15, 2007.
*Zhao et al. (2005), whose study period and base population overlaps that of Boice et al. (2006b), identified a larger number of subjects with potential TCE
exposures; 2,689 subjects with TCE score > 3, a group having medium to high cumulative TCE exposure.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.1.1.2. University of California at Los Angeles (UCLA) studies of Rocketdyne
workers.
B . 3 .1.1.1.2.1. Krishnadasan et al. (2007).
B.3.1.1.1.2.1.1. Author's abstract.
Background To date, little is known about the potential contributions of
occupational exposure to chemicals to the etiology of prostate cancer. Previous
studies examining associations suffered from limitations including the reliance on
mortality data and inadequate exposure assessment. Methods We conducted a
nested case-control study of 362 cases and 1,805 matched controls to examine the
association between occupational chemical exposures and prostate cancer
incidence. Workers were employed between 1950 and 1992 at a nuclear energy
and rocket engine-testing facility in Southern California. We obtained cancer
incidence data from the California Cancer Registry and seven other state cancer
registries. Data from company records were used to construct a job exposure
matrix (JEM) for occupational exposures to hydrazine, trichloroethylene (TCE),
polycyclic aromatic hydrocarbons (PAHs), benzene, and mineral oil.
Associations between chemical exposures and prostate cancer incidence were
assessed in conditional logistic regression models. Results With adjustment for
occupational confounders, including socioeconomic status, occupational physical
activity, and exposure to the other chemicals evaluated, the odds ratio for
low/moderate TCE exposure was 1.3; 95%CI=0.8 to 2.1, and for high TCE
exposure was 2.1; 95%CI=1.2 to 3.9. Furthermore, we noted a positive trend
between increasing levels of TCE exposure and prostate cancer (p-value for
trend=0.02). Conclusion Our results suggest that high levels of TCE exposure
are associated with prostate cancer among workers in our study population.
B.3 .1.1.1.2.2. Zhao et al. (2005).
B.3.1.1.1.2.2.1. Author's abstract.
Background A retrospective cohort study of workers employed at a California
aerospace company between 1950 and 1993 was conducted; it examined cancer
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-81 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
mortality from exposures to the rocket fuel hydrazine. Methods In this study, we
employed a job exposure matrix (JEM) to assess exposures to other known or
suspected carcinogens—including trichloroethylene (TCE), polycyclic aromatic
hydrocarbons (PAHs), mineral oils, and benzene—on cancer mortality
(1960-2001) and incidence (1988-2000) in 6,107 male workers. We derived
rate- (hazard-) ratios estimates from Cox proportional hazard models with time-
dependent exposures. Results High levels of TCE exposure were positively
associated with cancer incidence of the bladder (rate ratio (RR): 1.98, 95%
confidence interval (CI) 0.93-4.22) and kidney (4.90; 1.23-19.6). High levels of
exposure to mineral oils increased mortality and incidence of lung cancer (1.56;
1.02-2.39 and 1.99; 1.03-3.85), and incidence of melanoma (3.32; 1.20-9.24).
Mineral oil exposures also contributed to incidence and mortality of esophageal
and stomach cancers and of non-Hodgkin's lymphoma and leukemia when
adjusting for other chemical exposures. Lagging exposure measures by 20 years
changed effect estimates only minimally. No associations were observed for
benzene or PAH exposures in this cohort. Conclusions Our findings suggest that
these aerospace workers who were highly exposed to mineral oils experienced an
increased risk of developing and/or dying from cancers of the lung, melanoma,
and possibly from cancers of the esophagus and stomach and non-Hodgkin's
lymphoma and leukemia. These results and the increases we observed for TCE
and kidney cancers are consistent with findings of previous studies.
B.3.1.1.1.2.3. Study description and comment. The source population for Krishnadasen et
al. (2007) and Zhao et al. (2005) is the UCLA chemical cohort of6,044 male workers with 2
or more years of employment Rocketdyne between 1950 and 1993, who engaged in rocket
testing at SSFL before 1980 and who have never been monitored for radiation. Zhao et al.
(2005) examined cancer mortality between 1960-2001, an additional 7 years from earlier
analyses of the chemical subcohort (Morgenstern et al., 1999; Ritz et al., 1999), and cancer
incidence (5,049 subjects) between 1988-2000, matching cohort subjects to names in
California's Cancer Registry and eight other state cancer registries. Deaths before 1998 are
coded using ICD, 9th revision, and ICD-10 after this date; ICD-0 was used to code cancer
incidence with leukemia, lymphoma, and other lymphopoietic tumors grouped on the basis of
morphology codes. A total of 600 cancer deaths and 691 incident cancers were identified
during the study period.
Krishnadasen et al. (2007) adopted a nested case-control design to examine occupational
exposure to several chemicals and prostate cancer incidence in a cohort which included the SSFL
chemically-exposed subjects and an additional 4,607 workers in the larger cohort who were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
enrolled in the company's radiation monitoring program. A total of 362 incident prostate
cancers were identified between 1988 and 12-31-1999. Controls were randomly selected from
the original cohorts using risk-set sampling and a 5:1 matching ratio on age at start of
employment, age at diagnosis, and cohort.
Both studies are based on the same exposure assessment approach. Walk-through visits,
interviews with managers and workers, job descriptions manual, and historical facility reports
supported the development of a JEM with jobs ranked on a scale of 0 (no exposure) to 3 (highly
exposure) on presumptive exposure reflecting relative intensity of that exposure over 3 temporal
periods: 1950-1960, 1970s, 1980-1990. Of the 6,044 subjects, 2,689 had TCE exposure scores
of >3 and 2,643 with an exposure score 3 or greater for hydrazine. Workers with job titles
indicating technical or mechanical work on rocket engines were presumed to have high
hydrazine rocket fuel exposure and high TCE exposure, which was used in cleaning rocket
engines and parts. Although fewer subjects had exposure to benzene (819 subjects) or mineral
oil (1,499 subjects), a high percentage of these subjects were also exposed to TCE. TCE use was
widespread at the facility and other mechanics, maintenance and utility workers, and machinists
were presumed as having exposure. No details were provided for job titles other than rocket test
stand mechanics for assigning TCE exposure intensity and historical trends in TCE usage. Air
monitoring data was absent for any chemicals prior to 1985 and investigators could not link
study subjects to specific work locations and rocket-engine test stands. As a result, exposures
were probably substantially misclassified, particularly those with low to moderate TCE
exposure. Cumulative intensity score was the sum of the job-and time-specific intensity score
and years in job. Exposure classification was assigned blinded to survival status and cause of
death.
Proportional hazards modeling in calendar time with both fixed and time-depend
predictors was used by Zhao et al. (2005) to estimate exposure effects on site-specific cancer
incidence and mortality for a combined exposure group of medium and high exposure intensity
with workers with no to low exposure intensity as referents. Variables in the proportional hazard
model included time since first employment, socioeconomic status, age at diagnosis or death, and
exposure to other chemical agents including benzene, polycyclic aromatic hydrocarbons (PAHs)
mineral oil, and hydrazine. Krishnadasen et al. (2007) fit conditional logistic regression model
to their data adjusting of cohort, age at diagnosis, occupation physical activity, socioeconomic
status and all other chemical exposure levels. Both publications include exposure-response
analysis and present p-values for linear trend. Race was not controlled in either study given the
lack of recording on personnel records. Smoking histories was available for only a small
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
percentage of the cohort; for those subjects reporting smoking information, mean cumulative
TCE score did not differ between smokers and nonsmokers.
This study develops semiquantitative exposure levels and is strength of the exposure
assessment. However, potential for exposure misclassification exists and would be of a
nondifferential direction. Rocket engine test stand mechanics had likely exposure to TCE,
kerosene, and hydrazine fuels; no information is available as to exposure concentrations.
Statistical analyses in both Zhao et al. (2005) and Krishnadansan et al. (2007) present risk
estimates for TCE that were adjusted for these other chemical exposures. Other strengths of this
study include a long follow-up period for mortality, greater than an average time of 29 years of
which 16 at SSFL, use of internal referents and the examination of cancer incidence, although
under ascertainment of cases is likely given only 8 state cancer registries were used to identify
cases and incidence ascertained after 1981, 40 years after the cohort's initial definition date.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-84 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Krishnadasan A, Kennedy N, Zhao Y, Morgenstern H, Ritz B. (2007) . Nested case-control study of occupational chemical
exposures and prostate cancer in aerospace and radiation workers. Am J Ind Med 50:383-390.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Nested case-control study of the UCLA chemical and radiation cohorts (Morgenstern
et al.. 1997, 19919899) to assess occupational exposures including TCE and prostate
cancer.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
4,607 radiation cohort + 6,107 Santa Susana chemical cohort (Ritz et al., 1999; Zhao
et al., 2005), excluded 1,410 deaths before 1988 (date of cancer incidence follow-up).
Incident prostate cancer cases identified from eight State cancer registries (California,
Nevada, Arizona, Texas, Washington Florida, Arkansas, and Oregon). Controls were
randomly selected from the original cohorts using risk-set sampling.
362 cases and 1,805 controls (100% participation rate).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Prostate cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma

CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
TCE exposure assigned to cases and controls based on longest job held at company as
identified from personnel records. Cumulative exposure—ranked exposure intensity
score for TCE by 3 time periods—using method of Zhao et al. (2005).
Blinded ranking of exposure status.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Employment records were used to assign exposure. 734 subjects (249 cases and
485 controls, or 33% of all cases and controls) were interviewed via telephone or sent
a mailed questionnaire to obtain medical history, education and personal information
on physical activity level and smoking history.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
No proxy interviews.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Any TCE exposure: 135 cases (37%) and 668 controls (37%).
High cumulative TCE exposure: 45 cases (12%) and 124 controls (7%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Cohort, age at diagnosis, occupational physical activity, SES, other chemical
exposures (benzene, PAHs, mineral oil, hydrazine). No adjustment for race due to
lacking information; affect of race on OR examined using information from survey of
workers still alive in 1999. Few African American workers (n = 7), TCE levels did
not vary greatly with race.
Statistical methods
Crude and adjusted conditional logistic regression.
Exposure-response analysis presented in
published paper
p-walue for trend with exposure lag (0 yrs, 20 yr).
Documentation of results
Adequate.
-------

§•
to
s
a, 5
c> to
to s
Co
<3 Co'
s.	a
£
O	>r
S	vgs
£. >
s ?
^	>5
^ 5
Oq 5.
<3 TO
^ ^3
^3 5
S. ^
.TO Co
V; to
Co
0
s
1
a.
VO
DRAFT
OR=odds ratio. SES= socio-economic status.
to
o
EPA DELIBERATIVE - DO NOT QUOTE OR CITE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Zhao Y, Krishnadasan A, Kennedy N, Morgenstern H, Ritz B. (2005). Estimated effects of solvents and mineral oils on cancer
incidence and Mortality in a cohort of aerospace workers. Am J Ind Med 48:249-258.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From introduction "one aim of this new investigation was to determine whether
these aerospace workers also developed cancers from exposures to other chemicals
including trichloroethylene (TCE), polycyclic aromatic hydrocarbons (PAHs),
mineral oils, and benzene."
Selection and characterization in cohort studies
of exposure and control groups and of cases and
controls in case-control studies is adequate
6,107 male workers employed for 2 or more years and before 1980 at Santa Susana
Field Laboratory. Internal referents (no or low TCE exposure).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence between 1988-2000.
Mortality between 1950-2001.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-0 for cancer incidence. Leukemia, lymphomas, and other lymphopoietic
malignancies grouped on the basis of morphology codes.
Mortality: ICD-9, before 1998, and ICD-10 thereafter. Incidence: ICD-Oncology
Lymphoma and leukemia grouping includes lymphosarcoma and reticulosarcoma,
Hodgkin's disease, other malignant neoplasm of the lymphoid and histiocytic tissue,
multiple myeloma and immunoproliferative neoplasms, and all leukemias except
chronic lymphoid leukemia. The following incident tumors were also included:
Hodgkin's disease, leukemia, polycythemia vera, chronic myeloproliferative
disease, myelosclerosis, eosinophilic conditions, platelet diseases, and red blood cell
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o

diseases.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Cumulative exposure—ranked exposure intensity score for TCE by 3 time periods
Blinded ranking of exposure status.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
99% follow-up for mortality (6,044 of 6,107 subjects).
>50% cohort with full latency
Average latency = 29 yrs (Ritz et al., 1999).
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
600 cancer deaths, 621 cancer cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Time since first employment, SES, age (at incidence or mortality), exposure to other
carcinogens, including hydrazine. No adjustment for race. Indirectly assessment of
smoking through examination of smoking distribution by chemical exposure. Mean
TCE cumulative exposure scores of smokers and nonsmokers is not statistically
significant different.
Statistical methods
Cox proportional hazards modeling in calendar time with both fixed and
time-dependent predictors.
Exposure lagged 10 and 20 yrs.
Exposure-response analysis presented in
Test for monotonic trend of cumulative exposure, two-sided p-value for trend.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
published paper

Documentation of results
Liver cancer results are not reported in published paper.
SES = socio-economic status.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.1.1.3. Comment on the Santa Susanna Field Laboratory (SSFL) studies.
Rocketdyne workers at SSFL are subject of two separate and independent studies. Both
research groups draw subjects from the same underlying source population, Rocketdyne
workers including those at SSFL, however, the methods adopted to identify study subjects
and to define TCE exposure differ with each study. A subset of SSFL workers is common
to both studies; however, no information exist in final published reports (IEI, 2005;
Morgenstern et al., 1997,1999) to indicate the percentage overlap between cohorts or
between observed number of site-specific events.
1	Notable differences in both study design and analysis including cohort identification,
2	endpoint, exposure assessment approaches, and statistical methods exist between Zhao et al.
3	(2005) and Krishnadasan et al. (2007), whose source population is the UCLA cohort, and Boice
4	et al. (2006b) whose source population is the IEI cohort. A perspective of each study's
5	characteristics may be obtained from Table B-6, below.
6
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-93 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Table B-6. Characteristics of epidemiologic investigations of Rocketdyne workers
Study
Boice et al. (2006b)
Zhao et al. (2005)
Source
population
41,351 administrative/scientific and
nonadministrative male and female employees
between 1949-1999 at Rocketdyne SSFL and two
nearby facilities
-55,000 subjects of SSFL and two nearby facilities employed
between 1950 and 1993
TCE subcohort
1,111 male test stand mechanics with potential
TCE exposure
6,107 males working at SSFL before 1980 and identified as test
stand personnel, of whom 2,689 males had exposure scores
greater than no- to low-TCE exposure potential
Pay-type (hourly)
100% of TCE subcohort
11.3%
Job title with
potential TCE
exposure
Test stand mechanics identified with greatest
potential for TCE exposure
Other job titles with direct association with test
stand work—instrument mechanics, inspectors,
test stand engineers, and research engineers—
identified with lower exposure potential to TCE
and included in referent population
High potential exposure group included job titles as
propulsion/test mechanics or technicians; Medium potential
exposure group included propulsion/test inspector, test or
research engineer, and instrumentation mechanic; Low-exposure
potential included employees who, according to job title may
have been present during engine test firings but without direct
contact
Exposure metric
Qualitative, yes/no, and employment duration
Cumulative exposure score = £ (exposure score (0-3) x number
of years in job)
Endpoint
Mortality as of 1999
Mortality as of 2001 and Incidence as of 2000
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Statistical
analysis
Standardized mortality ratio
Proportional hazards modeling with covariates for
birth year, hire year, and potential exposure to
hydrazine.
Proportional hazards modeling with covariates for time since first
employment, socioeconomic status, age at event, and exposure to
all other carcinogens, including hydrazine
Observed
number of
deaths:


Total cancer
121
600
Lung
51
No/low, 99
Medium, 62
High, 33
Kidney
7
No/low, 7
Medium, 7
High, 3
Bladder
5
No/low, 8
Medium, 6
High, 3
NHL/Leukemia
6
No/low, 27
Medium, 27
High, 6
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
A number of strengths and limitations underlie these studies. First, the Zhao et al. (2005)
and Krishnadasan et al. (2007) analyses is of a larger population and of more cancer cases or
deaths; 600 cancer deaths and 691 cancer cases in Zhao et al. (2005) compared to 121 cancer
deaths in the TCE subcohort of Boice et al. (2006b), and for prostatic cancer among all
Rocketdyne workers, 362 incident prostatic cancer cases in Krishnandasan et al. (2007)
compared to 193 deaths in Boice et al. (2006b). Second, exposed populations appear
appropriately selected in the three studies although questions exist regarding the referent
population in Boice et al. (2006b) whose referent population included subjects with some direct
association with test stand work but whose job title was other than test stand mechanic. As a
result, it appears that these studies identify TCE exposure potential different for possibly similar
job titles. For example, jobs as instrument mechanics, inspectors, test stand engineers, and
research engineers are identified with medium potential exposure in Zhao et al. (2005). Boice et
al. (2006b) on the other hand included these subjects in the referent population and assumed they
had background exposure. TCE use at SSFL was also widespread and rocket engine cleaning
occurred at other locations besides at test sites (Morgenstern et al., 1999), locations from which
the referent population in Boice et al. (2006b) arose. If referents in Boice et al. (2006b) had
more than background exposure, the bias introduced leads to an underestimation of risk. Third,
Zhao et al. (2005) and Krishnadasan et al. (2007) studies include an examination of incidence,
and are likely to have a smaller bias associated with disease misclassification than Boice et al.
(2006b) who examines only mortality. Fourth, use of cumulative exposure score although still
subject to biases is preferred to qualitative approach for exposure assessment. Last, all three
studies adjusted for potentially confounding factors such as smoking, socioeconomic status, and
other carcinogenic exposures using different approaches either in the design of the study, such as
Boice et al. (2006b) limitation to only hourly workers, or in the statistical analysis such as Zhao
et al. (2005) and Krishnadansen et al. (2007). For this reason, the large difference in hourly
workers between the UCLA cohort and Boice et al. (2006b) is not likely to greatly impact
observations.
B.3.1.1.2. Blair et al. (1998), Radican et al. (2008).
B.3 .1.1.2.1. Radican et al. (2008)) abstract.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-96 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
OBJECTIVE: To extend follow-up of 14,455 workers from 1990 to 2000, and
evaluate mortality risk from exposure to trichloroethylene (TCE) and other
chemicals. METHODS: Multivariable Cox models were used to estimate relative
risk (RR) for exposed versus unexposed workers based on previously developed
exposure surrogates. RESULTS: Among TCE-exposed workers, there was no
statistically significant increased risk of all-cause mortality (RR = 1.04) or death
from all cancers (RR = 1.03). Exposure-response gradients for TCE were
relatively flat and did not materially change since 1990. Statistically significant
excesses were found for several chemical exposure subgroups and causes and
were generally consistent with the previous follow-up. CONCLUSIONS: Patterns
of mortality have not changed substantially since 1990. Although positive
associations with several cancers were observed, and are consistent with the
published literature, interpretation is limited due to the small numbers of events
for specific exposures.
B.3.1.1.2.2. Blair et al. (1998) abstract.
OBJECTIVES: To extend the follow up of a cohort of 14,457 aircraft
maintenance workers to the end of 1990 to evaluate cancer risks from potential
exposure to trichloroethylene and other chemicals. METHODS: The cohort
comprised civilians employed for at least one year between 1952 and 1956, of
whom 5727 had died by 31 December 1990. Analyses compared the mortality of
the cohort with the general population of Utah and the mortality and cancer
incidence of exposed workers with those unexposed to chemicals, while adjusting
for age, sex, and calendar time. RESULTS: In the combined follow up period
(1952-90), mortality from all causes and all cancer was close to expected
(standardized mortality ratios (SMRs) 97 and 96, respectively). Significant
excesses occurred for ischemic heart disease (SMR 108), asthma (SMR 160), and
cancer of the bone (SMR 227), whereas significant deficits occurred for
cerebrovascular disease (SMR 88), accidents (SMR 70), and cancer of the central
nervous system (SMR 64). Workers exposed to trichloroethylene showed non-
significant excesses for non-Hodgkin's lymphoma (relative risk (RR) 2.0), and
cancers of the oesophagus (RR 5.6), colon (RR 1.4), primary liver (RR 1.7),
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-97 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
breast (RR 1.8), cervix (RR 1.8), kidney (RR 1.6), and bone (RR 2.1). None of
these cancers showed an exposure-response gradient and RRs among workers
exposed to other chemicals but not trichloroethylene often had RRs as large as
workers exposed to trichloroethylene. Workers exposed to solvents other than
trichloroethylene had slightly increased mortality from asthma, non-Hodgkin's
lymphoma, multiple myeloma, and breast cancer. CONCLUSION: These
findings do not strongly support a causal link with trichloroethylene because the
associations were not significant, not clearly dose-related, and inconsistent
between men and women. Because findings from experimental investigations and
other epidemiological studies on solvents other than trichloroethylene provide
some biological plausibility, the suggested links between these chemicals and
non-Hodgkin's lymphoma, multiple myeloma, and breast cancer found here
deserve further attention. Although this extended follow up cannot rule out a
connection between exposures to solvents and some diseases, it seems clear that
these workers have not experienced a major increase in cancer mortality or cancer
incidence.
B.3 .1.1.2.3. Study description and comment. This historical cohort study of 14,457
(9,400 male and 3,138 female) civilian personnel employed at least one year between 1942
and 1956 at Hill Air Force Base in Utah examines mortality to the end of 1982 (Spirtas et
al., 1991) to the end of 1990 (Blair et al., 1998), or to the end of 2000 (Radican et al., 2008).
About half of the cohort was identified with exposure to TCE (6,153 white men and 1,051
white women). One-fourth of subjects were born before 1909 with an attained age of 43
years at cohort's identification date of 1952 and whose first exposure could have been as
early as 1939, a cohort considered as a "survivor cohort."
As of December 2008, the end of follow-up in Radican et al. (2008), 8,580 deaths (3,628
in TCE subcohort) were identified, an increase of 2,853 deaths with the additional 8 years
follow-up period compared to Blair et al. (1998) (5,727 total deaths, 2,813 among TCE
subcohort subjects), with a larger proportion deaths among non-TCE exposed subjects (58%) as
of December 2008 compared to the December 2000 (51%). Approximately 50% of
TCE-exposed subjects and 60% of all cohort subjects had died, with mean age of 75 years for
TCE-exposed subjects still alive and 45 or more years since the cohort's definition (1953 to
1955), a time period longer than that typically considered for an induction or latent window for
detecting an adverse outcome like cancer. Blair et al. (1998) additionally examined cancer
incidence among white TCE-exposed workers alive on 1-1-1973, a period of 31 years after the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-98 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
cohort's inception date, to the end of 1990. Incident cancer cases are likely under ascertained for
this reason.
Statistical analyses in Spirtas et al. (1991) and Blair et al. (1998) focus on site-specific
mortality for white subjects or subjects with unknown race who were assumed to as white since
97% of all subjects with know race were white. SMRs are presented with expected numbers of
deaths based upon age-, race- and year-specific mortality rates of the Utah population (Blair et
al., 1998; Spirtas et al., 1991) or rate ratios for mortality or cancer incidence for the TCE
subcohort from Poisson regression models, adjusting for date of birth, calendar year of death,
and sex where appropriate, and an internal standard of mortality rates of the cohort's
nonchemical exposed subjects (internal referents) (Blair et al., 1998). Blair et al. (1998), in
addition to their presentation in the published papers of risk estimates associated with TCE
exposure, also, presented risk estimates for subjects with an aggregated category of "any solvent
exposure" (ever exposed) and for exposure to 14 solvents. To compare with risk ratios from
Poisson regression models of Blair et al. (1998), Radican et al. (2008) adopted Cox proportional
hazard models to reanalyze mortality observations of follow-up through 1990. For most site-
specific cancers, Radican et al. (2008) did not observe large differences between the Cox hazard
ratio and Poisson rate ratio of Blair et al. (1998), although difference between risk estimates from
Cox proportional hazard and Poisson regression of 20% or larger was observed for kidney cancer
(increased risk estimate) and primary liver cancer (decreased risk estimate). Radican et al.
(2008), furthermore, noted hazard ratios for all subjects were similar to results for white subjects
only; therefore, their analyses of follow-up through 2000 included all subjects.
The original exposure assessment of Stewart et al. (1991) who conducted a detailed
exposure assessment of TCE exposures at Hill Air Force Base was used by Radican et al. (2008),
Blair et al. (1998), and Spirtas et al. (1991). Their was limited for linking subjects with
exposures principally because solvent exposures were associated with work in "shops," but work
records listed only broad job titles and administrative units. As a result, exposures were
probably substantially misclassified, particularly in "mixed solvent group." Trichloroethylene
was used principally for degreasing and hand cleaning in work areas during 1955-1968. TCE
was the predominant solvent used in the few available vapor degreasers located in the
electroplating (main hanger), propeller, and engine repair shops before the mid-1950 and,
afterwards, as a cold state solvent, replacing Stoddard solvent. Solvents, notably TCE after
1955, were used primarily by aircraft mechanics with short but high exposures and sheet metal
workers for spot clean aircraft surfaces. The investigators determined that 32% had "frequent"
exposures to peak concentrations (one or two daily peaks of about 15 minutes to
trichloroethylene at 200-600 ppm) during vapor degreasing. Work areas were located in very
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-99 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
large buildings with few internal partitions, which aided dispersion of trichloroethylene. While
TCE exposures were less controlled in the 1950s, by the end of 1960s, TCE exposure had been
reduced significantly. Only a small number of subjects with "high" exposure had long-duration
exposures, no more than 16%. Few workers were exposed only to trichloroethylene; most had
mixed exposures to other chlorinated and nonchlorinated solvents. Person-years of exposure
were computed from date of first exposure, which could have been as early as 1939, to the end of
1982.
Overall, Blair et al. (1998) and Radican et al. (2008) are studies with approximately half
of the larger cohort identified as having some potential for TCE exposure (the TCE subcohort)
and calculation of cancer risk estimates for TCE exposure, either risk ratios in Blair et al. (1998)
or hazard ratios in Radican et al. (2008), using workers in the cohort without any chemical
exposures as referent population, superior to standardized mortality ratios of Spirtas et al. (1991)
who first reported on mortality and TCE exposure. Use of an internal referent population of
workers from the same company or plant, but lacking the exposure of interest, is considered to
reduce bias associated with the healthy worker effect. For follow-up in Radican et al. (2008)
who examined mortality 45 years after first exposure and likely at the tail of or beyond a window
for cancer induction time, any influence on exposure on disease development or detection times
would be diminshed or less evident if exposures like TCE shortened induction time, e.g., if
exposure shortened the natural course of disease development, which would become evident in
an unexposed subjects with longer follow-up periods. The induction time of 35 years in Blair et
al. (1998)may also fall outside a cancer induction window; however, it is more consistent with
cancer induction times observed with other chemical carcinogens such as aromatic amines
(Weistenhofer et al., 2008) and vinyl chloride (Du and Wang, 1998). A strong exposure
assessment was performed, but precision in the exposure assignment was limited by vague
personnel data. The cohort had a modest number of highly exposed (about 100 ppm) subjects,
but overall most were exposed to low concentrations (about 10 ppm) of trichloroethylene.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-100 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Radican L, Blair A, Stewart P, Wartenberg D. (2008). Mortality of aircraft maintenance workers exposed to
trichloroethylene and other hydrocarbons and chemicals: extended follow-up. J Occup Environ Med 50:1306-1319.
Blair A, Hartge P, Stewart PA, McAdams M, Lubin J. (1998). Mortality and cancer incidence of aircraft maintenance
workers exposed to trichloroethylene and other organic solvents and chemicals: extended follow-up. Occup Environ Med
55:161-171.
Spirtas R, Stewart PA, Lee JS, Marano DE, Forbes CD, Grauman DJ, Pettigrew HM, Blair A, Hoover RN, Cohen JL. (1991).
Retrospective cohort mortality study of workers at an aircraft maintenance facility. I. Epidemiological results. Br J Ind Med
48:515-530.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Abstract: "...to evaluate cancer risks from potential exposure to trichloroethylene and
other chemicals."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
All civilians employed at Hill AFB for >1 yr between 1-1-1952 and 12-31-1956;
cohort of 14,457 workers identified form earnings records.
TCE subcohort—7,204 white males and females (50%).
External referents, all civilian cohort—Utah population rates, 1953-1990.
Internal referents, TCE subcohort analysis of mortality (Blair et al., 1998); Radican et
al. (2008) and incidence (Blair et al., 1998)—workers without chemical exposures.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality, all civilian cohort and TCE subcohort.
Incidence, TCE subcohort.
-------
DRAFT
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Underlying and contributing causes of deaths as coded to ICDA 8.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Detailed records on setting and job activities, worker interviews; work done in large
open shops; shops not recorded in personnel records, link of job with IH data was
weak. Limited exposure IH measurements for TCE between 1960-1990. Plant JEM,
rank order assignments by history; determined exposure duration during vapor
degreasing tasks about 2,000 ppm-h and hard degreasing about 20 ppm-h. Median
exposure were about 10 ppm for rag and bucket (cold degreasing process);
100-200 ppm for vapor degreasing (Stewart et al., 1991). Cherrie et al. (2001)
estimated long-term exposure as ~50 ppm with short-term excursion up to
-600 ppm. NRC (2006) concluded the cohort had a modest number of highly
exposed (about 100 ppm) subjects, but overall most were exposed to low TCE
concentrations (about 10 ppm).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
97%o of cohort traced successfully to 12-31-1982.
>50% cohort with full latency
Yes, all subjects followed minimum of 35 yrs (Blair et al., 1998) or 45 yrs (Radican et
al., 2008).
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Number of deaths in cohort mortality studies; TCE subcohort—2,813 deaths (39%), 528 cancer deaths, and 549 incident cancers
numbers of total cancer incidence studies;	(1973-1990) (Blair et al., 1998); 3,628 deaths (50%). 729 cancer deaths (Radican et
numbers of exposed cases and prevalence of	al., 2008).
exposure in case-control studies
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
SMR analysis evaluates age, sex, and calendar year (Spirtas et al., 1991).
Date of hire, calendar year of death, and sex in Poisson regression analysis (Blair et
al., 1998).
Age, gender, and race (to compare with RR of Blair et al. (1998), or age and gender
for follow-up to 2000] in Cox proportional hazard analysis (Radican et al., 2008).
Statistical methods
External analysis is restricted to Caucasian subjects—Life table analysis for mortality
(Spirtas et al., 1991).
Internal analysis restricted to Caucasian subjects or subject of unknown race assumed
to be Caucasian and followed to 1990—Poisson regression (Blair et al., 1998) or Cox
Proportional Hazard (Radican et al., 2008).
Internal analysis—all subjects followed to 2000 (Radican et al., 2008).
Exposure-response analysis presented in
published paper
Risk ratios from Poisson regression model and hazard ratios from Cox Proportional
Hazard model for exposure rankings but no formal statistical trend test presented in
papers.
Documentation of results
Adequate.
RR = relative risk.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B 3 1 1 3 Boice et al. (19991989).
B.3.1.1.3.1. Author's abstract.
OBJECTIVES: To evaluate the risk of cancer and other diseases among workers
engaged in aircraft manufacturing and potentially exposed to compounds
containing chromate, trichloroethylene (TCE), perchloroethylene (PCE), and
mixed solvents. METHODS: A retrospective cohort mortality study was
conducted of workers employed for at least 1 year at a large aircraft
manufacturing facility in California on or after 1 January 1960. The mortality
experience of these workers was determined by examination of national, state,
and company records to the end of 1996. Standardized mortality ratios (SMRs)
were evaluated comparing the observed numbers of deaths among workers with
those expected in the general population adjusting for age, sex, race, and calendar
year. The SMRs for 40 causes of death categories were computed for the total
cohort and for subgroups defined by sex, race, and position in the factory, work
duration, year of first employment, latency, and broad occupational groups.
Factory job titles were classified as to likely use of chemicals, and internal
Poisson regression analyses were used to compute mortality risk ratios for
categories of years of exposure to chromate, TCE, PCE, and mixed solvents, with
unexposed factory workers serving as referents. RESULTS: The study cohort
comprised 77,965 workers who accrued nearly 1.9 million person-years of follow
up (mean 24.2 years). Mortality follow-up, estimated as 99% complete, showed
that 20,236 workers had died by 31 December 1996, with cause of death obtained
for 98%. Workers experienced low overall mortality (all causes of death SMR
0.83) and low cancer mortality (SMR 0.90). No significant increases in risk were
found for any of the 40 specific causes of death categories, whereas for several
causes the numbers of deaths were significantly below expectation. Analyses by
occupational group and specific job titles showed no remarkable mortality
patterns. Factory workers estimated to have been routinely exposed to chromate
were not at increased risk of total cancer (SMR 0.93) or of lung cancer (SMR
1.02). Workers routinely exposed to TCE, PCE, or a mixture of solvents also were
not at increased risk of total cancer (SMRs 0.86, 1.07, and 0.89, respectively), and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-106 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
the numbers of deaths for specific cancer sites were close to expected values.
Slight to moderately increased rates of non-Hodgkin's lymphoma were found
among workers exposed to TCE or PCE, but none was significant. A significant
increase in testicular cancer was found among those with exposure to mixed
solvents, but the excess was based on only six deaths and could not be linked to
any particular solvent or job activity. Internal cohort analyses showed no
significant trends of increased risk for any cancer with increasing years of
exposure to chromate or solvents.
The results from this large scale cohort study of workers followed up for over
3 decades provide no clear evidence that occupational exposures at the aircraft
manufacturing factory resulted in increases in the risk of death from cancer or
other diseases. Our findings support previous studies of aircraft workers in which
cancer risks were generally at or below expected levels.
B.3 .1.1.3 .2. Study description and comment. This study was conducted on an aircraft
manufacturing worker cohort employed at Lockheed-Martin in Burbank, California with
exposure assessment described by Marano et al. (2000). This large cohort study of
77,965 subject workers with at least 1 year employment on or after 1-1-1960, examined
causes of mortality in the entire cohort, but also by broad job titles and for selected
chemical exposures including TCE. Mortality was assessed as of 12-31-1996, with subjects
lacking death certificates presumed alive at end of follow-up. Exposure assessment
developed using a method of exposure assignment by job categories based on job histories
(Kardex cards) and the judgment of long-term employees. Job histories were not available
for every worker, and, if missing, auxiliary sources of job information were used to broadly
classify workers into various job categories. Only subjects with job histories as recorded
on Kardex cards are included in exposure duration analyses. TCE was used for vapor
degreasing on routine basis prior to 1966 and, given the cohort beginning date of 1960, only
a small percentage of the total cohort was identified as having potential TCE exposure.
The investigators determined that 5,443 factory workers had potential TCE exposure. Of
these subjects, 3% (2,267 out of 77,965 subjects) had "routine" defined as use of TCE as
part of daily job activities and an additional 3,176 subjects (4%) had potential
"intermittent" based upon job title and judgment of nonroutine or nondaily TCE usage
and were included in the mortality analysis. No information was provided on building and
working conditions or the frequency of exposure-related tasks, and no atmospheric
monitoring data were available on TCE, although some limited data were available after
1970 on other solvents such as perchloroethylene, which replaced TCE in 1966 in vapor
degreasing, methylene chloride, and 1,1,1-trichloroethane. Without more information, it is
not possible to determine the quality of some of the TCE assignments. This study had
limited ability to detect exposure-related effects given its use of duration of exposure, a
poor exposure metric given subjects may have differing exposure intensity with similar
exposure duration (NRC, 2006). Lacking monitoring information, analyses examining the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09
B-107
DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
number of years of routine and intermittent TCE exposure are likely biased due to
exposure misclassification related to inability to account for changes in process and
chemical usage patterns over time. Stewart et al. (1991) show atmospheric TCE
concentrations decreased over time. Similarly, an observation of inverse relationship
between some site-specific causes of death and duration of exposure may be due to selection
bias or to misallocation of person-years of follow-up (NYSDOH, 2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-108 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Boice JD, Marano DE, Fryzek JP, Sadler CJ, McLaughlin JK. (1999). Mortality among aircraft manufacturing workers.
Occup Environ Med 56:581-597.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract: "To evaluate the risk of cancer and other diseases among workers
engaged in aircraft manufacturing and potentially exposed to compounds containing
chromate, trichloroethylene (TCE), perchloroethylene (PCE), and mixed solvents."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
All workers employed on or after 1-1-1960 for at least 1 yr at Lockheed Martin
aircraft manufacturing factories in California.
Control population: U.S. mortality rates or factory workers no exposed to any solvent
(internal referents).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD code in use at the time of death.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Qualitative. Few exposure measurements existed prior to the late 1970s, a period
after TCE had been discontinued at Lockheed-Martin aircraft manufacturing
factories.
Subjects are categorized as potentially TCE exposed received on a routine basis
(2,075 subjects), daily job activity, or routine and intermittent basis (3,016 subjects),
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
nonroutine or nondaily TCE usage, based on information on Service Record and
Permanent Employment Record (Kardex) and other sources of job history
information for subjects lacking Kardex cards.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
Co
s
o
to
o
s
Co
3"
rS"
>9
to
S
"tS
§•
TO
£
a
TO
S
Co
>T
<§>
>s
>S
TO
TO'
*
S
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
This study does not adopt methods to verify vital status of employees. All workers
for which death certificate were not found are assumed to be alive until end of
follow-up.
>50% cohort with full latency
Average follow-up of TCE cohort was 29 yrs.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,100 total deaths and 277 cancer deaths in TCE subcohort.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
SMR analysis—age, sex and calendar-time.
Poisson regression using internal referents—birth date, date first employed, date of
finishing employment, race, and sex.
Statistical methods
SMR for routine TCE exposure subcohort.
Poisson regression for routine and intermittent TCE exposure subcohort.
Exposure-response analysis presented in
published paper
Duration of exposure for subjects with Kardex cards only—
2-sides test for linear trend.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Documentation of results
Adequate.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.1.4. Morgan et al. (2000a; 1998).
B.3.1.1.4.1. Author's abstract.
We measured mortality rates in a cohort of 20,508 aerospace workers who were
followed up over the period 1950-1993. A total of 4,733 workers had
occupational exposure to trichloroethylene. In addition, trichloroethylene was
present in some of the washing and drinking water used at the work site. We
developed a job-exposure matrix to classify all jobs by trichloroethylene exposure
levels into four categories ranging from "none" to "high" exposure. We calculated
standardized mortality ratios for the entire cohort and the trichloroethylene
exposed subcohort. In the standardized mortality ratio analyses, we observed a
consistent elevation for nonmalignant respiratory disease, which we attribute
primarily to the higher background rates of respiratory disease in this region. We
also compared trichloroethylene-exposed workers with workers in the "low" and
"none" exposure categories. Mortality rate ratios for nonmalignant respiratory
disease were near or less than 1.00 for trichloroethylene exposure groups. We
observed elevated rare ratios for ovarian cancer among those with peak exposure
at medium and high levels] relative risk (RR) = 2.74; 95% confidence interval
(CI) = 0.84-8.99] and among women with high cumulative exposure (RR = 7.09;
95% CI = 2.14-23.54). Among those with peak exposures at medium and high
levels, we observed slightly elevated rate ratios for cancers of the kidney (RR =
1.89; 95% CI = 0.85-4.23), bladder (RR = 1.41; 95% CI = 0.52-3.81), and
prostate (RR = 1.47; 95% CI = 0.85-2.55). Our findings do not indicate an
association between trichloroethylene exposure and respiratory cancer, liver
cancer, leukemia or lymphoma, or all cancers combined.
Erratum:
One of the authors of the article entitled Mortality of aerospace workers exposed
to trichloroethylene, by Robert W. Morgan, Michael A. Kelsh, Ke Zhao, and
Shirley Heringer, published in Epidemiology (1998);9:424-431, informed us of
some errors in one of the tables. In Table 5, the authors had inadvertently included
both genders in counting person-years, rather than presenting gender-specific risk
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-l 13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
ratios for prostate and ovarian cancer. In addition, one subject, in the high
trichloroethylene (TCE) exposure category, had been incorrectly classified with a
diagnosis of ovarian cancer, instead of other female genital cancer. The authors
report that correction of these errors did not change the overall conclusions of the
study. The correct estimates of effect for prostate and ovarian cancer are
presented in the Table below.
B.3 .1.1.4.2. Study description and comment. This study of a cohort of 20,508 aircraft
manufacturing workers employed for at least 6 months between 1950 and 1985 at Hughes
Aircraft in Arizona was followed through 1993 for mortality. Cause-specific SMRs are
resented for the entire cohort and the TCE-subcohort using U.S. Mortality rates from
1950-1992 as referents. Additionally, internal cohort analyses fitting Cox proportional
hazards models are presented comparing risks for those with TCE exposure to never-
exposed subjects. Morgan et al. (2000a; 1998) do not identify job titles of individuals in the
never-exposed group; however, it is assumed these individuals were likely white-collar
workers, administrative staff, or other blue-collar worker with chemical or solvents
exposures other than TCE.
The company conducted a limited semiquantitative assessment of TCE exposure based
on the judgment of long-term employees. Most TCE exposure occurred in vapor degreasing
units between 1952 and 1977. No details were provided on the protocol for processing the jobs
in the work histories into job classifications; no examples were provided. Additionally, no
information is provided other chemical exposures that may also have been used in the different
jobs. Of the 20,508 subjects, 4,733 were identified with TCE exposure. Exposure categories
were assigned to job classifications: high = worked on degreasers (industrial hygiene reported
exposures were >50 ppm); medium = worked near degreasers; and low = work location was
away from degreasers but "occasional contact with (trichloroethylene)." There was also a "no
exposure" category. No data were provided on the frequency of exposure-related tasks. Without
more information, it is not possible to determine the quality of some of these assignments. Only
the high category is an unambiguous setting. Depending on how the degreasers were operated,
operator exposure to trichloroethylene might have been substantially greater than 50 ppm.
Furthermore, TCE intensity likely changed over time with changes in degreaser operations and
exposure assignment based on job title only is able to correctly place subjects with a similar job
title but held at different time periods. Furthermore, there are too many possible situations in
which an exposure category of medium or low might be assigned to determine whether the
ranking is useful. Therefore, the medium and low rankings are likely to be highly misclassified.
Deficiencies in job rankings are further magnified in the cumulative exposure groupings.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-l 14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Internal analyses examine TCE exposed, defined as low and high cumulative exposure,
compared to never-TCE exposed subjects. Low cumulative exposure group includes any
workers with the equivalent of up to 5 years of exposure at jobs at low exposure or 1.4 years of
medium exposure; all other workers were placed in the high cumulative exposure grouping.
Ambiguity in low and medium job rankings and the lack of exposure data to define "medium"
and "low" precludes meaningful analysis of cumulative exposure, specifically, and
exposure-response, generally.
The development of exposure assignments in this study was insufficient to define
exposures of the cohort and bias related to exposure misclassification is likely great. The
inability to account for changes in TCE use and exposure potential over time introduces bias and
may dampen observed risks. This study had limited ability to detect exposure-related effects
and, overall, limited ability to provide insight on TCE exposure and cancer outcomes.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-l 15 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Morgan RW, Kelsh MA, Zhao K, Heringer S. (1998). Mortality of aerospace workers exposure to trichloroethylene.
Epidemiol 9:424-431.
Morgan RW, Kelsh MA, Zhao K, Heringer S. (2000a). Mortality of aerospace workers exposed to trichloroethylene.
Erratum. Epidemiology 9:424-431.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
"measured mortality rates in a cohort of aerospace workers, comparing TCE workers
with workers in low and none exposure categories."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
20,508 male and female workers are identified using company records and who were
employed at plant for at least 6 mos between 1-1-1950 and 12-31-1985.
TCE subcohort—4,733 (23%) male and female subjects.
External referents—U.S. population rates, 1950-1992.
Internal referents—Analysis of peak exposure, Low or no TCE exposure; analysis of
cumulative exposure, never exposed to TCE. Internal referents are likely white-collar
workers, administrative staff, and blue-collar workers with chemical exposure other
than TCE. White-collar and administrative staff subjects are not representative of
blue-collar workers due to SES and sex differences. Also, the never-TCE exposed
blue-collar workers may potentially have other chlorinated solvents exposures,
exposures that may be associated with a similar array of targets as TCE. These
individuals may not be representative of a nonchemical exposed population as that
used in Blair et al. (1998).
CATEGORY B: ENDPOINT MEASURED
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Levels of health outcome assessed
Mortality
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
No, ICD in use at time of death (ICD 7, 8, 9).
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Semiquantitative. Limited IH measurements before 1975. Jobs ranked into high,
medium, or low intensity exposure categories; categories are undefined as to TCE
intensity. Jobs with high intensity exposure rating involved work on degreaser
machines with TCE exposure equivalent to 50 ppm; assigned exposure score of 9.
Job with medium rating were near (distance undefined in published paper) degreasing
area and a score of 4. Jobs with low rating were away (undefined distance) from
degreasing area and assigned score of 1. Cumulative exposure score = £ (duration
exposure x score). Peak exposure defined by job with highest ranking score.
CATEGORY D: FOLLOW-UP (Cohort)
More than 10% loss to follow-up
No, 27 subjects were excluded from analysis due to missing information.
>50% cohort with full latency
Average 22 yrs of follow-up for TCE subcohort.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
TCE subcohort—917 total deaths (19%>) of subcohort, 270 cancer deaths.
CATEGORY H: ANALYSIS
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
o
o
s
Co
I?
Oq ^
53
S"4
>}'
§•
to
s
a
to
s
S"
a
a,
>r
<§>
>s
Control for potential confounders in statistical
analysis
Age, race, sex, and calendar year in SMR analysis.
Internal analysis- age (for bladder, prostate, ovarian cancers) and, age and sex (liver,
kidney cancers).
Statistical methods
Life table analysis (SMR).
Cox proportional hazards modeling (unexposed subjects as internal referents)—peak
and two-levels of cumulative exposure (EHS, 1997; Morgan et al., 1998); any TCE
exposure (EHS, 1997).
Exposure-response analysis presented in
published paper
Qualitative presentation, only; no formal statistical test for linear trend.
Documentation of results
Adequate.
to
*
«
s
TO
Co
SES = socio-economic status.
I
vo
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.1.5. Costa et al. (1989).
B.3.1.1.5.1. Author's abstract.
Mortality in a cohort of 8626 workers employed between 1954 and 1981 in an
aircraft manufacturing factory in northern Italy was studied. Total follow up was
132,042 person-years, with 76% accumulated in the age range 15 to 54. Median
duration of follow up from the date of first employment was 16 years. Vital status
was ascertained for 98.5% of the cohort. Standardized mortality ratios were
calculated based on Italian national mortality rates. Altogether 685 deaths
occurred (SMR = 85). There was a significant excess of mortality for melanoma
(6 cases, SMR = 561). Six deaths certified as due to pleural tumors occurred. No
significant excess of mortality was found in specific jobs or work areas.
B.3.1.1.5.2. Study description and comment. This study assesses mortality in a small
cohort of 8,626 aircraft manufacturing workers employed between 1954 and the end of
follow-up in June, 1981. A period of minimum employment duration before accumulating
person-years was not a prerequisite for cohort definition. The cohort included employees
identified as blue collar workers, technical staff, administrative clerks, and white-collar
workers. Blue-collar workers comprised 7,105 of the 8,626 cohort subjects. Mortality was
examined for all workers and included job title of blue collar workers, technical staff
members, administrative clerks, and white collar workers- not otherwise specified. No
exposure assessment was used and the published paper does not identify chemical
exposures. In fact, Costa et al. (1989) do not even mention TCE in the paper.
Overall, the lack of exposure assessment, the inability to identify TCE as an exposure to
this cohort, and the inclusion of subjects who likely do not have potential TCE exposure are
reasons why this study is not useful for determining whether trichloroethylene may cause
increased risk of disease.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-120 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Costas G, Merletti F, Segnan N. (1989). A mortality study in a north Italian aircraft factory. Br J Ind Med 46:738-743.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The 1st paragraph of the paper identified this study was carried out to investigate an
apparently high number of malignant tumors among employees that were brought to
the attention of the local health authority by staff representative. This study was not
designed to examine TCE exposure and cancer outcomes.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cohort is defined as all workers every employed between 1-1-1954 and 6-30-1981
(end of follow-up) at a north Italian aircraft manufacturing factory. Cohort include
8.626 subjects: 950 women (636 clerks, 314 blue-collar workers/technical staff) and
7,676 men (5,625 blue collar workers, 965 technical staff, 571 administrative clerks,
and 515 white collar workers).
External referent—Age, year (5-yr periods over 1955-1981)-sex and cause-specific
death rates of Italian population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Causes and underlying causes of death coded to ICD rule in effect at the time of
death and grouped into categories consistent with ICD 8th revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Exposure is defined as employment in the factory. TCE is not mentioned in
published paper and no exposure assessment was carried out by study investigators.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
Vital status ascertained for 98% of cohort; 2% could not be traced (1% unknown and
1% had emigrated).
>50% cohort with full latency
Average mean follow-up: males, 17 yrs; females, 13 yrs.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
642 total deaths, 168 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex and calendar year.
Statistical methods
SMR.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Adequate.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.1.6. Garabrant et al. (1988).
B.3.1.1.6.1. Author's abstract.
A retrospective cohort mortality study was conducted among men and women
employed for four or more years, between 1958 and 1982, at an aircraft
manufacturing company in San Diego County. Specific causes of death under
investigation included cancer of the brain and nervous system, malignant
melanoma, and cancer of the testicle, which previous reports have suggested to be
associated with work in aircraft manufacturing. Follow-up of the cohort of 14,067
subjects for a mean duration of 15.8 yr from the date of first employment resulted
in successful tracing of 95% of the cohort and found 1,804 deaths through 1982.
Standardized mortality ratios (SMRs) were calculated based on U. S. national
mortality rates and separately based on San Diego County mortality rates.
Mortality due to all causes was significantly low (SMR = 75), as was mortality
due to all cancer (SMR = 84). There was no significant excess of cancer of the
brain, malignant melanoma, cancer of the testicle, any other cancer site, or any
other category of death. Additional analyses of cancer sites for which at least ten
deaths were found and for which the SMR was at least 110 showed no increase in
risk with increasing duration of work or in any specific calendar period. Although
this study found no significant excesses in cause-specific mortality, excess risks
cannot be ruled out for those diseases that have latency periods in excess of 20 to
30 yr, or for exposures that might be restricted to a small proportion of the cohort.
B.3.1.1.6.2. Study description and comment. This study reported on the overall
mortality of a cohort of workers in the aircraft manufacturing industry in southern
California who had worked 1 day at the facility and had at least 4 years duration of
employment. Fifty-four (54) percent of cohort entered cohort at beginning date (1-1-1958).
This is a survivor cohort. This study lacks exposure assessment for study subjects. The
only exposure metric was years of work. Examination of jobs held by 70 study subjects, no
details provided in paper on subject selection criteria, identified 37% as having possible
trichloroethylene TCE exposure, but no information was presented on how they were
exposed, frequency or duration of exposure, or job titles associated with exposure. No
information is provided on possible trichloroethylene exposure to the remaining -14,000
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-124 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
subjects in this cohort. The exposure assignment in this study was insufficient to define
exposures of the cohort and the frequency of exposures was likely low. Given the enormous
misclassification on exposure, the effect of exposure would have to be very large to be
detected as an overall risk for the population. Null findings are to be expected due to bias
likely associated with a survivor cohort and to exposure misclassification. Therefore, this
study provides little information on whether trichloroethylene is related to disease risk.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-125 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Garabrant DH, Held J, Langholz B, Bernstein L. (1988). Mortality of Aircraft Manufacturing Workers in Southern
California. Am J Ind Med 13:683-693.
Langholz B, Goldstein L. (1996). Risk Set Sampling in Epidemiologic Cohort Studies. Stat Sci 11:35-53.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
"Our objects were to evaluate the oval mortality among the [aircraft manufacturing]
workers and to test the hypotheses that brain tumors, malignant melanoma, and
testicular neoplasms are associated with work in this industry." [Introduction]
This study was not designed to evaluate any specific exposure, but rather
employment in aircraft manufacturing industry.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
14,067 males and females working at least 4 yrs with a large aircraft manufacturing
company and who had worked for at least 1 day at a factory in San Diego County,
CA. Person-year accrued from the anniversary date of an individual's 4th yr of
service or from 1-1-1958 to end of follow-up 12-31-1982.
External referents—age-, race-, sex-, calendar year- and cause-specific mortality
rates of United States population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD revision in effect at the date of death. Lymphomas in 4 groupings:
lymphosarcoma and reticulosarcoma, HD, leukemia and aleukemia, and other.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD revision in effect at the date of death. Lymphomas in 4 groupings:
lymphosarcoma and reticulosarcoma, HD, leukemia and aleukemia, and other.
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Exposure assessment is lacking for all subjects except 70 deaths (14 esophageal and
56 others) who were included in a nested case-control study. Of the 362 jobs held by
these 70 subjects, 37% were identified as having potential for TCE exposure.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
4.7%) with unknown vital status.
>50% cohort with full latency
Average 16 yr follow-up.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,804 deaths (12.8%> of cohort), 453 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, race, sex, and calendar year.
Statistical methods
SMR.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
a,	s
Cf	^
to	s
Co
<3	Co'
8.	Si
§	t
&.	>
s	?
^	5
!>q	S.
TO
^	^3
^	5
s.	^
• TO	Co
V;	to
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
SMR analysis, adequate; Published paper lacks documentation of nested case-control
study of esophageal cancer.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.2. Cancer Incidence Studies Using Biological Monitoring Databases
Finland and Denmark historically have maintained national databases of biological
monitoring data obtained from workers in industries where toxic exposures are a concern.
Legislation required that employers provide workers exposed to toxic hazards with regular health
examinations, which must include biological monitoring to assess the uptake of toxic chemicals,
including trichloroethylene. In Sweden, the only local producer of trichloroethylene operated a
free exposure-surveillance program for its customers, measuring U-TCA. These programs used
the linear relationship found for average inhaled trichloroethylene versus U-TCA:
3	3
trichloroethylene (mg/m ) = 1.96; U-TCA (mg/L) = 0.7 for exposures lower than 375 mg/m
(69.8 ppm) (Ikeda et al., 1972). This relationship shows considerable variability among
individuals, which reflects variation in urinary output and activity of metabolic enzymes.
Therefore, the estimated inhalation exposures are only approximate for individuals but can
provide reasonable estimates of group exposures. There is evidence of nonlinear formation of
U-TCA above about 400 mg/m3 or 75 ppm of trichloroethylene. The half-life of U-TCA is about
100 hours. Therefore, the U-TCA value represents roughly the weekly average of exposure from
all sources, including skin absorption. The Ikeda et al. (1972) relationship can be used to convert
urinary values into approximate airborne concentration, which can lead to misclassification if
tetrachloroethylene and 1,1,1-trichloroethane are also being used because they also produce
U-TCA. In most cases, the Ikeda et al. relationship (1972) provides a rough upper boundary of
exposure to trichloroethylene.
B.3 .1.2.1. Hansen et al. (2001).
B.3.1.2.1.1. Author's abstract.
Human evidence regarding the carcinogenicity of the animal carcinogen
trichloroethylene (TCE) is limited. We evaluated cancer occurrence among 803
Danish workers exposed to TCE, using historical files of individual air and
urinary measurements of TCE-exposure. The standardized incidence ratio (SIR)
for cancer overall was close to unity for both men and women who were exposed
to TCE. Men had significantly elevated SIRs for non-Hodgkin's lymphoma (SIR
= 3.5; n = 8) and cancer of the esophagus (SIR = 4.2; n = 6). Among women, the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-129 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
SIR for cervical cancer was significantly increased (SIR = 3.8; n = 4). No clear
dose-response relationship appeared for any of these cancers. We found no
increased risk for kidney cancer. In summary, we found no overall increase in
cancer risk among TCE-exposed workers in Denmark. For those cancer sites
where excesses were noted, the small numbers of observed cases and the lack of
dose-related effects hinder etiological conclusions.
B.3.1.2.1.2. Study description and comment. This Danish study evaluated cancer
incidence in a small cohort of individuals (n = 803) who had been monitored for
trichloroethylene exposures in a national surveillance program between 1947 and 1989 for
U-TCA or TCE in breath since 1974. In all, 2,397 samples were analyzed for U-TCA of
workers at 275 companies and 472 breathing zone samples of TCE from workers at 81
companies. Individual workers could not be identified for roughly one-third of the U-TCA
measurements and 50% of breathing zone measurements; many of the individuals most
likely had died prior to 1968, the start of the Central Population Registry from which
workers were identified and follow-up for cancer incidence. A cohort of 658 males and 145
females were identified from the remaining 1,519 U-TCA and 245 air-TCE measurements.
Only two of 803 cohort subjects had both urine and air measurements. Follow-up for
cancer incidence ended as of 12-31-1996.
The retirement and measurement records contained general information about the type of
employer and the subject's job. The subjects in this study came predominantly from the iron and
metal industry with jobs such as metal-product cleaner. Each subject had 1 to 27 measurements
of U-TCA measurements, an average of 2.2 per subject, going back to 1947. Using the linear
relationship from Ikeda et al. (1972), the historic median exposures estimated from the U-TCA
concentrations were low: 9 ppm for 1947 to 1964, 5 ppm for 1965 to 1973, 4 ppm for 1974 to
1979, and 0.7 ppm for 1980 to 1989. However, the distributions were highly skewed.
Additionally, 5% of the cohort had urine or air samples below the limit of detection. Overall,
median exposure in this cohort was 4 ppm and suggests that, in general, workers in a wide
variety of industry and job groups and identified as "exposed" in this study had low TCE
intensity exposures. Overall, the cohort in this study is small, drawn from a wide variety of
industries, predominantly degreasing and metal cleaning, and had generally low exposures (most
less than 20 ppm). The study has a lower power to examine TCE exposure and cancer for these
reasons.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-130 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Hansen J, Raaschou-Nielsen O, Christensen JM, Johansen I, McLaughlin JK, Lipworth L, Blot WJ, Olsen JH. (2001). Cancer
incidence among Danish workers exposed to trichloroethylene. J Occup Environ Med 43:133-139.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From introduction—A study of incidence was carried out to address shortcomings in
earlier TCE studies related to the lack of direct exposure information and to
assessment of mortality as opposed to incidence.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
803 subjects identified from biological monitoring of urine TCA from 1947-1989
(1,519 measurements) or breathing zone TCE since 1974 (245 measurements) and
who were alive as of 1968, followed to 1996.
External referents—cancer incidence rates of Danish population (age-, sex-, calendar
years-, and site-specific).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 7th revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Biological marker of TCE in urine or in breath used to assign TCE exposure to
cohort subject. Historic median exposures estimated from the U-TCA were low:
9 ppm for 1947 to 1964, 5 ppm for 1965 to 1973, 4 ppm for 1974 to 1979, and
0.7 ppm for 1980 to 1989. Overall, median TCE exposure to cohort was 4 ppm
(arithmetic mean, 12 ppm).
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No.
>50% cohort with full latency
Unable to determine given insufficient information in paper; however, text notes
follow-up for most subjects achieved a full latency.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% Face-to-Face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
128 incident cancers among 804 cohort subjects (15%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex and calendar year.
Statistical methods
SIR, Life table analysis.
Exposure-response analysis presented in
published paper
Yes, as dichotomous variable for mean exposure (<4 ppm, 4+ ppm) and for
cumulative exposure.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.2.2. Anttila et al. (1995).
B.3.1.2.2.1. Author's abstract.
Epidemiologic studies and long-term carcinogenicity studies in experimental
animals suggest that some halogenated hydrocarbons are carcinogenic. To
investigate whether exposure to trichloroethylene, tetrachloroethylene, or
1,1,1-trichloroethane increases carcinogenic risk, a cohort of 2050 male and 1924
female workers monitored for occupational exposure to these agents was followed
up for cancer incidence in 1967 to 1992. The overall cancer incidence within the
cohort was similar to that of the Finnish population. There was an excess of
cancers of the cervix uteri and lymphohematopoietic tissues, however. Excess of
pancreatic cancer and non-Hodgkin lymphoma was seen after 10 years from the
first personal measurement. Among those exposed to trichloroethylene, the
overall cancer incidence was increased for a follow-up period of more than 20
years. There was an excess of cancers of the stomach, liver, prostate, and
lymphohematopoietic tissues combined. Workers exposed to 1,1,1-trichloroethane
had increased risk of multiple myeloma and cancer of the nervous system. The
study provides support to the hypothesis that trichloroethylene and other
halogenated hydrocarbons are carcinogenic for the liver and lymphohematopoietic
tissues, especially for non-Hodgkin lymphoma. The study also documents excess
of cancers of the stomach, pancreas, cervix uteri, prostate, and the nervous system
among workers exposed to solvents.
B.3.1.2.2.2. Study description and comment. This Finnish study evaluated cancer risk in
a small cohort of individuals (2,050 males and 1,924 females) who had been monitored
between 1965 and 1982 for exposures to trichloroethylene by measuring their U-TCA. The
main source of exposure was identified as degreasing or cleaning metal surfaces. Some
workplaces identified rubber work, gluing, and dry-cleaning. There was an average of 2.7
measurements per person. Using the Ikeda et al. (1972) conversion relationship, the
exposure for trichloroethylene was approximately 7 ppm in 1965, which declined to
approximately 2 ppm in 1982; the 75th percentiles for these dates were 14 and 7 ppm,
respectively. The maximum values for males were approximately 380 ppm during 1965 to
1974 and approximately 96 ppm during 1974 to 1982. Females showed a similar pattern
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-134 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
over time but had somewhat higher exposures than males before the 1970s. Median TCE
exposure for females of 4 ppm compared to 3 ppm for males; maximum values were
similar for both sexes. Duration of exposure was counted from the first measurement of U-
TCA, which might underestimate the length of exposure. Without job histories, the length
of exposure is uncertain. Another concern is the sampling strategy; it was not reported
how the workers were chosen for monitoring. Therefore, it is not clear what biases might
be present, especially the possibility of under sampling highly exposed workers.
1	Overall, this TCE exposed cohort drawn from a wide variety of industries was twice the
2	size of other Nordic biomonitoring studies (Axelson et al., 1994; Hansen et al., 2001) with urine
3	TCA measurements from a more recent period, 1965 to 1982, compared to other Nordic studies
4	of Danish cohorts, 1947 to 1980s, or Swedish cohorts, 1955 to 1975 (Axelson et al., 1994;
5	Hansen et al., 2001; Raaschou-Nielsen et al., 2002). Exposures to trichloroethylene were
6	generally low, less than 14 ppm for the 75th percentile of all measurements, and median TCE
7	exposures decreasing from 7 ppm to 2 ppm over the 17-year period. The medians are similar to
8	estimated exposures to Danish workers with biological markers of U-TCA (Hansen et al., 2001;
9	Raaschou-Nielsen et al., 2001). The duration of exposure was uncertain.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-135 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Anttila A, Pukkala E, Sallmen M, Hernberg S, Hemminki K. (1995). Cancer incidence among Finnish workers exposed to
halogenated hydrocarbons. J Occup Environ Med 37:797-806.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes, study aim was to assess cancer incidence among workers biologically monitored
for exposure to TCE, PERC, and 1,1,1-trichloroethane.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
3, 976 subjects identified from biological monitoring of urine TCA between 1965 to
1982; PERC in blood, 1974 to 1983; and, 1,1,1-trichloroethane in blood, 1975 to
1983 (a total of 10.743 measurements). 109 of cohort subjects with TCE poisoning
report between 1965 to 1976. Follow-up for mortality between 1965 to 1991 and for
cancer between 1967 to 1992.
TCE subcohort—3,089 (1,698 males, 1,391 females).
External referents—age-, sex-, calendar year-, and site-specific cancer incidence rates
of the Finnish population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality and cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 7th revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Exposure assessment approach, including Biological marker of TCE in urine used to assign TCE exposure for TCE subcohort.
adoption of JEM and quantitative exposure There were on average 2.5 U-TCA measurements per individual. 6% of cohort had
estimates	measurements for 2 or all three solvents. The overall median of U-TCA for females
was 8.3 mg/L and 6.3 mg/L for males, and before 1970, 10 to 13 mg/L for females
and 13 to 15 mg/L for males. Using Ikeda et al. (1972) relationship for U-TCA
and TCE concentration, median TCE exposures over the period of study were
roughly <4-9 ppm (median, 4 ppm; arithmetic mean, 6 ppm).
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No.
>50% cohort with full latency
Yes, 18 yr mean follow-up period.
CATEGORY E: INTERVIEW TYPE
<90% Face-to-Face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
208 cancers among 3,089 TCE-exposed subjects (7%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and calendar year.
Statistical methods
SMR and SIR, Life table analysis.
Exposure-response analysis presented in
published paper
Yes, U-TCA as dichotomous variable (<6 ppm, 6+ ppm).
Documentation of results
Adequate for SIR analysis; details on SMR analysis of TCE subcohort are few.
PERC = perchloroethylene, SIR = standardized incidence ratio.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.2.3. Axelson et al. (1994).
B.3.1.2.3.1. Author's abstract.
There is limited evidence for mutagenicity and carcinogenicity of
trichloroethylene (TRI) in experimental test systems. Whether TRI is a human
carcinogen is unclear, however. This paper presents an update and extension of a
previously reported cohort of workers exposed to TRI, in total 1670 persons.
Among men (n = 1421), the overall standardized mortality ratio (SMR) and
cancer morbidity ratio (SIR) were close to the expected, with SMR, 0.97; 95%
confidence interval (CI), 0.86 to 1.10; and SIR, 0.96; 95% CI, 0.80 to 1.16,
respectively. The cancer mortality was significantly lower than expected (SMR,
0.65; 95% CI, 0.47 to 0.89), whereas an increased mortality from circulatory
disorders (cardiovascular, cerebrovascular) was of borderline significance (SMR,
1.17; 95% CI, 1.00 to 1.37). No significant increase of cancer of any specific site
was observed, except for a doubled incidence of nonmelanocytic skin cancer
without correlation with the exposure categories. In the small female subcohort
(n = 249), a nonsignificant increase of cancer and circulatory deaths was observed
(SMR, 1.53 and 2.02, respectively). For both genders, however, excess risks were
largely confined to groups of workers with lower exposure levels or short duration
of exposure or both. It is concluded that this study provides no evidence that TRI
is a human carcinogen, i.e., when the exposure is as low as for this study
population.
B.3.1.2.3.2. Study description and comment. This Swedish study evaluated cancer risk
in a small cohort of individuals (1,421 males and 249 females), who were monitored for U-
TCA as part of a surveillance system by the trichloroethylene producer during 1955 to
1975. Both mortality between 1955 and 1986 and cancer morbidity between 1958 and 1987
are assessed in males only due to the small number of female subjects. Eighty-one percent
of the male subjects had low exposures (<50 mg/L), corresponding to an airborne
concentration of trichloroethylene of approximately 20 ppm. There was uncertainty about
the beginning and end of exposure. Exposure was assumed to begin with the first urine
sample and to end in 1979 (the reason for this date is unclear). Because the investigators
did not have job histories, there is considerable uncertainty about the duration of exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-139 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
No information is, additionally, presented to evaluate if a large proportion of the cohort
had a full latency period for cancer development. Most subjects appear to have had short
durations of exposure, but these might have been underestimated. Another concern is the
sampling strategy. It was not reported how the workers were chosen for monitoring.
Therefore, it is not clear what biases could be present in the data, especially the possibility
of under sampling highly exposed workers.
1	Overall, this study had a small cohort drawn from a wide variety of industries,
2	predominantly from industries involving degreasing and metal cleaning. Exposure to
3	trichloroethylene was generally low (most less than 20 ppm). The duration of exposure was
4	uncertain and bias related to under sampling of higher exposed workers is possible but can not be
5	evaluated.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-140 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Axelson O, Selden A, Andersson K, Hogstedt C. (1994). Updated and expanded Swedish cohort study on trichloroethylene
and cancer risk. J Occup Environ 36:556-562.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes- "This paper present an update and extension of a previously reported cohort of
workers exposure to TCE."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
1,670 subjects (1,421 males, 249 females) with records of biological monitoring of
urine TCA from 1955 and 1975.
Analysis restricted to 1,421 males.
External referents—age-, sex-, calendar year-, site-specific mortality or cancer
incidence rates of Swedish population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence from 1958 to 1987 and all-cause mortality from 1955 to 1986.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 7th revision.
ICD, 8l revision from 1975 onward for all lympho-hematopoietic system cancers.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Biological marker of TCE in urine used to assign TCE exposure to cohort subject.
No extrapolation of U-TCA data to air-TCE concentration. Roughly % of cohort
had U-TCA concentrations equivalent to <20 ppm TCE.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
Sf
<§>
>S
TO
*
«
s
TO
>;
0
s
1
a.
VO
to
o
>50% cohort with full latency
Insufficient to estimate for full cohort; however, 42% of person years in subjects
with 2+ exposure years also had 10+ yrs of latency.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
229 deaths (16% of male subjects).
107 incident cancer cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and calendar year.
Statistical methods
SMR—age, sex, and calendar-year.
SIR—analyses restricted to males—age and calendar-year.
Exposure-response analysis presented in
published paper
Yes, by 3 categories of U-TCA concentration.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.3. Studies in the Taoyuan Region of Taiwan
B. 3.1.3.1. Sung et al. (2007; 2008).
B.3.1.3.1.1. Sung et al.(2008) abstract.
There is limited evidence on the hypothesis that maternal occupational exposure
near conception increases the risk of cancer in offspring. This study is to
investigate whether women employed in an electronics factory increases
childhood cancer among first live born singletons. We linked the databases of
Birth Registration and Labor Insurance, and National Cancer Registry, which
identified 40,647 female workers ever employed in this factory who gave 40,647
first live born singletons, and 47 of them developed cancers during 1979-2001.
Mothers employed in this factory during their periconceptional periods (3 months
before and after conception) were considered as exposed and compared with those
not employed during the same periods. Poisson regression model was constructed
to adjust for potential confounding by maternal age, education, sex, and year of
birth. Based on 11 exposed cases, the rate ratio of all malignant neoplasms was
increased to 2.26 [95% confidence interval (CI), 1.12-4.54] among children
whose mothers worked in this factory during periconceptional periods. The RRs
were associated with 6 years or less (RR=3.05; 95% CI, 1.20-7.74) and 7-9 years
(RR=2.49; 95% CI, 1.26-4.94) of education compared with 10 years or more. An
increased association was also found between childhood leukemia and exposed
pregnancies (RR=3.83; 95% CI, 1.17-12.55). Our study suggests that maternal
occupation with potential exposure to organic solvents during periconception
might increase risks of childhood cancers, especially for leukemia.
B.3.1.3.1.2. Sung et al. (2007) abstract.
Background In 1994, a hazardous waste site, polluted by the dumping of
solvents from a former electronics factory, was discovered in Taoyuan, Taiwan.
This subsequently emerged as a serious case of contamination through chlorinated
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-144 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
hydrocarbons with suspected occupational cancer. The objective of this study was
to determine if there was any increased risk of breast cancer among female
workers in a 23-year follow-up period. Methods A total of 63,982 female
workers were retrospectively recruited from the database of the Bureau of Labor
Insurance (BLI) covering the period 1973-1997; the data were then linked with
data, up to 2001, from the National Cancer Registry at the Taiwanese Department
of Health, from which standardized incidence ratios (SIRs) for different types of
cancer were calculated as compared to the general population. Results There
were a total of 286 cases of breast cancer, and after adjustment for calendar year
and age, the SIR was close to 1. When stratified by the year 1974 (the year in
which the regulations on solvent use were promulgated), the SIR of the cohort of
workers who were first employed prior to 1974 increased to 1.38 (95%
confidence interval, 1.11-1.70). No such trend was discernible for workers
employed after 1974. When 10 years of employment was considered, there was a
further increase in the SIR for breast cancer, to 1.62. Those workers with breast
cancer who were first employed prior to 1974 were employed at a younger age
and for a longer period. Previous qualitative studies of interviews with the
workers, corroborated by inspection records, showed a short-term high exposure
to chlorinated alkanes and alkenes, particularly trichloroethylene before 1974.
There were no similar findings on other types of cancer. Conclusions Female
workers with exposure to trichloroethylene and/or mixture of solvents, first
employed prior to 1974, may have an excess risk of breast cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-145 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.3.1.3. Study description and comment. Sung et al. (2007) examine breast cancer
incidence among females in a cohort of electronic workers with employment at one factory
in Taoyuan, Taiwan between 1973 and 1992, date of factory closure and followed to 2001.
Some female subjects in Sung et al. (2007) overlap those in Chang et al. (2003; 2005) who
included workers from the same factory whose employment dates were between 1978 and
1997, the closing date of the study a date of vital status ascertainment. A total of 64,000
females were identified with 63,982 in the analysis after the exclusion of 15 women with less
than one full day of employment and three women with cancer diagnoses prior to the time
of first employment; approximately 6,000 fewer female subjects compared to Chang et al.
(21989005) (70,735 females). Cancer incidence between 1979 and 2001 as identified using
the National Cancer Registry which contained 80% of all cancer cases in Taiwan (Parkin et
al., 2002) is examined using life table methods with exposure lag periods of 5-15 years,
depending on the cancer site, and cancer rates from the larger Taiwanese population as
referent.
Company employment records were lacking and the cohort was constructed using the
Bureau of Labor Insurance database that contained computer records since 1978 and paper
records for the period 1973 to 1978. Duration of employment was calculated from the beginning
of coverage of labor insurance and is likely an underestimate. Labor insurance hospitalization
data and a United Labor Association list of names were used to verify cohort completeness.
While these sources may have been sufficient to identified current employees, their ability to
identify former employees may be limited, particularly from the hospitalization data if the
subject's current employer was listed.
This study assumes all employees in the factory were exposed to chlorinated organic
solvent vapors and the primary exposure index was duration of employment at the plant. Most
subjects had employment durations of <1 year (65%). Durations of exposure were likely
underestimated as dates of commencement and termination of insurance coverage were
incomplete, 7.5% and 6%, respectively. There is little to no information on chemical usage and
exposure assignment to individual cohort subjects. As reported in Chang et al. (2003; 2005),
records of the Department of Labor Inspection ad Bureau of International Trade, in addition, to
recall of former industrial hygienists were used to identify chemicals used after 1975 in the
plants. No information is available prior to this date.
Sung et al. (2008) presents an analysis of childhood cancer incidence (1979-2001)
among first liveborn singleton births (1978 and 2001) of female subjects employed at the plant
during a period 3 months before and after beginning of pregnancy, an estimate derived by Sung
et al. (2008) from the date of birth and estimated length of gestation plus 14 days. Sung et al.
(2007) used Poisson regression methods and cancer incidence among first liveborn births of all
other women in Taiwan in the same time to calculate relative risks associated with leukemia risk
among exposed offspring. Poisson models were adjusted for maternal age, maternal educational
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-146 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1	level, child's sex, and year of birth. A total of 8,506 first born singleton births among
2	63,982 female subjects were identified from the Taiwan Birth Registry database, and 11 cancers,
3	including 6 leukemia cases and no brain/central nervous system (CNS) cases identified from the
4	National Cancer Registry database.
5	Overall, these studies do not provide substantial weight for determining whether
6	trichloroethylene may cause increased risk of disease. The lack of TCE-assessment to individual
7	cohort subjects; grouping cohort subjects with different exposure potential, both to different
8	solvents and different intensities; and deficiencies in the record system used to construct the
9	cohort introduce uncertainty.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-147 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Sung T-I, Chen P_C, Lee L J-H, Lin Y-P, Hsieh G-Y, Wang J-D. (2007). Increased standardized incidence ratio of breast
cancer in female electronics workers. BMC Public Health 7:102. http://www.biomedcentral.com/content/pdf/1471-2458-7-
102.pdf.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or hypothesis
From abstract "This study is to investigate whether women employed in an
electronics factory increases childhood cancer among first live born singletons."
This study was not able to evaluate TCE exposures uniquely.
Selection and characterization in cohort studies of
exposure and control groups and of cases and
controls in case-control studies is adequate
63,982 females, some who were also subjects were also in cohort of Chang et al.
(2003; 2005) with 70,735 females.
Cohort initially established using labor insurance records (computer records after
1978 and paper records from 1973 and 1978) in the absence of company records.
Cohort definition dates are not clearly identified. Cohort identified from records
covering period 1973 and 1997 with vital status ascertained as of 2001. Factory
closed in 1992.
External referents: age-, calendar-, and sex-specific incidence rates of the
Taiwanese general population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence as ascertained from National (Taiwan) Cancer Registry (80% of
all cancers reported to Registry).
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's lymphoma
ICD-Oncology, a supplement to ICD-9.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including adoption
of JEM and quantitative exposure estimates
All employees assumed to be potentially exposed to chlorinated organic solvent
vapors; study does not assign potential chemical exposures to individual subjects.
No information on specific chemical exposures or intensity. Limited identification
of solvents used in manufacturing process from the period after 1975 inferred from
records of Department of Labor Inspection, Bureau of International Trade, and
former industrial hygienists recall. No information on solvent usage was available
before 1975.
Exposure index defined as duration of exposure which was likely underestimated.
21%) of cohort with >10 yrs duration of employment and 53% with <1 yr duration.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No information on loss to follow-up. Subject was assumed disease free at end of
follow-up if lacking cancer diagnosis as recorded in the National Cancer Registry.
>50% cohort with full latency
No, 57%o of cohort employed after November 21, 1978.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies; numbers
of exposed cases and prevalence of exposure in
case-control studies
1,311 cancer cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age-, calendar-, and sex-specific incidence rates.
Statistical methods
SIR, analyses include a lag period of 5, 10, or 15 yrs since first employment (as
indicated by labor insurance record).
Exposure-response analysis presented in published
paper
Cancer incidence examined by duration of employment; however, employment
durations were likely underestimates as dates of commencement and termination
dates on of insurance coverage date were incomplete and misclassification bias is
likely present.
Documentation of results
Inadequate—analyses that do not include a lag are not presented nor discussed in
published paper or in supplemental documentation.
SIR = standardized incidence ratio.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Sung T-I, Wang J-D, Chen P_C. (2008). Increased risk of cancer in the offspring of female electronics workers. Reprod
Toxicol 25:115-119.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract "The study was designed to examine whether breast cancer risk in
females was increased, as had been observed in Chang et al. (2003; 2005) in a cohort
with earlier employment dates." This study was not able to evaluate TCE exposure.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
11 cancers among 8,506 first born singleton births between 1978-2001 in
63,982 female subjects of Sung et al. (2007). Cancers identified from National
Cancer Registry and births identified from Taiwan Birth Registration database.
External referents: cancer incidence among all other first birth singleton births
among Taiwanese females over the same time period.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence as ascertained from National (Taiwan) Cancer Registry (80% of all
cancers reported to Registry).
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-Oncology, a supplement to ICD-9, specific leukemia subtypes not identified in
paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
All births were among subjects with employment at factory during a period 3 mos
before and after beginning of pregnancy. All mothers were assumed potentially
exposed to chlorinated organic solvent vapors; specific solvents are not identified nor
assigned to individual subjects. Limited identification of solvents used in
manufacturing process from the period after 1975 inferred from records of
Department of Labor Inspection, Bureau of International Trade, and former industrial
hygienists recall. No information on solvent usage was available before 1975.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No information on loss to follow-up for females in Sung et al. (2007).
>50% cohort with full latency
66%o of births would have been 16 yrs of age as of 2001, the date cancer incidence
ascertainment ended.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
11 cancer cases among 8,506 first born singleton births.
CATEGORY H: ANALYSIS
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
>3
o
o
s
Co
§•
to
£
a
to
s
Co
>T
<§>
>s
"tS
TO
*
«
8
TO
Co
0
S
1
a.
VO
to
o
Control for potential confounders in statistical
analysis
Maternal age, maternal educational level, child's sex and child's year of birth.
Statistical methods
Poisson regression using childhood cancer incidence among all other first live born
children in Taiwan during same time period.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.3.2. Chang et al. (2003; 2005).
B.3 .1.3 .2.1. Chang et al. (2005) abstract.
A retrospective cohort morbidity study based on standardized incidence ratios
(SIRs) was conducted to investigate the possible association between exposure to
chlorinated organic solvents and various types of cancers in an electronic factory.
The cohort of the exposure group was retrieved from the Bureau of Labor
Insurance (BLI) computer database records dating for 1978 through December 31,
1997. Person-year accumulation began on the date of entry to the cohort, or
January 1, 1979 (whichever came later), and ended on the closing date of the
study (December 31, 1997), if alive with out contracting any type of cancers, or
the date of death, or the date of the cancer diagnosis. Vital status and cases of
cancer of study subjects were determined from January 1, 1979 to December 31,
1997 by linking cohort data with the National Cancer Registry Database. The
cancer incidence of the general population was used fro comparison. After
adjustment for age and calendar year, only SIR for breast cancer in the exposed
female employees were significantly elevated when compared with the Taiwanese
general population, based on the entire cohort without exclusion. The SIR of
female breast cancer also showed a significant trend of period effect, but no
significant dos-response relationship on duration of employment. Although the
total cancer as well as the cancer for the trachea, bronchus[,] and lung for the
entire female cohort was not significantly elevated, trend analysis by calendar-
year interval suggested an upward trend. However, when duration of employment
or latency was taken into consideration, no significantly elevated SIR was found
for any type of cancer in either male or female exposed workers. In particular, the
risk of female breast cancer was not indicated to be increased. No significant
dose-response relationship on duration of employment and secular trend was
found for the above-mentioned cancers. This study provides no evidence that
exposure to chlorinated organic solvents at the electronics factory was associated
with elevated human cancers.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-154 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.3.2.2. Chang et al. (2003) abstract.
PURPOSE: A retrospective cohort mortality study based on standardized
mortality ratios (SMRs) was conducted to investigate the possible association
between exposure to chlorinated organic solvents and various types of cancer
deaths. METHODS: Vital status and causes of death of study subjects were
determined from January 1, 1985 to December 31, 1997, by linking cohort data
with the National Mortality Database. Person-year accumulation began on the
date of entry to the cohort, or January 1, 1985 (whichever came later), and ended
on the closing date of the study (December 31, 1997), if alive; or the date of
death. RESULTS: This retrospective cohort study examined cancer mortality
among 86,868 workers at an electronics factory in the northern Taiwan. Using
various durations of employment and latency and adjusting for age and calendar
year, no significantly elevated SMR was found for any cancer in either male or
female exposed workers when compared with the general Taiwanese population.
In particular, the risk of female breast cancer was not found to be increased.
Although ovarian cancer suggested an upward trend when analyzed by length of
employment, ovarian cancer risk for the entire female cohort was not elevated.
CONCLUSIONS: It is concluded that this study provided no evidence that
exposure to chlorinated organic solvents was associated with human cancer risk.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-155 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.3.2.3. Study description and comment. Both Chang et al. (20198903) and Chang et
al. (2005) studied a cohort of 86,868 subjects employed at an electronics factory between
1985 and 1997, and both administrative and nonadministrative (blue-collar) workers were
included in the cohort. Cancer incidence between 1979 and 1997 was presented by Chang
et al. (2005) and cancer mortality from 1985 to 1997 in Chang et al. (2003). The cohort was
predominately composed of females. The factory operated between 1968 and 1992, and the
inclusion in the cohort of subjects after factory closure is questionable. Incidence was
ascertained from the Taiwan National Cancer Registry which contains 80% of all cancer
cases in Taiwan (Parkin et al., 2002). The factory could be divided into three plants by
manufacturing process: manufacture of television remote controls, manufacture of solid
state and integrated circuit products, and manufacture of printed circuit boards.
Furthermore, a factory waste disposal site was found to have contaminated the
underground water supply of area communities with organic solvents, however, Chang et
al. (2005) does not provide information on possible exposure to factory employees through
ingestion. The analysis of communities adjacent to the factory is described in Lee et al.
(2003).
Company employment records were lacking and the cohort was constructed using the
Bureau of Labor Insurance database that contained computer records since 1978. Labor
insurance hospitalization data and a United Labor Association list of names were used to verify
cohort completeness. While these sources may have been sufficient to identified current
employees, their ability to identify former employees may be limited, particularly from the
hospitalization data if the subject's currently employer was listed.
All employees in the factory were assumed with potential exposure to chlorinated organic
solvent vapors with duration of employment at the factory as the exposure surrogate. Subjects
had varying exposure potentials and employment durations of <1 year (65% of cohort in Chang
et al. (2005)). Durations of exposure were likely underestimated as dates of commencement and
termination of insurance coverage were incomplete, 7.5 and 6%, respectively. Three plants
comprised the factory and with different production processes. A wide variety of organic
solvents were used in each process including dichloromethane, toluene, and methyl ethyl
alcohol, used at all three plants, and perchloroethylene, propanol, and dichloroethylene which
was used at one of the 3 plants Chang et al. (2005). Records of the Department of Labor
Inspection and Bureau of International Trade, in addition, to recall of former industrial hygienists
were used to identify chemicals used after 1975 in the plants. No information is available prior
to this date. These sources documented the lack of TCE use between 1975 and 1991 and
perchloroethylene was after 1981. No information was available on TCE and perchloroethylene
usage during other periods. Given the period of documented lack of TCE usage is before the
cohort start date of 1978 and factory closure, there is great uncertainty of TCE exposure to
cohort subjects.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-156 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1	Overall, both studies are not useful for determining whether trichloroethylene may cause
2	increased risk of disease. The lack of TCE-assessment to individual cohort subjects and
3	uncertainty of TCE usage in the factory; potential bias likely introduced through missing
4	employment dates; and, examination of incidence using broad organ-level categories, i.e.,
5	lymphatic and hematopoietic tissue cancer together, decrease the sensitivity of this study for
6	examining trichloroethylene and cancer. Furthermore, few cancers are expected, 1% of the
7	cohort expected with cancer, and results in low statistical power from the cohort's young average
8	age of 39 years.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-157 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Chang Y-M, Tai C-F, Yang S-C, Lin R, Sung F-C, Shin T-S, Liou S-H. (2005). Cancer Incidence among Workers Potentially
Exposed to Chlorinated Solvents in An Electronics Factory. J Occup Health 47:171-180.
Chang Y-M, Tai C-F, Yang S-C, Chan C-J, S Shin T-S, Lin RS, Liou S-H. (2003). A cohort mortality study of workers
exposed to chlorinated organic solvents in Taiwan. Ann Epidemiol 13:652-660.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The study was not designed to uniquely evaluate TCE exposure but rather
chlorinated solvents exposures. From abstract: "... to investigate the possible
association between chlorinated organic solvents and various types of cancer in an
electronics factory."
This study is quite limited to meet stated hypothesis by the inclusion of all factory
employees in the cohort and lack of exposure assessment on individual study
subjects to TCE, specifically, and to chlorinated solvents, generally.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Selection and characterization in cohort
n = 86,868 in cohort. Cohort initially established using labor insurance records in the
studies of exposure and control groups and of
absence of company records.
cases and controls in case-control studies is
Cohort definition dates are not clearly identified. Cohort identified from labor
adequate
insurance records covering period 1978 and 1997; yet, plant closed in 1992. All

subjects followed through 1997.

Paper states cohort was completely identified; however, former workers who were

eligible for cohort membership may not have been identified if validation sources did

not identify former employer. Duration of employment reconstructed from insurance

records: -40% of subjects had employment durations <3 mos, 9% employed >5 yrs,

0.7% employed >10 yrs.

External referents: Age-, calendar-, and sex-specific incidence rates of the Taiwanese

general population.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence as ascertained from National (Taiwan) Cancer Registry (80% of all
cancers reported to Registry) (Chang et al., 2005).
Mortality. ICD revision is not identified other than that used in 1981 (Chang et al.,
2003).
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-Oncology, a supplement to ICD-9 (Chang et al., 2005).
ICD, 9th revision was in effect in 1981, but paper does not identify to which ICD
revision used to assign cause of death (Chang et al., 2003).
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
All employees assumed to be potentially exposed to chlorinated organic solvent
vapors. No information on specific chemical exposures or intensity. Limited
identification of solvents used in manufacturing process from the period after 1975
inferred from records of Department of Labor Inspection, Bureau of International
Trade, and former industrial hygienists recall. No information on solvent usage was
available before 1975.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No information on loss to follow-up. Subject was assumed disease free at end of
follow-up if lacking cancer diagnosis as recorded in the National Cancer Registry.
>50% cohort with full latency
Average 16-yr follow-up (incidence) and 12 yrs (mortality).
Other
Subject's age determined by subtracting year of birth from 1997; however, insurance
records did not contain DOB for 6% of subjects. Furthermore, commencement and
termination dates were incomplete on insurance records, 7 and 6%, respectively.
CATEGORY E: INTERVIEW TYPE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,031 cancer cases.
1,357 total deaths (1.6% of cohort), 316 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age-, calendar-, and sex-specific incidence rates (Chang et al., 2005) or age-,
calendar-, and sex-specific mortality rates (Chang et al., 2003).
Statistical methods
SIR (Chang et al., 2005) and SMR (Chang et al., 2003).
Exposure-response analysis presented in
published paper
Cancer incidence and mortality examined by duration of employment; however,
employment durations were likely underestimates as dates of commencement and
termination dates on of insurance coverage date were incomplete and calculated from
date on insurance records. Misclassification bias is likely present.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4. Studies of Other Cohorts
B.3 .1.4.1. Clapp and Hoffman (2008).
B.3.1.4.1.1. Author's abstract.
BACKGROUND: In response to concerns expressed by workers at a public
meeting, we analyzed the mortality experience of workers who were employed at
the IBM plant in Endicott, New York and died between 1969 - 2001. An
epidemiologic feasibility assessment indicated potential worker exposure to
several known and suspected carcinogens at this plant. METHODS: We used the
mortality and work history files produced under a court order and used in a
previous mortality analysis. Using publicly available data for the state of New
York as a standard of comparison, we conducted proportional cancer mortality
(PCMR) analysis. RESULTS: The results showed significantly increased
mortality due to melanoma (PCMR = 367; 95% CI: 119, 856) and lymphoma
(PCMR = 220; 95% CI: 101, 419) in males and modestly increased mortality due
to kidney cancer (PCMR = 165; 95% CI: 45, 421) and brain cancer (PCMR =
190; 95% CI: 52, 485) in males and breast cancer (PCMR = 126; 95% CI: 34,
321) in females. CONCLUSION: These results are similar to results from a
previous IBM mortality study and support the need for a full cohort mortality
analysis such as the one being planned by the National Institute for Occupational
Safety and Health.
B.3.1.4.1.2. Study description and comment. This proportional cancer mortality ratio
study of deaths between 1969 and 2001 among employees at an IBM facility in Endicott,
NY, who were included on the IBM Corporate Mortality File compared the observed
number of site-specific cancer deaths are compared to the expected proportion, adjusted
for age, using 10-year rather than 5-year grouping, and sex, of site-specific cancer deaths
among New York residents during 1979 to 1998. Of the 360 deaths identified of Endicott
employees, 115 deaths were due to cancer, 11 of these with unidentified site of cancer.
Resultant proportional mortality ratios estimates do not appear adjusted for race nor does
the paper identify whether referent rates excluded deaths among New York City residents
or are for New York deaths. The IBM Corporate Mortality File contained names of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-163 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
employees who had worker >5 years, who were actively employed or receiving retirement
or disability benefits at time of death, or whose family had filed a claim with IBM for death
benefits and Endicott plant employees were identified using worker employment data from
the IBM Corporate Employee Resource Information System. Study investigators had
previously obtained the IBM Corporate Mortality file through a court order and litigation.
The Endicott plant began operations in 1991 and manufactured a variety of products
including calculating machines, typewriters, guns, printers, automated machines, and chip
packaging. The most recent activities were the production of printed circuit boards. It was
estimated from a National Institute of Occupational Safety and Health (NIOSH) feasibility study
that a larger percentage of the plant's employee were potentially exposure to multiple chemicals,
including asbestos, benzene, cadmium, nickel compounds, vinyl chloride, tetrachloroethylene,
TCE , PCBs, and o-toluidine. Chlorinated solvents were used at the plant until the 1980s. The
study does not assign exposure potential to individual study subjects.
This study provides little information on cancer risk and TCE exposure given its lack of
worker exposure history information and absence of exposure assignment to individual subjects.
Other limitations in this study which reduces interpretation of the observations included
incomplete identification of deaths, the analysis limited to only vested employees or to those
receiving company death benefits, incomplete identification of all employees at the plant, the
inherent limitation of the PMR method and instability of the effect measure particularly in light
of bias resulting of excesses or deficits in deaths, and observed differences in demographic (race)
between subjects and the referent (New York) population.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-164 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Clapp RW, Hoffman K. (2008). Cancer mortality in IBM Endicott plant workers, 1969-2001: an update on a NY production
plant. Environ health 7:13.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract . .In response to concerns expressed by workers at a public meeting,
we analyzed the mortality experience of workers who were employed at the IBM
plant in Endicott, New York and died between 1969-2001."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Deaths among IBM workers identified in IBM Corporate Mortality File; workers
with >5 yrs employment, who were actively employed or receiving retirement or
disability benefits at time of death, or whose family had filed a claim with IBM for
death benefits. Expected number of site-specific cancer deaths calculated from
proportion of cancer deaths among New York residents. Paper does not identify if
referent included all New York residents or those living upstate.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD 9.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
This study lacks exposure information. TCE and other chemicals were used at the
factory and inclusion on the employee list served as a surrogate for TCE exposure of
unspecified intensity and duration.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
Sf
<§>
>S
TO
*
«
s
TO
>;
0
s
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
Not able to evaluate given inability to identify complete cohort.
>50% cohort with full latency
Not able to evaluate given lack of work history records.
Other

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
360 deaths, 115 due to cancer, between 1969-2001.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and gender. No information was available on race and PMRs are unadjusted for
race.
Statistical methods
Proportionate mortality ratio.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4.2. Agency for Toxic Substances and Disease Registry (2004).
B.3.1.4.2.1. Author's abstract.
The View-Master stereoscopic slide viewer has been a popular children's toy
since the 1950s. For nearly half a century, the sole U.S. manufacturing site for the
View-Master product was a factory located on Hall Boulevard in Beaverton,
Oregon. Throughout this period, an on-site supply well provided water for
industrial purposes and for human consumption. In March 1998, chemical
analysis of the View-Master factory supply well revealed the presence of the
degreasing solvent trichloroethylene (TCE) at concentrations as high as 1,670
micrograms per liter (Cg/L)—the U.S. Environmental Protection Agency
maximum contaminant level is 5 fg/L. Soon after the contamination was
discovered, the View-Master supply well was shut down. Up to 25,000 people
worked at the plant and may have been exposed to the TCE contamination. In
September of 2001, the Oregon Department of Human Services (ODHS) entered
into a cooperative agreement with the Agency for Toxic Substances and Disease
Registry (ATSDR) to determine both the need for and the feasibility of an
epidemiological study of the View-Master site. In this report, ODHS compiles the
findings of the feasibility investigation of worker exposure to TCE at the View-
Master factory.
On the basis of the levels of TCE found in the supply well, the past use of the
well as a source of drinking water, and the potential for adverse health effects
resulting from past exposure to TCE, ODHS determined that the site posed a
public health hazard to people who worked at or visited the plant prior to the
discovery of the contamination. Because the use of the View-Master supply well
was discontinued when the contamination was discovered in March 1998, the
View-Master supply well does not pose a current public health hazard. No other
drinking water wells tap into the contaminated aquifer, and the long-term
remediation efforts appear to be containing the contamination.
ATSDR and ODHS obtained a list of 13,700 former plant workers from the
Mattel Corporation. In collaboration with ATSDR, ODHS conducted a
preliminary analysis of mortality and identified excesses in the proportions of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-167 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
deaths due to kidney cancer and pancreatic cancer among the factory's former
employees. Although this analysis was limited by the lack of information about
the entire worker population and individual exposures to TCE, the preliminary
findings underscore the need to fully investigate the impact of TCE exposure on
the population of former View-Master workers.
The findings of this feasibility investigation are:
•	TCE appears to have been the primary contaminant of the drinking water
at the plant;
•	Contamination was likely present for a long period of time (estimated to
have been present in the groundwater since the mid-1960s);
•	A large number were likely exposed to the contamination:
•	The primary route of exposure (for the last 18 years the factory operated)
was through contaminated drinking water;
•	Levels of TCE contamination were 300 time the maximum contaminant
levels; and
•	A significant portion of the former workers of their next of kin can indeed
be located and invited to participate in a public health evaluation of their
exposures.
Therefore, ODHS recommends further investigation to include the following:
1.	A fate and transport assessment to better establish when TCE reached the
supply well, and to provide a historical understanding of the concentration of
TCE in the well, and
2.	Epidemiological studies among former workers to determine their exposure
and whether they have experienced adverse health and reproductive outcomes
associated with TCE exposure at the plant, to determine the mortality
experience of the population, and to document the cancer incidence in this
population.
B.3.1.4.2.2. Study description and comment. This proportionate mortality ratio study of
deaths between 1995-2001 among 13,697 former employees at a View-Master toy factory in
Beaverton, Oregon contains no exposure information on individual study subjects. The
PMR analysis was conducted as a feasibility study for further epidemiologic investigations
of these subjects by Oregon Department of Health on behalf of ATSDR, and findings have
not been published in the peer-reviewed literature. A former plant owner provided a
listing of former employees; however, employees were not identified using IRS records and
the roster was known to be incomplete. Additionally, work history records were not
available and not information was available on employment length or job title. The goal of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-168 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
the feasibility analysis was to evaluate ability to identify completeness of death
identification using several sources.
Monitoring of a water supply well in March 1998 showed detectable concentrations of
TCE, and this study assumes all subjects had exposure to TCE in drinking water. TCE had been
used in large quantities for metal degreasing at the factory between 1952 and 1980; this activity
mostly occurred in the paint shop located in one building. At the time metal degreasing ceased,
company records suggested historical use of TCE was up to 200 gallons per month. Historical
practices resulted in releases of hazardous substances at the factory site and former employees
reported waste TCE from the degreased was transported to other sites on the premises, and
discharged to the ground (ATSDR, 2004). Additionally, chemical spills allegedly occurred in
the paint shop and one report in 1964 of an inspection of the degreaser indicated atmospheric
TCE concentrations above occupational limits. TCE was detected at concentrations between
1,220-1,670 |ig/L in four water samples and the Oregon Department of Environmental Quality
estimated the well had been contaminated for over 20 years. Other volatile organic compounds
(VOCs) besides TCE detected in the supply well water in March 1998 included
cis-l,2-dichloroethylene at levels up to 33 |ig/L and perchloroethylene at concentrations up to
56-|ig/L. The 160-foot-deep supply well was on the property since original construction in 1950
and it supplied water for drinking, sanitation, fire fighting, and industrial use. Connection to
municipal water supply occurred in 1956; however, although municipal water was directed to
some parts of the plant, the supply well continued to serve the facility's needs, including most of
the drinking and sanitary water (ATSDR, 2003a).
This study provides little information on cancer risk and TCE exposure given the absence
of monitoring data beyond a single time period, absence of estimated TCE concentrations in
drinking water, and exposure pathways other than ingestion. Other limitation in this study which
reduces interpretation of the observations included incomplete identification of employees with
the result of missing deaths likely, the inherent limitation of the PMR method and instability of
the effect measure particularly in light of bias resulting of excesses or deficits in deaths, and
observed differences in demographic (age and male/female ratio) between subjects and the
referent (Oregon) population.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-169 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
ATSDR (Agency for Toxic Substances and Disease Registry). (2004). Feasibility investigation of worker exposure to
trichloroethylene at the View-Master Factory in Beaverton, Oregon. Final Report. Submitted by Environmental and
Occupational Epidemiology, Oregon Department of Human Services. December 2004.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The goal of this feasibility investigation for a cohort epidemiologic study of former
employees at a plant manufacturing stereoscopic slide viewers examined the ability
to identify former employees and ascertain vital status.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Name of-13,000 former employee names were provided to ATSDR by the former
plant owner. The current list of employees was known to be incomplete. The
proportion of site-specific mortality among workers between 1989-2001 was
compared to the proportion expected using all death in Oregon for a similar time
period.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD 9 and ICD 10.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
This study lacks actual exposure information; work history records were not
available. TCE was used at the factory and inclusion on the employee list served as a
surrogate for TCE exposure of unspecified intensity and duration.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
Not able to evaluate given inability to identify complete cohort.
>50% cohort with full latency
Not able to evaluate given lack of work history records.
Other

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
616 deaths between 1989-2001.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and gender. No information was available on race and PMRs are unadjusted for
race.
Statistical methods
Proportionate mortality ratio.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.4.3. Raaschou-Nielsen et al. (2003).
B.3.1.4.3.1. Author's abstract.
Trichloroethylene is an animal carcinogen with limited evidence of
carcinogenicity in humans. Cancer incidence between 1968 and 1997 was
evaluated in a cohort of 40,049 blue-collar workers in 347 Danish companies with
documented trichloroethylene use. Standardized incidence ratios for total cancer
were 1.1 (95% confidence interval (CI): 1.04, 1.12) in men and 1.2 (95% CI: 1.14,
1.33) in women. For non-Hodgkin's lymphoma and renal cell carcinoma, the
overall standardized incidence ratios were 1.2 (95% CI: 1.0, 1.5) and 1.2 (95% CI:
0.9, 1.5), respectively; standardized incidence ratios increased with duration of
employment, and elevated standardized incidence ratios were limited to workers
first employed before 1980 for non-Hodgkin's lymphoma and before 1970 for
renal cell carcinoma. The standardized incidence ratio for esophageal
adenocarcinoma was 1.8 (95% CI: 1.2, 2.7); the standardized incidence ratio was
higher in companies with the highest probability of trichloroethylene exposure. In
a subcohort of 14,360 presumably highly exposed workers, the standardized
incidence ratios for non-Hodgkin's lymphoma, renal cell carcinoma, and
esophageal adenocarcinoma were 1.5 (95% CI: 1.2, 2.0), 1.4 (95% CI: 1.0, 1.8),
and 1.7 (95% CI: 0.9, 2.9), respectively. The present results and those of previous
studies suggest that occupational exposure to trichloroethylene at past higher
levels may be associated with elevated risk for non-Hodgkin's lymphoma.
Associations between trichloroethylene exposure and other cancers are less
consistent.
B.3.1.4.3.2. Study description and comment. Raaschous-Nielsen et al. (2003) examine
cancer incidence among a cohort of workers drawn from 347 companies with documented
trichloroethylene. Almost half of these companies were in the iron and metal industry.
The cohort was identified using the Danish Supplementary Pension Fund, which includes
type of industry of a company and a history of employees, for the years 1964 to 1997.
Altogether, 152,726 workers were identified of whom 39,074 were white-collar and
assumed not to have TCE exposure, 56,970 workers were of unknown status, and 56,578
blue-collar workers, of which 40,049 had been employed at the company for more than 3
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-172 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
months and are the basis of the analysis. The cohort was relatively young, 56% were 38 to
57 years old at end of follow-up, and 29% of subjects were older than 57 years of age.
Cancer rates typically increase with increasing ages; thus, the lower age of this cohort
likely limits the ability of this study to fully examine TCE and cancer, particularly cancers
that may be associated with aging. Observed number of site-specific incident cancers are
obtained from 4-1-1968 to the end of 1997 and compared to expected numbers of site-
specific cancers based on incidence rates of the Danish population.
A separate exposure assessment was conducted using regulatory agency data from 1947
to 1989 (Raaschou-Nielsen et al., 2002). This assessment identified three factors as increasing
potential for TCE exposure, duration of employment, year of first employment, and number of
employees, to increase the likelihood of cohort subjects as TCE exposed. The percentage of
exposed workers was found to decrease as company size increased: 81% for <50 workers, 51%
for 50-100 workers, 19% for 100-200 workers, and 10% for >200 workers. About 40% of the
workers in the cohort were exposed (working in a room where trichloroethylene was used).
Smaller companies had higher exposures. Median exposures to trichloroethylene were
40-60 ppm for the years before 1970, 10-20 ppm for 1970 to 1979, and approximately 4 ppm
for 1980 to 1989. Additionally, an assessment of TCA concentrations in urine of Danish
workers suggested a similar trend over time; mean concentrations of 58 mg/L for the period
between 1960 and 1964 and 14 mg/L in sample taken between 1980 and 1985 (Raaschou-
Nielsen et al., 2001).
Only a small fraction of the cohort was exposed to trichloroethylene. The highest
exposures occurred before 1970 at period in which 21.2% of blue-collar workers had begun
employment in a TCE-using company. The iron and metal industry doing degreasing and
cleaning with trichloroethylene had the highest exposures, with a median concentration of
60 ppm and a range up to about 600 ppm. Overall, strengths of this study include its large
numbers of subjects; however, the younger age of the cohort and the small fraction expected with
TCE exposure limit the ability of the study to provide information on cancer risk and TCE
exposure. For these reasons, positive associations observed in this study are noteworthy.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-173 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Raaschou-Nielsen O, Hansen J, McLaughlin JK, Kolstad H, Christensen JM, Tarone RE, Olsen JH. (2003). Cancer risk
among workers at Danish companies using trichloroethylene: a cohort study. Am J Epidemiol 158:1182-1192.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study was designed to evaluate associations observed in Hansen et al. (2001)
with TCE exposure and NHL, esophageal adenocarcinoma, cervical cancer, and
liver-biliary tract cancer.
Selection and characterization in cohort studies
of exposure and control groups and of cases
and controls in case-control studies is adequate
Cohort of 40,049 blue-collar workers employed in 1968 or after with >3 mo
employment duration identified by linking 347 companies, who were considered as
having a high likelihood for TCE exposure, with the Danish Supplementary Pension
Fund to identify employees and with Danish Central Population Registry.
External referents are age-, sex-, calendar year-, site-specific cancer incidence rates
of the Danish population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence between 4-1-1968 and 12-31-1997 as identified from records of
Danish Cancer Registry.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 7th revision.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Qualitative exposure assessment. A previous industrial hygiene survey of Danish
companies identified several characteristics increase likelihood of TCE
exposure-duration of employment, year of 1st employment, and number of employees
in company (Raaschou-Nielsen et al., 2002). Exposure index defined as duration of
employment.
Median exposures to trichloroethylene were 40-60 ppm for the years before
1970,10-20 ppm for 1970 to 1979, and approximately 4 ppm for 1980 to 1989.
Additionally, an assessment of TCA concentrations in urine of Danish workers
suggested a similar trend over time; mean concentrations of 58 mg/L for the
period between 1960 and 1964 and 14 mg/L in sample taken between 1980 and
1985 (Raaschou-Nielsen et al., 2001).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
Danish Cancer Registry is considered to have a high degree of reporting and accurate
cancer diagnoses.
>50% cohort with full latency
Yes, average follow-up was 18 yrs.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
3.244 cancers (8% of cohort had developed a cancer over the period from 1968 to
1997). Although of a large number of subjects, this cohort is of a young age, 29% of
cohort was >57 years of age at end of follow-up.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and calendar year.
Statistical methods
SIR using life-table analysis.
Exposure-response analysis presented in
published paper
Yes, duration of employment.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4.4. Ritz (1999a, b).
B.3.1.4.4.1. Author's abstract.
Data provided by the Comprehensive Epidemiology Data Resource allowed us to
study patterns of cancer mortality as experience by 3814 uranium-processing
workers employed at the Fernald Feed Materials Production Center in Fernald,
Ohio. Using risk-set analyses for cohorts, we estimated the effects of exposure to
trichloroethylene, cutting fluids, and kerosene on cancer mortality. Our results
suggest that workers who were exposed to trichloroethylene experienced an
increase in mortality from cancers of the liver. Cutting-fluid exposure was found
to be strongly associated with laryngeal cancers and, furthermore, with brain,
hemato- and lymphopoietic system, bladder, and kidney cancer mortality.
Kerosene exposure increased the rate of death from several digestive-tract cancers
(esophageal, stomach, pancreatic, colon, and rectal cancers) and from prostate
cancer. Effect estimates for these cancers increased with duration and level of
exposure and were stronger when exposure was lagged.
B.3.1.4.4.2. Study description and comment. This study of 3,814 white male uranium
processing workers employed for at least 3 months between 1-1-1951 and 12-31-1972 at the
Fernald Feed Materials Production Center in Fernald, Ohio, was of deaths as of 1-1-1990.
Subjects were part of a larger cohort study of Fernald workers with potential uranium and
products of uranium decay exposures that observed associations with lung cancer and
lymphatic/hematopoietic cancer (Ritz, 1999b). Average length of follow-up time was 31.5
years. During this period, 1,045 deaths were observed with expected numbers of deaths
based upon age- and calendar-specific U.S. white male mortality rates and age- and
calendar-specific white male mortality rates from the NIOSH Computerized Occupational
Referent Population System (CORPS) (Zahm, 1992). Internal analyses based upon risk-set
sampling and Cox proportional hazards modeling compared workers with differing
exposure intensity rankings (light and moderate) and a category for no- TCE exposure/<2
year duration TCE exposure.
Fernald produced uranium metal products for defense programs (Hornung et al., 2008).
Subjects had potential exposures to uranium, mainly as insoluble compounds and varying from
depleted to slight enriched, small amounts of thorium, an alpha particle emitter, respiratory
irritants such as tributyl phosphate, ammonium hydroxide, sulfuric acid and hydrogen fluoride,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-177 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
trichloroethylene, and cutting fluids (Ritz, 1999a, b). Exposure assessment for analysis of
chemical exposures utilized a job-exposure matrix (JEM) to assign intensity of TCE, cutting
fluids, and kerosene to individual jobs from the period 1952 to 1977. Industrial hygienists, a
plant foreman, and an engineer during the late 1970s and early 1980s determined the likelihood
of exposure to TCE, cutting fluids, and kerosene for each job title and plant area. Based on work
records, the workforce appeared stable and 54% were employed >5 years and had held only one
job title during employment. Both intensity or exposure level and duration of exposure in years
were used to rank subjects into 4 categories of no exposure (level 0), light exposure (level 1),
moderate exposure (level 2), and heavy exposure (level 3). Seventy eight (78) percent of the
cohort was identified with some potential for TCE exposure, 2,792 subjects were identified with
low TCE exposure (94%), 179 with moderate exposure (6%), and no subjects were identified
with heavy TCE exposure. TCE exposure was highly correlated with other chemical exposures
and with alpha radiation (Hornung et al., 2008; Ritz, 1999a, b). Fernald subjects had higher
exposures to radiation compared to those of radiation-exposed Rocketdyne workers (Ritz, 1999b;
Ritz et al., 2000; Ritz et al., 1999). Atmospheric monitoring information is lacking on TCE
exposure conditions as is information on changes in TCE usage over time. The cohort was
identified from company rosters and personnel records and it is not known whether these were
sources for a subject's job title information. Analysis of TCE exposure carried out using
conditional logistic regression adjusting for pay status, time since first hired, external and
internal radiation dose and previous chemical exposure. Relative risks for TCE exposure are
also presented with a lag time period of 15 years.
Overall, strengths of this study are the long follow-up time and a large percentage of the
cohort who had died by the end of follow-up. TCE exposure intensity is low in this cohort, 94%
of TCE exposed subjects were identified with "light" exposure intensity, and all subjects had
potential for radiation exposure, which was highly correlated with chemical exposures. No
information is presented on the definition of "light" exposure and monitoring data are lacking.
Only 179 subjects were identified with TCE exposure above "light" and the number of cancer
deaths not presented. The published paper reported limited information on site-specific cancer
and TCE exposure; risk estimates are reported for lymphatic and hematopoietic cancers,
esophageal and stomach cancer, liver cancer, prostate cancer and brain cancer. Risk estimates
for bladder and kidney cancer and TCE exposure are found in NRC (2006). Few deaths were
observed with moderate TCE exposure and exposure durations of longer than 2 years: 1 death
due to lymphatic and hematopoietic cancer, 0 deaths due to kidney or bladder cancer (as noted in
NRC (2006)), and 2 liver cancer deaths among these subjects. Low statistical power reflecting
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-178 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1	few cases with moderate TCE exposure and multicolinearity of chemical and radiation exposures
2	greatly limits the support this study provides in an overall weight-of-evidence analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-179 DRAFT—DO NOT CITE OR QUOTE
-------
Ritz B. (1999aa).
41:556-566.
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Cancer mortality among workers exposed to chemicals during uranium processing. J Occup Environ Med

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The hypothesis in this study was to examine the influence of chemical exposures in
the work environment of the Fernald Feed Materials Production Center (FFMPC) in
Fernald, Ohio, on cancer mortality with a focus on the effects of TCE, cutting fluids,
and a combination of kerosene exposure with carbon (graphite) and other solvents.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
3,814 white male subjects identified from company rosters and personnel records,
hired between 1951 and 1972 and who were employed continuously for 3 mos and
monitored for radiation. 2,971 subjects identified as exposed to TCE at "light" and
"moderate" exposures. Subjects were identified in a previous study of cancer
mortality and radiation exposure and most subjects had radiation exposures above
10+ mSV (Ritz, 1999b).
External analysis: U.S. white male mortality rates and NIOSH-Computerized
Occupational Referent Population System mortality rates.
Internal analysis: cohort subjects according to level and duration of chemical
exposure.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Vital status searched through Social Security Administration records, before 1979,
and National Death Index for the period 1979-1989.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Changes in diagnostic coding systems for
External analysis: ICDA, 8th revision.
lymphoma, particularly non-Hodgkin's
Internal analysis: aggregation of several subsite causes of deaths into larger
lymphoma
categories based on ICD, 9th revision.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Semiquantitative approach and development of job-exposure matrix. JEM developed
by expert assessment by plant employees to classify jobs into four levels of chemical
exposures for the period 1952 to 1977. Intensity using the four-level scale and
duration of exposure to TCE, cutting fluids and kerosene were assigned to individual
cohort subjects using JEM. 73% of cohort identified as TCE exposed (2,971 male
with TCE exposure in cohort of 3,814 subjects). Only 4% of TCE-exposed subjects
with exposure identified as "moderate" and no subjects with "high" exposure. High
correlation between TCE and other chemical exposure and radiation exposure.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
All workers without death certificate assumed alive at end of follow-up.
>50% cohort with full latency
Average follow-up time, 31.5 yrs.
Other

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,045 deaths (27% of cohort), 328 due to cancer. No information on number of all-
cancer deaths among TCE exposed subjects, although reported numbers for specific
sites reported by Ritz (1999a) or NRC (2006): >2 year exposure duration, hemato-
and lymphopoietic cancer (n= 18 with light exposure, 1 with moderate exposure),
esophageal and stomach cancer (n= 15 with light exposure, 0 with moderate
exposure), liver cancers (n = 3 with light exposure, 1 with moderate exposure),
kidney and bladder cancers, (n = 7 with light exposure, 0 with moderate exposure)
prostate cancers (n= 10 with light exposure, 1 with moderate exposure), and brain
cancers (n = 6 with light exposure, 1 with moderate exposure).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
External analysis: age- and calendar-specific mortality rates for white males.
Internal analysis: pay status, time since first hired, and cumulative time-dependent
external- and internal-radiation doses (continuous); indirect assessment of smoking
through examination of smoking distribution by chemical exposure.
Statistical methods
SMR (external analysis) and RR (internal analysis).
Exposure-response analysis presented in
published paper
Yes, RR presented for exposure to TCE (level 1 and level 2, separately) by duration
of exposure.
Documentation of results
Adequate.
RR = relative risk.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3 .1.4.5. Henschler et al. (1995).
B.3.1.4.5.1. Author's abstract.
A retrospective cohort study was carried out in a cardboard factory in Germany to
investigate the association between exposure to trichloroethene (TRI) and renal
cell cancer. The study group consisted of 169 men who had been exposed to TRI
for at least 1 year between 1956 and 1975. The average observation period was 34
years. By the closing day of the study (December 31, 1992) 50 members of the
cohort had died, 16 from malignant neoplasms. In 2 out of these 16 cases, kidney
cancer was the cause of death, which leads to a standard mortality ratio of 3.28
compared with the local population. Five workers had been diagnosed with
kidney cancer: four with renal cell cancers and one with an urothelial cancer of
the renal pelvis. The standardized incidence ratio compared with the data of the
Danish cancer registry was 7.97 (95% CI: 2.59-18.59). After the end of the
observation period, two additional kidney tumors (one renal cell and one
urothelial cancer) were diagnosed in the study group. The control group consisted
of 190 unexposed workers in the same plant. By the closing day of the study 52
members of this cohort had died, 16 from malignant neoplasms, but none from
kidney cancer. No case of kidney cancer was diagnosed in the control group. The
direct comparison of the incidence on renal cell cancer shows a statistically
significant increased risk in the cohort of exposed workers. Hence, in all types of
analysis the incidence of kidney cancer is statistically elevated among workers
exposed to TRI. Our data suggest that exposure to high concentrations of TRI
over prolonged periods of time may cause renal tumors in humans. A causal
relationship is supported by the identity of tumors produced in rats and a valid
mechanistic explanation on the molecular level.
B.3 .1.4.5 .2. Study description and comment. This was a cohort study of workers in a
cardboard factory in the area of Arnsberg, Germany. Trichloroethylene was used in this
area until 1975 for degreasing and solvent needs. Plant records indicated that
2,800-23,000 L/year was used. Small amounts of tetrachloroethylene and 1,1,1-
trichloroethane were used occasionally, but in much smaller quantities than
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-184 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
trichloroethylene. Trichloroethylene was used in three main areas: cardboard machine,
locksmith's area, and electrical workshop. Cleaning the felts and sieves and cleaning
machine parts of grease were done regularly every 2 weeks, in a job that required 4-5
hours, plus whatever additional cleaning was needed. Trichloroethylene was available in
open barrels and rags soaked in it were used for cleaning. The machines ran hot
(80-120°C) and the cardboard machine rooms were poorly ventilated and warm (about
50°C), which would strongly enhance evaporation. This would lead to very high
concentrations of airborne trichloroethylene. Cherrie et al. (2001) estimated that the
machine cleaning exposures to trichloroethylene were greater than 2,000 ppm. Workers
reported frequent strong odors and a sweet taste in their mouths. The odor threshold for
trichloroethylene is listed as 100 ppm (ATSDR, 1997b). Workers often left the work area
for short breaks "to get fresh air and to recover from drowsiness and headaches." Based
on reports of anesthetic effects, it is likely that concentrations of trichloroethylene exceeded
200 ppm (Stopps and McLaughlin, 1967). Those reports, the work setting description, and
the large volume of trichloroethylene used are all consistent with very high concentrations
of airborne trichloroethylene. The workers in the locksmith's area and the electrical
workshop also had continuous exposures to trichloroethylene associated with degreasing
activities; parts were cleaned in cold dip baths and left on tables to dry. Trichloroethylene
was regularly used to clean floors, work clothes, and hands of grease, in addition to the
intense exposures during specific cleaning exercises, which would produce a background
concentration of trichloroethylene in the facility. Cherrie et al. (2001) estimated the long-
term exposure to trichloroethylene was approximately 100 ppm.
1	The subjects in this study clearly had substantial peak exposures to trichloroethylene that
2	exceeded 2,000 ppm and probably sustained long-term exposures greater than 100 ppm, which
3	are not confounded by concurrent exposures to other chlorinated organic solvents.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-185 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Henschler D, Vamvakas S, Lammert M, Dekant W, Kraus B, Thomas B, Ulm K. (1995). Increased incidence of renal cell
tumors in a cohort of cardboard workers exposed to trichloroethene. Arch Toxicol 69:291-299.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract "... retrospective cohort study was carried out in a cardboard factory I
Germany to investigate the association between exposure to trichloroethene and renal
cell cancer."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Employee records were used to identify 183 males employed in a cardboard factory
for at least 1 yr between 1956 and 1975 and with presumed TCE exposure and a
control group of 190 male workers at same factory during the same period of time
but in jobs not involving possible TCE exposure.
Mortality rates from German population residing near factory used as referent in
mortality analysis.
Renal cancer incidence rates from Danish Cancer Registry used to calculate expected
number of incident cancer. The age-standardized rate in the late 1990s among men
in Denmark was 10.6 and in Germany it was 1.2 (Ferlay et al., 2004). If these
differences in rates apply when the study was carried out, this would imply that the
expect number of deaths would have been inflated by about 14% (and the rate ratio
underestimated by that amount).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality and renal cell cancer incidence.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-9 for deaths.
Hospital pathology records were used to verify diagnosis of renal cell carcinoma.
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Walkthrough survey and interviews with long-term employees were used to identify
work areas and jobs with potential TCE exposure. The workers in the locksmith's
area and the electrical workshop also had continuous exposures to trichloroethylene
associated with degreasing activities; parts were cleaned in cold dip baths and left on
tables to dry. Cherrie et al. (2001) estimated that the machine cleaning
exposures to trichloroethylene were greater than 2,000 ppm with average
long-term exposure as 10-225 ppm. Estimated average chronic exposure to
TCE was -100 ppm to subjects using TCE in cold degreasing processes.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
14 exposed subjects (8%>) were excluded from life-table analysis and no information
is presented in paper on loss-to-follow-up among control subjects.
>50% cohort with full latency
Median follow-up period was over 30 yrs for both exposed and control subjects.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
50 total deaths (30%) and 15 cancer death among exposed subjects.
52 deaths (27%) and 15 cancer deaths among control subjects.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and calendar-year.
Statistical methods
SMR and SIR. Analysis excludes person-years of subjects excluded from exposed
population with the number of person-years underestimated and an underestimate of
the expected numbers of deaths and incident renal carcinoma cases.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Exposure-response analysis presented in
published paper
No.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4.6. Greenland et al. (1994).
B.3.1.4.6.1. Author's abstract.
To address earlier reports of excess cancer mortality associated with employment
at a large transformer manufacturing plant each plant operation was rated for
seven exposures: Pyranol (a mixture of polychlorinated biphenyls and
trichlorobenzene), trichloroethylene, benzene, mixed solvents, asbestos, synthetic
resins, and machining fluids. Site-specific cancer deaths among active or retired
employees were cases; controls were selected from deaths (primarily
cardiovascular deaths) presumed to be unassociated with any of the study
exposures. Using job records, we then computed person-years of exposure for
each subject. All subjects were white males. The only unequivocal association
was that of resin systems with lung cancer (odds ratio = 2.2 at 16.6 years of
exposure, P = 0.0001, in a multiple logistic regression including asbestos, age,
year of death, and year of hire). Certain other odds ratios appeared larger, but no
other association was so robust and remained as distinct after considering the
multiplicity of comparisons. Study power was very limited for most associations,
and several biases may have affected our results. Nevertheless, further
investigation of synthetic resin systems of the type used in the study plant appears
warranted.
B.3.1.4.6.2. Study discussion and comment. This nested case-control study at General
Electric's Pittsfield, MA, plant was of deaths reported to the GE pension fund among
employees vested in the pension fund. The cohort from which cases and controls were
identified was defined as plant employees who worked at the facility before 1984; whose
date of deaths was between 1969, the date pension records became available, and 1984; and
existence of a job history record. The size of the underlying employee cohort was unknown
because work history records did not exist for a large fraction of former employees,
especially in the earlier years of deaths. All deaths were identified from records
maintained by GE's pension office; other record sources such as the Social Security
Administration and National Death Index were not utilized. Requirements for eligibility or
"vestment" for a pension varied over time, but for most of the study period, required 10 to
15 years employment with the company. The analysis was restricted to white males
because of few deaths among females and nonwhite males. A total of 1,911 deaths were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-190 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
identified from pension records and cases and controls, with 90 deaths excluded as possible
cases and controls due to several reasons. Cases were identified as site-specific deaths and
controls were selected from the remaining noncancer deaths due to circulatory disease,
respiratory disease, injury, and other causes. No information was available on the number
of controls selected per case. Controls were not matched to cases, were slightly older than
cases, and were from earlier birth cohorts which have a lower job history availability or
greater frequency of missing exposure ratings in work history records (Salvan, 1990).
Statistical analysis of the data included covariates for age and year of death.
The company's job history record served as the source for exposure rating. The JEM
linked possible exposures to over 1,000 job title from 50 separate departments and 100 buildings.
A categorical ranking was developed for exposure to seven exposures (Pyranol, TCE, benzene,
other solvents, asbestos, resin systems, machining fluids) from 1901 to 1984 based upon on-site
interviews with 18 long-term employees and knowledge of one of the study investigators who
was an industrial hygienist. Two categories were used for potential TCE exposure: Level 1,
duration of indirect exposure (TCE in workplace but does not work directly with TCE) and
Level 2, duration of direct work with TCE, with the continuous exposure scores rescaled to the
97th percentile of controls (Salvan, 1990). Statistical analyses in Greenland et al. (1994)
collapsed these two categories into a dichotomous ranking of no exposure or any exposure. In
many instances, exposure levels were inaccurately estimated and some exposures were highly
correlated (Salvan, 1990). Although of low correlation, TCE exposure was statistically
significantly correlated with exposure to other solvents (/' = 0.11), benzene (r = 0.22) and
machining fluids (r = 0.28) (Salvan, 1990). Industrial hygiene monitoring data were not
available before 1978 and limited production and purchase records did not extend far back in
time (Salvan, 1990). TCE was used as a degreaser since the 1930s and discontinued between
1966 and 1975, depending on department. In all, fewer than 10% of jobs were identified as have
TCE exposure potential, primarily through indirect exposure and not directly working with TCE.
In fact, few subjects were identified with as working directly with TCE (Salvan, 1990). It is not
surprising that exposure score distributions were highly skewed towards zero (Salvan, 1990). No
details were provided on the protocol for processing the jobs in the work histories into job
classifications.
Job history information was missing for roughly 35% of the cases and controls,
particularly from subjects with earlier years of death. The highest percentage of missing
information among cases was for leukemia deaths (43% of deaths) and the lowest percentage for
rectal deaths (11%). Moreover, work history records did not exist for a large fraction of former
employees, especially in the earlier years of death. Bias resulting from exposure
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-191 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
misclassification is likely high due to the lack of industrial monitoring to support rankings and
the inability of the JEM to account for changes in TCE exposure concentrations over time.
This study had a number of weaknesses with the likely result of dampening observed
risks. Deaths were underestimated given nonpensioned employees are not included in the
analysis; possible differences in exposure potential between pensioned and nonpensioned
workers may introduce bias, particularly if a subject leaves work as a consequence of a
precondition related to exposure, and would dampen observed associations (Robins, 1987).
Misclassification bias related to exposure is highly likely given missing job history records for
over one-third of deaths, mostly among deaths from the earlier study period, a period when TCE
was used. Salvan (1990) noted "exposure measurements should be regarded as heavily
nondifferentially misclassified relative to the true exposure does" and exposure associations with
outcomes will be underestimated. For TCE specifically, the development of exposure
assignments in this study was insensitivity to define TCE exposures of the cohort-industrial
hygiene data were not available for the time period of TCE use, exposure rates applied to a job-
building-operation time matrix and may not reflect individual variation, and exposure ratings
obtained by employee interview are subject to subjective assessment and measurement error.
NRC (2006) also noted a low likelihood of exposure potential to subjects in this nested case-
control study. Last, the lymphoma category includes Hodgkin's lymphoma, in addition to
traditional NHL forms such as reticulosarcoma and lymphosarcoma. Overall, the sensitivity of
this study for evaluating cancer and TCE exposure is quite limited. The inability of this study to
detect associations for two known human carcinogens, benzene and leukemia and asbestos and
lung cancer, provides ancillary support for the study's low sensitivity and statistical power.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-192 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Greenland S, Salvan A, Wegman DH, Hallock MF, Smith TH. (1994). A case-control study of cancer mortality at the
transformer-assembly facility. Int Arch Occup Environ Health 66:49-54.
Greenland S. (1992). A semi-Bayes approach to the analysis of correlated multiple associations with an application to an
occupational cancer-mortality study. Stat Med 11:219-230.
Salvan A. (1990). Occupational exposure and cancer mortality at an electrical manufacturing plant: A case-control study.
Ph.D. Dissertation, University of California, Los Angeles.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The study was carried out to reevaluate an earlier observation from a PMR study of
GE employment and excess leukemia and colorectal cancer risks.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Selection of cases and controls is not adequate because only deaths among pensioned
workers were included in the analysis. Also, the size of the underlying cohort was
not known and potential for selection bias is likely given cases and controls are
drawn from a select population.
Cases were identified from deaths among white males employed before 1984, who
had died between 1969 and 1984, and for whom a job history record was available.
Controls selected from noncancer deaths due to cardiovascular disease, circulatory
disease, respiratory disease, injury, or other causes. Controls are not matched to
cases on covariates such as age, or date of hire.
In total, 2,653 subjects were identified as meeting criteria for inclusion in subject,
either as a case or as a control. Job history records were available for 1,714 (512
cases, 1,202 controls) of these subjects (65%).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICDA, 8th revision. Lymphomas, Codes 200-202 and includes Hodgkin's
lymphoma.
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Dichotomous ranking, not exposed/exposed, for indirect and direct exposure
potential. Most subjects identified with indirect TCE exposure. The company's job
history record served as the source for exposure rating. The JEM linked possible
exposures to over 1,000 job title from 50 separate departments and 100 buildings.
Potential TCE exposure assigned to 10%> of all job titles. The seven exposures were
highly correlated. NRC (2006) noted a low likelihood of TCE exposure potential to
subjects in this nested case-control study.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers
Record study.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
220 of 732 cases and 1,202 or 1,921 possible controls had job history records; job
history records are missing for 35% of all possible cases and controls.
Any potential TCE exposure prevalence among cases:
Laryngeal, pharyngeal cancer, 38%
Liver and biliary passages, 22%
Pancreas, 45%
Lung, 33%)
Bladder, 30%
Kidney, 33%
Lymphoma, 27%
Leukemias, 36%
Brain, 31%
Control exposure prevalence, 34%.
Control for potential confounders in statistical
analysis
Age and year of death. Other unidentified covariates are included if risk estimate is
altered by more than 20%.
Statistical methods
Logistic regression with (1) dichotomous exposure (Greenland et al., 1994) (2)
continuous exposure (Salvan, 1990), (3) epoch analysis (Salvan, 1990), and (4)
empirical bayes models (Greenland, 1992).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Adequate.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4.7. Sinks et al. (1992).
B.3.1.4.7.1. Author's abstract.
A physician's alert prompted us to investigate workers' can cancer risk at a
paperboard printing manufacturer. We conducted a retrospective cohort mortality
study of all 2,050 persons who had worked at the facility for more than 1 day,
calculated standardized incidence ratios (SIRs) for bladder and renal cell cancer,
and conducted a nested case-control study for renal cell cancer. Standardized
mortality ratios (SMRs) from all causes [SMR = 1.0, 95% confidence interval
(CI) = 0.9 - 1.2] and all cancers (SMR = 0.6, 95% CI = 0.3 - 1.0) were not greater
than expected. One bladder cancer and one renal cell cancer were included in the
mortality analysis. Six incident renal cell cancers were observed, however,
compared with less than two renal cell cancers expected (SIR = 3.7, 95% CI = 1.4
- 8.1). Based on a nested case-control analysis, the risk of renal cell carcinoma
was associated with overall length of employment but was not limited to any
single department or work process. Although pigments containing congeners of
dichlorobenzidine and o-toluidine had been used at the plant, environmental
sampling could not confirm any current exposure. Several limitations and a
potential selection bias limit the inferences that can be drawn.
B.3.1.4.7.2. Study description and comment. Sinks et al. (1992) is the published report of
analyses examining morbidity and mortality among employees at a James River
Corporation plant in Newnan, GA. This plant manufactured paperboard (cardboard)
packaging. The study was carried out as a National Institute of Occupational Safety and
Health, Health Hazard Evaluation to investigate a possible cluster of urinary tract cancers
and work in the plant's Finishing Department (NIOSH, 1992)\. A cohort of 2,050 white
and nonwhite, male and female, subjects were identified from company personnel and
death records, considered complete since 1-1-1957, and were follows for site-specific
mortality and cancer morbidity to 6-30-1988. Records of an additionally 36 subjects were
missing hire dates or birth dates, indicated employment duration of less than 1 day, and or
employment outside the study period and these subjects were excluded from the analysis.
This study suffers from missing information. A large percentage of personnel records did
not identify a subject's race and these subjects were considered as white in statistical
analyses. Additionally, vital status was unknown for approximately 10% of the cohort.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09
B-197
DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Life-table analyses are based upon U.S. population age-, race-, sex-, calendar- and cause-
specific mortality rates. Expected numbers of incident bladder and kidney cancers for
white males were derived using white male age-specific bladder and renal cell incidence
rates from the Atlanta-Surveillance, Epidemiology, and End Results (SEER) registry for
the years 1973 to 1977.
A nested case-control analysis of the incident renal carcinoma cases was also undertaken.
This analysis is based on 6 renal cell carcinoma cases and 48 controls (1:8 matching) who were
selected by risk set sampling of all employees born within 5 years of the case, the same sex as
the case, and having attained the age at which the case was diagnosed or died if date of diagnosis
was not known. A diagnosis of renal carcinoma was confirmed for 4 of the 6 cases through
pathologic examination. Both the nested case-control analysis and the life-table analyses of
morbidity included a renal carcinoma case from the original cluster.
Exposures are poorly defined in this study assessing renal cancer among paper board
printing workers. Trichloroethylene was mentioned in material-safety data sheets for one or
more materials used by the process but no information was provided regarding TCE usage and
use by job title. It was not possible to assess the degree of contact with trichloroethylene or the
printing inks which were identified as containing benzidine. Furthermore, the lack of monitoring
data precludes evaluation of possible exposure intensity. This study is limited for assessing risks
associated with exposures to trichloroethylene due to the large percentage of missing information
and due to its exposure assessment approach.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-198 DRAFT—DO NOT CITE OR QUOTE
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
Sinks T, Lushniak B, Haussler BJ, Sniezek J, Deng J-F, Roper P, Dill P, Coates R. (1992). Renal cell cancer among
paperboard printing workers. Epidemiol 3:483-489.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The purpose of the cohort and nested case-control investigations was to determine
whether an excess of bladder or renal cell cancer had occurred among workers in a
paperboard packaging plant and, if so, to determine whether it was associated with
any specific exposure or work-related process.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cohort analysis: 2 050 males and females employed at the plant between 1-1-1957
and 6-30-1988. External referents for mortality analysis were age-, sex-, race-, and
calendar- cause specific mortality rates of the U.S. population. External referents for
morbidity analysis were age-specific bladder and renal-cell cancer rate for white
males from the Atlanta-SEER registry for the years 1973-1977.
Nested case-control analysis: Cases were all subjects with renal cell cancer;
8 nonrenal cell carcinoma controls chosen from a risk set of all employees matched
to case on date of birth (within 5 yrs), sex and attained age of cancer diagnosis or
death, if diagnosis date unknown.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD revision in effect at the time of death; incident cases of renal cell carcinoma
diagnoses confirmed with pathology reports for 4 of 6 cases.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
5r-
&
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
3'
Co
a
>r
<§>
>S
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Exposure in cohort analysis defined broadly at level of the plant and, in case-control
study, department worked as identified on company's personnel.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
Yes, 10%o of cohort with unknown vital status (n = 204).
P-Y for these workers were censored at the date of last follow-up.
>50% cohort with full latency
18 yr average follow-up.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Department assignment based on company personnel records.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
141 total deaths (7% of cohort had died by end of follow-up), 16 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Mortality analysis: Age, race, sex, and calendar year.
Morbidity analysis limited to white males: age.
Nested case-control analysis: Risk set sampling matching controls to cases on date of
birth (within 5 yrs), sex, and attained age at diagnosis.
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
o
o
S"4
>}'
§•
to
s
a
to
s
S"
a
a,
>r
<§>
Statistical methods
SIR.
Conditional logistic regression used for nested case-control analysis.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Adequate.
SIR = standardized incidence ratio.
s
TO
TO'
*
«
s
TO
Co
0
S
1
a.
VO
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
B.3.1.4.8. Blair et al. (1989).
B.3.1.4.8.1. Author's abstract.
Work history records and fitness reports were obtained for 1 767 marine
inspectors of the U.S. Coast Guard between 1942 and 1970 and for a comparison
group of 1 914 officers who had never been marine inspectors. Potential exposure
to chemicals was assessed by one of the authors (RP), who is knowledgeable
about marine inspection duties. Marine inspectors and noninspectors had a deficit
in overall mortality compared to that expected from the general U.S. population
(standardized mortality ratios [SMRs = 79 and 63, respectively]). Deficits
occurred for most major causes of death, including infectious and parasitic
diseases, digestive and urinary systems, and accidents. Marine inspectors had
excesses of cirrhosis of the liver (SMR = 136) and motor vehicle accidents (SMR
= 107, and cancers of the lymphatic and hematopoietic system (SMR = 157,
whereas noninspectors had deficits for these causes of death. Comparison of
mortality rates directly adjusted to the age distribution of the inspectors and
noninspectors combined also demonstrated that mortality for these causes of death
was greater among inspectors than noninspectors (directly adjusted ratio ratios of
190, 145, and 198) for cirrhosis of the liver, motor vehicle accidents, and
lymphatic and hematopoietic system cancer, respectively. The SMRs rose
with increasing probability of exposure to chemicals for motor vehicle accidents,
cirrhosis of the liver, liver cancer, and leukemia, which suggests that contact with
chemicals during inspection of merchant vessels may be involved in the
development of these diseases among marine inspectors, physician's alert
prompted us to investigate workers' can cancer risk at a paperboard printing
manufacturer. We conducted a retrospective cohort mortality study of all 2,050
persons who had worked at the facility for more than 1 day, calculated
standardized incidence ratios (SIRs) for bladder and renal cell cancer, and
conducted a nested case-control study for renal cell cancer. Standardized
mortality ratios (SMRs) from all causes [SMR = 1.0, 95% confidence interval
(CI) = 0.9 - 1.2] and all cancers (SMR = 0.6, 95% CI = 0.3 - 1.0) were not greater
than expected. One bladder cancer and one renal cell cancer were included in the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-202 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
mortality analysis. Six incident renal cell cancers were observed, however,
compared with less than two renal cell cancers expected (SIR = 3.7, 95% CI = 1.4
- 8.1). Based on a nested case-control analysis, the risk of renal cell carcinoma
was associated with overall length of employment but was not limited to any
single department or work process. Although pigments containing congeners of
dichlorobenzidine and o-toluidine had been used at the plant, environmental
sampling could not confirm any current exposure. Several limitations and a
potential selection bias limit the inferences that can be drawn.
B.3.1.4.8.2. Study description and comment. This cohort of 1,767 U. S. Coast Guard
male officers and enlisted personnel performing marine inspection duties between 1942 and
1970 and 1,914 noninspectors matched to inspectors for registry, rank and year that rank
was achieved examined mortality as of January 1,1980. Standardized mortality ratios
compared the observed number of site-specific deaths among marine inspectors (n = 483,
27%) to that expected of the total U. S. white male population and to standardized
mortality ratios of noninspectors (n = 369,19%). The cohort was predominantly white
(91%), race was unknown for the remaining 8% of subjects, considered in the statistical
analysis as white, with a large percentage (69%) of the marine inspectors having >20 year
employment duration. The minimum latent period was 10 years, calculated from the end
date of cohort identification to the date of vital status ascertainment.
This study lacks exposure information on potential exposures of marine inspectors, who
enter cargo tanks, void spaces, cofferdams, and pump rooms during inspections. TCE is
identified in the paper as a possible exposure along with nine other agents. One authors
acquainted with Coast Guard processes estimated the level of exposure to general chemical
exposures during a marine inspection. A four-point rating scales was developed: nonexposed,
person generally held administrative position; low exposed, assigned to staff with duties that
occasionally required vessel inspections; moderate exposed, assign to inspection duties that did
not regularly include hull structures, and regular inspection of hull structures in geographic areas
where chemicals were not major items of cargo; and, high exposed, assigned to subjects who
performed hull inspections at ports were vessels transported chemicals. A cumulative exposure
score was calculated by summing the product of the four-point rating scale and the duration in
each job.
Overall, the exposure assessment in this study is insufficient for examining TCE
exposure and cancer mortality. Furthermore, the few site-specific deaths among marine
inspectors greatly limits statistical power.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-203 DRAFT—DO NOT CITE OR QUOTE
-------
Blair A, Haas T, Prosser R, Morrissette M, Blackman, Grauman D, van Dusen P, Morgan F. (1989). Mortality among United
States Coast Guard marine Inspectors. Arch Environ Health 44:150-156.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The purpose of the cohort study was to examine mortality patterns among Coast
Guard marine inspectors. This study was not designed to examine specific
exposures, including TCE.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
1,767 U. S. Coast Guard male officers and enlisted personnel performing marine
inspections between 1942 and 1970 and 1,914 noninspectors matched to inspectors
on registry, rank, and year that rank was achieved.
External referents: age-specific mortality rates of the U. S. white male population
and noninspectors.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICDA, 8th revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
TCE identified in paper as one of ten potential exposures; however, no exposure
assessment to TCE to individual subjects. Exposure in cohort analysis defined
broadly at level of the plant and, in case-control study, department worked as
identified on company's personnel. A cumulative exposure surrogate developed from
duration in each job and a four-point rating scale: nonexposed, person generally held
administrative position; low exposed, assigned to staff with duties that occasionally
required vessel inspections; moderate exposed, assign to inspection duties that did
not regularly include hull structures, and regular inspection of hull structures in
geographic areas where chemicals were not major items of cargo; and, high exposed,
assigned to subjects who performed hull inspections at ports were vessels transported
chemicals.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No
>50% cohort with full latency
Not reported; minimum latent period was 10 years.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
483 deaths among marine inspectors (27% of cohort), 103 cancer deaths.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Mortality analysis: Age, race, sex, and calendar year. Directly adjusted rate ratios
compared cause-specific SMR of marine inspectors to that of noninspectors.
Statistical methods
SMR and RR.
Exposure-response analysis presented in
published paper
Yes, using a ranked cumulative exposure surrogate.
Documentation of results
Adequate.
RR = relative risk. SMR = standardized mortality ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3.1.4.9. Shannon et al. (1988).
B.3.1.4.9.1. Author's abstract.
A historical prospective study of cancer in lamp manufacturing workers in one
plant was conducted. All men and women who worked for a total of at least 6
months and were employed at some time between 1960 and 1975 were included.
Work histories were abstracted and subjects were divided according to whether
they had worked in the coiling and wire drawing area (CWD). Cancer morbidity
from 1964 to 1982 was ascertained via the provincial registry, and was compared
with the site-specific incidence in Ontario, adjusting for age, sex and calendar
period. Of particular interest were primary breast and gynecological cancers in
women.
The cancers of a priori concern were significantly increased in women in CWD,
but not elsewhere in the plant. The excess was greatest in those with more than 5
yr exposure (in CWD) and more than 15 yr since first working in CWD, with
eight cases of breast and gynecological cancers observed in this category
compared with 2.67 expected. Only three cancers occurred in men in CWD.
Environmental measurements had not been made in the past and little information
was available on substances used in the 1940s and 1950s, the period when the
women with the highest excess began employment. It is known that methylene
chloride and trichloroethylene have been used, but not enough is known about the
dates and patterns.
B.3.1.4.9.2. Study description and comments. This cohort of 1,770 workers (1,044
females, 826 males) employed >6 months and working between 1960 and 1975 at a General
Electric plant in Ontario, Canada, in the lamp manufacturing department identified cancer
incidence cases from a regional cancer registry from 1964, the first date of high quality
information, to 1982. Office workers were included in the study population. The study
was carried out in response to previous reports of excess breast and gynecological cancer in
women employed in the CWD area. Standardized incidence ratios (SIR) compared the
observed number of site-specific incident cancers to that expected of the Ontario
population and supplied by the regional cancer registry. SIR estimates were calculated for
all lamp department workers, and for two subgroups defined by job title, workers in the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-207 DRAFT—DO NOT CITE OR QUOTE
-------
coil and wire-drawing area (CWD) and workers in all other areas. The cohort was
successfully traced, with low rates of lost to follow-up (6% among CWD workers, 7 all
other workers). A total of 98 incident cancer cases were identified (58 in females, 40 in
males) and over half of the incident cancers in females (n = 31) due to breast and
gynecological cancers. The number of incident cancers is likely underestimated given the
4-year period between cohort identification and the first date of high quality information in
the cancer registry. Additionally, cancer cases among workers who moved from the
province would not be found in the registry, leading to underascertainment of cases. This
is likely a small number given follow-up tracing identified 2% of workers had left the
province.
1	This study lacks exposure information on individual study subjects. Exposures in CWD
2	were of concern given previous reports. The study lacks exposure monitoring data and potential
3	exposures in CWD area were identified using purchase records. A number of chemicals were
4	identified including methylene chloride from 1959 onward and trichloroethylene, which records
5	suggested may have been used beforehand.
6	Overall, the exposure assessment in this study is insufficient for examining TCE
7	exposure and cancer mortality. The inclusion of office workers, who likely have low potential
8	exposure, would introduce a downward bias. Furthermore, the few site-specific deaths among
9	CWD and all other workers greatly limits statistical power.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-208 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Shannon HS, Haines T, Bernholz C, Julian JA, Verma DK, Jamieson E, Walsh C. (1988). Cancer morbidity in lamp
manufacturing workers. Am J Ind Med 14:281-290.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study was undertaken in response to previous report of apparent excess breast
and gynecological cancers in women employed in the coil and wire drawing area of a
lamp manufacturing plant.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cohort analysis: 1,770 workers (1,044 females, 826 males)in the lamp manufacturing
department of a GE plant in Ontario Province, Canada.
External referents: Age-, sex- and race-specific site-specific cancer incidence rates
for Ontario Province population
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not reported.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
This study does not assign TCE exposure to individual subjects. Job title and work
in the CWD area used to assign exposure potential and chemical usage in CWD
identified from purchase records. Methylene chloride used from 1959 onward, with
one report from 1955 indicating TCE used as degreasing solvent.
CATEGORY D: FOLLOW-UP (COHORT)
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
More than 10% loss to follow-up
No, follow-up was complete for 6% of CWD workers and 7% for all other workers.
>50% cohort with full latency
Not reported
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
98 incident cancer cases
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, race, sex, and calendar year.
Statistical methods
SIR.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Adequate.
CWD = coil and wire drawing area. SIR = standardized incidence ratio.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B.3.1.4.10. Shindell and Ulrich (1985).
B.3.1.4.10.1. Author's abstract.
A prospective study was conducted of 2,646 employees who worked three months
or more during the period January, 1957, through July, 1983, in a manufacturing
plant that used trichloroethylene as a degreasing agent throughout the study
period. Ninety-eight percent of the study cohort were traced; they accounted for
16,388 person-years of employment and 38,052 person-years of follow-up.
Mortality experience was found to be generally more favorable than that of the
comparable segment of the U.S. population over the same period of time. For the
white male cohort there were fewer deaths than expected from heart disease,
cancer, and trauma (standard mortality rate for all causes = 0.79, p less than .01).
Reports by current and former employees of health problems requiring medical
treatment showed that there were only one third as many persons with heart
disease or hypertension as were reported in a comparable reference population
studied over the past five years.
B.3.1.4.10.2. Study description and comment. This study of 2, 546 current and former
office and production employees at a manufacturing plant in northern Illinois compares
broad groupings of cause-specific mortality between 1957 and 1983 to expected number of
deaths based on U.S. population mortality rates for the period. The published paper lacks
an assessment of TCE exposure other than noting TCE was used as a degreasing agent at
the plant. No information is presented on quantity used, job titles with potential exposure,
or likely exposure concentrations Not all study subjects had the same potential for
exposure and the inclusion of office workers who had a very low exposure potential
decreased the study's detection sensitivity. Deaths were identified from company records
or from direct or indirect contact with former employees or next-of-kin for subjects not
known to the company to be deceased instead of using national-based registries such as
Social Security listings or National Death Index for identifying vital status. There were few
deaths in this cohort, a total of 141 among male and female subjects; vital status could not
be ascertained for 52 subjects. The few numbers of cancer deaths (21 total) precluded
examination of cause-specific cancer mortality. Overall, this study provides no information
on TCE and cancer; it lacked exposure assessment to TCE and the few cancer deaths
observed greatly limited its detection sensitivity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-211 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Shindell S, Ulrich S. (1985). A cohort study of employees of a manufacturing plant using trichloroethylene. J Occup Med
27:577-579.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study was designed to assess mortality patterns of office and production
employees at an Illinois manufacturing plant.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
2,646 males and female workers employed from 1-1-1957 to 7-31-1983. Mortality
rates of U.S. population used as referent. The paper lacks information on source for
identifying cohort subjects and if company records were complete.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
The paper does not identify TCE usage other than as a degreaser. Conditions of
exposure and jobs potentially exposure are not identified in paper. This study lacks
an assessment of TCE exposure.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
2%.
>50% cohort with full latency
No information provided in paper.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
&
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
This study does not use standard approaches to verify deaths and vital status. Deaths
are self-reported in response to contact by employer representative. 141 deaths (6%)
were reported to employer, 9 deaths lacked a death certificate.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Sex and race.
Statistical methods
SMR.
Exposure-response analysis presented in
published paper
No.
Documentation of results
The paper lacks discussion of process used to contact former employees to verify
vital status and methods used to identify subjects.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B 3 1 4 11 Wilcosky et al. (1984).
B.3.1.4.11.1. Author's abstract.
Some evidence suggests that solvent exposures to rubber industry workers may be
associated with excess cancer mortality, but most studies of rubber workers lack
information about specific chemical exposure. In one large rubber and tire-
manufacturing plant, however, historical documents allowed a classification of
jobs based on potential exposures to all solvents that were authorized for use in
the plant. A case-control analysis of a 6,678 member cohort compared the solvent
exposure histories of a 20% age-stratified random sample of the cohort with those
of cohort members who died during 1964-1973 for stomach cancer, respiratory
system cancer, prostate cancer, lymphosarcoma, or lymphatic leukemia. Of these
cancers, only lymphosarcoma and lymphatic leukemia showed significant positive
associations with any other potential solvents exposures. Lymphatic leukemia
was especially strongly related to carbon tetrachloride (OR = 1.3, p< .0001) and
carbon disulfide (OR = 8.9, p = .0003). Lymphosarcoma showed similar, but
weaker, association with these two solvents. Benzene, a suspected carcinogen,
was not significantly associated with any of the cancers.
B.3.1.4.11.2. Study description and comment. Exposure was assessed in this nested
case-control study of four site-specific cancers among rubber workers at a plant in Akron,
OH through use of a JEM originally used to examine benzene specifically, but had the
ability to assess 24 other solvents, including TCE, or solvent classes. Exposure was inferred
using information on production operations and product specifications that indicated
whether solvents were authorized for use during tire production, and by process area and
calendar year. A subject's work history record was linked to the JEM to assign exposure
potential to TCE. Overall, a low prevalence of TCE exposure, ranging from 9 to 20% for
specific cancers was observed among cases.
The JEM was developed originally to assign exposure to benzene and other aromatic
solvents in a nested case-control study of lymphocytic leukemia (Arp et al., 1983). Details of
exposure potential to TCE are not described by either Arp et al. (1983) or Wilcosky et al. (1984).
No data were provided on the frequency of exposure-related tasks. Without more information, it
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-215 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
is not possible to determine the quality of some of the assignments. Similarly, the lack of
industrial hygiene monitoring data precluded validation of the JEM.
Cases of respiratory, stomach and prostate cancers; lymphosarcoma and reticulum cell
sarcoma; and lymphatic leukemia were identified from a previous study which had observed
associations with these site-specific cancers among a cohort of rubber workers employed at a
large tire manufacturing plant in Akron, OH. Statistical power is low in this study, particularly
for evaluation of lymphatic cancer for which there were 9 cases of lymphosarcoma and 10 cases
of lymphatic leukemia. Controls were chosen from a 20% age-stratified random sample of the
cohort. The published paper does not identify if subjects with other diseases associated with
solvents or TCE were excluded as controls. If no exclusion criteria were adopted, a bias may
have been introduced which would dampen observed associations towards the null.
The few details provided in the paper on exposure assessment and JEM developments,
few details of control selection, the low prevalence of TCE exposure and the few lymphatic
cancer cases greatly limit the ability of this study for assessing risks associated with exposures to
tri chl oroethyl ene.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-216 DRAFT—DO NOT CITE OR QUOTE
-------
Wilcosky TC, Checkoway H, Marshall EG, Tyroler HA. (1984). Cancer mortality and solvent exposure in the rubber
industry. Am Ind Hyg Assoc J 45:809-811.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study was identified as "exploratory" to examine several site-specific cancer
and specific solvents, primarily benzene.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Underlying population at risk was a cohort of 6,678 male workers employed in the
rubber industry in 1964. Cases are deaths due to respiratory, stomach and prostate
cancers; lymphosarcoma; and lymphatic leukemia observed in the cohort analysis—
30 deaths due to stomach cancer, 333 deaths from prostate cancer, 9 deaths from
lymphosarcoma, and 10 deaths from lymphatic leukemia.
Controls were a 20% age-stratified random sample of the cohort (exclusion criteria
not identified in paper).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICDA, 8th revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

>3
o
o
s
Co
I?
"tS
3'
§•
to
£
a
to
s
Co
>T
<§>
>s
>S
TO
TO'
*
^S
S
s
TO
>;
0
s
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Procedure to assign TCE and other solvent exposures based upon JEM developed
originally to assess benzene and other solvent exposures (Arp et al., 1983). The JEM
was linked to a detailed work history as identified from a subject's personnel record
to assign TCE exposure potential. Details of JEM for TCE not well-described in
Wilcosky et al. (1984). Multiple solvent exposures likely (McMichael et al., 1976).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------

>3
o
o
s
Co
I?
"tS
3'
§•
to
£
a
to
s
Co
>T
<§>
>s
>S
TO
TO'
*
^S
S
s
TO
>;
0
s
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Record study with exposure assignment using JEM and personnel records.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
N/A
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
TCE exposure prevalence:
Stomach cancer, 5 exposed cases (17% exposure prevalence)
Prostate cancer, 3 exposed cases (9% exposure prevalence)
Lymphosarcoma, 3 exposed cases (33% exposure prevalence)
Lymphatic leukemia, 2 exposed cases (20% exposure prevalence).
No information presented in paper on exposure prevalence among control subjects.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age.
Statistical methods
Not described in published paper.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Methods and analyses not fully described in published paper.
1
2
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
B.3.2. Case-Control Studies
B.3.2.1. Bladder Cancer Case-Control Studies
B.3.2.1.1. Pesch et al. (2000a)
B.3.2.1.1.1. Author's abstract.
BACKGROUND: This multicentre population-based case-control study was
conducted to estimate the urothelial cancer risk for occupational exposure to
aromatic amines, polycyclic aromatic hydrocarbons (PAH), and chlorinated
hydrocarbons besides other suspected risk factors. METHODS: In a population-
based multicentre study, 1035 incident urothelial cancer cases and 4298 controls
matched for region, sex, and age were interviewed between 1991 and 1995 for
their occupational history and lifestyle habits. Exposure to the agents under study
was self-assessed as well as expert-rated with two job-exposure matrices and a job
task-exposure matrix. Conditional logistic regression was used to calculate
smoking adjusted odds ratios (OR) and to control for study centre and age.
RESULTS: Urothelial cancer risk following exposure to aromatic amines was
only slightly elevated. Among males, substantial exposures to PAH as well as to
chlorinated solvents and their corresponding occupational settings were associated
with significantly elevated risks after adjustment for smoking (PAH exposure,
assessed with a job-exposure matrix: OR = 1.6, 95% CI: 1.1-2.3, exposure to
chlorinated solvents, assessed with a job task-exposure matrix: OR =1.8, 95% CI:
1.2-2.6). Metal degreasing showed an elevated urothelial cancer risk among males
(OR = 2.3, 95% CI: 1.4-3.8). In females also, exposure to chlorinated solvents
indicated a urothelial cancer risk. Because of small numbers the risk evaluation
for females should be treated with caution. CONCLUSIONS: Occupational
exposure to aromatic amines could not be shown to be as strong a risk factor for
urothelial carcinomas as in the past. A possible explanation for this finding is the
reduction in exposure over the last 50 years. Our results strengthen the evidence
that PAH may have a carcinogenic potential for the urothelium. Furthermore, our
results indicate a urothelial cancer risk for the use of chlorinated solvents.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-220 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
B.3 .2.1.1.2. Study description and comment. This multicenter study of urothelial
(bladder, ureter, and renal pelvis) and renal cell carcinoma in Germany included the five
regions (West Berlin, Bremen, Leverkusen, Halle, Jena), identified two case series from
participating hospitals, 1,035 urothelial cancer cases and 935 renal cell carcinoma cases
with a single population control series matched to cases by region, sex, and age (1:2
matching ratio to urothelial cancer cases and 1:4 matching ratio to renal cell carcinoma
cases). Findings in Pesch et al. (Pesch et al., 2000a) are from analyses of urothelial cancer
analysis and Pesch et al. (Pesch et al., 2000b) from analyses of renal cell carcinoma. In all,
1,035 (704 males, 331 females) urothelial carcinoma cases were interviewed face-to-face
using with a structured questionnaire in the hospital within 6 months of first diagnosis and
4,298 randomly selected population controls were interviewed at home. Logistic regression
models were fit separately to for males and females conditional on age (nine 5-year
groupings), study region, and smoking, to examine occupational chemical exposures and
urothelial carcinoma.
Two general JEMs, British and German, were used to assign exposures based on
subjects' job histories reported in an interview. This approach was the same as that described for
the renal cell carcinoma analysis of Pesch et al. (2000b). Researchers also asked about job tasks
associated with exposure, such as metal degreasing and cleaning, and use of specific agents
(organic solvents chlorinated solvents, including specific questions about carbon tetrachloride,
trichloroethylene, and tetrachloroethylene) to evaluate TCE potential using a JTEM. A category
of "any use of a solvent" mixes the large number with infrequent slight contact with the few
noted earlier who have high intensity and prolonged contact. Analyses examining
trichloroethylene exposure using either the JEM of JTEM assigned a cumulative TCE exposure
index of none to low, medium high and substantial, defined as the product of exposure
probability x intensity x duration with the following cutpoints: none to low, <30th percentile of
cumulative exposure scores; medium, 30th-<60th percentile; high, 60th-<90th percentile; and,
substantial, >90th percentile. The use of the German JEM identified approximately twice as
many cases with any potential TCE exposure (44%) compared to the JTEM (22%) and, in both
cases, few cases identified with substantial exposure, 7% by JEM and 5% by JTEM. Pesch et al.
(2000a) noted "exposure indices derived from an expert rating of job tasks can have a higher
agent-specificity than indices derived from job titles." For this reason, the JTEM approach with
consideration of job tasks is considered a more robust exposure metric for examining TCE
exposure and urothelial carcinoma due to likely reduced potential for exposure misclassification
compared to TCE assignment using only job history and title.
While this case-control study includes a region in the North Rhine-Westphalia region
where the Arnsberg area is also located, several other regions are included as well, where the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-221 DRAFT—DO NOT CITE OR QUOTE
-------
1	source of the trichloroethylene and chlorinated solvent exposures are expected as much less well
2	defined. Few cases were identified as having substantial exposure to TCE and, as a result, most
3	subjects identified as exposed to trichloroethylene probably had minimal contact, averaging
4	concentrations of about 10 ppm or less (NRC, 2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-222 DRAFT—DO NOT CITE OR QUOTE
-------
Pesch B, Haerting H, Ranft U, Klimpel A, Oelschlagel B, Schill W, and the MURC Study Group. 2000a. Occupational risk
factors for urothelial carcinoma: agent-specific results from a case-control study in Germany. Int J Epidemiol 29:238-247.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes, this case-control study was conducted to estimate urothelial carcinoma risk for
exposure to occupational-related agents; chlorinated solvents including trichloroethylene
were identified as exposures of a priori interest.
Selection and characterization in cohort
studies of exposure and control groups
and of cases and controls in case-control
studies is adequate
1,035 urothelial (bladder, ureter, renal pelvis) carcinoma cases were identified from
hospitals in a five-region area in Germany between 1991 and 1995. Cases were
confirmed histologically. 4,298 population controls identified from local residency
registries in the five-region area were frequency matched to cases by region, sex and age
comprised the control series for both the urothelial carcinoma cases and the RCC cases,
published as Pesch et al. (2000a).
Participation rate: cases, 84%; controls, 71%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
No information in paper.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
s^
Sf
<§>
>S
TO
*
^s
«
s
TO
>;
0
s
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative
exposure estimates
A trained interviewer interviewed subjects using a structured questionnaire which covered
occupational history and job title for all jobs held longer than one yr, medical history, and
personal information. Two general JEMs, British and German, were used to assign
exposures based on subjects' job histories reported in an interview. Researchers also
asked about job tasks associated with exposure, such as metal degreasing and cleaning,
and use of specific agents (organic solvents chlorinated solvents, including specific
questions about carbon tetrachloride, trichloroethylene, and tetrachloroethylene) and
chemical-specific exposure were assigned using a JTEM. Exposure index for each
subject is the sum over all jobs of duration x probability x intensity. A four category
grouping was used in statistical analyses defined by exposure index distribution of
controls: no-low; medium, 30th percentile; high, 60th percentile; substantial, 90th
percentile.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Interviewers carried out face-to-face interview with all cases and controls. All cases were
interviewed in the hospital within 6 mos of initial diagnosis. All controls had home
interviews.
Blinded interviewers
No, by nature of interview location.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No, all cases and controls were alive at time of interview.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
s^
Sf
<§>
>S
TO
*
^s
«
s
TO
>;
0
s
1
a.
VO
<&
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality
studies; numbers of total cancer incidence
studies; numbers of exposed cases and
prevalence of exposure in case-control
studies
JEM: 460 cases with TCE exposure index of medium or higher (44% exposure prevalence
among cases), 71 cases with substantial exposure (7% exposure prevalence).
JTEM: 157 cases with TCE exposure index of medium or higher (22% exposure
prevalence among cases), and 36 males assigned substantial exposure (5% exposure
prevalence).
No information is presented in paper on control exposure prevalence.
CATEGORY H: ANALYSIS
Control for potential confounders in
statistical analysis
Age, study center, and smoking.
Statistical methods
Conditional logistic regression.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
B.3.2.1.2. Siemiatycki et al. (1994), Siemiatycki (1991).
B.3.2.1.2.1. Author's abstract.
A population-based case-control study of the associations between various
cancers and occupational exposures was carried out in Montreal, Quebec, Canada.
Between 1979 and 1986, 484 persons with pathologically confirmed cases of
bladder cancer and 1,879 controls with cancers at other sites were interviewed, as
was a series of 533 population controls. The job histories of these subjects were
evaluated by a team of chemist/hygienists for evidence of exposure to a list of 294
workplace chemicals, and information on relevant non-occupational confounders
was obtained. On the basis of results of preliminary analyses and literature
review, 19 occupations, 11 industries, and 23 substances were selected for in-
depth multivariate analysis. Logistic regression analyses were carried out to
estimate the odds ratio between each of these occupational circumstances and
bladder cancer. There was weak evidence that the following substances may be
risk factors for bladder cancer: natural gas combustion products, aromatic amines,
cadmium compounds, photographic products, acrylic fibers, polyethylene,
titanium dioxide, and chlorine. Among the substances evaluated which showed no
evidence of an association were benzo(a)pyrene, leather dust, and formaldehyde.
Several occupations and industries were associated with bladder cancer, including
motor vehicle drivers and textile dyers.
B.3.2.1.2.2. Study description and comment. Siemiatycki et al. (1994) and Siemiatycki
(1991) reported data from a case-control study of occupational exposures and bladder
cancer conducted in Montreal, Quebec (Canada) and part of a larger study of 10 other site-
specific cancers and occupational exposures. The investigators identified 617 newly
diagnosed cases of primary bladder cancer, confirmed on the basis of histology reports,
between 1979 and 1985; 484 of these participated in the study interview (78%
participation). One control group (n = 1,295) consisted of patients with other forms of
cancer (excluding lung and kidney cancer) recruited through the same study procedures
and time period as the bladder cancer cases. A population-based control group (n = 533,
72% response), frequency matched by age strata, was drawn using electoral lists and
random digit dialing. Face-to-face interviews were carried out with 82% of all cancer cases
with telephone interview (10%) or mailed questionnaire (8%) for the remaining cases.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09
B-226
DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Twenty percent of all case interviews were provided by proxy respondents. The
occupational assessment consisted of a detailed description of each job held during the
working lifetime, including the company, products, nature of work at site, job activities,
and any additional information that could furnish clues about exposure from the
interviews.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Siemiatycki et al.
(1994) presents observations of analyses examining job title, occupation, and some chemical-
specific exposures, but not TCE. Observations on TCE are found in the original report of
Siemiatycki (1991). Any exposure to TCE was 2% among cases (n = 8) but <1% for substantial
TCE exposure (n = 5); "substantial" is defined as >10 years of exposure for the period up to
5 years before diagnosis. Logistic regression models adjusted for age, ethnicity, socioeconomic
status, smoking, coffee consumption, and status of respondent (Siemiatycki et al., 1994) or
Mantel-Henszel % stratified on age, family income, cigarette smoking, coffee, and respondent
status (Siemiatycki, 1991). Odds ratios for TCE exposure are presented in Siemiatycki (1991)
with 90% confidence intervals.
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of bladder cancer. However, the use
of the general population (rather than a known cohort of exposed workers) reduced the likelihood
that subjects were exposed to TCE, resulting in relatively low statistical power for the analysis.
The job exposure matrix, applied to the job information, was very broad since it was used to
evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-227 DRAFT—DO NOT CITE OR QUOTE
-------
Siemiatycki J, Dewar R, Nadon L, Gerin M. (1994). Occupational risk factors for bladder cancer: results from a case-control
study in Montreal, Quebec, Canada. Am J Epidemiol 140:1061-1080.
Siemiatycki J. (1991). Risk Factors for Cancer in the Workplace. Baca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
617 bladder cancer cases were identified among male Montreal residents between
1979 and 1985 of which 484 were interviewed.
740 eligible male controls identified from the same source population using random
digit dialing or electoral lists; 533 were interviewed. A second control series
consisted of all other cancer controls excluding lung and kidney cancer cases.
Participation rate: cases, 78%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 188 (Malignant neoplasm of bladder).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 300 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were conducted
either at home or in the hospital; all population control interviews were conducted at
home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
484 cases (78% response), 533 population controls (72%).
Exposure prevalence: Any TCE exposure, 2% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), <1% cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, income, index for cigarette smoking, coffee, and respondent status
(Siemiatycki, 1991).
Age, ethnicity, socioeconomic status, smoking, coffee consumption, and status of
respondent (Siemiatycki et al., 1994).
-------
a,	s
c>	to
to	s
Co
<3	Co'
8.	Si
§	t
£.	>
s	?
^	>5
!>q	S.
<3	TO
S	^
^	^3
^	5
S.	^
• TO	Co
V;	TO
VO
to
o
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Siemiatycki et al., 1994).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
B.3.2.2. Central Nervous System Cancers Case-Control Studies
B.3.2.2.1. De Roos et al. (2001a).
B.3.2.2.1.1. Author's abstract.
To evaluate the effects of parental occupational chemical exposures on incidence
of neuroblastoma in offspring, the authors conducted a multicenter case-control
study, using detailed exposure information that allowed examination of specific
chemicals. Cases were 538 children aged 19 years who were newly diagnosed
with confirmed neuroblastoma in 1992-1994 and were registered at any of 139
participating hospitals in the United States and Canada. One age-matched control
for each of 504 cases was selected through random digit dialing. Self-reported
exposures were reviewed by an industrial hygienist, and improbable exposures
were reclassified. Effect estimates were calculated using unconditional logistic
regression, adjusting for child's age and maternal demographic factors. Maternal
exposures to most chemicals were not associated with neuroblastoma. Paternal
exposures to hydrocarbons such as diesel fuel (odds ratio (OR) = 1.5; 95%
confidence interval (CI): 0.8, 2.6), lacquer thinner (OR = 3.5; 95% CI: 1.6, 7.8),
and turpentine (OR = 10.4; 95% CI: 2.4, 44.8) were associated with an increased
incidence of neuroblastoma, as were exposures to wood dust (OR = 1.5; 95% CI:
0.8, 2.8) and solders (OR = 2.6; 95% CI: 0.9, 7.1). The detailed exposure
information available in this study has provided additional clues about the role of
parental occupation as a risk factor for neuroblastoma.
B.3.2.2.1.2. Study description and comment. De Roos et al. (2001a) a large multicenter
case-control study of neuroblastoma in offspring and part of the pediatric collaborative
clinical trial groups, the Children's Cancer Group and the pediatric Oncology Group,
examined parental and maternal chemical exposures, focusing on solvent exposures,
expanding the exposure assessment approach of Olshan et al. (1999) who examined
parental occupational title among cases and controls. Neuroblastoma in patients under the
age of 19 years was identified at one of 139 participating hospitals in the United States and
Canada from 1992 to 1996. One population control per case s was using a telephone
random digit dialing procedure and matched to the case on date of birth (+6 months for
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-232 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
cases 3 years old or younger and +1 year for cases old than 3 years of age). A total of 741
cases and 708 controls were identified with direct interviews by telephone obtained from
538 case mothers (73% participation), 405 case fathers, 504 control mothers (71%
participation), and 304 control fathers. Mothers served as proxy respondents for paternal
information for 67 cases (12%) and 141 controls (28%).
A strength of the study was its use of industrial hygienist review of self-reported
occupational exposure to increase specificity, reduce the number of false-positive information
from self-reported exposures, and to minimize exposure misclassification bias. A parent was
coded as having been exposed to individual chemicals or chemical group (halogenated
hydrocarbons, paints, metals, etc.) if the industrial hygiene review determined probable exposure
in any job. Individual chemicals in the halogenated hydrocarbons grouping included carbon
tetrachloride, chloroform, Freon, methylene chloride, perchloroethylene and TCE. Typical of
population case-control studies, reported TCE exposure was uncommon among cases and
controls. Only 6 case and 8 control mothers were identified by industrial hygiene review of
occupational information to have probable exposure to halogenated hydrocarbons. The few
numbers prevented examination of specific chemical exposure. Of the 538 cases and
504 controls, paternal exposure to TCE was self-reported for 22 cases (5%) and 12 controls (4%)
were identified with paternal TCE exposure with fewer fathers with probable TCE exposure
confirmed from industrial hygiene expert review, 9 cases (2%) and 7 controls (2%).
Overall, this study has a low sensitivity and statistical power for evaluating parental TCE
exposure and neuroblastoma in offspring due to the low exposure prevalence to TCE. Although
study investigators took effort to reduce false positive reporting, exposure misclassification bias
may still be possible from false negative reporting of occupational information. As discussed by
study authors, job duty information reported by parents was best used to infer exposure to
chemical categories but was not detailed sufficiently to infer specific exposures. The study's
reported risk estimates for TCE exposure are imprecise and do not provide support for or against
an association.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-233 DRAFT—DO NOT CITE OR QUOTE
-------
De Roos AJ, Olshan AF, Teschke K, Poole Ch, Savitz DA, Blatt J, Bondy ML, Pollock BH. (2001a). Parental occupational
exposure to chemicals and incidence of neuroblastoma in offspring. Am J Epidemiol 154:106-114.
Olshan AF, De Roos AJ, Teschke K, Neglin JP, Stram DO, Pollock BH, Castleberry RP. (1999). Neuroblastoma and parental
occupation. Cancer Causes Control 10:539-549.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or hypothesis
This multicenter population case-control study examined parental
chemical-specific occupational exposures using detailed exposure information.
Selection and characterization in cohort studies of
exposure and control groups and of cases and
controls in case-control studies is adequate
538 cases of neuroblastoma in children <19 years of age and diagnosed between
1992 and 1994 at any of 139 United States or Canadian hospitals participating in
the Children's Cancer Group and Pediatric Oncology Group studies.
504 population controls were selected through random digit dialing and matched
(1:1) with cases on date of birth. Controls could not be located for 34 cases.
538 of 741 potentially eligible cases (73% participation rate).
504 of 681 potentially eligible controls (74% participation rate).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's lymphoma

CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
s
s
TO
>;
0
s
1
a.
VO
to
o
Exposure assessment approach, including adoption
of JEM and quantitative exposure estimates
Self-reported exposure to any of 65 chemicals, compounds, or broad categories
was obtained from structured questionnaire. An industrial hygienist confirmed
each respondent's self-reported chemical exposure responses. Exposures were
not assigned using JEM.
TCE exposure examined in analysis as separate exposure and as one of several
chemicals in the broader category of "halogenated hydrocarbons."
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Telephone interview with mother and father of each case and control.
Blinded interviewers
Not identified in paper.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No proxy information on maternal exposure; direct interview with mother was
obtained for 537 cases and 503 controls.
Analysis of paternal chemical exposures did not include information on paternal
exposure from proxy interviews.
CATEGORY G: SAMPLE SIZE
-------
ir-
>9
to
S
"tS
§•
TO
£
a
to
s
>3
o
«••*.
o
s
Co
I?
f? Ng
^ 3
Co
S
S*
<§>
>S
TO
*
^S
s
s
TO
>;
0
s
•?
1
a.
VO
to
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies; numbers
of exposed cases and prevalence of exposure in
case-control studies
Self-reported TCE exposure: 22 cases (5% exposure prevalence) and 12 controls
(4% exposure prevalence).
IH-reviewed TCE exposure: 9 cases (2% exposure prevalence) and 7 controls
(2% exposure prevalence).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Analyses of maternal and paternal occupational exposure each adjusted for
child's age, maternal race, maternal age, and maternal education.
Statistical methods
Separate analyses are conducted for maternal and paternal exposure using
logistic regression methods.
Exposure-response analysis presented in published
paper
No.
Documentation of results
Yes, results are well documented.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
B.3.2.2.2. Heineman et al. (1994).
B.3.2.2.2.1. Author's abstract.
Chlorinated aliphatic hydrocarbons (CAHs) were evaluated as potential risk
factors for astrocytic brain tumors. Job-exposure matrices for six individual
CAHs and for the general class of organic solvents were applied to data from a
case-control study of brain cancer among white men. The matrices indicated
whether the CAHs were likely to have been used in each industry and occupation
by decade (1920-1980), and provided estimates of probably and intensity of
exposure for "exposed" industries and occupations. Cumulative exposure indices
were calculated for each subject.
Associations of astrocytic brain cancer were observed with likely exposure to
carbon tetrachloride, methylene chloride, tetrachloroethylene, and
trichloroethylene, but were strongest for methylene chloride. Exposure to
chloroform or methyl chloroform showed little indication of an association with
brain cancer. Risk of astrocytic brain tumors increase with probability and
average intensity of exposure, and with duration of employment in jobs
considered exposed to methylene chloride, but not with a cumulative exposure
score. These trends could not be explained by exposures to the other solvents.
B.3.2.2.2.2. Study description and comment. Heineman et al. (1994) studied the
association between astrocytic brain cancer (ICD-9 codes 191,192, 225, and 239.7) and
occupational exposure to chlorinated aliphatic hydrocarbons. Cases were identified using
death certificates from southern Louisiana, northern New Jersey, and the Philadelphia
area. This analysis was limited to white males who died between 1978 and 1981. Controls
were randomly selected from the death certificates of white males who died of causes other
than brain tumors, cerebrovascular disease, epilepsy, suicide, and homicide. The controls
were frequency matched to cases by age, year of death, and study area.
Next-of-kin were successfully located for interview for 654 cases and 612 controls,
which represents 88 and 83% of the identified cases and controls, respectively. Interviews were
completed for 483 cases (74%) and 386 controls (63%). There were 300 cases of astrocytic
brain cancer (including astrocytoma, glioblastoma, mixed glioma with astrocytic cells). The
ascertainment of type of cancer was based on review of hospital records which included
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-237 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
pathology reports for 229 cases and computerized tomography reports for 71 cases. After
excluding 66 controls with a possible association between occupational exposure to chlorinated
aliphatic hydrocarbons and cause of death (some types of cancer, cirrhosis of the liver), the final
analytic sample consisted of 300 cases and 320 controls.
In the next-of-kin interviews, the work history included information about each job held
since the case (or control) was 15 years old (job title, description of tasks, name and location of
company, kinds of products, employment dates, and hours worked per week). Occupation and
industry were coded based on four digit Standard Industrial Classification and Standard
Occupational Classification (Department of Commerce) codes. The investigators developed
matrices linked to jobs with likely exposure to six chlorinated aliphatic hydrocarbons (carbon
tetrachloride, chloroform, methyl chloroform, methylene dichloride, tetrachloroethylene, and
trichloroethylene), and to organic solvents (Gomez et al., 1994). This assessment was done
blinded to case-control status. Exposure was defined as the probability of exposure to a
substance (the highest probability score for that substance among all jobs), duration of
employment in the exposed occupation and industry, specific exposure intensity categories,
average intensity score (the three-level semiquantitative exposure concentration assigned to each
job multiplied by duration of employment in the job, summed across all jobs), and cumulative
exposure score (weighted sum of years in all exposed jobs with weights based on the square of
exposure intensity [1, 2, 3] assigned to each job). Secular trends in the use of specific chemicals
were considered in the assignment of exposure potential. Exposures were lagged 10 or 20 years
to account for latency. Thus, this exposure assessment procedure was quite detailed.
The strengths of this case-control study include a large sample size, detailed work
histories including information not just about usual or most recent industry and occupation, but
also about tasks and products for all jobs held since age 15, and comprehensive exposure
assessment and analysis along several different dimensions of exposure. The major limitation
was the lack of direct exposure information and potential inaccuracy of the description of work
histories that was obtained from next-of-kin interviews. The authors acknowledge this limitation
in the report, and in response to a letter by Norman (1968) criticizing the methodology and
interpretation of the study with respect to the observed association with methylene chloride,
Heineman et al. (1994) noted that while the lack of direct exposure information must be
interpreted cautiously, it does not invalidate the results. Differential recall bias between cases
and controls was unlikely because work histories came from next-of-kin for both groups and, the
industrial hygienists made their judgments blinded to disease status. Nondifferential
misclassification is possible due to underreporting of job information by next of kin and would,
on average, attenuate true associations.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-23 8 DRAFT—DO NOT CITE OR QUOTE
-------
Heineman EF, Cocco P, Gomez MR, Dosemeci M, Stewart PA, Hayes RB, Zahm SH, Thomas TL, Blair A. (1994).
Occupational exposure to chlorinated aliphatic hydrocarbons and risk of astrocytic brain cancer. Am J Ind Med 26:155-169.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes, study further examines six specific solvents including trichloroethylene in a
previous study of brain cancer which reported association with electrical equipment
production and repair.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Brain cancer deaths among white males in southern Louisiana, northern New Jersey
and Philadelphia, Pennsylvania, were identified using death certificates (n = 741).
Controls were randomly selected (source not identified in paper) among other
cause-specific deaths among white male residents of these areas and matched to
cases by age, year of death and study area (n = 741).
Participation rate, 483 of 741 (65% of cases with brain cancer); 386 of 741 controls
(52%). Of the 483, 300 deaths were due to astrocytic brain cancer.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 9th revision, Codes 191, 192, 225, 239.7.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
The job-exposure-matrix of Gomez et al. (1994) was used to assign potential
exposure to 6 solvents including trichloroethylene.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% Face-to-Face
Interview with next-of-kin but paper does not identify whether telephone or face-to-
face.
Blinded interviewers
Interviewer was blinded as to case and control status.
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
Proxy information was obtained from 100% of cases and controls.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
TCE exposure prevalence: 128 cases (43%) and 125 controls (39%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Stratified analysis controlled for age, year of death and study area; employment in
electronics-related occupations was included in addition in logistic regression
analyses.
Statistical methods
Stratified analysis using 2x2 tables and logistic regression.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
B.3.2.3. Colon and Rectal Cancers Case-Control Studies
B.3.2.3.1. Goldberg et al. (2001), Siemiatycki (1991).
B.3.2.3.1.1. Author's abstract.
BACKGROUND: We conducted a population-based case-control study in
Montreal, Canada, to explore associations between hundreds of occupational
circumstances and several cancer sites, including colon. METHODS: We
interviewed 497 male patients with a pathologically confirmed diagnosis of colon
cancer, 1514 controls with cancers at other sites, and 533 population-based
controls. Detailed job histories and relevant potential confounding variables were
obtained, and the job histories were translated by a team of chemists and
industrial hygienists into a history of occupational exposures. RESULTS: We
found that there was reasonable evidence of associations for men employed in
nine industry groups (adjusted odds ranging from 1.1 to 1.6 per a 10-year increase
in duration of employment), and in 12 job groups (OR varying from 1.1 to 1.7). In
addition, we found evidence of increased risks by increasing level of exposures to
21 occupational agents, including polystyrene (OR for "substantial" exposure
(OR(subst) = 10.7), polyurethanes (OR(subst) = 8.4), coke dust (OR(subst) = 5.6),
mineral oils (OR(subst) = 3.3), polyacrylates (OR(subst) = 2.8), cellulose nitrate
(OR(subst) = 2.6), alkyds (OR(subst) = 2.5), inorganic insulation dust (OR(subst)
= 2.3), plastic dusts (OR(subst) = 2.3), asbestos (OR(subst) = 2.1), mineral wool
fibers (OR(subst) = 2.1), glass fibers (OR(subst) = 2.0), iron oxides (OR(subst) =
1.9), aliphatic ketones (OR(subst) = 1.9), benzene (OR(subst) = 1.9), xylene
(OR(subst) = 1.9), inorganic acid solutions (OR(subst) = 1.8), waxes, polishes
(OR(subst) = 1.8), mononuclear aromatic hydrocarbons (OR(subst) = 1.6),
toluene (OR(subst) = 1.6), and diesel engine emissions (OR(subst) = 1.5). Not all
of these effects are independent because some exposures occurred
contemporaneously with others or because they referred to a group of substances.
CONCLUSIONS: We have uncovered a number of occupational associations
with colon cancer. For most of these agents, there are no published data to support
or refute our observations. As there are few accepted risk factors for colon cancer,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-242 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
we suggest that new occupational and toxicologic studies be undertaken focusing
on the more prevalent substances reported herein.
B.3.2.3.1.2. Study description and comment. Goldberg et al. (2001), and Siemiatycki
(1991) reported data from a case-control study of occupational exposures and colon cancer
conducted in Montreal, Quebec (Canada) and part of a larger study of 10 other site-specific
cancers and occupational exposures. The investigators identified 607 newly diagnosed
cases of primary colon cancer (ICD9,153), confirmed on the basis of histology reports,
between 1979 and 1985; 497 of these participated in the study interview (81.9%
participation). One control group (n = 1,514) consisted of patients with other forms of
cancer (excluding cancers of the lung, peritoneum, esophagus, stomach, small intestine,
rectum, liver and intrahepatic bile ducts, gallbladder and extrahepatic bile ducts and
pancreas) recruited through the same study procedures and time period as the colon
cancer cases. A population-based control group (n = 533, 72% response), frequency
matched by age strata, was drawn using electoral lists and random digit dialing. Face-to-
face interviews were carried out with 82% of all cancer cases with telephone interview
(10%) or mailed questionnaire (8%) for the remaining cases. Twenty percent of all case
interviews were provided by proxy respondents. The occupational assessment consisted of
a detailed description of each job held during the working lifetime, including the company,
products, nature of work at site, job activities, and any additional information that could
furnish clues about exposure from the interviews.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Goldberg et al.
(2001) presents observations of analyses examining industries, occupation, and some
chemical-specific exposures, but not TCE. Observations on TCE are found in the original report
of Siemiatycki (1991). Any exposure to TCE was 2% among cases (n = 12) and 1% for
substantial TCE exposure (n = 7); "substantial" is defined as >10 years of exposure for the
period up to 5 years before diagnosis.
Logistic regression models adjusted for a number of nonoccupational variables including
age, ethnicity, birthplace, education, income, parent's occupation, smoking, alcohol
consumption, tea consumption, respondent status, heating source and cooking source in
childhood home, consumption of nonpublic water supply, and body mass index (Goldberg et al.,
2001) or Mantel-Haenszel x stratified on age, family income, cigarette smoking, coffee, ethnic
origin, and beer consumption (Siemiatycki, 1991). Odds ratios for TCE exposure are presented
in Siemiatycki (1991) with 90% confidence intervals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-243 DRAFT—DO NOT CITE OR QUOTE
-------
1	The strengths of this study were the large number of incident cases, specific information
2	about job duties for all jobs held, and a definitive diagnosis of colon cancer. However, the use of
3	the general population (rather than a known cohort of exposed workers) reduced the likelihood
4	that subjects were exposed to TCE, resulting in relatively low statistical power for the analysis.
5	The job exposure matrix, applied to the job information, was very broad since it was used to
6	evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-244 DRAFT—DO NOT CITE OR QUOTE
-------
Goldberg MS, Parent M-E, Siemiatycki J, Desy M, Nadon L, Richardson L, Lakhani R, Lateille B, Valois M-F. (2001). A
case-control study of the relationship between the risk of colon cancer in men and exposure to occupational agents. Am J Ind
Med 39:5310-546.
Siemiatycki J. (1991). Risk Factors for Cancer in the Workplace. Baca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on
possible association between 11 site-specific cancers and occupational title or
chemical exposures.
Selection and characterization in cohort studies of
exposure and control groups and of cases and
controls in case-control studies is adequate
607 colon cancer cases were identified among male Montreal residents between
1979 and 1985 of which 497 were interviewed.
740 eligible male controls identified from the same source population using
random digit dialing or electoral lists; 533 were interviewed. A second control
series consisted of all other cancer controls excluding lung peritoneum and other
digestive cancers.
Participation rate: cases, 81.9%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's lymphoma
ICD-9, 153 (Malignant neoplasm of colon).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job
history with supplemental questionnaire for jobs of a priori interest (e.g.,
machinists, painters). Team of chemist and industrial hygienist assigned exposure
using job title with a semiquantitative scale developed for 294 exposures, including
TCE. For each exposure, a 3-level ranking was used for concentration (low or
background, medium, high) and frequency (percent of working time: low, 1 to 5%;
medium, >5 to 30%; and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were
conducted either at home or in the hospital; all population control interviews were
conducted at home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
497 cases (81.9%> response), 533 population controls (72%>).
Exposure prevalence: Any TCE exposure, 2% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), 1% cases.
CATEGORY H: ANALYSIS
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
I
a.
VO
to
o
Control for potential confounders in statistical
analysis
Age, ethnicity, birthplace, education, income, parent's occupation, smoking,
alcohol consumption, tea consumption, respondent status, heating source and
cooking source in childhood home, consumption of nonpublic water supply, and
body mass index (Goldberg et al., 2001).
Age, family income, cigarette smoking, coffee, ethnic origin, and beer consumption
(Siemiatycki, 1991).
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Goldberg et al., 2001).
Exposure-response analysis presented in published
paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3.2.3.2. Dumas et al.(2000), Siemiatycki (1991).
B.3.2.3.2.1. Author's abstract.
In 1979, a hypothesis-generating, population-based case-control study was
undertaken in Montreal, Canada, to explore the association between occupational
exposure to 294 substances, 130 occupations and industries, and various cancers.
Interviews were carried out with 3,630 histologically confirmed cancer cases, of
whom 257 had rectal cancer, and with 533 population controls, to obtain detailed
job history and data on potential confounders. The job history of each subject was
evaluated by a team of chemists and hygienists and translated into occupational
exposures. Logistic regression analyses adjusted for age, education, cigarette
smoking, beer consumption, body mass index, and respondent status were
performed using population controls and cancer controls, e.g., 1,295 subjects with
cancers at sites other than the rectum, lung, colon, rectosigmoid junction, small
intestine, and peritoneum. We present here the results based on cancer controls.
The following substances showed some association with rectal cancer: rubber
dust, rubber pyrolysis products, cotton dust, wool fibers, rayon fibers, a group of
solvents (carbon tetrachloride, methylene chloride, trichloroethylene, acetone,
aliphatic ketones, aliphatic esters, toluene, styrene), polychloroprene, glass fibers,
formaldehyde, extenders, and ionizing radiation. The independent effect of many
of these substances could not be disentangled as many were highly correlated with
each other.
B.3.2.3.2.2. Study description and comment. Dumas et al. (2000) and Siemiatycki (1991)
reported data from a case-control study of occupational exposures and rectal cancer
conducted in Montreal, Quebec (Canada) and part of a larger study of 10 other site-specific
cancers and occupational exposures. The investigators identified 304 newly diagnosed
cases of primary rectal cancers, confirmed on the basis of histology reports, between
1979 and 1985; 257 of these participated in the study interview (84.5% response). One
control group (n = 1,295) consisted of patients with other forms of cancer (excluding lung
cancer and other intestinal cancers) recruited through the same study procedures and time
period as the rectal cancer cases. A population-based control group (n = 533), frequency
matched by age strata, was drawn using electoral lists and random digit dialing (72%
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-249 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
response). The occupational assessment consisted of a detailed description of each job held
during the working lifetime, including the company, products, nature of work at site, job
activities, and any additional information that could furnish clues about exposure from the
interviews. The percentage of proxy respondents was 15.2% for cases, 19.7% for other
cancer controls, and 12.6% for the population controls.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Any exposure to
TCE was 5% among cases (n = 12) and 1% for substantial TCE exposure (n = 3); "substantial" is
defined as >10 years of exposure for the period up to 5 years before diagnosis.
Logistic regression models adjusted for age, education, respondent status, cigarette
smoking, beer consumption and body mass index (Dumas et al., 2000) or Mantel-Haenszel %
stratified on age, family income, cigarette smoking, coffee, ethnic origin, and beer consumption
(Siemiatycki, 1991). Dumas et al. (2000) presents observations of analyses examining
industries, occupation, and some chemical-specific exposures, including TCE. Observations on
TCE from Mantel-Haenszel analyses are found in the original report of Siemiatycki (1991).
Odds ratios for TCE exposure are presented in Siemiatycki (1991) with 90% confidence intervals
and 95% confidence intervals in Dumas et al. (2000).
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of rectal cancer. However, the use of
the general population (rather than a known cohort of exposed workers) reduced the likelihood
that subjects were exposed to TCE, resulting in relatively low statistical power for the analysis.
The job exposure matrix, applied to the job information, was very broad since it was used to
evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-250 DRAFT—DO NOT CITE OR QUOTE
-------
Dumas S, Parent M-E, Siemiatycki J, Brisson J. (2000). Rectal cancer and occupational risk factors: a hypothesis-generating,
exposure-based case-control study. Int J Cancer 87:874-879.
SiemityckiJ. (1991). Risk Factors for Cancer in the Workplace. Boca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
304 rectal cancer cases were identified among male Montreal residents between 1979
and 1985 of which 294 were interviewed.
740 eligible male controls identified from the same source population using random
digit dialing or electoral lists; 533 were interviewed. A second control series
consisted of all other cancer controls excluding lung and other intestinal cancer cases.
Participation rate: cases, 84.5%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 154 (Malignant neoplasm of rectum, rectosigmoid junction and anus).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 294 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face to face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were conducted
either at home or in the hospital; all population control interviews were conducted at
home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
294 cases (78% response), 533 population controls (72% response).
Exposure prevalence: Any TCE exposure, 5% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), 1% cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, education, respondent status, cigarette smoking, beer consumption and body
mass index (Dumas et al., 2000).
Age, family income, cigarette smoking, coffee, ethnic origin, and beer consumption
(Siemiatycki, 1991).
-------
a,	s
c>	to
to	s
Co
<3	Co'
8.	Si
§	t
£.	>
s	?
^	>5
!>q	S.
<3	TO
S	^
^	^3
^	5
S.	^
• TO	Co
V;	TO
VO
to
o
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Dumas et al., 2000).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3 .2.3 .3. Fredriksson et al. (1989).
B.3.2.3.3.1. Author's abstract.
A case-control study on colon cancer was conducted encompassing 329 cases and
658 controls. Occupations and various exposures were assessed by questionnaires.
A decreased risk was found in persons with physically active occupations. This
effect was most pronounced in colon descendens and sigmoideum with an odds
ratio (OR) of 0.49 whereas no reduced risk was found for right-sided colon
cancer. Regarding specific jobs, reduced ORs were found for agricultural,
forestry, and saw mill workers and increased OR for railway employees. High-
grade exposure to asbestos or to organic solvents gave a two-fold increased risk.
Regarding exposure to trichloroethylene in general, a slightly increased risk was
found whereas such exposure among dry cleaners gave a 7-fold increase of the
risk.
B.3.2.3.3.2. Study description and comment. Fredriksson et al. (1989) reported data
from a population case-control study of occupational and nonoccupational exposures and
rectal cancer conducted in Urea, Sweden. The investigators identified 329 diagnosed cases
of rectal cancers (ICD 8,153), between 1980 and 1983, confirmed on the basis of histology
reports and alive at the time of data collect between 1984 and 1986; 302 (165 males and 165
females) of these participated in the study interview (92% response). A population-based
control group (n = 658), matched by a 1:2 ratio to cases on age sex and county residence,
was drawn using the Swedish National Population Register list; 623 (306 males and 317
females) returned mailed questionnaires and participated in the study (95% response).
The occupational assessment consisted of a detailed description of each job held during
the working lifetime, including details on specific occupations and exposures. Occupation
information was provided directly from each case and control given the study's eligibility
requirement of being alive at the time of data collection. A team of experts independently
classified three exposures of interest (asbestos, organic solvents, and impregnating agents) into
two categories, low grade exposure and high grade exposure and other chemical-specific
exposures, including TCE, as either "exposed" or "unexposed." Fredriksson et al. (1989) do not
define these categories nor do they provide information on exposure potential, frequency of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-255 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
exposure, or concentration of exposure. No information is provided whether experts were
blinded as to disease status.
Statistical analysis examining occupation and agent-specific exposures was carried out
using Mantel-Haenszel % stratified on age, sex, and an index of physical activity. Odds ratios
associated with specific chemical exposure are presented with their 95% confidence intervals.
The strengths of this study were its specific information about job duties for all jobs held
and a definitive diagnosis of rectal cancer. However, the study's assignment of exposure
potential from information using mailed questionnaires is considered inferior to information
obtained directly from trained interviewers and expert assessment because of greater uncertainty
and misclassification (Fritschi et al., 1996). The degree of potential exposure misclassification
bias in this population case-control study of colon cancer is not known. Furthermore, exposure
prevalence to TCE appears low, as judged by the wide confidence interval around the odds ratio.
This study is considered as having decreased sensitivity for examining colon cancer and TCE
given the apparent lower exposure prevalence and likely exposure misclassification bias
associated with mailed questionnaire information.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-256 DRAFT—DO NOT CITE OR QUOTE
-------
Fredriksson M, Bengtsson N-O, Hardell L, Axelson O. (1989). Colon cancer, physical activity, and occupational exposure. A
case-control study. Cancer 63:1838-1842.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Abstract—to evaluate occupational and nonoccupational exposures as risk factors for
colon cancer.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
302 (165 males and 165 females) cases participated in study out of 329 eligible cases
reported to the Swedish Cancer Registry between 1980 and 1983, among resident of
Umea, Sweden, alive at time of data collection 1984 and 1986, and with
histological-confirmed diagnosis of colon cancer.
623 (306 males and 317 females) identified from Swedish Population Registry and
matched for age, sex, and county of residence.
Participation rate: cases, 92%; population controls, 95%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-8, 153 (Malignant neoplasm of large intestine, except rectum).
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Self-reported information on occupational exposure as obtained from a mailed
questionnaire to study participants. Questionnaire sought information on complete
working history, other exposures, and dietary habits. Procedure for assigning
chemical exposures from job title information not described in paper.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Mailed questionnaire.
Blinded interviewers
No information in published paper.
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
No proxy respondents, all cases and controls alive at time of data collection.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
302 cases (92% response), 623 population controls (95% response).
Exposure prevalence not calculated, published paper lacks number of TCE exposed
cases and controls.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Yes, age, sex, and index of physical activity.
Statistical methods
Mantel-Haenszel.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
B.3.2.4. Esophageal Cancer Case-Control Studies
B.3.2.4.1. Parent et al. (2000a) , Siemiatycki (1991).
B.3.2.4.1.1. Parent et al. (2000a) abstract.
OBJECTIVES: To describe the relation between oesophageal cancer and many
occupational circumstances with data from a population based case-control study.
METHODS: Cases were 99 histologically confirmed incident cases of cancer of
the oesophagus, 63 of which were squamous cell carcinomas. Various control
groups were available; for the present analysis a group was used that comprised
533 population controls and 533 patients with other types of cancer. Detailed job
histories were elicited from all subjects and were translated by a team of chemists
and hygienists for evidence of exposure to 294 occupational agents. Based on
preliminary results and a review of literature, a set of 35 occupational agents and
19 occupations and industry titles were selected for this analysis. Logistic
regression analyses were adjusted for age, birthplace, education, respondent (self
or proxy), smoking, alcohol, and beta-carotene intake. RESULTS: Sulphuric acid
and carbon black showed the strongest evidence of an association with
oesophageal cancer, particularly squamous cell carcinoma. Other substances
showed excess risks, but the evidence was more equivocal-namely chrysotile
asbestos, alumina, mineral spirits, toluene, synthetic adhesives, other paints and
varnishes, iron compounds, and mild steel dust. There was considerable overlap
in occupational exposure patterns and results for some of these substances may be
mutually confounded. None of the occupations or industry titles showed a clear
excess risk; the strongest hints were for warehouse workers, food services
workers, and workers from the miscellaneous food industry. CONCLUSIONS:
The data provide some support for an association between oesophageal cancer
and a handful of occupational exposures, particularly sulphuric acid and carbon
black. Many of the associations found have never been examined before and
warrant further investigation.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-260 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
B.3.2.4.1.2. Study description and comment. Parent et al. (2000a) and Siemiatycki (1991)
reported data from a case-control study of occupational exposures and esophageal cancer
conducted in Montreal, Quebec (Canada) and part of a larger study of 10 other site-specific
cancers and occupational exposures. The investigators identified 129 newly diagnosed
cases of primary esophageal cancers, confirmed on the basis of histology reports, between
1979 and 1985; 99 of these participated in the study interview (76.7% response). One
control group consisted of patients with other forms of cancer recruited through the same
study procedures and time period as the esophageal cancer cases. A population-based
control group (n = 533), frequency matched by age strata, was drawn using electoral lists
and random digit dialing (72% response). Face-to-face interviews were carried out with
82% of all cancer cases with telephone interview (10%) or mailed questionnaire (8%) for
the remaining cases. Twenty percent of all case interviews were provided by proxy
respondents.
The occupational assessment consisted of a detailed description of each job held during
the working lifetime, including the company, products, nature of work at site, job activities, and
any additional information that could furnish clues about exposure from the interviews. A team
of industrial hygienists and chemists blinded to subject's disease status translated jobs into
potential exposure to 294 substances with three dimensions (degree of confidence that exposure
occurred, frequency of exposure, and concentration of exposure). Each of these exposure
dimensions was categorized into none, any, or substantial exposure. Any exposure to TCE was
1% among cases (n = 1) and 1% for substantial TCE exposure (n = 1); "substantial" is defined as
>10 years of exposure for the period up to 5 years before diagnosis.
Logistic regression models adjusted for age, education, respondent status, birthplace,
cigarette smoking, beer consumption spirits consumption and beta-carotene intake (Parent et al.,
2000a) or Mantel-Haenszel x stratified on age, family income, cigarette smoking, coffee, and an
index for alcohol consumption (Siemiatycki, 1991). Parent et al. (2000a) presents observations
of analyses examining industries, occupation, and some chemical-specific exposures, including
solvents, but not TCE. Observations on TCE from Mantel-Haenszel analyses are found in the
original report of Siemiatycki (1991). Odds ratios for TCE exposure are presented in
Siemiatycki (1991) with 90% confidence intervals and 95% confidence intervals in Parent et a.
(2000b).
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of esophageal cancer. However, the
use of the general population (rather than a known cohort of exposed workers) reduced the
likelihood that subjects were exposed to TCE, resulting in relatively low statistical power for the
analysis. The job exposure matrix, applied to the job information, was very broad since it was
used to evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-261 DRAFT—DO NOT CITE OR QUOTE
-------
Parent M-E, Siemiatycki J, Fritschi L. (2000b). Workplace exposures and oesophageal cancer. Occup Environ Med
57:325-334.
SiemityckiJ. (1991). Risk Factors for Cancer in the Workplace. Boca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
129 esophageal cancer cases were identified among male Montreal residents between
1979 and 1985 of which 99 were interviewed.
740 eligible male controls identified from the same source population using random
digit dialing or electoral lists; 533 were interviewed. A second control series
consisted of all other cancer controls.
Participation rate: cases, 76.7%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 150 (Malignant neoplasm of esophagus).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 294 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were conducted
either at home or in the hospital; all population control interviews were conducted at
home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
99 cases (16.1% response), 533 population controls (12%).
Exposure prevalence: Any TCE exposure, 1% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), 1% cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, education, respondent status, birthplace, cigarette smoking, beer consumption
spirits consumption and beta-carotene intake (Parent et al., 2000b).
Age, family income, cigarette smoking, and index for alcohol consumption
(Siemiatycki, 1991).
-------
a,	s
c>	to
to	s
Co
<3	Co'
8.	Si
§	t
£.	>
s	?
^	>5
^	5
!>q	S.
<3	TO
S	^
^	^3
^	5
s.	^
^	o
• TO	Co
V:	TO
VO
to
o
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Parent et al., 2000b).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
B.3.2.5. Liver Cancer Case-Control Studies
B.3.2.5.1. Lee et al. (2003).
B.3.2.5.1.1. Author's abstract.
Aims: To investigate the association between cancer mortality risk and exposure
to chlorinated hydrocarbons in groundwater of a downstream community near a
contaminated site. Methods: Death certificates inclusive for the years 1966-97
were collected from two villages in the vicinity of an electronics factory operated
between 1970 and 1992. These two villages were classified into the downstream
(exposed) village and the upstream (unexposed) according to groundwater flow
direction. Exposure classification was validated by the contaminant levels in 49
residential wells measured with gas chromatography/mass spectrometry.
Mortality odds ratios (MORs) for cancer were calculated with cardiovascular-
cerebrovascular diseases as the reference diseases. Multiple logistic regressions
were performed to estimate the effects of exposure and period after adjustment for
age. Results: Increased MORs were observed among males for all cancer, and
liver cancer for the periods after 10 years of latency, namely, 1980-89, and 1990-
97. Adjusted MOR for male liver cancer was 2.57 (95% confidence interval 1.21
to 5.46) with a significant linear trend for the period effect. Conclusion: The
results suggest a link between exposure to chlorinated hydrocarbons and male
liver cancer risk. However, the conclusion is limited by lack of individual
information on groundwater exposure and potential confounding factors.
B.3.2.5.1.2. Study description and comment. Exposure potential to chlorinated
hydrocarbons was assigned in this community case-control study of liver cancer in males
>30 years of age using residency as coded on death certificates obtained from local
household registration offices. No information is available to assess the completeness of
death reporting to the local registration office. Of the 1,333 deaths between 1966 and 1997
in two villages surrounding a hazardous waste site, an electronics factory operating
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-266 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
between 1970 and 1992 in Taoyuan, Taiwan,3 266 cancer deaths were identified; 53 liver
cancer deaths, 39 stomach cancer deaths, 26 colorectal deaths, and 41 lung cancer deaths.
Controls were identified from 344 deaths due to cardiovascular and cerebrovascular
diseases, without arrhythmia; 286 were included in the statistical analysis. Residents from
a village north and northeast of the plant were considered exposed and residents living
south considered unexposed to chlorinated hydrocarbons. Statistical analyses are limited
to Mantel-Haenszel chi-square approaches stratified by sex and age and, for male cases and
controls, logistic regression with age as a covariate. Socioeconomic characteristics were
similar between residents of the two villages (Wang, 2004). The study does not include
control for potential confounding from hepatitis virus; high rates of hepatitis B and C are
endemic to Taiwan and northern Taiwan, the location of this study, has a high prevalence
of hepatitis C virus infection (Lee et al., 2003). Confounding would be introduced if the
prevalence of hepatitis C differed between the two villages.
Exposure assessment is quite limited and misclassification bias likely high using
residence address as recorded on the death certificate as a surrogate for consumption of
contaminated drinking water. The paper not only lacks information on intensity and duration of
hydrocarbon exposures to individual cases and controls, but no information is available on an
estimate of the amount of TCE ingested. Information on residence length, population mobility,
and chemical usage at the plant are lacking. Similarly, well water monitoring is sparse, based on
seven chlorinated hydrocarbons monitored over a 7 month period between 1999-2000 in
69 groundwater samples from 44 wells to the north and northeast, or downstream from the
factory, and in 5 groundwater samples from 2 wells to the south or upstream from the factory.
Monitoring from other time periods is lacking with no information available to judge if current
monitoring are representative of past concentrations. Median concentrations ((J,g/L or ppb) and
ranges ((J,g/L or ppb) for these seven chemicals are identified in the table below. Highest
concentration of contaminants was from wells closest to the factory boundary with
concentrations detected at or close to maximum contaminant levels in wells located 0.5 mile
(1,000 meters) away. A municipal system supplied water to upstream village residents (start date
no identified); however, wells served as source for water to of the north or downstream village
residents. The exposure assessment does not consider potential occupational exposure.
Chemical
Downstream
Upstream
Median
Range
Median
Range
Trichl oroethyl ene
28
N.D.-1,791
0.1
0.1-0.1
3 The factory's workers were subjects in the cohort studies of Chang et al. (2003, 2005) and Sung et al. (2007,
2008).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-267 DRAFT—DO NOT CITE OR QUOTE
-------
Perchl oroethyl ene
3
N.D.-5,228
0.05
N.D.-0.1
cis-1,2-dichl oroethyl ene
3
N.D.-1,376
N.D.
N.D.
1,1-dichloroethane
2
N.D.-228
0.05
N.D.-0.1
1,1 -dichl oroethyl ene
1
N.D.-1,240
N.D.
N.D.
Vinyl chloride
0.003
N.D.-72
N.D.
N.D.
1
2	N.D. = not detected
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-268 DRAFT—DO NOT CITE OR QUOTE
-------
Lee L J-H, Chung C-W, Ma Y-C, Wang G-S, Chen P-C, Hwang Y-H, Wang J-D. (2003). Increased mortality odds ratio of
male liver cancer in a community contaminated by chlorinated hydrocarbons in groundwater. Occup Environ Med
60:364-369.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Study hypothesis of investigating cancer mortality risk and exposure to chlorinated
hydrocarbons in groundwater.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Deaths in 1966-1997 identified from local housing registration offices among
residents in two villages were the source for case and control series. The two villages
were north (contaminated community) and south (unexposed) of an electronics
factory declared as a hazardous waste site. No information if all death among
residents were reported to registration office.
Cases: 53 liver cancer deaths in males and females, 51 included in statistical analysis
(96%); stomach cancer deaths (n = 39), colon and rectum deaths (n = 26), and lung
cancer deaths (n = 41). Paper does not present numbers of stomach, colo-rectal and
lung cancer deaths used in statistical analyses.
Controls: 344 cardiovascular-cerebrovascular CV-CB disease deaths, 286 CV-CB
deaths without arrhythmia included in statistical analysis (83%).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-9.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Exposure potential to chlorinated hydrocarbons in drinking water was inferred from
residence address on deaths certificate.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
NA, Record based information.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
NA
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Liver cancer case exposure prevalence [downstream village resident], 53% (n = 24
males, n = 4 females).
Control exposure prevalence [upstream village resident], 30% (n = 44 males, n = 41
females).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Sex and age (categorical). No control for potential confounding due to hepatitis virus
(for liver cancer) or smoking (for lung cancer analyses).
Statistical methods
Mantel-Haenszel Chi square.
Multiple logistic regressions (males deaths only).
Exposure-response analysis presented in
published paper
No, MORs presented by time period.
Documentation of results
Inadequate, the paper does not discuss mobility patterns of residents, percentage of
population who may have moved from area, pr completeness of death ascertainment
using certificates obtained from local housing registration offices.
-------
MOR = mortality odds ratio.
t-
to
s
a, s
c> to
to s
Co
<3 Co'
s.	a
£
O	>r
S	vgs
£. >
s ?
^ 5
Oq 5.
<3 TO
^ ^3
^3 5
S. ^
.TO Co
V; to
Co
0
s
1
a.
VO
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
B.3.2.6. Lymphoma Case-Control Studies
B.3 .2.6.1. (Gold et al., In Press), Purdue et al. (2011)
1.1.1.1.1.1.	(Gold et al., In Press) abstract.
Objectives Few studies have examined whether exposure to chlorinated solvents
is associated with multiple myeloma. We evaluated associations between multiple
myeloma and occupational exposure to six chlorinated solvents: 1,1,1-
trichloroethane, trichloroethylene (TCE), methylene chloride (DCM),
perchloroethylene, carbon tetrachloride and chloroform. Methods In-person
interviews obtained occupational histories and information on jobs with likely
solvent exposure. We assigned exposure metrics of probability, frequency,
intensity and confidence using job-exposure matrices modified by job-specific
questionnaire information. We used logistic regression to estimate ORs and 95%
CIs for associations between multiple myeloma and ever exposure to each, and
any, chlorinated solvent and analysed whether associations varied by duration and
cumulative exposure. We also considered all occupations that were given the
lowest confidence scores as unexposed and repeated all analyses. Results Risk of
multiple myeloma was elevated for subjects ever exposed to 1,1,1-trichloroethane
(OR (95% CI): 1.8 (1.1 to 2.9)). Ever exposure to TCE or DCM also entailed
elevated, but not statistically significant, risks of multiple myeloma; these became
statistically significant when occupations with low confidence scores were
considered unexposed (TCE: 1.7 (1.0 to 2.7); DCM: 2.0 (1.2 to 3.2)). Increasing
cumulative exposure to perchloroethylene was also associated with increasing
multiple myeloma risk. We observed non-significantly increased multiple
myeloma risks with exposure to chloroform; however, few subjects were exposed.
Conclusions Evidence from this relatively large case-control study suggests that
exposures to certain chlorinated solvents may be associated with increased
incidence of multiple myeloma; however, the study is limited by relatively low
participation (52%) among controls.
1.1.1.1.1.2.	Purdue et al (2011) abstract.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-273 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
BACKGROUND: Previous epidemiologic findings suggest an association
between exposure to trichloroethylene (TCE), a chlorinated solvent primarily used
for vapor degreasing of metal parts, and non-Hodgkin lymphoma (NHL).
OBJECTIVES: We investigated the association between occupational TCE
exposure and NHL within a population-based case-control study using detailed
exposure assessment methods. METHODS: Cases (n = 1,189; 76% participation
rate) and controls (n = 982; 52% participation rate) provided information on their
occupational histories and, for selected occupations, on possible workplace
exposure to TCE using job-specific interview modules. An industrial hygienist
assessed potential TCE exposure based on this information and a review of the
TCE industrial hygiene literature. We computed odds ratios (ORs) and 95%
confidence intervals (CIs) relating NHL and different metrics of estimated TCE
exposure, categorized using tertiles among exposed controls, with unexposed
subjects as the reference group. RESULTS: We observed associations with NHL
for the highest tertiles of estimated average weekly exposure (23 exposed cases;
OR = 2.5; 95% CI, 1.1-6.1) and cumulative exposure (24 exposed cases; OR =
2.3; 95% CI, 1.0-5.0) to TCE. Tests for trend with these metrics surpassed or
approached statistical significance (p-value for trend = 0.02 and 0.08,
respectively); however, we did not observe dose-response relationships across the
exposure levels. Overall, neither duration nor intensity of exposure was associated
with NHL, although we observed an association with the lowest tertile of
exposure duration (OR = 2.1; 95% CI, 1.0-4.7). CONCLUSIONS: Our findings
offer additional support for an association between high levels of exposure to
TCE and increased risk of NHL. However, we cannot rule out the possibility of
confounding from other chlorinated solvents used for vapor degreasing and note
that our exposure assessment methods have not been validated.
1.1.1.1.1.3. Gold et al. description and comment.
The population case-control study of multiple myeloma in men and women who were
residents of two SEER reporting sites, the Seattle-Puget Sound, WA region and the Detroit, MI
metropolitan area, evaluated occupational risk factors in relation to the risk of multiple myeloma
(MM). Detailed exposure information obtained from job-specific questionnaires allowed
evaluation of association between 1,1, 1-trichloroethane, trichloroethylene, dichloromethane,
perchloroethylene carbon tetrachloride, and chloroform. Histologically-confirmed incident cases
of MM (ICD-O-2/3, Codes 9731, 9732) in men and women without a previous diagnosis of MM,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-274 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
NHL or HIV, between 35 and 74 years of age, and diagnosed between 2000 and 2002 were
eligible as cases, with population controls having Seattle-Puget Sound, WA or Detriot, MI
metropolitan area addresses identified from random digit dialing if <65 years of age, or by
random selection from Medicare or Medicaid files for controls 65-74 years of age. Controls for
this study were the same as those participating in the population-based case-control study of
NHL carried out at the same time in these SEER areas, in addition, to two other SEER areas. A
greater proportion of controls than cases were from Seattle-Puget Sound area. Face-to-face
interviews were completed for 181 cases (71% participation rate) and 418 (52% participation
rate).
In-person interviews were conducted using a computer-assisted interview program with
modules focused specifically on solvent exposures for jobs held >2 years in 20 occupations.
Proxy interviews were not permitted but were allowed to aid in recalling occupational details.
All jobs were coded according to the Standard Occupational Classification system. For each of
the six solvents, exposure metrics of probability, frequency, intensity and confidence were
assigned by modifying JEMs based on the subjects' answers to the questionnaire's sections on
work history and job module. The JEMs were developed for each decade for specific industries,
occupational and tasks by an industrial hygienist after reviewing published paper and reports on
chlorinated solvents (e.g., 2007 for TCE). The assignment of exposure probability defined as the
theoretical percentage of workers reporting the same information that would have been likely to
have had exposure to the solvent is one strength of the study. For all jobs with probability scores
of at least 1 (>\% of subjects were likely to have had exposure), frequency and intensity scores
were also assigned, with values of 1, 2, 3, or 4 for each variable. Additionally, depending on the
information source for assigning the probability, frequency, and intensity score, whether from
literature review or self-reported, a confidence level was assigned on a scale of 1 to 4. Exposure
surrogates developed for each of the six solvents were ever exposed and cumulative exposure,
defined as the sum over all jobs of the product of intensity, exposure duration, and frequency. Of
the 180 cases, 66 (37%) were identified as having been ever exposed to TCE (confidence scores
of 1 or higher) with 24 of the TCE exposed cases (13% of all cases) assigned to the highest
cumulative exposure group. Moreover, roughly one-third of the TCE-exposed cases were
identified as having a low confidence level score (no information was available on probability,
frequency or intensity or contradictory information exists in the literature), suggesting a greater
potential for exposure misclassification bais in TCE assignment.
Assocation between MM and individual occupational solvents exposure was
assessed using unconditional logistic regression to estimate ORs and 95% confidence
intervals. Jobs with probability score of 2 or higher (10% or more subjects in that job
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-275 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
were likely to have had TCE exposure) were defined as ever exposed to TCE. A lag
period of 10 years, e. g., summing TCE exposures up to a period 10 years before disease
diagnosis, was also examined in analyses of cumulative exposure. All statistical models
included covarates for sex, age (3 categories), race (4 categories), education (3
categories) and SEER site. Each of the continuous exposure metrics was categorized into
four groups according to quartiles of the control exposure distribution. For TCE,
cumulative exposure scores were 2,218 ppm-year (median) [range, 1-50,000 ppm-year].
Test of trend were conducted using a linear term for the median duration and cumulative
scores among controls in each category. Gold et al. further reported findings from
sensitivity analyses considering all cases and controls with confidence scores of 1 as
unexposed to address potential misclassification bias resulting from the identification of
of unexposed individuals as exposed. In studies with low exposure prevalences like Gold
et al. (in press), this misclassification bias would diminish observed associations between
TCE and multiple myeloma (Stewart and Correa-Villaseor, 1991).
1.1.1.1.1.4. Purdue et al (2011) description and comment.
This population case-control study of non-Hodgkin's lymphoma in four SEER reporting
areas was designed to investigate the association between NHL and occupational factors and
focused on TCE exposures with a detailed exposure assessment method. Histologically-
confirmed incident cases of NHL n men and women between 20 and 74 years of age, diagnosed
between 1998 and 2000, and without know HIV infection were identified from four SEER
reporting areas - the State of Iowa, the Seattle, Was and Detroit, MI metropolitan areas, and Los
Angeles County, CA-with populations controls having addresses in the four SEER reporting
areas identified from random digit dialing for men and women <65 years of age, or by random
selection from Medicare files, for men and women 65 - 74 years of age. NHLs were classified
using according to the ICD-O-2 (converted to ICD-O-3, Codes 967-972): B-cell lymphomas,
including small B-cell lymphoma, large diffuse B-cell lymphoma, follicular, or precursor
lymphoblastic leukemia, and T-cell lymphoma, including analplastic T-cell, N/K, and
lymphblastic leukemia. Subjects with chronic lymphocytic leukemia were ineligible; however,
28 recruited cases of small lymphocytic lymphoma were later identified by pathology review to
be cases of chronic lymphocytic leukemia and were retained because the two diagnoses comprise
the same disease. Face-to-face interviews were completed for 1,321 NHL cases (76%
participation rate) and 1,057 controls (52% participation rate). Of these, 132 cases and 75
controls that were never employed or had unknown occupation were excluded, leaving 1,189
cases and 982 controls for the analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-276 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Subjects provided information on residential and occupation history from a mailed
calendar, with an in-person interview and home visit using a computer- assisted interview
program with modules on solvent exposure, added one-year after the study's start date. Of the
computer-assisted personal interviews, 682 cases and 640 controls included the solvent-focused
modules. The occupational history gathered information on each job held by the subject for
1 year or longer since the age of 16. For selected occupations, one of 32 job- or industry-specific
modules was administered based on information collected in the occupational histories. The
information collected in the modules included the average frequency of various solvent-related
tasks, the average length of time it took to perform given solvent-related tasks, sensory
descriptions, dermal exposure, work practices, engineering controls, and personal protective
equipment use. Information was also sought from subjects who reported jobs that could involve
degreasing on the usual number of hours per instance spent degreasing, the identity of the
chemical used for degreasing, the percentage of time each chemical was used, whether the
degreasing solvent was heated or at room temperature, and the manner in which parts were
cleaned.
The 23 exposure matrices developed by the industrial hygienist using information from
the literature reiview, including Bakke et al. (2007), the subject's occupational history, and the
information collected in the job modules, an expert industrial hygienist assessed levels of
probability, frequency, and intensity of TCE exposure for each job. The assignment of exposure
probability defined as the theoretical percentage of workers reporting the same information that
would have been likely to have had exposure to the solvent is one strength of this study. For all
jobs with probability scores of at least 1 (>1% of subjects were likely to have had exposure),
frequency and intensity scores were also assogined on a scale of 1 to 4 for frequency and 1-5 for
intensity. The intensity score also reflected dermal exposure. The jo-specifid estimates of
frequency and intensity for each subject were integrated to develop several metrics of TCE
exposure. A subject was identified as "unexposed" if all jobs had been assigned an exposure
probability of 0%, "possibly exposed" if one or more jobs had been assigned an exposure
probability of <50% (probability scores of 1, 2, or 3, and "probably exposed" if at least one job
had been assigned an exposure probability of >50% (probability scores of 4 or 5). For subjects
defined as probably exposed, the following additional exposure metrics were calculated:
exposure duration; cumulative exposure, defined as the sum, across all jobs with exposure
probability scores of 4 or 5, of the product of intensity midpoint, the frequency midpoint, and the
duration in weeks; average week exposure, defined as the cumulative exposure divided by
exposure duration; and average exposure intensity defined as the duration-weighted average
intensity level across all jobs with probability scores of 4 or 5. Of the 1,189 cases, 545 (46%)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-277 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
were assigned an exposure level of "possible" and 45 cases (4%) an exposure level of
"probable." Among subjects with probable confidence TCE exposure, the median cumulative
exposure score was 150 ppm-year [range, 1-.=>234,000 ppm-year],
Assocation between NHL and TCE exposure metrics was assessed using unconditional
logistic regression to estimate ORs and 95% confidence intervals. Other than the ever/never
analysis, all analyses include subjects with probable TCE exposure, those with probability scores
of 4 or 5. The observed exposure prevalence among subjects assigned possible exposure,
defined as holding a job with a confidence score of 1, 2, or 3, suggested poor specificity and was
inconsistent with the narrow set of occupational applications for TCE from the literature review.
The higher likelihood for possible exposure misclassification bias and the importance of high
specificity exposure assessment, further analysis of this measure was judged as unlikely to be
informative. All statistical analyses included covariates for age (3 categories), sex, race (4
categories), education (3 categories) and SEER area. The exposure metrics were categorized
using tertiles among probably exposed controls as cut-points. In addition, ORs and 95%
confidence intervals were reported for exposure defined as the difference between the second
and third tertiles among exposed controls. Test of trend were performed by modeling exposure
the exposure metrics as continuous variables. Last, the association between TCE exposure and
specific histolgoically-defined NHL subtypes (diffuse large B-cell, follicular lymphoma, and
small lymphocytic lymphoma/chronic lymphocytic leukemia, were reported using polytomous
regression to explore possible heterogeneity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-278 DRAFT—DO NOT CITE OR QUOTE
-------
1	Gold LS, Stewart PA, Milliken K, Purdue M, Severson R, Seixas N, Blair A, Hartge P, Davis S, Dr Roos AJ. . The
2	relationship between multiple myeloma and occupational exposure to six chlorinated solvents. Occup Environ Med [Online
3	September 10, 2010 doi:10.1136/oem.2009.054809].
4
5	Purdue MP, Bakke B, Stewart P, De Roos AJ, Schenk M, Lynch CF, Bernstein L, Morton LM, Cerhan JR, Severson RK,
6	Cozen W, Davis S, Rothman N, Martge P, Colt JS. (2011). A case-control study of occupational exposure to trichloroethylene
7	and non-Hodgkin lymphoma. Environ Health Perspect 119:232-238 doi:10.1289/ehp.l002106 [Online 2 November 2010]
8

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Study hypotheses of investigating association between TCE exposure and NHL using
detailed exposure assessment methods (Purdue et al., 2011) and evaluating
associations between multiple myeloma (Gold et al., In Press) and occupational
exposure to six chlorinated solvents: 1,1, 1-trichloroethane, methylene chloride,
perchloroethylene, carbon tetrachloride, and chloroform.
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-279 DRAFT—DO NOT CITE OR QUOTE
-------
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cases: 1,321 (2,248 eligible) histologically-confirmed NHL cases in males and
females, 20-74 years of age, 1998-2000, and residents of four SEER reporting areas -
Iowa, Los Angeles County, CA, Seattle, WA metropolitan area and Detroit, MI
metropolitan area (Purdue et al.,Siemiatycki, 1991 2011); 181 (255 eligible)
histologically-confirmed multiple myeloma cases in males and females, 35-74 years
of age, 2000-2002, and residents of two SEER reporting areas - Seattle-Puget Sound,
WA area and Detroit, MI metropolitan area (Gold et al., In Press).
Controls: 1,057 (2,409 eligible) controls identified from RDD (<65 years old) or
Medicare file (65-75 years old) who were residents in the 4 SEER areas (Purdue et
al., 2011); 481 (1,133 eligible) controls identified from Purdue et al. (2011) who
were 35-74 years of age, no previous diagnosis of HIV, MM, plasmacytoma, or
NHL, spoke English, and residents of Seattle-Puget Sound, WA area and Detroit, MI
metropolitan area (Gold et al., In Press).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
NHL and multiple myeloma incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-0-2 [Codes 967-972, NHL; 9731-9732, MM],
CATEGORY C: TCE-EXPOSURE CRITERIA
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-280 DRAFT—DO NOT CITE OR QUOTE
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Literautre review, exposure matrices occupational histories and information collected
in the job module supported assignment by expert industrial hygienist of probability,
frequency, and intensity of TCE for each job held >12 months (Purdue et al., 2011)
or >2 years (Gold et al., In Press).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
In-person interview using questionnaire or computer-assisted personal interview (682
of 1,321 cases and 640 of 1,057 controls in Purdue et al. (2011) with modules for
jobs of interest.
Blinded interviewers
Interviewer not blinded. Exposure assessment assigned blinded.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No proxy interviews.
CATEGORY G: SAMPLE SIZE
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-281 DRAFT—DO NOT CITE OR QUOTE
-------
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,321 cases (76% participation rate); 1,051 controls (52% participation rate) (Purdue
et al., 2011). Of these, 132 cases and 75 controls that were never employed or had
unknown occupation were excluded, leaving 1,189 cases and 982 controls for the
analysis.
181 cases (71% participation rate); 1,113 controls (52% participation rate) (Gold et
al., in press).
Exposure prevalence, ever exposed to TCE (>50% of subiects in iob probablv
exposed), 27 (2.8%) NHL cases; 0.7% of cases in highest cumulative exposure
category and 2.3% in highest average exposure intensity category (Purdue et al.,
2011); ever exposed to TCE (>10% of subiects in iob with probable exposure) (Gold
et al., In PressX
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, SEER center, race and education (Gold et al., In Press; Purdue et al., 2011).
Statistical methods
Unconditional logistic regression.
Exposure-response analysis presented in
published paper
Test for trend performed by modeling the exposure metrics as continuous variable
(Purdue et al., 2011) or using median duration and cumulative scores among controls
for each exposure category.
Documentation of results
Yes, study was well documented with supplemental material on publisher's webpage
(Purdue et al., 2011).
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-282 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
B.3.2.6.2. Cocco et al. (2010).
B.3.2.6.2.1. Author's abstract.
BACKGROUND: Several studies have suggested an association between
occupational exposure to solvents and lymphoma risk. However, findings are
inconsistent and the role of specific chemicals is not known. Objective To
investigate the role of occupational exposure to organic solvents in the aetiology
of B-cell non-Hodgkin's lymphoma (B-NHL) and its major subtypes, as well as
Hodgkin's lymphoma and T-cell lymphoma. METHODS: 2348 lymphoma cases
and 2462 controls participated in a case-control study in six European countries.
A subset of cases were reviewed by a panel of pathologists to ensure diagnostic
consistency. Exposure to solvents was assessed by industrial hygienists and
occupational experts based on a detailed occupational questionnaire. RESULTS:
Risk of follicular lymphoma significantly increased with three independent
metrics of exposure to benzene, toluene and xylene (BTX) (combined p=4 x 10(-
7)) and to styrene (p=l x 10(-5)), and chronic lymphocytic leukaemia (CLL) risk
increased with exposure to solvents overall (p=4 x 10(-6)), BTX (p=5 x 10(-5)),
gasoline (p=8 x 10(-5)) and other solvents (p=2 x 10(-6)). Risk of B-NHL for ever
exposure to solvents was not elevated (OR=l.l, 95% CI 1.0 to 1.3), and that for
CLL and follicular lymphoma was 1.3 (95% CI 1.1 to 1.6) and 1.3 (95% CI 1.0 to
1.7), respectively. Exposure to benzene accounted, at least partially, for the
association observed with CLL risk. Hodgkin's lymphoma and T-cell lymphoma
did not show an association with solvent exposure. CONCLUSION: This analysis
of a large European dataset confirms a role of occupational exposure to solvents
in the aetiology of B-NHL, and particularly, CLL. It is suggested that benzene is
most likely to be implicated, but we cannot exclude the possibility of a role for
other solvents in relation to other lymphoma subtypes, such as follicular
lymphoma. No association with risk of T-cell lymphoma and Hodgkin's
lymphoma was shown.
B.3 .2.6.3. Study description and comment. This population case control study of non-
Hodgkin's lymphoma in the Czech Republic, France, Germany, Italy, Ireland, and Spain
was designed to examine possible personal and occupational risk factors for lymphoma
subtypes as defined using the World Health Organization's classification (the Epilymph
study). Observations in German subjects are reported separately in Seidler et al. (2007)
(see B.3.2.6.6.). The publication of Cocco et al. (2010) examined solvents and adopted
expert assessment to assign exposure potential to organic solvents, specifically, chlorinated
aliphatic hydrocarbons, benzene, toluene, xylene, gasoline, mineral spirts, styrene, and
trichloroethylene. Cases of lymphoma in adults, >17 years of age, and diagnosed in 22
centers in 1998 and 2004 with population controls selected by sampling from the general
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09
B-283
DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
population, and matched to cases on sex, age, and residence area, in Germany and Italy, or
matched hospital controls limited to diagnoses other than cancer, infectiou diseases, and
immunodeficient diseases in the Czech Republic, France, Ireland, and Spain. The
lymphoma diagnosis was classified according to the 2001 World Health Organization's
classification of lymphoma, and slides of about 20% of ceases from each center were
reviewed centrally by a panel of pathologists and reclassified when necessary. Lymphoma
cases included in this study were B-cell lymphomas, including B-cell subtypes, T-cell
lymphomas, and Hodgkin's lymphoma. Informed consent was obtained for 2,348
lymphoma cases (88%) and 2,462 controls (81% hospital controls, 52% population
controls) who participated in the study. Most cases were B-cell lymphomas (n=l,869) with
fewer T-cell (n=133) and Hodgkin's (n=339) lymphoma.
Trained interviewers administered a structured questionnaire through in-person
interviews with cases and controls to collect information on sociodemographic factors, lifestyle,
health history, and complete work history for all ful-time jobs held for 1 year or longer. Special
questionnaire modules for specific ooccupations gathered additional details on jobs and exposure
of a priori interest. Industrial hygienists in each center reviewed the general and specific
questionnaires and assessed exposure to 43 agents, including organic solvents according to
confidence, intensity, and frequency of exposure. The paper does not report if proxy or next-of-
kin provided information if the case or control was deceased. Confidence represented the degree
of certainty that the worker had been exposed to the agent and was based both on probability of
exposure and on the proportion of workers exposed in a give job, <40% (possible exposure), 40-
90%, (probable exposure), and >90% (certain/definite exposure). Intensity of exposure was
defined as a rank-ordered variable, unexposed (0), low (1), medium (2), high (3), with agent-
specific cut-off points defined based on current threshold limit values, likely half the TLV (low),
51%) - 150%) (medium), and >150%> (high) (Kiran et al., 2010). Exposure frequency expressed
the proportion of work time involving contact with the agent: unexposed (coded as 0), 1-5% of
the work time (coded as 1), >5-30%> of the work time (coded as 2), and >30%> of the work time
(coded as 3). Exposure potential to trichloroethylene for cases and controls was based surrogates
for overall exposure and cumulative exposure score. The cumulative exposure score was the
sum over a subjects work history of the product of duration and frequency/3 to the power of
intensity and results in a log distribution of exposure scores. Exposure prevalence to TCE is low
in this study; Cocco et al. (2010) identifies 71 cases of B-cell lymphoma (4% exposure
prevalence) and 117 controls (5% exposure prevalence) with high confidence overall TCE
exposure and of these exposed subjects, 29 cases (2%) and 37 (2%) with a high confidence high
cumulative exposure score.
Association between B-cell lymphoma and B-cell lymphoma subtypes and individual
occupational solvent exposures was assessed using unconditional logistic regression which
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-284 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
adjusted for age, sex, education and center. Alcohol and smoking were not included as a
potential confounder as previous analysis of the Epilymph data showed no association (Besson et
al., 2006). Statistical analyses are limited to subjects whose jobs TCE exposure was assessed
with high degree of confidence, defined as >90%of worker exposed in a given job. Lymphoma
subtypes examined included diffuse large B-cell lymphoma, follicular lymphoma, chronic
lymphocytic leukemia, and multiple myeloma. There were few cases of T-cell lymphomas with
high confidence TCE exposure; 6 cases with overall exposure, 2 of which with high confidence
high cumulative score. Two-tailed 95% confidence intervals of the odds ratio were calculated
with the Wald statistics and trend test defining cumulative exposure score as a continuous
variable using Wald's test for trend. As common to epidemiological studies, the many statistical
analyses and comparisons in Cocco et al. (2010) increases the potential for false positive errors
and Cocco et al. (2010) used Bonferroni correction of individual confidence intervals and trend
tests as an attempt to reduce this type of bias.
This study adopted a detailed exposure assessment, current classification system for
lymphomas, and was of a large number of cases and controls, although exposure prevalence ot
TCE was less than 5%, typical of population case-control studies. This study defines the
cumulative exposure score using a log scale, in addition, to using a rank-order value for intensity
instead of a midpoint of an range of exposure concentrations. Other cohort and case-control
studies of TCE and NHL, e. g., Purdue et al. (2011), define their cumulative exposure score as a
product of intensity, frequency, and duration. Each approach will produce a slightly different
rank ordering (personal communication). In the cumulative exposure formula of Cocco et al.
(2010), exposure duration contributes the greatest weigh in light of the formula's treatment of
1/3 the value of frequency (Cocco et al., 2010). The direction of bias in estimated trends of
disease risk by cumulative exposure depends on the variation of duration, with large variation in
durations between exposure exposures leading to downward bias (Steenland et al., 2001). Cocco
et al. (2010), also, reported odds ratios and confidence intervals for high confidence TCE
exposure, assigned to a job title when over 90% of workers were exposed. In comparison, both
Purdue et al. (2011) or Gold et al. defined probable exposure if at least one job has been
assigned an exposure probability of >50%. Any differences in reported findings between Cocco
et al. (2010) and the other NHL studies of Miligi et al. (2006), Wang et al. (2009) and Purdue et
al. (2011) may be due to these differences.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-285 DRAFT—DO NOT CITE OR QUOTE
-------
1	Cocco P, Mannetje A, Fadda D, Melis M, Becker N, Sanjose S, Foretova L, Marekova J, Staines A, Kleefeld S, Maynadie M,
2	Nieters A, Brennan P, Boffetta P. (2010). Occupational exposure to solvents and risk of lymphoma subtypes: results from the
3	Epilymph case-control study. Occup Environ Med 67:341-347.
4

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study evaluated occupational exposure to organic solvents as risk factors of
NHL in a population-based case-control study of men and women in.6 European
countries.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
2,348 hospital cases of NHL diagnosed between 1998 and 2004 among men and
women, >17 years of age, and residents of Czech Republic, France, Germany,
Ireland, Italy, and Spain; 2,462 population and hospital controls, identified from
census lists in Germany and Italy or smale hospitals as the cases, in all other
countries, and matched to cases on age, sex, and study center.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Lymphoma incidence - B-cell lymphoma (CLL, follicular, and diffuse large B-cell),
T-cell lymphoma, Hodgkin's lymphoma, and multiple myeloma. Postransplant
lymphoproliferative disorder or monoclonal gammopathies of undetermined
significance were excluded as cases.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
WHO classification system
CATEGORY C: TCE-EXPOSURE CRITERIA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-286 DRAFT—DO NOT CITE OR QUOTE
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
All jobs held for >1 yr assigned to standardized occupation (5-digit code). Industrial
hygienists at each center assigned exposure to 43 agents, including TCE and other
solvents (benzene, toluene, xylene, chlorinated aliphatic hydrocarbons, and gasoline)
to subjects according to confidence (possible, probable, certain), intensity
(unexposed, low, medium, high), and frequency. Exposure surrogates for overall
exposure and cumulative exposure (low, medium, high).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Face-to-face interview with questionnaire for information about medical history,
lifestyle factors, lifetime occupational history (all jobs held >1 yr) and supplemental
modules for specific occupations to gather additional details on jobs and exposures of
a priori interest.
Blinded interviewers
Unblinded interviews. Blinded exposure assessment.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Not reported in published paper.
CATEGORY G: SAMPLE SIZE
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-287 DRAFT—DO NOT CITE OR QUOTE
-------
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
2,348 cases (88% participation rate) and 2,462 controls (81% participation rate,
hospital controls, 52% participation rate, population controls).
Exposure prevalence, subjects with high confidence overall TCE exposure, 71 (4%)
all B-cell lymphoma, 6 (7%) T-cell lymphoma, and 48 (6%) NHL (B-cell diffuse and
follicular subtypes and T-cell); subjects with high confidence high cumulative TCE
exposure, 29 (2%) all B-cell lymphomas, 2 (2%) T-cell lymphoma, 14 (2%) NHL (B-
cell diffuse and follicular subtypes and T-cell).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, education and center.
Statistical methods
Unconditional logistic regression.
Exposure-response analysis presented in
published paper
Yes, using cumulative exposure defined as low, medium, high.
Documentation of results
Yes.
1
2
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-288 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
B.3.2.6.4. Wang et al. (2009).
B.3.2.6.4.1. Author's abstract.
A population-based case-control study involving 601 incident cases of non-
Hodgkin lymphoma (NHL) and 717 controls was conducted in 1996-2000 among
Connecticut women to examine associations with exposure to organic solvents. A
job-exposure matrix was used to assess occupational exposures. Increased risk of
NHL was associated with occupational exposure to chlorinated solvents (odds
ratio (OR) = 1.4, 95% confidence interval (CI): 1.1, 1.8) and carbon tetrachloride
(OR = 2.3, 95% CI: 1.3, 4.0). Those ever exposed to any organic solvent in work
settings had a borderline increased risk of NHL (OR = 1.3, 95% CI: 1.0, 1.6);
moreover, a significantly increased risk was observed for those with average
probability of exposure to any organic solvent at medium-high level (OR = 1.5,
95% CI: 1.1, 1.9). A borderline increased risk was also found for ever exposure to
formaldehyde (OR = 1.3, 95% CI: 1.0, 1.7) in work settings. Risk of NHL
increased with increasing average intensity (P = 0.01), average probability (p<
0.01), cumulative intensity (P = 0.01), and cumulative probability (p < 0.01) level
of organic solvent and with average probability level (P = 0.02) and cumulative
intensity level of chlorinated solvent (P = 0.02). Analyses by NHL subtype
showed a risk pattern for diffuse large B-cell lymphoma similar to that for overall
NHL, with stronger evidence of an association with benzene exposure. Results
suggest an increased risk of NHL associated with occupational exposure to
organic solvents for women.
B.3.2.6.4.2. Study description and comment. This population case-control study of
non-Hodgkin's lymphoma in Connecticut women was designed to examine possible
personal and occupational risk factors for NHL. The publication of Wang et al. (2009)
examined solvent exposure and adopted a job-exposure matrix to assign exposure potential
to nine chemicals—benzene, formaldehyde, chlorinated solvents, chloroform, carbon
tetrachloride, dichloromethane, methyl chloride and trichloroethylene. Histologically-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-289 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
confirmed incident cases of NHL in women aged between 21 and 84 years of age and
diagnosed in Connecticut between 1996 and 2000 were identified from the Connecticut
Cancer Registry, a SEER reporting site, with population controls having Connecticut
address identified from random digit dialing for women <65 years of age, or by random
selection from Centers for Medicare and Medicaid Service files for women aged 65 year or
older. Controls were frequency matched to cases within 5-year age groups. Face-to-face
interviews were completed for 601 (72%) cases and 717 controls (69% of those identified
from random digit dialing and 47% identified using Health Care Financing Administration
files).
Trained interviewers administered a structured questionnaire through in-person
interviews with cases and controls to collect information on diet, nutrition, and alcohol intake;
reproductive factors; hair dye use; and lifetime occupational history of all jobs held >1 year.
Jobs were coded to standardized occupational classification and standardized industry
classification titles and assigned probability and intensity of exposure to formaldehyde and nine
other solvents (benzene, any chlorinated solvents, dichloroethylene, chloroform, methylene
chloride, dichloroethane, methyl chloride, TCE and carbon tetrachloride) using a job-exposure
matrix developed by the National Cancer Institutes (Dosemeci et al., 1994; Gomez et al., 1994).
All jobs held up to a year before cancer diagnosis were assigned blinded as to disease status
potential exposure to each exposure of interest. Lifetime exposure potential for cases and
controls was based on exposure duration and a weighted score for exposure intensity and
probability of each occupational and industry and defined as a cumulative exposure metric,
average metric, or ever/never metric. Of the 601 cases, 77 (13%) were assigned with potential
TCE exposure over their lifetime; eight cases were assigned potential for high intensity exposure,
but with low probability and the 31 cases identified with medium and high probability of
exposure were considered as having low intensity exposure potential. The low exposure
prevalence to TCE, overall, and few subjects identified with confidence with high TCE exposure
intensity or probability implies exposure misclassification bias is likely, and likely
nondifferential, notably for high exposure categories (Dosemeci et al., 1990).
Association between NHL and individual occupational solvent exposure was assessed
using unconditional logistic regression model which adjusted for age, family history of
hematopoietic cancer, alcohol consumption and race. Statistical analyses treated exposure
defined as a categorical variable, divided into tertiles based on the distribution of controls, in
logistic regression analyses and as a continuous variable, whenever possible, to test for linear
trend. Polytomous logistic regression was used to evaluate the association between histologic
subtypes of NHL (DLBCL, follicular lymphoma, or chronic lymphocytic leukemia/small
lymphocytic lymphoma) and exposure. The largest number of cases was of the cell type
DLBCL.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-290 DRAFT—DO NOT CITE OR QUOTE
-------
1	Strength of this study is assignment of TCE exposure potential to individual subjects
2	using a validated job-exposure matrix, although uncertainty accompanied exposure assignment
3	and TCE exposure was largely of low intensity/low probability, and no cases with medium to
4	high intensity/probability. Resultant misclassification bias would dampen observed associations
5	for high exposure potential categories. Low prevalence of high intensity TCE exposure would
6	reduce the study's statistical power.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-291 DRAFT—DO NOT CITE OR QUOTE
-------
Wang R, Zhang Y, Lan Q, Holford TR, Leaderer B, Zahm SH, Boyle P, Dosemeci M, Rothman N, Zhu Y, Qin Q, Zheng T.
(2009). Occupational exposure to solvents and risk of non-Hodgkin lymphoma in Connecticut women. Am J Epidmiol
189:176-185.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study evaluated multiple potential risk factors of NHL in a population-based
case-control study of Connecticut women. Occupational exposure to TCE was not an
a priori hypothesis.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
601 (832 eligible) cases of NHL, diagnosed between 1996 and 2000 among women,
age 20 to 84 yrs and residents of Connecticut and histologically-confirmed, were
identified from the Yale Comprehensive Cancer Center's Rapid Case Ascertainment
Shared Resource, a component of the Connecticut Tumor Registry; 717 (number of
eligible controls not identified) population controls were randomly identified using
random digit dialing, if age <65 yrs, or from Medicare and Medicaid Service files,
for women aged 65 yrs or older and stratified by sex and 5-yr age groups.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
NHL and chronic lymphatic leukemia incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O-2 [Codes, M-9590-9642, 9690-9701, 9740-9750],
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
All jobs held for >1 yr were assigned to standardized occupation and industry
classifications. Using job exposure matrix of NCI (Dosemeci et al., 1994; Gomez et
al., 1994), probability of exposure level (low, medium and high) and intensity (very
low, low, medium and high) to TCE and other solvents (benzene, any chlorinated
solvents, dichloroethylene, chloroform, methylene chloride, dichloroethane, methyl
chloride, carbon tetrachloride, and formaldehyde) was assigned blinded as to case or
control status.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Face-to-face interview with questionnaire for detailed information about medical
history, lifestyle factors, education, lifetime occupational history (all jobs held >1 yr).
Blinded interviewers
Unblinded interviews.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
None.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
601 cases (72% participation) and 717 controls (69% participation for random digit
dialing controls and 47% participation for HCFA controls).
Exposure prevalence, ever exposed to TCE, 77 (13%) NHL cases; medium to high
TCE intensity, 13 NHL cases (2%); medium to high TCE probability, 34 cases (6%).
All 34 cases with medium to high TCE probability assigned low intensity exposure.
-------
>3
o
o
s
Co
I?
S"4
>}'
§•
TO
s
a
to
s
S"
a
a,
>r
<§>
>s
to
*
^s
«
s
TO
Co
0
s
1
a.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, family history of hematopoietic cancer, alcohol consumption and race.
Statistical methods
Unconditional logistic regression.
Exposure-response analysis presented in
published paper
Yes, by exposure intensity and by exposure probability.
Documentation of results
Yes.
VO
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
B.3.2.6.5. Costantini et al.(2008), Miligi et al. (2006).
B.3 .2.6.5.1. Costantini et al.(2008) abstract.
Background While there is a general consensus about the ability of benzene to
induce acute myeloid leukemia (AML), its effects on chronic lymphoid leukemia
and multiple myeloma (MM) are still under debate. We conducted a population-
based case-control study to evaluate the association between exposure to organic
solvents and risk of myeloid and lymphoid leukemia and MM.
Methods Five hundred eighty-six cases of leukemia (and 1,278 population
controls), 263 cases of MM (and 1,100 population controls) were collected.
Experts assessed exposure at individual level to a range of chemicals.
Results We found no association between exposure to any solvent and AML.
There were elevated point estimates for the associations between medium/high
benzene exposure and chronic lymphatic leukemia (OR: 1.8, 95% CF/40.9-3.9)
and MM (OR: 1.9, 95% CI: 0.9-3.9). Risks of chronic lymphatic leukemia were
somewhat elevated, albeit with wide confidence intervals, from medium/high
exposure to xylene and toluene as well.
Conclusions We did not confirm the known association between benzene and
AML, though this is likely explained by the strict regulation of benzene in Italy
nearly three decades prior to study initiation. Our results support the association
between benzene, xylene, and toluene and chronic lymphatic leukemia and
between benzene and MM with longer latencies than have been observed for
AML in other studies.
B.3.2.6.5.2. Miligi et al. (2006) abstract.
BACKGROUND: A number of studies have shown possible associations between
occupational exposures, particularly solvents, and lymphomas. The present
investigation aimed to evaluate the association between exposure to solvents and
lymphomas (Hodgkin and non-Hodgkin) in a large population-based, multicenter,
case-control study in Italy. METHODS: All newly diagnosed cases of malignant
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-295 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
lymphoma in men and women age 20 to 74 years in 1991-1993 were identified in
8 areas in Italy. The control group was formed by a random sample of the general
population in the areas under study stratified by sex and 5-year age groups. We
interviewed 1428 non-Hodgkin lymphoma cases, 304 Hodgkin disease cases, and
1530 controls. Experts examined the questionnaire data and assessed a level of
probability and intensity of exposure to a range of chemicals. RESULTS: Those
in the medium/high level of exposure had an increased risk of non-Hodgkin
lymphoma with exposure to toluene (odds ratio = 1.8; 95% confidence interval =
1.1-2.8), xylene 1.7 (1.0-2.6), and benzene 1.6 (1.0-2.4). Subjects exposed to all 3
aromatic hydrocarbons (benzene, toluene, and xylene; medium/high intensity
compared with none) had an odds ratio of 2.1 (1.1-4.3). We observed an increased
risk for Hodgkin disease for those exposed to technical solvents (2.7; 1.2-6.5) and
aliphatic solvents (2.7; 1.2-5.7). CONCLUSION: This study suggests that
aromatic and chlorinated hydrocarbons are a risk factor for non-Hodgkin
lymphomas, and provides preliminary evidence for an association between
solvents and Hodgkin disease.
B.3.2.6.5.3. Study description and comment. This series of papers of a population
case-control study of lymphomas in 11 areas in Italy (Costantini et al., 2008) and
occupation examines author's assigned exposure to TCE and other solvents using job-
specific or industry-specific questionnaires and expert rating to cases and controls. Miligi
et al. (2006) reported findings for non-Hodgkin lymphoma, a category which included
chronic lymphocytic leukemia, NHL subtypes, and Hodgkin lymphoma in 8 regions and
Constantini et al. (2008) presented observations for specific leukemia subtypes and multiple
myeloma in 7 regions (8 regions for chronic lymphocytic leukemia). Exclusion of the
regions in the original study does not appear to greatly reduce study power or to introduce
a selection bias. For example, Miligi et al. (2006) included 1,428 of the 1,450 total NHL
cases, the largest percentage of all lymphoma subtypes. The number of other lymphoma
subtypes was much smaller compared to NHL; 304 cases of Hodgkin disease, 586 cases of
leukemia, and 263 cases of multiple myeloma. All cases were identified from participating
study centers and controls were randomly selected from the each area's population using
stratified sampling for sex and age.
A face-to-face unblinded interview was conducted primarily at the interviewee's home
with a high proportion of proxy responses among cases (19%) but not controls (5%). Bias is
likely introduced by the lack of blinding of interviewers and from the high proportion of proxy
interviews. A questionnaire was used to obtain information on medical history, lifestyle factors,
occupational exposure and nonoccupational solvent exposures. Industrial hygiene professionals
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-296 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
assessed the probability and intensity of exposure to individual and classes of solvents using
information provided by questionnaire. Probability was classified into 3 levels (low, medium,
and high) with a 4-category scale for intensity (very low, low, medium, and high). These
qualitative scales lacked information on exposure concentrations and likely introduces
misclassification bias that can either dampen or inflate observed risks given the study's use of
multiple exposure groupings. "Very low level" was used for subjects with occupational
exposure intensities judged to be comparable to the upper end of the normal range for the general
population; "low-level intensity" when workplace exposure was judged to be low because of
control measures but higher than background; "medium exposure" for occupational
environments with moderate or poor control measures; and "high exposure" for workplaces
lacking any control measures. Groupings of "very low/low" and "medium/high" exposure was
used to examine association with NHL. Prevalence of medium to high TCE exposure among
NHL cases was low, 3% for NHL cases and 2% for all leukemia subtypes. Whether temporal
changes in TCE exposure concentrations were considered in assigning level and intensity is not
known. Overall, this study has low sensitivity for examining TCE and lymphoma given the low
prevalence of exposure, particularly to medium to high TCE intensity, the high proportion of
proxy interviews among cases, particularly NHL cases (15%), and qualitative exposure
assessment approach.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-297 DRAFT—DO NOT CITE OR QUOTE
-------
Costantini AS, Benvenuti A, Vineis P, Kriebel D, Tumino R, Ramazzotti V, Rodella S, et al. (2008). Risk of leukemia and
multiple myeloma associated with exposure to benzene and other organic solvents: evidence from the Italian multicenter case-
control study. Am J Ind Med 51:803-811.
Miligi L, Costantini AS, Benvenuti A, Kreibel D, Bolejack V, et al. (2006). Occupational exposure to solvents and the risk of
lymphomas. Epidemiol 17:552-561.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study evaluated TCE and other solvent exposures and lymphoma in a large
population-based, multicenter, case-control study.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
1,732 (2,066 eligible) cases of NHL, chronic lymphatic leukemia, and Hodgkin
lymphoma, diagnosed between 1991 and 1993 among men and women, age 20 to
74 yrs and residents of 8 regions in Italy, were identified from; 1,530 (2,086 eligible)
population controls were randomly selected from demographic files or from
sampling of National Health Service files and stratified by sex and 5-yr age groups.
586 leukemia and 263 multiple myeloma among men and women, age 20 to 74 in the
period 1991-1993, from 7 regions (8 regions for chronic lymphocytic leukemia) in
Italy, were identified from hospital or pathology department records or a regional
cancer registry; and 1,100 population controls selected from demographic files or
from sampling of National Health Service files and stratified by sex and 5-yr age
groups.
CATEGORY B: ENDPOINT MEASURED
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Levels of health outcome assessed
NHL and Hodgkin's lymphoma incidence (Miligi et al., 2006).
Leukemia and multiple myeloma (Costantini et al., 2008).
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
All NHL cases were defined following NCI Working Formulation Workgroup
classification and Hodgkin lymphomas defined following the Rye classification.
NHL diagnosis confirmed for 334 of 1,428 cases (23%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
IH experts from each region using information collected on questionnaires assigned
the probability of exposure level (low, medium and high) and intensity (very low,
low, medium and high) to TCE and other solvents. Exposure was assigned blinded
as to case or control status.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Face-to-face interview with questionnaire for detailed information about medical
history, lifestyle factors, education, occupational history (period is not identified in
published paper), and nonoccupational exposures including solvent exposure.
Blinded interviewers
Unblinded interviews.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
19%o of all lymphoma cases and 5% of controls were with proxy respondents
(Costantini et al., 2008).
CATEGORY G: SAMPLE SIZE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
0
1
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,732 cases (83% participation) and 1,530 controls (73% participation) (Miligi et al.,
2006); no information on participation rate for leukemia or multiple myeloma cases
or their controls in Costantini et al. (2008).
Exposure prevalence, medium to high TCE intensity, 35 NHL cases (3%) (Miligi et
al., 2006); 11 leukemia cases (2%) and 5 multiple myeloma cases (2%) (Costantini et
al., 2008).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, region, education, and region.
Statistical methods
Multiple logistic regressions.
Exposure-response analysis presented in
published paper
Yes, by exposure intensity and by duration (years) of exposure.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
B.3.2.6.6. Seidler et al. (2007).
B.3.2.6.6.1. Author's abstract.
AIMS: To analyze the relationship between exposure to chlorinated and aromatic
organic solvents and malignant lymphoma in a multi-centre, population-based
case-control study. METHODS: Male and female patients with malignant
lymphoma (n = 710) between 18 and 80 years of age were prospectively recruited
in six study regions in Germany (Ludwigshafen/Upper Palatinate,
Heidelberg/Rhine-Neckar-County, Wiirzburg/Lower Frankonia, Hamburg,
Bielefeld/Giitersloh, and Munich). For each newly recruited lymphoma case, a
gender, region and age-matched (+/-1 year of birth) population control was drawn
from the population registers. In a structured personal interview, we elicited a
complete occupational history, including every occupational period that lasted at
least one year. On the basis of job task-specific supplementary questionnaires, a
trained occupational physician assessed the exposure to chlorinated hydrocarbons
(trichloroethylene, tetrachloroethylene, dichloromethane, carbon tetrachloride)
and aromatic hydrocarbons (benzene, toluene, xylene, styrene). Odds ratios (OR)
and 95% confidence intervals (CI) were calculated using conditional logistic
regression analysis, adjusted for smoking (in pack years) and alcohol
consumption. To increase the statistical power, patients with specific lymphoma
subentities were additionally compared with the entire control group using
unconditional logistic regression analysis. RESULTS: We observed a statistically
significant association between high exposure to chlorinated hydrocarbons and
malignant lymphoma (Odds ratio = 2.1; 95% confidence interval 1.1-4.3). In the
analysis of lymphoma subentities, a pronounced risk elevation was found for
follicular lymphoma and marginal zone lymphoma. When specific substances
were considered, the association between trichloroethylene and malignant
lymphoma was of borderline statistical significance. Aromatic hydrocarbons were
not significantly associated with the lymphoma diagnosis. CONCLUSION: In
accordance with the literature, this data point to a potential etiologic role of
chlorinated hydrocarbons (particularly trichloroethylene) and malignant
lymphoma. Chlorinated hydrocarbons might affect specific lymphoma subentities
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-302 DRAFT—DO NOT CITE OR QUOTE
-------
2
3
1
differentially. Our study does not support a strong association between aromatic
hydrocarbons (benzene, toluene, xylene, or styrene) and the diagnosis of a
malignant lymphoma.
4
B.3 .2.6.6.2. Study description and comment. This population case-control study of NHL
and Hodgkin's lymphoma patients in six Germany regions is part of a larger multiple-
center and -country case-control study of lymphoma and environmental exposures, the
EPILYMPH study [see Cocco et al. (2010) in B.3.2.6.3.). A total of 710 cases and 710
controls that were matched to cases on age, sex, and region, participated in this study.
Participation rates were 88% for cases and 44% for controls. Potential for selection bias
may exist given the low control response rate. Strength of this study is the use of WHO
classification scheme for classifying lymphomas and the high percentage of cases with
histologically-confirmed diagnoses. An industrial physician blinded to case and control
status assigned exposure to specific solvents (i.e., TCE, perchloroethylene, carbon
tetrachloride, etc.) using a JEM developed for the EPILYMPH investigators, a
modification of Bolm-Audorff et al.(1988). Exposure prevalence to TCE among cases was
13%. A cumulative exposure score was calculated and was the sum for every job held of
intensity of solvent exposure, frequency of exposure, and duration of exposure. High
exposure to TCE was defined as >35 ppm-years; 3% of cases had high cumulative exposure
to TCE. Intensity of TCE exposure was assessed on a semiquantitative scale with the
following categories: low intensity, 2.5 ppm (0.5 to 5); medium intensity, 25 ppm (>5 to 50),
high intensity, 100 ppm (>50). The frequency of exposure was the percentage of working
time during which the exposure occurred based upon a 40-hour week. A semiquantitative
scale was adopted for frequency of exposure with the following categories: low frequency,
3% of working time (range, 1 to 5%), medium frequency, 17.5 % (range, >5 to 30%), high
frequency, 65% of working time (>30%). A cumulative Prevalence of TCE exposure
among cases was 13% overall with 3% of cases identified with cumulative exposure
>35 ppm-years.
5	Overall, the use of expert assessment for exposure and WHO classification for disease
6	coding likely reduce misclassification bias in this study. This population case-control study, like
7	other population case-control studies of lymphoma and TCE, has a low prevalence of TCE
8	exposure and limits statistical power to detect risk factors.
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
9	B-303 DRAFT—DO NOT CITE OR QUOTE
-------
Seidler A, Mohner M, Berger J, Mester B, Deeg E, Eisner G, Neiters A, Becker N. (2007). Solvent exposure and malignant
lymphoma: a population-based case-control study in Germany. J Occup Med Toxicol 2:2. Accessed August 27, 2007,
http://www.occup-med.eom/content/2/l/2.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This case-control study of NHL and Hodgkin lymphomas was designed to investigate
association between specific exposure and distinct lymphoma classifications which
are defined by REAL and WHO classifications.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
812 male and female lymphoma patients between the ages of 18 and 80 yrs were
identified from a six German study regions from 1999 to 2003. 1,602 controls were
identified from population registers and matched (1:1) to cases on sex, region and
age. 710 cases and 710 controls were interviewed.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
NHL and Hodgkin's lymphoma incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
WHO classification. Diagnosis confirmed by pathological report for 691 cases.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Blinded assignment of intensity and frequency of exposure to specific chlorinated
hydrocarbons (includes TCE) and to aromatic hydrocarbons based upon
questionnaire information on complete occupational history for all jobs of >1 yr
duration. Exposure assessment approach based on a modification of Bolm-Audorff
etal. (1988)
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Face-to-face interview with questionnaire for detailed information about medical
history, lifestyle factors, and occupation. Job-task-specific supplementary
questionnaire administered to subjects having held jobs of interest; e.g., painters,
metal workers and welders, dry cleaners, chemical workers, shoemakers and leather
workers, and textile workers.
Blinded interviewers
Unblinded interviews.
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
No information provided in paper.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
710 cases (87.4%) and 710 controls (44.3%).
Exposure prevalence: Any TCE exposure, Cases, 13%, Controls, 15%.
High cumulative exposure (>35 ppm-yr), Cases, 3%, Controls, 1%.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, region, pack years of smoking, and # grams of alcohol consumed per day.
Statistical methods
Conditional logistic regression.
Exposure-response analysis presented in
published paper
Yes, by ppm-yr as continuous variable.
Documentation of results
Yes.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
B.3.2.6.7. Persson and Fredrikson (1999), Persson et al. (1989; 1993).
B.3.2.6.7.1. Author's abstract.
Non-Hodgkin's lymphoma (NHL) has been subject to several epidemiological
studies and various occupational and non-occupational exposures have been
identified as determinants. The present study is a pooled analysis of two earlier
methodologically similar case-referent studies encompassing 199 cases of NHL
and 479 referents, all alive. Exposure information, mainly on occupational agents,
was obtained by mailed questionnaires to the subjects. Exposure to white spirits,
thinner, and aviation gasoline as well as work as a painter was connected with
increased odds ratios, whereas no increased risk was noted for benzene. Farming
was associated with a decreased odds ratio and exposure to phenoxy herbicides,
wood preservatives, and work as a lumberjack showed increased odds ratios.
Moreover, exposure to plastic and rubber chemicals and also contact with some
kinds of pets appeared with increased odds ratios. Office employment and
housework showed decreased odds ratios. This study indicates the importance of
investigating exposures not occurring very frequently in the general population.
Solvents were studied as a group of compounds but were also separated into
various specific compounds. The present findings suggest that the carcinogenic
property of solvents is not only related to the aromatic ones or to the occurrence
of benzene contamination, but also to other types of compounds.
B.3.2.6.7.2. Study description and comment. The exposure assessment approach of
Persson and Fredriksson (1999), a pooled analysis of NHL cases and referents in Persson
et al. (1993), and Persson et al. (1989), was based upon self-reported information obtain
from a mailed questionnaire to cases and controls. Ten of 17 main questions of the detailed
multiple-page questionnaire concerned occupational exposure, with additional questions on
specific job and exposure details. These studies of the Swedish population considered
exposure durations of 1 or more years and those received 5 to 45 years before NHL
diagnosis for cases and before the point in time of selection for controls. The period of
TCE exposure assessed in the between 1964 and 1986, a time period similar to that of
Axelson et al. (1994). Semiqualitative information about solvent exposure was obtained
directly from the questionnaires. Assignment of exposure potential to individual solvents
such as TCE and white spirit is not described nor does the paper describe whether
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-307 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
assignment was done blinded as to case or control status. A five-category classification for
intensity was developed although statistical analyses grouped the TCE categories as
intensity scores of >2 compared to 0/1. TCE exposure prevalence among cases was 8% (16
of 199) and 7% among referents (32 of 479).
This small study of 199 NHL cases diagnosed between 1964 and 1986 at a regional
Swedish hospital (Orebro) and alive at the time of data acquisition in 1986 was similar in design
to other lymphoma (chronic lymphocytic leukemia, multiple myeloma) and occupation studies
from these investigators (Flodin et al., 1987). A series of 479 referents from the same catchment
area and from the same time period, identified previously from the multiple myeloma and
chronic lymphocytic leukemia studies, served as the source for controls in Persson and
Fredrikson (1999) for the NHL analysis and in Persson et al. (1989; 1993) for the Hodgkin's
lymphoma analysis. Given the study's entrance date as 1964, with interviews carried out in the
1980s, some cases were deceased with information likely provided by proxy respondents. The
paper does not identify the percentage of deceased cases and the magnitude of potential bias
associated with proxy respondents can not be determined. Little information is provided in the
published paper on controls; however, the paper notes 17% of eligible controls were not able or
unwilling to respond to the questionnaire. Case and control series appear to differ given only
subjects 40 to 80 years of age were included in the statistical analysis. Cases in Perrson et al.
(1993) were histologically confirmed diagnosis of NHL; this was not so for Persson et al. (1989).
Misclassification associated with misdiagnosis is not expected to be large given observation in
Perrson et al. (1993) of 2% of lymphoma cases were misclassified.
Overall, the study's 20-year period between initial case and control identification and
interview suggests some subjects were either survivors or information was obtained from proxy
respondents. In both instances, misclassification bias is likely. No information is provided on
job titles or the nature of TCE exposure, which was defined in the exposure assessment as
"exposed or unexposed." Exposure prevalence to TCE in this study is higher than that found in
community population studies of Miligi et al.(2006), Seidler et al. (2007) and Costantini et al.
(2008).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-308 DRAFT—DO NOT CITE OR QUOTE
-------
Persson B, Fredrikson M. (1999). Some risk factors for non-Hodgkin's lymphoma. Int J Occup Med Environ Health
12:135-142.
Persson B, Fredriksson M, Olsen K, Boeryd B, Axelson O. (1993). Some occupational exposure as risk factors for malignant
lymphomas. Cancer 72:1773-1778.
Persson B, Dahlander A-M, Fredriksson M, Brage HN, Ohlson C-G, Axelson O. (1989). Malignant lymphomas and
occupational exposures. Br J Ind Med 46:516-520.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
These studies of Hodgkin's Lymphoma and NHL investigated occupational
associations. Examination of TCE is not stated as a priori hypothesis.
Selection and characterization in cohort
studies of exposure and control groups and
of cases and controls in case-control studies
is adequate
Incident NHL and Hodgkin's lymphoma cases reported to a regional cancer registry
between 1975 and 1984, n = 148 (Persson et al., 1993), or identified from hospital
records (Orebro Medical Center Hospital) for the period 1964 and 1986, n = 175
(Persson et al., 1989). Population controls from the same geographical area as cases
were identified from previous case-control studies of leukemia and multiple myeloma
and matched on age and sex. Analysis of NHL and Hodgkin's lymphoma each used the
same set of controls.
Persson and Fredrikson (1999)—199 cases of NHL, 479 controls.
Persson et al. (1993)—93 NHL and 31 Hodgkin's lymphoma (90% participation);
204 controls.
Persson et al. (1989)—106 NHL and 54 Hodgkin's lymphoma (91%); 275 controls.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Classification system not identified in papers.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Self-reported occupational exposures as obtained from a mailed questionnaire.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Mailed questionnaire, only.
Blinded interviewers
N/A
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No information provided in paper.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality
studies; numbers of total cancer incidence
studies; numbers of exposed cases and
prevalence of exposure in case-control
studies
Exposure prevalence to TCE—
Persson and Fredrikson (1999)—16 NHL cases (8%>) and 32 controls (7%).
Persson et al. (1993)—8 NHL cases (8%>) and 5 Hodgkin's lymphoma cases (16%>); 18
controls (9%>).
Persson et al. (1989)—8 NHL cases (8%>) and 7 Hodgkin's lymphoma cases (13%>); 14
controls (5%).
CATEGORY H: ANALYSIS
Control for potential confounders in
statistical analysis
Cases and controls are matched on age and sex. Statistical analyses do not control for
other possible confounders.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
h
to
s
"tS
§*
3'
Co
a
>r
<§>
TO
*
^S
«
8
TO
Co
0
S
1
a.
VO
to
o
Statistical methods
Only crude odds ratios are presented for TCE exposure, although logistic regression was
used to examine other occupational exposure and NHL/Hodgkin's lymphoma.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Poor, unable to determine response rate in control population, if controls were similar to
cases on demographic variables such as sex and age, and whether controls were
identified from same time period as cases.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
B.3.2.6.8. Nordstrom et al. (1998).
B.3.2.6.8.1. Author's abstract.
To evaluate occupational exposures as risk factors for hairy cell leukemia (HCL),
a population-based case-control study on 121 male HCL patients and 484 controls
matched for age and sex was conducted. Elevated odds ratio (OR) was found for
exposure to farm animals in general: OR 2.0, 95% confidence interval (CI) 1.2-
3.2. The ORs were elevated for exposure to cattle, horse, hog, poultry and sheep.
Exposure to herbicides (OR 2.9, CI 1.4-5.9), insecticides (OR 2.0, CI 1.1-3.5),
fungicides (OR 3.8, CI 1.4-9.9) and impregnating agents (OR 2.4, CI 1.3-4.6) also
showed increased risk. Certain findings suggested that recall bias may have
affected the results for farm animals, herbicides and insecticides. Exposure to
organic solvents yielded elevated risk (OR 1.5, CI 0.99-2.3), as did exposure to
exhaust fumes (OR2.1, CI 1.3-3.3). In an additional multivariate model, the ORs
remained elevated for all these exposures with the exception of insecticides. We
found a reduced risk for smokers with OR 0.6 (CI 0.4-1.1) because of an effect
among non-farmers.
B.3 .2.6.8.2. Study description and comment. This population case-control of hairy cell
leukemia, a B-cell lymphoid neoplasm and NHL, examined occupational organic solvent
and pesticide exposures among male cases reported to the Swedish Cancer Registry
between 1987 and 1992. A total of 121 cases, including 1 case one case, originally thought
to have a diagnosis within the study's window, but latter learned as in 1993, and four
controls per case matched on age and county of residence from the Swedish Population
Registry. Occupational exposure was assessed based upon self-reported information
provided in a mailed questionnaire with telephone follow-up by trained interviewer blinded
to case or control status. Chemical-specific exposures of at least 1 day duration and
occurring one year prior to case diagnosis were assigned to study subjects; however, the
procedure for doing this was not described in the paper. Potential for organic solvents
exposure included exposure received during leisure activities and work-related activities.
Exposure prevalence to TCE among cases is 8 and 7% among controls. The low exposure
prevalence and study size limit the statistical power of this study for detecting relative risks
smaller than 2.0.
Odds ratios and 95% confidence intervals are presented for chemical-specific exposures,
including TCE, from logistic regression models in two separate analyses, univariate analysis and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-313 DRAFT—DO NOT CITE OR QUOTE
-------
1	multivariate analysis adjusting for age. The odds ratio for TCE exposure is presented only from
2	univariate analysis. Age may not greatly confound or bias the observed association; an
3	examination of risk estimates from univariate and multivariate analyses of the aggregated
4	exposure category for organic solvents showed similar odds ratios, indicating age was not a
5	significant source of bias in the statistical analyses because age was controlled in the study's
6	design, a control was matching to a case on age.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-314 DRAFT—DO NOT CITE OR QUOTE
-------
Nordstrom M, Hardell L, Hagberg H, Rask-Andersen A. (1998). Occupational exposures, animal exposure and smoking as
risk factors for hairy cell leukemia evaluated in a case-control study. Br J Cancer 77:2048-2052.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Abstract—To evaluate occupational exposure as risk factors for hairy cell leukemia.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
121 cases of HCL in males reported to the Swedish Cancer Registry between 1987
and 1992.
484 controls (1:4 matching) identified from Swedish Population Registry and
matched for age and county of residence.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper, likelv ICD-9 (http://www.socialstvrelsen.se/, accessed
February 6, 2009).
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Self-reported information on occupational exposure as obtained from a mailed
questionnaire to study participants. Questionnaire sought information on complete
working history, other exposures, and leisure time activities with telephone interview
in cases of incomplete information. Paper does not describe the procedure for
assigning chemical exposures from job title information.
CATEGORY D: FOLLOW-UP (COHORT)
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Mailed questionnaire.
Blinded interviewers
Follow-up telephone interview and job/exposure coding were done blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Proxy responses: 4%>, cases; 1% controls.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Ill HCL cases, 400 controls.
Response rate: 91% cases and 83% controls.
Exposure prevalence among cases is 8 and 7% among controls.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Cases and controls are matched for age, sex, and county of residence. Effect measure
for TCE exposure from univariate analysis presented in paper; other possible
confounders or covariates not included in statistical analysis.
Statistical methods
Logistic regression.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
HCL = hairy cell leukemia.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
B.3.2.6.9. Fritschi and Siemiatycki (1996a), Siemiatycki (1991).
B.3.2.6.9.1. Author's abstract.
The known risk factors for lymphoma and myeloma cannot account for the
current incidence rates of these cancers, and there is increasing interest in
exploring occupational causes. We present results regarding lymphoma and
myeloma from a large case-control study of hundreds of occupational exposures
and 19 cancer sites. We examine in more detail those exposures previously
considered to be related to these cancers, as well as exposures which were
strongly related in our initial analyses. Lymphoma was not associated in our data
with exposure to solvents or pesticides, or employment in agriculture or wood-
related occupations, although numbers of exposed cases were sometimes small.
Hodgkin's lymphoma was associated with exposure to fabric dust, and non-
Hodgkin's lymphoma was associated with exposure to copper dust, ammonia and
a number of fabric and textile-related occupations and exposures. Employment as
a sheet metal worker was associated with development of myeloma.
B.3 .2.6.9.2. Study description and comment. This population study of several cancer
sites included histologically-confirmed cases of NHL, Hodgkin's lymphoma and myeloma
ascertained from 16 Montreal-area hospitals between 1979 and 1985 and part of a larger
study of 10 other cancer sites. This study relies on the use of expert assessment of
occupational information on a detailed questionnaire and face-to-face interview. Fritschi
and Siemiatycki (1996a) present observations of analyses examining industries, occupation,
and some chemical-specific exposures, including solvents, but not TCE. Observations on
TCE are found in the original report of Siemiatycki (1991).
A total of 215 NHL cases (83% response) were identified from 19 Montreal-area
hospitals and while this case group is larger than that in Swedish lymphoma case-control studies,
there are fewer NHL cases than other multicenter studies published since 2000. The
533 population controls (72% response), identified through the use of random digit dialing, and
were used for each site-specific cancer case analyses. All controls were interviewed using
face-to-face methods; however, 20% of the NHL cases were either too ill to interview or had
died and, for these cases, occupational information was provided by a proxy respondent. The
quality of interview conducted with proxy respondents was much lower, increasing the potential
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-317 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
for misclassification bias, than that with the subject. The direction of this bias would diminish
observed risk towards the null. Interviewers were unblinded, although exposure assignment was
carried out blinded as to case and control status. The questionnaire sought information on the
subject's complete job history and included questions about the specific job of the employee and
work environment. Occupations considered with possible TCE exposure included machinists,
aircraft mechanics, and industrial equipment mechanics. An additional specialized questionnaire
was developed for certain job title of a prior interest that sought more detailed information on
tasks and possible exposures. For example, the supplemental questionnaire for machinists
included a question on TCE usage.
A team of industrial hygienists and chemicals assigned exposures blinded based on job
title and other information obtained by questionnaire. A semiquantitative scale was developed
for 294 exposures and included TCE (any, substantial). Any exposure to TCE was 3% among
cases but <1% for substantial TCE exposure; "substantial" is defined as >10 years of exposure
for the period up to 5 years before diagnosis. The TCE exposure frequencies in this study are
lower than those in more recent NHL case-control studies examining TCE. The expert
assessment method is considered a valid and reliable approach for assessing occupational
exposure in community-base studies and likely less biased from exposure misclassification than
exposure assessment based solely on self-reported information (Fritschi et al., 2003; IOM, 2003;
Siemiatycki et al., 1997).
Logistic regression models adjusted for age, ethnicity, income, and respondent status
(Fritschi and Siemiatycki, 1996a) or Mantel-Haenszel x stratified on age, family income, and
cigarette smoking (Siemiatycki, 1991). Odds ratios for TCE exposure are presented with 90%
confidence intervals in Siemiatycki (1991) and with 95% confidence intervals in Fritschi and
Siemiatycki (1996b).
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of NHL. However, the use of the
general population (rather than a known cohort of exposed workers) reduced the likelihood that
subjects were exposed to TCE, resulting in relatively low statistical power for the analysis. The
job exposure matrix, applied to the job information, was very broad since it was used to evaluate
294 chemicals. Overall, a reasonably good exposure assessment is found in this analysis;
however, examination of NHL and TCE exposure is limited by statistical power considerations
related to low exposure prevalence, particularly for "substantial" exposure. For the exposure
prevalence found in this study to TCE and for NHL, the minimum detectable odds ratio was 3.0
when p = 0.02 and a = 0.05 (one-sided). The low statistical power to detect a doubling of risk
and an increased possibility of misclassification bias associated with case occupational histories
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-318 DRAFT—DO NOT CITE OR QUOTE
-------
1	resulting from proxy respondents suggests this study is less sensitive than other NHL case-
2	controls published since 2000 for examining NHL and TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-319 DRAFT—DO NOT CITE OR QUOTE
-------
Fritschi L, Siemiatycki J. (1996a). Lymphoma, myeloma and occupation: Results of a case-control study. Int J Cancer 67:
498-503.
Siemitycki J. (1991). Risk Factors for Cancer in the Workplace. J Siemiatycki, Ed. Baca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study of NHL was designed to investigate association
between specific exposure and cancers at 20 sites using expert assessment method for
exposure assignment.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
258 histologically-confirmed NHL cases were identified among Montreal area males,
aged 35 to 70 yrs, diagnosed in 16 Montreal hospitals between 1979 and 1985.
740 male population controls were identified from the same source population using
random digit dialing methods.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
NHL.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICDO-O, 200 and 202, International Statistical Classification of Diseases for
Oncology (WHO, 1977).
ICDO-O is based upon rubrics of ICD, 9th Revision.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 300 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Yes, 82%) of case interviews were face-to-face; 100%> of control interviews were with
subject.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, -20%) of cases had proxy respondents. Interviews were completed with all
control subjects.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
215 cases (83%> response), 533 population controls (71%>).
Exposure prevalence: Any TCE exposure, 3% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), <1% cases.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq S.
^ ^3
^ 5
S. ^
^ o
• TO Co
V; to
Co
>r
<§>
>S
VO
to
o
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, income, index for cigarette smoking (Siemiatycki, 1991).
Age, proxy status, income, ethnicity (Fritschi and Siemiatycki, 1996a).
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Unconditional logistic regression (Fritschi and Siemiatycki, 1996a).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
B.3.2.6.10. Hardell et al. (1994; 1981).
B.3.2.6.10.1. Author's abstract.
Results on 105 cases with histopathologically confirmed non-Hodgkin's
lymphoma (NHL) and 335 controls from a previously published case-control
study on malignant lymphoma are presented together with some extended
analyses. No occupation was a risk factor for NHL. Exposure to phenoxyacetic
acids yielded, in the univariate analysis, an odds ratio of 5.5 with a 95%
confidence interval of 2.7-11. Most cases and controls were exposed to a
commercial mixture of 2, 4-dichlorophenoxyacetic acid and 2, 4, 5-
trichlorophenoxyacetic acid. Exposure to chlorophenols gave an odds ratio of 4.8
(2.7-8.8) with pentachlorophenol being the most common type. Exposure to
organic solvents yielded an odds ratio of 2.4 (1.4-3.9). These results were not
significantly changed in the multivariate analysis.
Dichlorodiphenyltrichloroethane, asbestos, smoking, and oral snuff were not
associated with an increased risk for NHL. The results regarding increased risk
for NHL following exposure to phenoxyacetic acids, chlorophenols, or organic
solvents were not affected by histopathological type, disease stage, or anatomical
site of disease presentation. Median survival was somewhat longer in cases
exposed to organic solvents than the rest. This was explained by more prevalent
exposure to organic solvents in the group of cases with good prognosis NHL
histopathology.
A number of men with malignant lymphoma of the histiocytic type and
previous exposure to phenoxy acids or chlorophenols were observed and reported
in 1979. A matched case-control study has therefore been performed with cases of
malignant lymphoma (Hodgkin's disease and non-Hodgkin lymphoma). This
study included 169 cases and 338 controls. The results indicate that exposure to
phenoxy acids, chlorophenols, and organic solvents may be a causative factor in
malignant lymphoma. Combined exposure of these chemicals seemed to increase
the risk. Exposure to various other agents was not obviously different in cases and
in controls.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-323 DRAFT—DO NOT CITE OR QUOTE
-------
B.3.2.6.10.2. Study description and comment. Exposure in these case-control studies of
histologically-confirmed lymphoma (NHL and Hodgkin's lymphoma) (Hardell et al., 1981)
or only the NHL cases only (Hardell et al., 1994) over a 4-year period, 1974-1978, in Umea,
Sweden was assessed based upon information provided in a self-administered
questionnaire. The questionnaire obtained information on a complete working history
over the life of the subjects along with information on various other exposures and leisure
time activities. Organic solvent exposures were examined secondary to this study's
primary hypothesis examining phenoxy acid or chlorophenol exposures and lymphoma.
The extent of recall bias related to self-reported information can not be determined nor is
information provided in the published papers misclassification bias resulting from next-of-
kin interviews. Occupations were classification according to the Nordic Working
Classification system. Chemical specific exposures assignment was not described but
appears to have been carried out blinded as to case or control status. A semiquantitative
classification scheme based on intensity and duration of exposure was used to categorize
solvent exposure into two groupings: low grade—less than 1 week continuously or less than
1 month in total—and high grade for all other exposure scenarios. TCE exposure
prevalence is similar in both studies; 4% for cases and 1% for controls. The low exposure
prevalence and small numbers of cases with TCE exposure (n = 4) limits the statistical
power of these analyses and results in wide confidence intervals around the estimated odds
ratio for TCE exposure (95% Confidence Interval, 1.3-42).
1	The Rappaport Classification was used to identify non-Hodgkin's and Hodgkin's
2	Lymphoma cases. The Rappaport Classification was in widespread use until the 1970s and was
3	based on a cell's pathologic characteristics. Equivalence of non-Hodgkin's lymphoma groupings
4	according to Rappaport Classification system to ICDA-8 groupings, also in use during this time
5	period, is 200 "Lymphosarcoma and reticulum-cell sarcoma" and 202 "Other neoplasms of
6	lymphoid tissue."
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-324 DRAFT—DO NOT CITE OR QUOTE
-------
Hardell L, Eriksson M, Degerman A. (1994). Exposure to phenoxyacetic acids, chlorophenols, or organic solvents in relation
to histopathology, stage, and anatomical localization of non-Hodgkin's lymphoma. Cancer Res 54:2386-2389.
Hardell L, Eriksson M, Lenner P, Lundgren E. (1981). Malignant lymphoma and exposure to chemicals, especially organic
solvents, chlorophenols and phenoxy acids: a case-control study. Br J Cancer 43:169-176.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
NHL cases from a case-control study of lymphoma (NHL and Hodgkin's lymphoma)
are analyzed separately to evaluate herbicide and organic solvents exposure.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
105 cases of histologically-confirmed NHL among males aged 25-85 yrs admitted to
local hospital's oncology department between 1974 and 1978.
A total of 335 male controls identified from the Swedish Population Registry, for
living cases, and from the Swedish Registry for Causes of Death, for dead cases.
Controls matched to cases by age, residence municipality, and year of death, for dead
cases.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Rappaport Classification; equivalent to ICDA-8 Codes, 200, and 202.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Self-reported information on occupational exposure as obtained by questionnaire,
with a telephone interview for incomplete or unclear information. Questionnaire
sought information on complete working history, other exposures and leisure time
activities. Paper does not describe the procedure for assigning chemical exposures
from job title information.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
No information in paper.
Blinded interviewers
Follow-up telephone interview was done blinded as to case and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No information in paper.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
105 NHL cases, 332 controls.
Response rates could not be calculated given insufficient information in paper.
Prevalence of TCE exposure, 4% cases, 1% controls.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Cases and controls matched on sex, age, place of residence and vital status. For
deceased controls are matched to deceased cases on year of death.
Statistical methods
Mantel-Haenszel stratified by age and vital status.
-------

§•
to
s
a, 5
o a
TO 3
s
o
o
s
Co
I?
f? Ng
^ 3
>9
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
8
TO
Co
0
S
1
a.
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
B.3.2.7. Childhood Leukemia
B.3.2.7.1. Shu et al. (2004; 1999)
B.3.2.7.1.1. Author's abstract.
Ras proto-oncogene mutations have been implicated in the pathogenesis of many
malignancies, including leukemia. While both human and animal studies have
linked several chemical carcinogens to specific ras mutations, little data exist
regarding the association of ras mutations with parental exposures and risk of
childhood leukemia. Using data from a large case control study of childhood
acute lymphoblastic leukemia (ALL; age <15 years) conducted by the Children's
Cancer Group, we used a case-case comparison approach to examine whether
reported parental exposure to hydrocarbons at work or use of specific medications
are related to ras gene mutations in the leukemia cells of children with ALL. DNA
was extracted from archived bone marrow slides or cryopreserved marrow
samples for 837 ALL cases. We examined mutations in K-ras and N-ras genes at
codons 12, 13, and 61 by PCR and allele-specific oligonucleotide hybridization
and confirmed them by DNA sequencing. We interviewed mothers and, if
available, fathers by telephone to collect exposure information. Odds ratios (ORs)
and 95% confidence intervals (CIs) were derived from logistic regression to
examine the association of parental exposures with ras mutations. A total of 127
(15.2%) cases had ras mutations (K-ras 4.7% and N-ras 10.68%). Both maternal
(OR 3.2, 95% CI 1.7-6.1) and paternal (OR 2.0, 95% CI 1.1-3.7) reported use of
mind-altering drugs were associated with N-ras mutations. Paternal use of
amphetamines or diet pills was associated with N-ras mutations (OR 4.1, 95% CI
1.1-15.0); no association was observed with maternal use. Maternal exposure to
solvents (OR 3.1, 95% CI 1.0-9.7) and plastic materials (OR 6.9, 95% CI 1.2-
39.7) during pregnancy and plastic materials after pregnancy (OR 8.3, 95% CI
1.4-48.8) were related to K-ras mutation. Maternal ever exposure to oil and coal
products before case diagnosis (OR 2.3, 95% CI 1.1-4.8) and during the postnatal
period (OR 2.2, 95% CI 1.0-5.5) and paternal exposure to plastic materials before
index pregnancy (OR2.4, 95% CI 1.1-5.1) and other hydrocarbons during the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-328 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
postnatal period (OR 1.8, 95% CI 1.0-1.3) were associated with N-ras mutations.
This study suggests that parental exposure to specific chemicals may be
associated with distinct ras mutations in children who develop ALL.
Parental exposure to hydrocarbons at work has been suggested to increase the
risk of childhood leukemia. Evidence, however, is not entirely consistent. Very
few studies have evaluated the potential parental occupational hazards by
exposure time windows. The Children's Cancer Group recently completed a large-
scale case-control study involving 1842 acute lymphocytic leukemia (ALL) cases
and 1986 matched controls. The study examined the association of self-reported
occupational exposure to various hydrocarbons among parents with risk of
childhood ALL by exposure time window, immunophenotype of ALL, and age at
diagnosis. We found that maternal exposure to solvents [odds ratio (OR), 1.8;
95% confidence interval (CI), 1.3-2.5] and paints or thinners (OR, 1.6; 95% CI,
1.2-2.2) during the preconception period (OR, 1.6; 95% CI, 1.1-2.3) and during
pregnancy (OR, 1.7; 95% CI, 1.2-2.3) and to plastic materials during the postnatal
period (OR, 2.2; 95% CI, 1.0-4.7) were related to an increased risk of childhood
ALL. A positive association between ALL and paternal exposure to plastic
materials during the preconception period was also found (OR, 1.4; 95% CI, 1.0-
1.9). The ALL risk associated with parental exposures to hydrocarbons did not
vary greatly with immunophenotype of ALL. These results suggest that the effect
of parental occupational exposure to hydrocarbons on offspring may depend on
the type of hydrocarbon and the timing of the exposure.
B.3.2.7.1.2. Study description and comment. Parent hydrocarbon occupational exposure
in this case-control study of acute lymphatic leukemia in children less than 15 years of age
was assessed from telephone questionnaire to mothers and, whenever available, fathers of
cases and controls who were part of the large-scale incidence study by the Children's
Cancer/Oncology Group. A recent paper examines hydrocarbon exposures and
relationship with the ras proto-oncogene (Shu et al., 2004). Nearly 50% of childhood
leukemia cases in the United States were treated by a Children's Cancer Group hospital or
institution and between January 1,1989 and June 15,1993, the study period, a total of
2,081 incident childhood leukemia cases were identified with 1,914 interviews with mothers.
Controls were randomly selected using a random digit dialing procedure and matched to
cases on age, race, and geographic location. Using structured questionnaires, parents or a
surrogate when unavailable were asked about job title, industry, duties, starting and
stopping date for all jobs held by the father for more than 6 months beginning at age 18
years and by the mother for all jobs held at least 6 months in the period from 2 year prior
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-329 DRAFT—DO NOT CITE OR QUOTE
-------
to the index pregnancy to date of diagnosis of leukemia case or the reference date of the
controls. The questionnaire sought information on specific exposures to solvents (carbon
tetrachloride, TCE, benzene, toluene, and xylene), plastic materials, paints, pigments or
thinners, and oil or coal products. Exposure quantitative was not possible. Statistical
analyses use self-reported exposure to specific hydrocarbons as defined as a dichotomous
variable (yes/no). The potential for misclassification bias is greater with exposure
assessment based upon self-reports compared to that by expert assessment (Teschke et al.,
2002). Exposure information was linked to start and stop data of the relevant job to
determine the timing of exposure related to specific windows of possible susceptibility for
acute lymphoblastic leukemia (ALL). The author's do not describe jobs associated with
possible TCE exposure.
1	The father's questionnaire was completed for 1,801 of the 2,081 eligible cases and 1,813
2	of the 2,597 eligible controls. Of the 1,618 matched sets, direct interview with fathers were
3	obtained for 83% of cases and 68% of controls. Maternal interview were completed for 1,914 of
4	the 2,081 eligible cases (92%). The low prevalence of any exposure to TCE, 1% for mothers
5	(15 cases of 1,842 matched pairs with maternal exposure information) and 8% for fathers
6	(136 cases out 1,618 matched pairs), limits the statistical power of this study to detect low to
7	moderate risk.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-330 DRAFT—DO NOT CITE OR QUOTE
-------
Shu Xo, Perentesis JP, Wen W, Buckley JD, Boyle E, Ross, JA, Robison LL. (2004). Parental exposure to medications and
hydrocarbons and ras mutations in children with acute lymphoblastic leukemia: A report from the Children's Oncology
Group. Cancer Epidemiol Biomarkers Prev 13:1230-1235.
Shu XO, Stewart P, Wen W-Q, Han D, Potter JD, Buckley JD, Heineman E, Robison LL. (1999). Parental occupational
exposure to hydrocarbons and risk of acute lymphocytic leukemia in offspring. Cancer Epidemiol Markers Prev 8:783-291.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Shu et al. (2004; 1999) examine possible association with a number of maternal and
paternal exposures among cases and controls identified from the Children's
Cancer/Oncology Group. The Children's Cancer/Oncology Group is an association
of more than 120 centers in the United States, Canada, and Australia who
collaboratively carry out research on risk factors and treatment of childhood cancers.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
848 children with acute lymphatic leukemia of ages 0-9 yrs of age at diagnosis from
1980-1993 and <14 yrs old at diagnosis between 1994 and 2000 were identified from
cancer care centers in Quebec, Canada.
Controls are concurrently identified from population, from 1980-1993, from family
allowance files and from 1994-2000, from universal health insurance files; and,
matched (1:1 matching ratio) to cases on sex and age at the time of diagnosis
(calendar date).
Participation rates- 93.1% cases (790 of 849 eligible cases); 86.2% controls (790 of
916 eligible controls).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Childhood leukemia incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD, 9th revision, Code 204.0.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Telephone interviews of mothers of cases and controls using structured questionnaire
were administered to obtain information on general risk factors and potential
confounders. Questionnaire also sought information on a complete job history, for
the mother from 18 years of age to the end of pregnancy and included for each job,
job title, dates of employment, type of industry, and location of employer. Statistical
analyses based on self-reported occupational exposure to hydrocarbons as defined by
broad groups and individual hydrocarbons.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE

<90% face-to-face
Telephone interview, >99%> response.
Blinded interviewers
Telephone interviews were not blinded, but exposure assignment and coding was
carried out blinded to case and control status by chemists and industrial hygienists.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
100%) of cases and controls had maternal history provided by direct interview with
mothers.
13%o of cases and 30% of controls had paternal information provided by proxy
respondent (e.g., through maternal interview).
CATEGORY G: SAMPLE SIZE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
0
1
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
15 cases (2% exposure prevalence) and 9 controls (1% exposure prevalence) with
maternal TCE exposure.
136 cases (8% exposure prevalence) and 104 controls (13% exposure prevalence)
with paternal TCE exposure.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Child's age at time of diagnosis , sex, and calendar year of diagnosis, maternal age
and level of schooling.
Statistical methods
Conditional logistic regression—
By two time periods; 2 yrs before pregnancy up to birth, during specific
pregnancy period.
By level of exposure; Level 1 (some exposure) compared to no exposure, and
Level 2 (greater exposure potential) compared to no exposure.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
B.3.2.7.2. Costas et al. (2002), MADPH (1997a).
B.3.2.7.2.1. Author's abstract.
A 1981 Massachusetts Department of Public Health study confirmed a childhood
leukemia cluster in Woburn, Massachusetts. Our follow-up investigation attempts
to identify factors potentially responsible for the cluster. Woburn has a 130-year
industrial history that resulted in significant local deposition of tannery and
chemical manufacturing waste. In 1979, two of the city's eight municipal drinking
water wells were closed when tests identified contamination with solvents
including trichloroethylene. By 1986, 21 childhood leukemia cases had been
observed (5.52 expected during the seventeen year period) and the case-control
investigation discussed herein was begun. Nineteen cases and 37 matched
controls comprised the study population. A water distribution model provided
contaminated public water exposure estimates for subject residences. Results
identified a non-significant association between potential for exposure to
contaminated water during maternal pregnancy and leukemia diagnosis, (odds
RATIO=8.33, 95% CI 0.73-94.67). However, a significant dose-response
relationship (P<0.05) was identified for this exposure period. In contrast, the
child's potential for exposure from birth to diagnosis showed no association with
leukemia risk. Wide confidence intervals suggest cautious interpretation of
association magnitudes. Since 1986, expected incidence has been observed in
Woburn including 8 consecutive years with no new childhood leukemia
diagnoses.
B.3.2.7.2.2. Study description and comment. Exposure in this case-control study of
childhood leukemia over a 20-year period in Woburn, MA was assessed based upon the
potential for a residence at the time of diagnosis to receive water from wells G and H, wells
with a hydraulic mixing model of Murphy (Murphy, 1990) which described the town's
water distribution system. Monitoring of wells G and H in 1979 showed the presence of
several VOCs; TCE and perchloroethylene (PERC) were found to exceed drinking water
guidelines, at 267 ppb and 21 ppb, respectively. Low levels of other contaminates were
detected including chloroform, 1,2-dichloroethylene methyl chloroform,
trichlorotrifluoroethane, and inorganic arsenic. The Murphy model described the water
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-335 DRAFT—DO NOT CITE OR QUOTE
-------
flow through Woburn during the lifetime of wells G and H. The model uses data
describing the physical layout of Woburn's municipal water system and information
regarding the pumping cycles of wells G and H and other active uncontaminated wells that
supplied the municipal water system. Model accuracy showed distribution of water from
wells G and H to a block area with predicted mixture concentrations with an average error
within 10% of the know concentration. Nearly 70% of the model predictions were within
20% of the know validation concentrations. An exposure value for cases and controls by
exposure period was the sum of the model-predicted water concentration for each
residence in Woburn as assigned to a hydrologically-distinct area along the water
distribution network. Both cumulative and average exposure estimates were derived using
the model.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-336 DRAFT—DO NOT CITE OR QUOTE
-------
Costas K, Knorr RS, Condon SK. (2002). A case-control study of childhood leukemia in Woburn, Massachusetts: the
relationship between leukemia incidence and exposure to public drinking water. Sci Total Environ 300:23-25.
Massachusetts Department of Public Health (MADPH). (1997a). Woburn Childhood Leukemia Follow-up Study. Volumes I
and II. Final Report.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes, "this follow-up investigation attempts to identify factors potentially responsible
for the leukemia cluster in Woburn, MA" and the primary exposure of concern for
investigation is "the potential consumption of contaminated water from Wells G and
H by Woburn residents."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
21 cases of leukemia diagnosed in children <19 yrs between 1969 and 1989 who
were residents of Woburn MA. Cases diagnosed from 1982 and latter were provided
by the Massachusetts Cancer Registry. Cases diagnosed prior to 1982 were
identified from local pediatric health professionals and by contacting all
greater-Boston childhood oncology centers that treated children with leukemia.
Two controls for each case were randomly selected from Woburn Public School
records on a geographically basis and matched to cases on race, sex and date of birth
(+3 mos).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Childhood leukemia incidence.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Changes in diagnostic coding systems for
ICD-0 (Acute Lymphatic Leukemia, Acute Myelogenous Leukemia, and Chronic
lymphoma, particularly non-Hodgkin's
Myelogenous Leukemia).
lymphoma

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
In-person interviewers with mothers and fathers of cases and controls using
questionnaire to gather information regarding demographics, residential information
for the mother and child, occupational history, maternal medical and reproductive
history, child's medical history, and life-style questions. The father's questionnaire
contained questions concerning military and occupational history and also included
duplicate questions on maternal occupational history, child's medical history and
life-style habits.
A hydraulic mixing computer model describing Woburn's water distribution system
was utilized to assign an exposure index expressed as cumulative number of months
a household received contaminated drinking water from Wells G and H.
Exposure Index = fraction of time during month when water from Wells G and H
reached the user area + fraction of water from Wells G and H supplied to user area.
No quantitative measures of TCE and other volatile organic solvents concentrations
were included in hydraulic mixing model.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Personal interviews with cases and controls; 19 of 21 cases (91%>) and 38 of possible
54 controls (70%>) were interviewed.
Blinded interviewers
Interviewers were not blinded as to case and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
One parent interviewed for 21% of cases and 11% of controls.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Participation rates- 93.1% cases (790 of 849 eligible cases); 86.2% controls (790 of
916 eligible controls).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq S.
^ 3i
^ ^3
^ 5
S. ^
^ o
• TO Co
V; to
Co
>r
<§>
>S
VO
to
o
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Composite covariates used to control for socioeconomic status, maternal smoking
during pregnancy, maternal age at birth of child, and maternal alcohol consumption
during pregnancy.
Statistical methods
Conditional logistic regression.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes and includes information in MADPH Final Report (1997a).
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
B.3.2.7.3. McKinney et al. (1991).
B.3.2.7.3.1. Author's abstract.
OBJECTIVE—To determine whether parental occupations and chemical and other
specific exposures are risk factors for childhood leukemia. DESIGN—Case-
control study. Information on parents was obtained by home interview.
SETTING—Three areas in north England: Copeland and South Lakeland (west
Cumbria); Kingston upon Hull, Beverley, East Yorkshire, and Holderness (north
Humberside), and Gateshead. SUBJECTS—109 children aged 0-14 born and
diagnosed as having leukemia or non-Hodgkin's lymphoma in study areas during
1974-88. Two controls matched for sex and date and district of birth were
obtained for each child. MAIN OUTCOME MEASURES—Occupations of
parents and specific exposure of parents before the children's conception, during
gestation, and after birth. Other adults living with the children were included in
the postnatal analysis. RESULTS—Few risk factors were identified for mothers,
although preconceptional association with the food industry was significantly
increased in case mothers (odds ratio 2.56; 95% confidence interval 1.32 to 5.00).
Significant associations were found between childhood leukemia and reported
preconceptional exposure of fathers to wood dust (2.73, 1.44 to 5.16), radiation
(3.23, 1.36 to 7.72), and benzene (5.81, 1.67 to 26.44); ionizing radiation alone
gave an odds ratio of 2.35 (0.92 to 6.22). Raised odds ratios were found for
paternal exposure during gestation, but no independent postnatal effect was
evident. CONCLUSION—These results should be interpreted cautiously because
of the small numbers, overlap with another study, and multiple exposure of some
parents. It is important to distinguish periods of parental exposures; identified risk
factors were almost exclusively restricted to the time before the child's birth.
B.3 .2.7.3 .2. Study description and comment. A population case-control study of ALL
and NHL in children of <14 years of age and residing in three areas in the United Kingdom
was carried out to identify possible risk factors for the region's observed increased
background childhood leukemia rates. The Sellafield nuclear reprocessing plant was
located in one of the areas and one hypothesis was an examination of parental radiation
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-342 DRAFT—DO NOT CITE OR QUOTE
-------
exposure and childhood lymphoma. Un-blinded face-to-face interviews with cases,
identified from regional tumor registries, and controls, identified using regional birth
registers, used a structured questionnaire to ascertain a complete history of employment
and exposure to specific substances and radiation from both child's biological parents,
preferred, although, in the absence of one parent, surrogate information by the other
parent was obtained from the date of first employment to end of the study period or, if
earlier, the date the parent ceased seeing the child. The questionnaire additionally sought
information on maternal and paternal exposure to 22 known chemical carcinogens.
McKinney et al. (1991) noted that exposures were highly correlated. Information on job
title and industry as reported in the questionnaire was coded independently by experts to
occupational groupings and titles using a national classification scheme from the Office of
Population Census and Surveys and is a strength of this study. The category of metal
refining industry and occupations was one of nine occupational groups identified a priori
for hypothesis testing. Statistical analyses are based on exposure as defined by industry,
occupational title, or chemical-specific exposure.
1	Interviewers with one or both parents were carried out for 109 of 151 eligible cases
2	(72%) and with 206 of 269 eligible controls (77%), and the low exposure prevalence; no
3	information was presented on the number of surrogate interviews, or, where only one parent
4	responded for both parents. The low prevalence of TCE exposure, 5 discordant pairs (one
5	subject with exposure and the matched subject without exposure) identified with maternal TCE
6	exposure and 16 discordant pairs with paternal preconceptional TCE exposure, greatly limited
7	the statistical power of this study.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-343 DRAFT—DO NOT CITE OR QUOTE
-------
McKinney PA, Alexander FE, Cartwright RA, Parker L. (1991). Parental occupations of children with leukemia in west
Cumbria, north Humberside, and Gateshead. BMJ 302:681-687.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study examines a number of risk factors (specific chemicals and occupational
groups) as possibly associated with the high background rate of acute lymphatic
leukemia and non-Hodgkin's lymphoma in children <14 yrs in the three regions.
22 individual chemicals and 7 occupational groups for a priori hypothesis testing.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
151 case children identified from two tumor registries (Yorkshire and Northern
Region). No information provided in paper on reporting accuracy of these registries.
269 population controls identified from District health authority birth registers and
matched to cases on age, sex, and region of residency at time of case diagnosis.
Participation rates- 72% of cases (n = 109) and 77% of controls (n = 206).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Childhood leukemia incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
No information provided in published paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including Face-to-face interviews of mothers of cases and controls using structured
adoption of JEM and quantitative exposure questionnaire were administered to obtain information on general risk factors and
estimates	potential confounders. Questionnaire also sought information on a maternal and
paternal complete job history, from first employment to end of study and included for
job title, dates of employment, and industry. Questionnaire administered to both
parents, and, if one parent was unavailable, information was provided by proxy.
Questionnaire also sought information on 22 specific chemicals. Expert assignment
of occupation based upon National classification system. Statistical analyses
industry of employment, job or occupation, and specific exposures.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
No, face-to-face interview with 72% of case parents and 77% of control parents.
Blinded interviewers
Face-to-face interviews were not blinded. Expert assignment of occupation was
carried out blinded.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No information provided in paper on percentage of proxy interviews.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Exposure prevalence to TCE—maternal exposure, 2 cases (2%) and 3 controls (2%);
paternal exposure, 9 cases (9%) and 7 controls (4%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Cases and control matched on age, sex, and region of residency at time of case
diagnosis.
Statistical methods
Discordant pair analysis.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Limited reporting of odds ratios for job title and occupations.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
B.3.2.7.4. Lowengart et al. (1987)
B.3.2.7.4.1. Author's abstract.
A case-control study of children of ages 10 years and under in Los Angeles
County was conducted to investigate the causes of leukemia. The mothers and
fathers of acute leukemia cases and their individually matched controls were
interviewed regarding specific occupational and home exposures as well as other
potential risk factors associated with leukemia. Analysis of the information from
the 123 matched pairs showed an increased risk of leukemia for children whose
fathers had occupational exposure after the birth of the child to chlorinated
solvents [odds ratio (OR) = 3.5, P = .01], spray paint (OR = 2.0, P = .02), dyes or
pigments (OR = 4.5, P = .03), methyl ethyl ketone (CAS: 78-93-3; OR = 3.0, P =
.05), and cutting oil (OR = 1.7, P = .05) or whose fathers were exposed during the
mother's pregnancy with the child to spray paint (OR = 2.2, P = .03). For all of
these, the risk associated with frequent use was greater than for infrequent use.
There was an increased risk of leukemia for the child if the father worked in
industries manufacturing transportation equipment (mostly aircraft) (OR = 2.5, P
= .03) or machinery (OR = 3.0, P = .02). An increased risk was found for children
whose parents used pesticides in the home (OR = 3 .8, P = .004) or garden (OR =
6.5, P = .007) or who burned incense in the home (OR = 2.7, P = .007). The risk
was greater for frequent use. Risk of leukemia was related to mothers'
employment in personal service industries (OR = 2.7, P = .04) but not to specified
occupational exposures. Risk related to fathers' exposure to chlorinated solvents,
employment in the transportation equipment-manufacturing industry, and parents'
exposure to household or garden pesticides and incense remains statistically
significant after adjusting for the other significant findings.
B.3.2.7.4.2. Study description and comment. Self-assessed parental exposure to chemical
classes and to individual chlorinated solvents was assigned in this case-control study of
leukemia in children 10 years or younger using information obtained through telephone
interviews with mothers and fathers of cases and controls. Interviews were carried out for
79% of case mothers (159 or 202 cases) and 81% (124 of 154) case fathers. The number of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-347 DRAFT—DO NOT CITE OR QUOTE
-------
potential controls was not identified in the paper, although it was reported that interviews
were carried out for 136 referent mothers and 87 referent fathers. Mothers served as
proxy respondents for paternal exposures in roughly 20% of cases and 30% of controls.
The complete occupational history was sought for the period 1 year before the case
diagnosis date, if the case was older than 2 years, 6 months before the diagnosis date, if the
case was between the ages of 1 and 2 years, and the same as the date of diagnosis of the case
was <1 year old. Questions on specific occupational exposures such as solvents or
degreasers, metals, and other categories were included on the questionnaire, with self-
reported information used to assign exposure potential. Exposure is defined only as a
dichotomous variable (yes/no). In this study using a matched-pair design in the statistical
analyses, there were six case-control pairs of paternal cases but not controls and 3 case-
control pairs with paternal controls but not cases with TCE exposure before pregnancy or
during pregnancy. Few mothers reported exposure to chlorinated solvents. A strength of
the study is the ability to examine exposure at a number of developmental periods,
preconception, during pregnancy, and postnatal. Misclassification bias is likely strong in
this study, introduced through the large number of proxy respondents and exposure
assessment based upon self-reported information. Misclassification resulting from proxy
information will dampen observed risks, where as, misclassification of self-reported
exposures may bias observed risks in either direction. For this reason and because of the
low prevalence of exposure nature of exposure assessment approach, this study provides
little information on childhood leukemia risks and TCE exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-348 DRAFT—DO NOT CITE OR QUOTE
-------
Lowengart RA, Peters JM, Cicioni C, Buckley J, Bernstein L, Preston-Martin S, Rappaport E. (1987). Childhood leukemia
and parents' occupational and home exposures. JNCI 79:39-46.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This case-control study of children <10 yrs of age was conducted to identify possible
risk factors of childhood leukemia. TCE exposure was one of many occupational
exposures assessed in this study.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
202 cases of acute lymphatic leukemia in children <10 yrs of age at time of diagnosis
from 1980 through 1984 were identified from the Los Angeles County Cancer
Surveillance Program, a population-based cancer registry. Controls were identified
from among friends of cases with additional controls selected using random digit
dialing from the same population as cases and were matched to cases on age, sex,
race, and Hispanic origin.
123 cases (61% response rate) and 123 controls (not able to calculate response rate
since number of possible controls not identified in paper).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Telephone questionnaire sought information on maternal and paternal preconception,
pregnancy, and postnatal (up to 1 yr before case diagnosis) exposures, including a
full occupational history (job title, employers, and dates of employments) and on the
child's exposure from birth to 1 yr before case diagnosis. Parents also provide self-
reported information on specific exposures or occupational activities. Occupations
grouped according to hydrocarbon exposure potential using definition of Zack et al.
(1980).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Telephone interview with 159 of 202 (79%) case mothers and 124 of 202 case fathers
(61%>). Of controls, interviews were obtained from 136 mothers (65 friends of cases,
71 population controls) and 87 fathers.
Blinded interviewers
Unblinded interviews.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 19%o of paternal exposure information on cases was provided by the mother. 43
of 130 control mothers provided information on paternal exposures (33%>).
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Paternal TCE exposure—
1 yr before pregnancy, 1/0 discordant pairs
During pregnancy, 6/3 discordant pairs
After delivery 8/3 discordant pairs.
No information is provided in paper on maternal TCE exposure.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, race, and Hispanic origin.
Statistical methods
Discordant pair analysis.
-------

§•
to
s
a, 5
o a
TO 3
s
o
o
s
Co
I?
f? Ng
^ 3
>9
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
8
TO
Co
0
S
1
a.
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
B.3.2.8. Melanoma Case-Control Studies
B.3.2.8.1. Fritschi and Siemiatycki (1996b), Siemiatycki (199Siemiatvcki, 19911).
B.3.2.8.1.1. Author's abstract.
OBJECTIVES: Associations between occupational exposures and the occurrence
of cutaneous melanoma were examined as part of a large population based case-
control study of 19 cancer sites. METHODS: Cases were men aged 35 to 70 years
old, resident in Montreal, Canada, with a new histologically confirmed cutaneous
melanoma (n = 103). There were two control groups, a randomly selected
population control group (n = 533), and a cancer control group (n = 533)
randomly selected from among subjects with other types of cancer in the large
study. Odds ratios for the occurrence of melanoma were calculated for each
exposure circumstance for which there were more than four exposed cases (85
substances, 13 occupations, and 20 industries) adjusting for age, ethnicity, and
number of years of schooling. RESULTS: Significantly increased risk of
melanoma was found for exposure to four substances (fabric dust, plastic dust,
trichloroethylene, and a group containing paints used on surfaces other than metal
and varnishes used on surfaces other than wood), three occupations (warehouse
clerks, salesmen, and miners and quarrymen), and two industries (clothing and
non-metallic mineral products). CONCLUSIONS: Most of the occupational
circumstances examined were not associated with melanoma, nor is there any
strong evidence from previous research that any of those are risk factors. For the
few occupational circumstances which were associated in our data with
melanoma, the statistical evidence was weak, and there is little or no supporting
evidence in the scientific literature. On the whole, there is no persuasive evidence
of occupational risk factors for melanoma, but the studies have been too small or
have involved too much misclassification of exposure for this conclusion to be
definitive.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-353 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3.2.8.1.2. Study description and comment. Fritschi and Siemiatycki (1996b) and
Siemiatycki (1991) reported data from a case-control study of occupational exposures and
melanoma conducted in Montreal, Quebec (Canada) and part of a larger study of 10 other
site-specific cancers and occupational exposures. The investigators identified 124 newly
diagnosed cases of melanoma (ICD-0,172), confirmed on the basis of histology reports,
between 1979 and 1985; 103 of these participated in the study interview (83.1%
participation). One control group (n = 533) consisted of patients with other forms of
cancer recruited through the same study procedures and time period as the melanoma
cancer cases. A population-based control group (n = 533, 72% response), frequency
matched by age strata, was drawn using electoral lists and random digit dialing. Face-to-
face interviews were carried out with 82% of all cancer cases with telephone interview
(10%) or mailed questionnaire (8%) for the remaining cases. Twenty percent of all case
interviews were provided by proxy respondents. The occupational assessment consisted of
a detailed description of each job held during the working lifetime, including the company,
products, nature of work at site, job activities, and any additional information that could
furnish clues about exposure from the interviews.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Fritschi and
Siemiatycki (1996b) present observations of logistic regression analyses examining industries,
occupation, and some chemical-specific exposures, but not TCE. Observations on TCE from
Mantel-Haenszel analyses are found in the original report of Siemiatycki (1991). Any exposure
to TCE was 6% among cases (n = 8) and 4% for substantial TCE exposure (n = 4); "substantial"
is defined as >10 years of exposure for the period up to 5 years before diagnosis.
Logistic regression models adjusted for age, ethnic origin, socioeconomic status, Quetlet
as an index of body mass, and respondent status (Fritschi and Siemiatycki, 1996b) or
Mantel-Haenszel % stratified on age, family income, cigarette smoking, Quetlet, ethnic origin,
and respondent status (Siemiatycki, 1991). Odds ratios for TCE exposure are presented with
90% confidence intervals in Siemiatycki (1991) and 95% confidence intervals in Fritschi and
Siemiatycki (1996b).
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of melanoma. However, the use of
the general population (rather than a known cohort of exposed workers) reduced the likelihood
that subjects were exposed to TCE, resulting in relatively low statistical power for the analysis.
The job exposure matrix, applied to the job information, was very broad since it was used to
evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-3 54 DRAFT—DO NOT CITE OR QUOTE
-------
Fritschi L, Siemiatycki J. (1996b). Melanoma and occupation: Results of a case-control study. 1996. Occup Environ Med
53:168-173.
Siemiatycki J. (1991). Risk Factors for Cancer in the Workplace. J Siemiatycki, Ed. Boca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
124 melanoma cases were identified among male Montreal residents between 1979
and 1985 of which 103 were interviewed.
740 eligible male controls identified from the same source population using random
digit dialing or electoral lists; 533 were interviewed. A second control series
consisted of other cancer cases identified in the larger study (n = 533).
Participation rate: cases, 83.1%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 172 (Malignant neoplasm of skin).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 294 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
I
<§>
>S
TO
*
^s
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were conducted
either at home or in the hospital; all population control interviews were conducted at
home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
99 cases (16.1% response), 533 population controls (12%).
Exposure prevalence: Any TCE exposure, 8%> cases (n = 8); Substantial TCE
exposure (Exposure for >10 yrs and up to 5 yrs before disease onset), 4% cases
(n = 4).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, education, and ethnic origin (Fritschi and Siemiatycki, 1996b).
Age, family income, cigarette smoking, and ethnic origin (Siemiatycki, 1991).
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Fritschi and Siemiatycki, 1996b).
-------
a,	s
o	a
to	5
Co
<5	Co*
a
8	^
3	^
£.	>
s	?
^	5
!>q	S.
(3	TO
^	^3
^	5
s.	^
^	o
• TO	Co
V;	TO
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
B.3.2.9. Pancreatic Cancer Case-Control Studies
B.3.2.9.1. Kernan et al. (1999).
B.3.2.9.1.1. Author's abstract.
Background The relation between occupational exposure and pancreatic cancer is
not well established. A population-based case-control study based on death
certificates from 24 U.S. states was conducted to determine if occupations/
industries or work-related exposures to solvents were associated with pancreatic
cancer death.
Methods The cases were 63,097 persons who died from pancreatic cancer
occurring in the period 1984±1993. The controls were 252,386 persons who died
from causes other than cancer in the same time period.
Results Industries associated with significantly increased risk of pancreatic cancer
included printing and paper manufacturing; chemical, petroleum, and related
processing; transport, communication, and public service; wholesale and retail
trades; and medical and other health-related services. Occupations associated with
significantly increased risk included managerial, administrative, and other
professional occupations; technical occupations; and sales, clerical, and other
administrative support occupations.
Potential exposures to formaldehyde and other solvents were assessed by using a
job exposure matrix developed for this study. Occupational exposure to
formaldehyde was associated with a moderately increased risk of pancreatic
cancer, with ORs of 1.2, 1.2, 1.4 for subjects with low, medium, and high
probabilities of exposure and 1.2, 1.2, and 1.1 for subjects with low, medium, and
high intensity of exposure, respectively.
Conclusions The findings of this study did not suggest that industrial or
occupational exposure is a major contributor to the etiology of pancreatic cancer.
Further study may be needed to confirm the positive association between
formaldehyde exposure and pancreatic cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-359 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
B.3.2.9.1.2. Study description and comment. Kernan et al. (1999) reported data from a
case-control study of occupational exposures and pancreatic cancer, coding usual
occupation as noted on death certificates to assign potential TCE exposure to cases and
controls. Deaths from pancreatic cancer from 1984-1993 were identified from 24 U. S.
state and frequency-matched to nonpancreatitis or other pancreatic disease deaths by state,
race, sex, and age (5-year groups); 63,097 pancreatic cancer deaths (case series) and
252,386 controls were selected for analysis.
Exposure assessment in this study group occupational (n = 509) and industry (n = 231)
codes into 16 broad occupational and 20 industrial categories. Additionally, a job exposure
matrix (JEM) of Gomez et al. (1994) was applied to develop exposure surrogates for
11 chlorinated hydrocarbons, including TCE, and two larger groupings, all chlorinated
hydrocarbons and organic solvents. A qualitative surrogate (ever exposed/never exposed) for
TCE exposure is developed and no information is provided on death certifications on
employment duration to examine exposure-response patterns. Kernan et al. (1999) report
mortality odds ratios from logistic regression for TCE exposure intensity and probability of
exposure.
Overall, this is a large study that examined specific exposures using a generic JEM.
Errors resulting from exposure misclassification are likely, not only introduced by the generic
JEM, but through the use of usual occupation as coded on death certificates, which may not fully
represent an entire occupational history.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-360 DRAFT—DO NOT CITE OR QUOTE
-------
Kernan GJ, Ji B-T, Dosemeci M, Silverman DT, Balbus J, Zahm SH. (1999). Occupational risk factors for pancreatic cancer:
A case-control study based on death certificates from 24 U. S. states. Am J Ind Med 36:260-270.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between pancreatic cancers and occupational title or chemical exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
63,097 pancreatic cancer cases were identified using death certificates from 24 U. S.
states between 1984 and 1993.
63,097 noncancer, nonpancreatitis or other pancreatic disease deaths (controls)
identified from the same source population and frequency-matched to cases by state,
race, sex, and age (1:4 matching).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-9, 157 (Malignant neoplasm of pancreas).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Usual occupation coded on death certificate coded to 1980 U. S. census classification
system for occupation and industry. 509 occupation codes and 231 industry codes
grouped into 16 broad occupational and 20 industrial categories based on similarity
of occupational exposures. Job exposure matrix of Gomez et al. (1994) used to
assign exposure surrogates for 11 chlorinated hydrocarbons, including TCE, and 2
broad categories, chlorinated hydrocarbons and organic solvents.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
This study did not use interviews, information reported on death certificate used to
infer potential exposure.
Blinded interviewers
No interviews were conducted in this study.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
Exposure prevalence: Any TCE exposure (Low intensity exposure or higher), 14%
cases (n = 9,068); High TCE exposure, 2% cases (n = 1,271).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, metropolitan status, region of residence, and martial status.
Statistical methods
Logistic regression.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
B.3.2.10. Prostatic Cancer Case-Control Studies
B.3 .2.10.1. Aronson et al. (1996), Siemiatycki (1991).
B.3.2.10.1.1. Author's abstract.
A population-based case-control study of cancer and occupation was carried out
in Montreal, Canada. Between 1979 and 1986, 449 pathologically confirmed
cases of prostate cancer were interviewed, as well as 1,550 cancer controls and
533 population controls. Job histories were evaluated by a team of
chemist/hygienists using a checklist of 294 workplace chemicals. After
preliminary evaluation, 17 occupations, 11 industries, and 27 substances were
selected for multivariate logistic regression analyses to estimate the odds ratio
between each occupational circumstance and prostate cancer with control for
potential confounders. There was moderate support for risk due to the following
occupations: electrical power workers, water transport workers, aircraft
fabricators, metal product fabricators, structural metal erectors, and railway
transport workers. The following substances exhibited moderately strong
associations: metallic dust, liquid fuel combustion products, lubricating oils and
greases, and polyaromatic hydrocarbons from coal. While the population
attributable risk, estimated at between 12% and 21% for these occupational
exposures, may be an overestimate due to our method of analysis, even if the true
attributable fraction were in the range of 5—10%, this represents an important
public health issue.
B.3.2.10.1.2. Study description and comment. Aronson et al. (1996) and Siemiatycki
(1991) reported data from a case-control study of occupational exposures and prostate
cancer conducted in Montreal, Quebec (Canada) and was part of a larger study of 10 other
site-specific cancers and occupational exposures. The investigators identified 557 newly
diagnosed cases of prostate cancer (ICD-0,185), confirmed on the basis of histology
reports, between 1979 and 1985; 449 of these participated in the study interview (80.6%
participation). One control group consisted of patients with other forms of cancer
recruited through the same study procedures and time period as the prostate cancer cases.
A population-based control group (n = 533, 72% response), frequency matched by age
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-364 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
strata, was drawn using electoral lists and random digit dialing. Face-to-face interviews
were carried out with 82% of all cancer cases with telephone interview (10%) or mailed
questionnaire (8%) for the remaining cases. Twenty percent of all case interviews were
provided by proxy respondents. The occupational assessment consisted of a detailed
description of each job held during the working lifetime, including the company, products,
nature of work at site, job activities, and any additional information that could furnish
clues about exposure from the interviews.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Aronson et al.
(1996) presents observations of logistic regression analyses examining industries, occupation,
and some chemical-specific exposures, but not TCE. Observations on TCE from Mantel-
Haenszel analyses are found in the original report of Siemiatycki (1991). Any exposure to TCE
was 2% among cases (n= 11) and <2% for substantial TCE exposure (n = 7); "substantial" is
defined as >10 years of exposure for the period up to 5 years before diagnosis.
Logistic regression models adjusted for age, education, and ethnicity (AroSiemiatycki,
1991nson et al., 1996) or Mantel-Haenszel % stratified on age, family income, cigarette smoking,
coffee, and ethnic origin (Siemiatycki, 1991). Odds ratios for TCE exposure are presented with
90% confidence intervals in Siemiatycki (1991) and 95% confidence intervals in Aronson et al.
(1996).
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of prostate cancer. However, the use
of the general population (rather than a known cohort of exposed workers) reduced the likelihood
that subjects were exposed to TCE, resulting in relatively low statistical power for the analysis.
The job exposure matrix, applied to the job information, was very broad since it was used to
evaluate 294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-365 DRAFT—DO NOT CITE OR QUOTE
-------
Aronson KJ, Siemiatycki J, Dewar R, Gerin M. (1996). Occupational risk factors for prostate cancer: Results from a case-
control study in Montreal, Canada. Am J Epidemiol 143:363-373.
Siemitycki J. (1991). Risk Factors for Cancer in the Workplace. J Siemiatycki, Ed. Boca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
557 prostate cancer cases were identified among male Montreal residents between
1979 and 1985 of which 449 were interviewed.
740 eligible male controls identified from the same source population using random
digit dialing or electoral lists; 533 were interviewed. A second control series
consisted of other cancer cases identified in the larger study.
Participation rate: cases, 83.1%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 185 (Malignant neoplasm of prostate).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 294 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
s.
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%>
telephone interview, and 8%> mailed questionnaire. Cases interviews were conducted
either at home or in the hospital; all population control interviews were conducted at
home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
449 cases (80.6%> response), 533 population controls (72%).
Exposure prevalence: Any TCE exposure, 2% cases (n= 11); Substantial TCE
exposure (Exposure for >10 yrs and up to 5 yrs before disease onset), <2% cases
(n = 7).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, ethnic origin, socioeconomic status, Quetlet as an index of body mass, and
respondent status (Aronson et al., 1996).
Age, family income, cigarette smoking, ethnic origin, and respondent status
(Siemiatycki, 1991).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
s.
3'
Co
a
I"
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Aronson et al., 1996).
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
B.3.2.11. Renal Cell Carcinoma Case-Control Studies—Arnsberg Region of Germany
A series of studies (including Henschler et al. (1995). discussed in cohort study section)
have been conducted in an area with a long history of trichloroethylene use in several industries.
The main importance of these studies is that there is considerable detail on the nature of
exposures, which made it possible to estimate the order of magnitude of exposure even though
there were no direct measurements.
B.3 .2.11.1. Briining et al. (2003).
B.3.2.11.1.1. Author's abstract.
BACKGROUND: German studies of high exposure prevalence have been
debated on the renal carcinogenicity of trichloroethylene (TRI). METHODS: A
consecutive hospital-based case-control study with 134 renal cell cancer (RCC)
cases and 401 controls was conducted to reevaluate the risk of TRI in this region
which were estimated in a previous study. Exposure was self-assessed to compare
these studies. Additionally, the job history was analyzed, using expert-based
exposure information. RESULTS: The logistic regression results, adjusted for
age, gender, and smoking, confirmed a TRI-related RCC risk in this region. Using
the database CAREX for a comparison of industries with and without TRI
exposure, a significant excess risk was estimated for the longest held job in TRI-
exposing industries (odds ratio (OR) 1.80, 95% confidence interval (CI) 1.01-
3.20). Any exposure in "metal degreasing" was a RCC risk factor (OR 5.57, 95%
CI 2.33-13.32). Self-reported narcotic symptoms, indicative of peak exposures,
were associated with an excess risk (OR 3.71, 95% CI 1.80-7.54).
CONCLUSIONS: The study supports the human nephrocarcinogenicity of
tri chl oroethyl ene.
B.3.2.11.1.2. Study description and comment. This study is a second case-control follow-
up of renal cell cancer in the Arnsberg area of Germany, which was intended to deal with
some of the methodological issues present in the two earlier studies. The major advantage
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-370 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
of studies in the Arnsberg area is the high prevalence of exposure to trichloroethylene
because of the large number of companies doing the same kind of industrial work. An
interview questionnaire procedure for self-assessment of exposures similar to the one used
by Vamvakas et al. (1998) was used to obtain detailed information about solvents used, job
tasks, and working conditions, as well as the occurrence of neurological symptoms. The
industry and job title information in the subjects' job histories were also analyzed by two
schemes of expert-rated exposure assignments for broad groups of jobs. The CAREX
database from the European Union, for industry categories, and the British JEM developed
by Pannett et al. (1985), for potential exposure to chemical classes or specific chemical, but
not TCE, was adopted in an attempt to obtain a potentially less biased assessment of
exposures.
Exposure prevalences for employment in industries with potential TCE and
perchloroethylene exposures was high in both cases (87%) and controls (79%) using the CAREX
approach but much lower using the JEM approach for potential exposure to degreasing agents
(12% cases, 9% controls), self-reported exposure to TCE (18% cases, 10% controls), and TCE
exposure with any symptom occurrence (14% cases, 4% controls). Both the CAREX and British
JEM rating approaches are very broad and they have potentially high rates of misclassification of
exposure intensity in job groupings and industry groupings. In an attempt to avoid reporting
biases associated with the legal proceeding for compensation, analyses were conducted on
self-reported exposure to selected agents (yes or no). The regional use of trichloroethylene and
perchloroethylene (tetrachloroethylene) were so widespread that most individuals recognized the
local abbreviations. If individuals claimed to be exposed when they were not, it would reduce
the finding of a relationship if one existed. Similarly, subjects were grouped by frequency of
perceived symptoms (any, less than daily, daily) associated with TCE or perchloroethylene
exposure. Overreporting would also introduce misclassification and reduce evidence of any
relationship. Self-reporting of exposure to chemicals in case-control studies, generally, is
considered unreliable since, within the broad population, workers rarely know specific chemicals
to which they have potential exposure. However, in cohort studies and case-control studies in
which one industry dominates a local population such as in this study, this is less likely because
the numbers of possible industries and job titles are much smaller than in a broad population.
The Arnsberg area studies focused on a small area where one type of industry was very
prevalent, and that industry used primarily just two solvents: trichloroethylene and
tetrachloroethylene. As a result, it was common knowledge among the workers what solvent an
individual was using, and, for most, it was trichloroethylene. Self-reported TCE exposure is
considered to be less biased compared to possible misclassification bias associated with using the
CAREX exposure assessment approach which identified approximately 90% of all cases as
holding a job in an industry using TCE or perchloroethylene (see above discussion).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-371 DRAFT—DO NOT CITE OR QUOTE
-------
1	Some subjects in Briining et al. (2003) are drawn from the underlying Arnsberg
2	population as studied by Vamvakas et al. (1998) (reviewed below) and TCE exposures to these
3	subjects would be similar—substantial, sustained high exposures to TCE at 400-600 ppm during
4	hot dip cleaning and greater than 100 ppm overall. However, the larger ascertainment area
5	outside the Arnsberg region for case and control identification may have resulted in a lower
6	exposure prevalence compared to Vamvakas et al. (1998).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-372 DRAFT—DO NOT CITE OR QUOTE
-------
Briining T, Pesch B, Wiesenhiitter B, Rabstein S, Lammert M, Baumiiller A, Bolt H. (2003). Renal cell cancer risk and
occupational exposure to trichloroethylene: results of a consecutive case-control study in Arnsberg, Germany. Am J Ind Med
23:274-285.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
From abstract—study aim was to "reevaluate the risk of TRI in this region which
were estimated in a previous study."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
162 renal cell carcinoma cases identified from September 1999 to April 2000 and
who had undergone nephrectomy between 1992 and 2000 [a time period preceding
that adopted in Vamvakas et al., (1998)] from a regional hospital urology department
in Arnsberg, Germany; 134 of the recruited cases were interviewed. 401 hospital
controls were interviewed between 1999 and 2000 from local surgery departments or
geriatric departments and frequency matched to cases by sex and age.
134 of 162 (83%) cases; response rate among controls could not be calculated
lacking information on the number of eligible controls.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
N/A
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Face-to-face interview with subjects or their next of kin using a structured
questionnaire with questions to obtain information on a complete job history by job
title, supplemental information on job tasks with suspected exposure to specific
agents, medical history, and personal habits. Questionnaires also sought
self-reported information on duration and frequency of exposure to TCE and
perchloroethylene, and, for these individuals, frequency of narcotic symptoms as a
marker of high peak exposure.
Jobs titles were coded according to a British classification of occupations and
industries with potential chemical-specific exposures identified for each occupation
using CAREX, a carcinogen exposure database or the British job-exposure matrix of
Pannett et al. (198Siemiatycki, 19915) for chemical groupings (e.g., degreasing
agents, organic solvents).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
100%o of cases or their NOK and 100%> controls with face-to-face interviews.
Blinded interviewers
No information on whether interviewers were blinded.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 17%o of case interviews with next-of-kin; all controls were alive at time of
interview.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
<&
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
CAREX Job-exposure-matrix
117 cases with TCE exposure (87% exposure prevalence among cases).
316 cases with TCE exposure (79% exposure prevalence among controls).
Self-reoorted TCE exposure
25 cases with TCE exposure (18% exposure prevalence among cases).
38 cases with TCE exposure (9.5% exposure prevalence among controls).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and tobacco smoking.
Statistical methods
Conditional logistic regression.
Exposure-response analysis presented in
published paper
Yes, duration of exposure as 4 categories (no, <10 yrs, 10-<20 years, and 20+ yrs).
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
B.3.2.11.2. Pesch et al. (2000b).
B.3.2.11.2.1. Author's abstract.
BACKGROUND: This case-control study was conducted to estimate the renal
cell cancer (RCC) risk for exposure to occupation-related agents, besides other
suspected risk factors. METHODS: In a population-based multicentre study, 935
incident RCC cases and 4298 controls matched for region, sex, and age were
interviewed between 1991 and 1995 for their occupational history and lifestyle
habits. Agent-specific exposure was expert-rated with two job-exposure matrices
and a job task-exposure matrix. Conditional logistic regression was used to
calculate smoking adjusted odds ratios (OR). RESULTS: Very long exposures in
the chemical, rubber, and printing industries were associated with risk for RCC.
Males considered as 'substantially exposed to organic solvents' showed a
significant excess risk (OR = 1.6, 95% CI: 1.1-2.3). In females substantial
exposure to solvents was also a significant risk factor (OR = 2.1, 95% CI: 1.0-
4.4). Excess risks were shown for high exposure to cadmium (OR = 1.4, 95% CI:
1.1-1.8, in men, OR = 2.5, 95% CI: 1.2-5.3 in women), for substantial exposure
to lead (OR = 1.5, 95% CI: 1.0-2.3, in men, OR = 2.6, 95% CI: 1.2-5.5, in
women) and to solder fumes (OR = 1.5, 95% CI: 1.0-2.4, in men). In females, an
excess risk for the task 'soldering, welding, milling' was found (OR = 3.0, 95% CI
: 1.1-7.8). Exposure to paints, mineral oils, cutting fluids, benzene, polycyclic
aromatic hydrocarbons, and asbestos showed an association with RCC
development.
CONCLUSIONS: Our results indicate that substantial exposure to metals and
solvents may be nephrocarcinogenic. There is evidence for a gender-specific
susceptibility of the kidneys.
B.3 .2.11.2.2. Study description and comment. This multicenter study of renal cell
carcinoma and bladder cancer and in Germany, which included the Arnsberg region plus
four others, identified two case series from participating hospitals, 1,035 urothelial cancer
cases and 935 renal cell carcinoma cases with a single population control series matched to
cases by region, sex, and age (1:2 matching ratio to urothelial cancer cases and 1:4
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-376 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
matching ratio to renal cell carcinoma cases). A strength of the study was the high
percentage of interviews with renal cell carcinoma cases within 2 months of diagnosis
(88.5%), reducing bias associated with proxy or next-of-kin interview, and few cases
diagnoses confirmed by sonography only (5%). In all, 935 (570 males, 365 females) renal
cell carcinoma cases were interviewed face-to-face with a structured questionnaire.
Two general JEMs, British and German, were used to assign exposures based on
subjects' job histories reported in an interview. Researchers also asked about job tasks
associated with exposure, such as metal degreasing and cleaning, and use of specific agents
(organic solvents chlorinated solvents, including specific questions about carbon tetrachloride,
trichloroethylene, and tetrachloroethylene) to evaluate TCE potential using a JTEM. A category
of "any use of a solvent" mixes the large number with infrequent slight contact with the few
noted earlier who have high intensity and prolonged contact. Analyses examining
trichloroethylene exposure using either the JEM of JTEM assigned a cumulative TCE exposure
index of none to low, medium high and substantial, defined as the product of exposure
probability x intensity x duration with the following cutpoints: none to low, <30th percentile of
cumulative exposure scores; medium, 30th-<60th percentile; high, 601h-<90th percentile; and,
substantial, >90th percentile. The use of the German JEM identified approximately twice as
many cases with any potential TCE exposure (42%) compared to the JTEM (17%) and, in both
cases, few cases identified with substantial exposure, 6% by JEM and 3% by JTEM. Pesch et al.
(2000b) noted "exposure indices derived from an expert rating of job tasks can have a higher
agent-specificity than indices derived from job titles." For this reason, the JTEM approach with
consideration of job tasks is considered as a more robust exposure metric for examining TCE
exposure and renal cell carcinoma due to likely reduced potential for exposure misclassification
compared to TCE assignment using only job history and title.
While this case-control study includes the Arnsberg area, several other regions are
included as well, where the source of the trichloroethylene and chlorinated solvent exposures are
much less well defined. Few cases were identified as having substantial exposure to TCE and, as
a result, most subjects identified as exposed to trichloroethylene probably had minimal contact,
averaging concentrations of about 10 ppm or less (NRC, 2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-377 DRAFT—DO NOT CITE OR QUOTE
-------
Pesch B, Haerting J, Ranft U, Klimpet A, Oelschagel, Schill W, and the MURC Study Group. (2000b). Occupational risk
factors for renal cell carcinoma: agent-specific results from a case-control study in Germany. Int J Epidemiol 29:1014-1024.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This case-control study was conducted to estimate RCC risk for exposure to
occupational-related agents; chlorinated solvents including trichloroethylene were
identified as exposures of a priori interest.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
935 RCC cases were identified from hospitals in a five-region area in Germany
between 1991 and 1995. Cases were confirmed histologically (95%) or by
sonography (5%) and selected without age restriction. 4,298 population controls
identified from local residency registries in the five-region area were frequency
matched to cases by region, sex, and age.
Participation rate: cases, 88%; controls, 71%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
N/A
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
A trained interviewer interviewed subjects using a structured questionnaire which
covered occupational history and job title for all jobs held longer than 1 yr, medical
history, and personal information. Two general JEMs, British and German, were
used to assign exposures based on subjects' job histories reported in an interview.
Researchers also asked about job tasks associated with exposure, such as metal
degreasing and cleaning, and use of specific agents (organic solvents chlorinated
solvents, including specific questions about carbon tetrachloride, trichloroethylene,
and tetrachloroethylene) and chemical-specific exposure were assigned using a
JTEM. Exposure index for each subject is the sum over all jobs of duration x
probability x intensity. A four category grouping was used in statistical analyses
defined by exposure index distribution of controls: no-low; medium, 30th percentile;
high, 60th percentile; substantial, 90th percentile.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Interviewers carried out face-to-face interview with all cases and controls. All cases
were interviewed in the hospital; 88.5%> of cases were interviewed within 2 mos after
diagnosis. All controls had home interviews.
Blinded interviewers
No , by nature of interview location.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No.
-------
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
JEM: 391 cases with TCE exposure index of medium or higher (42% exposure
prevalence among cases).
JTEM: 172 cases with TCE exposure index of medium or higher (18% exposure
prevalence among cases).
No information is presented in paper on control exposure prevalence.
-------
>3
o
o
s
Co
I?
S"4
>}'
§•
TO
s
a
to
s
S"
a
a,
>r
<§>
>s
to
*
^s
«
s
TO
Co
0
s
1
a.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, study center, and smoking.
Statistical methods
Conditional logistic regression.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes.
VO
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
B.3.2.11.3. Vamvakas et al. (1998).
B.3.2.11.3.1. Author's abstract.
A previous cohort-study in a cardboard factory demonstrated that high and
prolonged occupational exposure to trichloroethene (C2HC13) is associated with
an increased incidence of renal cell cancer. The present hospital-based
case/control study investigates occupational exposure in 58 patients with renal
cell cancer with special emphasis on C2HC13 and the structurally and
toxicologically closely related compound tetrachloroethene (C2C14). A group of
84 patients from the accident wards of three general hospitals in the same area
served as controls. Of the 58 cases, 19 had histories of occupational C2HC13
exposure for at least 2 years and none had been exposed to C2C14; of the 84
controls, 5 had been occupationally exposed to C2HC13 and 2 to C2C14. After
adjustment for other risk factors, such as age, obesity, high blood pressure,
smoking and chronic intake of diuretics, the study demonstrates an association of
renal cell cancer with long-term exposure to C2HC13 (odds ratio 10.80; 95% CI:
3.36-34.75).
B.3.2.11.3.2. Study description and comment. In a follow-up to Henschler et al. (1995)
(discussed below), a case-control study was conducted in the Arnsberg region of Germany
where there has long been a high prevalence of small enterprises manufacturing small
metal parts and goods, such as nuts, lamps, screws, and bolts. Both cases and controls were
identified from hospital records; cases from of a large regional hospital in North Rhine
Wetphalia during the period 1987 and 1992 and controls who were admitted to accident
wards during 1993 at three other regional hospitals. Control selection was carried out
independent of cases demographic risk factors, i.e., controls were not matched to cases.
Controls may not be fully representative of the case series (NRC, 2006); they were selected
from a time period after case selection which may introduce bias if TCE use changes over
time resulted in decreased potential for exposure among controls, and use of accident ward
patients may be representative of the target population.
Exposures to TCE resulted from dipping metal pieces into vats, with room temperatures
up to 60°C, and placing the wet parts on tables to dry. Some work rooms were noted to be small
and poorly ventilated. These conditions are likely to result in high inhalation exposure to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-3 82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
trichloroethylene (100-500 ppm). Cherrie et al. (2001) estimated the long-term exposures to be
approximately 100 ppm. Some of the cases included in this study were also pending legal
compensation. As a result, there had been considerable investigation of the exposure situation by
occupational hygienists from the Employer's Liability Insurance Association and occupational
physicians, including walk-through visits and interviews of long-term employees. The legal
action could introduce a bias, a tendency to overreport some of the subjective reports by the
subjects. However, the objective working conditions were assessed by knowledgeable
professionals, who corroborated the presence of the poorly controlled hot dip tanks, extensive
use of trichloroethylene for all types of cleaning, and the process descriptions.
NRC (2006) discussed a number of criticisms in the literature on Vamvakas et al. (1998)
by Green and Lash (1999), Cherrie et al. (2001), and Mandel (2001) and noted the direction of
possible bias would be positive or negative depending on the specific criticism. Overall, cases in
this study substantial, sustained exposures to high concentrations of trichloroethylene at
400-600 ppm during hot dip cleaning and greater than 100 ppm overall and observations can
inform hazard identification although the magnitude of observed association is uncertain give
possible biases.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-3 83 DRAFT—DO NOT CITE OR QUOTE
-------
Vamvakas S, Briining T, Thomasson B, Lammert M, Baumiiller A, Bolt HM, Dekant W, Birner G, Henschler D, Ulm K.
(1998). Renal cell cancer risk and occupational exposure to trichloroethylene: results of a consecutive case-control study in
Arnsberg, Germany. Am J Ind Med 23:274-285.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes. From introduction—study aim was designed to investigate further the role of
occupation exposure to TCE/perchloroethylene in the formation of renal cancer.
Selection and characterization in cohort studies
of exposure and control groups and of cases
and controls in case-control studies is adequate
73 renal cell carcinoma cases that had undergone nephrectomy between December
1987 and May 1992 from a hospital urology department in Arnsberg, Germany were
contacted by mail; 58 of the recruited cases were. 112 controls identified from
accident wards of three area hospitals were interviewed during 1993. Controls
underwent abdominal sonography to exclude kidney cancer.
62 of 73 (85%) cases and 84 of 112 (75%) of controls participated in study.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
N/A
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Face-to-face interview with subjects or, if deceased, with their next of kin or former
colleagues using a structured questionnaire with questions to obtain information on job
tasks with selected exposure to specific agents and to self-reported selected exposures.
A supplemental questionnaire on job conditions was administered to subjects reporting
exposure to TCE and perchloroethylene. Subjects with TCE exposures were primarily
exposed through degreasing operations in small businesses. Self-reported TCE
exposure was ranked using a semiquantitative scale based upon total exposure time and
frequency/duration of self-reported acute prenarcotic symptoms. Cherrie et al. (2001)
estimated that the machine cleaning exposures to trichloroethylene were
-400-600 ppm, with long-term average TCE exposure as -100 ppm.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Personal physicians interviewed 100% of cases or their NOK/former colleague and
100%) controls.
Blinded interviewers
Interviewers were not blinded nor was developments of exposure assessment
semiquantitative scale.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No information provided in paper on number of cases with NOK interviews or
interviews with former colleagues; all controls were alive and interviewed by their
personal physician.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
19 cases with TCE or perchloroethylene exposure (33% exposure prevalence) and
1 control with perchloroethylene exposure.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq S.
^ ^3
^ 5
S. ^
^ o
• TO Co
V; to
Co
>r
<§>
>S
VO
to
o
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, obesity, high blood pressure, smoking, and diuretic use.
Statistical methods
Mantel-Haenszel % .
Exposure-response analysis presented in
published paper
Yes, semiquantitative scale of 4 categories (no, +, ++, +++).
Documentation of results
No information on number of eligible controls or number interviews with case NOK or
former colleagues.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
B.3.2.12. Renal Cell Carcinoma Case-Control Studies—Arve Valley Region of France
A case-control study was conducted in the Arve Valley to examine the a priori
hypothesis of an association with renal cell carcinoma and trichloroethylene exposure. The Arve
Valley, like the Arnsburg Region in Germany, has a long history of trichloroethylene use in the
screw-cutting industry. The Arve Valley, situated in the Rhone-Alpes region of eastern France is
a major metalworking sector with around 800 small and medium-sized firms specializing in
"screw-cutting" or the machining of small mechanical parts from bars, in small, medium, and
large series on conventional automatic lathes or by digital control. This industry evolved around
the time of World War I from the region's expertise in clock-making. A major point of this
study is that it was designed as a follow-up study to the German renal cell cancer case-control
studies but in a different population with similar exposure patterns and with high prevalence of
exposure to trichloroethylene. For this reason, there is considerable detail on the nature of
exposure, which made it possible to estimate the order of magnitude of exposure, even though
there were not direct measurements.
B.3.2.12.1. Charbotel et al.(2009), Charbotel et al. (2007) Charbotel et al. (2006).
B.3.2.12.1.1. Charbotel et al. (2009) abstract.
Abstract Background- Several studies have investigated the association between
trichloroethylene (TCE) exposure and renal cell cancer (RCC) but findings were
inconsistent. The analysis of a case control study has shown an increased risk of
RCC among subjects exposed to high cumulative exposure. The aim of this
complementary analysis is to assess the relevance of current exposure limits
regarding a potential carcinogenic effect of TCE on kidney.
Methods- Eighty-six cases and 316 controls matched for age and gender were
included in the study. Successive jobs and working circumstances were described
using a detailed occupational questionnaire. An average level of exposure to TCE
was attributed to each job period in turn. The main occupational exposures
described in the literature as increasing the risk of RCC were assessed as well as
non-occupational factors. A conditional logistic regression was performed to test
the association between TCE and RCC risk. Three exposure levels were studied
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-388 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
(average exposure during the eight-hour shift): 35 ppm, 50 ppm and 75 ppm.
Potential confounding factors identified were taken into account at the threshold
limit of 10% ( p = 0.10) (body mass index [BMI], tobacco smoking, occupational
exposures to cutting fluids and to other oils).
Results- Adjusted for tobacco smoking and BMI, the odd-ratios associated with
exposure to TCE were respectively 1.62 [0.77-3.42], 2.80 [1.12-7.03] and 2.92
[0.85-10.09] at the thresholds of 35 ppm, 50 ppm and 75 ppm. Among subjects
exposed to cutting fluids and TCE over 50 ppm, the OR adjusted for BMI,
tobacco smoking and exposure to other oils was 2.70 [1.02-7.17],
Conclusion- Results from the present study as well as those provided in the
international literature suggest that current French occupational exposure limits
for TCE are too high regarding a possible risk of RCC.
B.3.2.12.1.2. Charbotel et al. (2007) abstract.
Background: We investigated the association between exposure to
trichloroethylene (TCE) and mutations in the von Hippel-Lindau (VHL) gene and
the subsequent risk for renal cell carcinoma (RCC).
Methods: Cases were recruited from a case-control study previously carried out in
France that suggested an association between exposures to high levels of TCE and
increased risk of RCC. From 87 cases of RCC recruited for the epidemiological
study, 69 were included in the present study. All samples were evaluated by a
pathologist in order to identify the histological subtype and then be able to focus
on clear cell RCC. The majority of the tumor samples were fixed either in
formalin or Bouin's solutions. The majority of the tumors were of the clear cell
RCC subtype (48 including 2 cystic RCC). Mutation screening of the 3 VHL
coding exons was carried out. A descriptive analysis was performed to compare
exposed and non exposed cases of clear cell RCC in terms of prevalence of
mutations in both groups.
Results: In the 48 cases of RCC, four VHL mutations were detected: within exon
1 (c.332G>A, p.Serl 11 Asn), at the exon 2 splice site (c.463+lG>C and
c.463+2T>C) and within exon 3 (c.506T>C, p.Leul69Pro). No difference was
observed regarding the frequency of mutations in exposed versus unexposed
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-3 89 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
groups: among the clear cell RCC, 25 had been exposed to TCE and 23 had no
history of occupational exposure to TCE. Two patients with a mutation were
identified in each group.
Conclusion: This study does not confirm the association between the number and
type of VHL gene mutations and exposure to TCE previously described.
B.3 .2.12.1.3. Charbotel et al. (2006) abstract.
Background: We investigated the association between exposure to
trichloroethylene (TCE) and mutations in the von Hippel-Lindau (VHL) gene and
the subsequent risk for renal cell carcinoma (RCC).
Methods: Cases were recruited from a case-control study previously carried out in
France that suggested an association between exposures to high levels of TCE and
increased risk of RCC. From 87 cases of RCC recruited for the epidemiological
study, 69 were included in the present study. All samples were evaluated by a
pathologist in order to identify the histological subtype and then be able to focus
on clear cell RCC. The majority of the tumor samples were fixed either in
formalin or Bouin's solutions. The majority of the tumors were of the clear cell
RCC subtype (48 including 2 cystic RCC). Mutation screening of the 3 VHL
coding exons was carried out. A descriptive analysis was performed to compare
exposed and non-exposed cases of clear cell RCC in terms of prevalence of
mutations in both groups.
Results: In the 48 cases of RCC, four VHL mutations were detected: within exon
1 (c.332G>A, p.Serl 11 Asn), at the exon 2 splice site (c.463+lG>C and
c.463+2T>C) and within exon 3 (c.506T>C, p.Leul69Pro). No difference was
observed regarding the frequency of mutations in exposed versus unexposed
groups: among the clear cell RCC, 25 had been exposed to TCE and 23 had no
history of occupational exposure to TCE. Two patients with a mutation were
identified in each group.
Conclusion: This study does not confirm the association between the number and
type of VHL gene mutations and exposure to TCE previously described.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-390 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
To test the effect of the exposure to trichloroethylene (TCE) on renal cell cancer
(RCC) risk, a case-control study was performed in the Arve Valley (France), a
geographic area with a high frequency and a high degree of such exposure. Cases
and controls were selected from various sources: local general practitioners and
urologists practicing in the area and physicians (urologists and oncologists) from
other hospitals of the region who might treat patients from this area. Blinded
telephone interviews with cases and controls were administered by a single
trained interviewer using occupational and medical questionnaires. The analysis
concerned 86 cases and 316 controls matched for age and gender. Three
approaches were developed to assess the link between TCE exposure and RCC:
exposure to TCE for at least one job period (minimum 1 year), cumulative dose
number of ppm of TCE per job period multiplied by the number of years in the
job period) and the effect of exposure to peaks. Multivariate analysis was
performed taking into account potential confounding factors. Allowing for
tobacco smoking and Body Mass Index, a significantly 2-fold increased risk was
identified for high cumulative doses: odds ratio (OR) = 2.16 (1.02-4.60). A dose-
response relationship was identified, as was a peak effect; the adjusted OR for
highest class of exposure-plus-peak being 2.73 (1.06-7.07). After adjusting for
exposure to cutting fluids the ORs, although still high, were not significant
because of lack of power. This study suggests an association between exposures
to high levels of TCE and increased risk of RCC. Further epidemiological studies
are necessary to analyze the effect of lower levels of exposure.
B.3.2.12.1.4. Study description and comment. Cases in the population-based case-control
study were obtained retrospectively from regional medical practitioners or from teaching
hospitals from 1993 to 2002, and prospectively from 2002 to mid-2003. One case was
excluded from analysis because it was not possible to find a control subject. Controls were
either selected from the same urology practice as cases or, for cases selected from teaching
hospitals, from among patients of the case's general practitioner. Telephone interviews of
87 renal cell carcinoma cases and 316 controls matched for age and sex by a trained
interviewer were used to obtain information on occupational and medical history for the
case-control analysis of Charbotel et al. (2006). Of the 87 RCC cases, 67 cases provided
consent for mutational analysis of which 48 cases were diagnosed with clear cell RCC,
suitable for mutational analysis of the von Hippel Lindau (VHL) gene (Charbotel et al.,
2007). Tissue samples were paraffin-embedded or frozen tissues and ability to fully
sequence the VHL gene depended on type of the fixative procedure; only 26 clear cell RCC
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-391 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
cases (34% of 73 clear cell RCC cases in the case-control study) could full sequencing of the
VHL gene occur.
Two occupational questionnaires were administered to both cases and controls, a
questionnaire developed specifically to evaluate jobs and exposure potential in the screw-cutting
industry and a more general one for any other jobs. Interviewers were essentially blinded to
subject status as case or control for the occupational questionnaires given the medical
questionnaire was administered afterwards (Fevotte et al., 2006). The medical questionnaire
included familial kidney disease and medical history, body mass index, and history of smoking.
A task/TCE-Exposure Matrix was designed using information obtained from questionnaires and
routine atmospheric monitoring of work shops or biological monitoring (U-TCA) of workers
carried out since the 1960s. Questionnaires were used to elicit from each subject the main tasks
associated with each job, working conditions, activities or jobs that might involve TCE
exposures and possible exposure to other occupational risk factors for renal cell carcinoma.
The JEM linked to corresponding TCE-exposure levels using available industrial hygiene
monitoring data on atmospheric TCE levels and from biological measurement on workers.
Estimates reflected task duration, use of protective equipment and distance from TCE source, as
well, as both dermal and inhalation exposure routes. Estimated TCE intensities for jobs
associated with open cold degreasing were 15-18 ppm, 120 ppm for jobs working near open hot
degreasing machines, with up to 300 ppm for work directly above tank and for job and intensities
of 300 to 600 ppm for emptying, cleaning and refilling degreasers. Eight local physicians with
knowledge of working conditions corroborated the working conditions for individual job periods
after 1980 in screw-cutting shops. Overall, there was good agreement (72%) between physician
and the JEM. Three exposure surrogates were assigned to each case and control: time-weighted-
average exposure (Charbotel et al., 2009), cumulative exposure (Charbotel et al., 2006), and
cumulative exposure with and without peak exposure (Charbotel et al., 2006).
An 8-hour time-weighted average (TWA) exposure concentration was developed for each
job period from 1924 to 2003 and was the product of the task-specific estimated TCE intensity
and duration of task. A subject's lifetime 8-hour TWA was the sum of each job period specific
estimated TWA. Exposure peak, daily exposure reaching >200 ppm for at least 15 minutes, was
assessed as an additive factor and was defined by frequency (seldom exposed, few times yearly
to frequently exposure, few time weekly).
Over the study period, 19% (295 of 1,486) job periods were assessed as having TCE
exposure with an 8-hour TWA of less than 35 ppm for 72% of exposed jobs and >75 ppm for 5%
of exposed jobs. Exposure prevalence to TCE peaked in the 1970s with roughly 20% of job
periods with TCE exposure and 8% of subjects identified with >75 ppm. By the 1990s, exposure
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-392 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
prevalence had not only decreased to 7% but also exposure intensity, only 5% of job periods
with >75 ppm.
Cumulative TCE exposure was the sum of 8-hour TWAs over all job periods with
statistical analysis using four categories: no, low, medium, and high. These were defined as low,
5-150 ppm-years; medium, 155-335 ppm-year; and high, >335 ppm-years (HSIA, 2005).
Analyses were also carried out examining peak exposure, classified as yes/no and without
assignment of quantitative level, as additional exposure to average TCE concentration;
33 subjects were exposed to peaks and very few to high peaks.
The high exposure prevalence and strong approach for exposure assessment provides
Charbotel et al. (2006; 2009) more statistical power and ability to assess association of renal cell
carcinoma and TCE exposure. However, the low participation rate, inability to fully sequence
the VHL gene in all clear cell RCC cases, the lower background prevalence of mutations (15% in
this study compared to roughly 50% in other series) in Charbotel et al. (2007) suggest a relative
insensitivity of assay used and lack of a positive control limits the mutational analysis. These
methodological limitations introduce bias with greater uncertainties for evaluating consistency of
findings with somatic VHL mutations observed in other TCE-exposed RCC cases (Brauch et al.,
1999; Briining et al., 1997b). TCE exposure prevalence (>5 ppm-year) in Charbotel et al. (2006)
was 43% among cases and is higher than that observed in other population-based case-control
studies of renal cell carcinoma and TCE (e.g., Pesch et al., 2000a). While some subjects had
jobs with exposures to high concentrations of TCE during the 1970s and 1980s, a large
percentage of jobs were to TCE concentrations of less than 35 ppm (8-hour TWA). Jobs with
high TCE concentrations also were identified as having frequent exposure to peak TCE
concentrations, particularly before 1980. Peak TCE estimates in this study were judged to be
lower than those in German studies of the Arnsberg region (Henschler et al., 1995; Vamvakas et
al., 1998) but higher than those of Hill Air Force Base civilian workers (Blair et al., 1998;
Stewart et al., 1991) due to a lower frequency of degreasing tasks in Blair et al. (1998) cohort
and to slower technological changes in degreasing process in the French case-control study
(Fevotte et al., 2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-393 DRAFT—DO NOT CITE OR QUOTE
-------
Charbotel B, Fevotte J, martin JL, Bergeret A. (2009). Cancer du rein et expositions au trichloroethylene: les valeurs limites
d'exposition professionnelle fra?aises en vigueur sont-elles adaptees. Rev Epidemiol Sante Publique 57:41-47.
Charbotel B, Fevotte J, Hours M, Martin J-L, Bergeret A. (2006). Case-control study on renal cell cancer and occupational
exposure to trichloroethylene. Part II: Epidemiological Aspects. Ann Occup Hyg 50:777-787.
Fevotte J, Charbotel B, Muller-Beaute P, Martin J-L, Hours, Bergeret A. (2006). Case-control study on renal cell cancer and
occupational exposure to trichloroethylene. Part I: Exposure assessment. Ann Occup Hyg 50:765-775.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes. From abstract—study aim was to "test the effect of TCE exposure on renal cell
cancer."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
117 cases of renal cell carcinoma patients were identified retrospectively from 1993
to June 2002, and prospectively from June 2002 to June 2003 from patients of
urology practices and hospital urology and oncology departments in the region of
Arve Valley, France. 404 controls were identified from the same urology practice or
from the same general practitioner, for cases identified from hospital records and
matched on residency in the geographic study area at time of case diagnosis, sex, and
year of birth. Controls sought medical treatment for conditions other than kidney or
bladder cancer. Case definition included clear cell and other subtypes of renal cell
carcinoma including chromophil, chromophobe and collecting duct carcinomas.
87 or 117 (74%) cases and 316 of 404 (78%) controls participated in study.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
N/A
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Occupational questionnaires sought information for each study subject a complete
job history and was followed-up with either a questionnaire specific for jobs and
exposures in the screw-cutting industry or a General Occupational Questionnaire,
which ever was more applicable to subject. Questionnaires also sought self-reported
information on potential TCE exposures. A medical questionnaire seeking
information on medical history and familial kidney disease was administered after
occupational questionnaires.
Jobs titles were coded according to standardized classification of occupations and
1,486 job periods grouped into 3 categories (screw-cutting, nonscrew-cutting but job
with possible TCE exposure, and no TCE exposure). An estimated 8-hour TWA was
assigned to each job and job period using a job-task-exposure matrix.
RCC and TCE was examined using three exposure approaches: exposure to at least
5 ppm for at least one job period (minimum 1 yr), cumulative dose or £ (TCE ppm
per job x years) using quantitative ranking levels (no exposure, low, medium, and
high), and potential for peak defined as any exposure 200+ ppm. TCE
concentrations associated with quantitative ranking are low, 5-150 ppm-yrs;
medium, 155-335 ppm-yrs; high, >335 ppm-yrs.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
<90% face-to-face
Telephone interviews were conducted by a trained interviewer.
Blinded interviewers
The paper notes interviewers were blinded "as far as possible" since medical
questionnaire was administered after the occupational questionnaires.
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
Yes, 22% of cases were dead at time of interview compared to 7% of controls.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
3E
to
o
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
37 cases with TCE exposure (43% exposure prevalence), 110 controls with TCE
exposure (35% exposure prevalence).
16 cases with high level confidence TCE exposure (27% exposure prevalence),
37 controls with high level confidence TCE exposure (16%).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, tobacco smoking and body mass index (Charbotel et al., 2006).
Age, sex tobacco smoking, body mass index, and exposure to cutting or petroleum
oils (Charbotel et al., 2009).
Statistical methods
Conditional logistic regression on matched pairs.
Exposure-response analysis presented in
published paper
Yes, cumulative exposure as 4 categories (no, low, medium and high exposure) and
cumulative exposure plus peaks.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
B.3.2.13. Renal Cell Carcinoma Case-Control Studies in Other Regions
B.3.2.13.1. Moore et al. (2010)
B.3.2.13.1.1. Author's abstract.
Trichloroethylene (TCE) is a suspected renal carcinogen. TCE-associated renal
genotoxicity occurs predominantly through glutathione S-transferase (GST)
conjugation and bioactivation by renal cysteine beta-lyase (CCBL1). We
conducted a case-control study in Central Europe (1,097 cases and 1,476 controls)
specifically designed to assess risk associated with occupational exposure to TCE
through analysis of detailed job histories. All jobs were coded for
organic/chlorinated solvent and TCE exposure (ever/never) as well as the
frequency and intensity of exposure based on detailed occupational
questionnaires, specialized questionnaires, and expert assessments. Increased risk
was observed among subjects ever TCE exposed [odds ratio (OR) = 1.63; 95%
confidence interval (95% CI), 1.04-2.54], Exposure-response trends were
observed among subjects above and below the median exposure [average intensity
(OR = 1.38; 95% CI, 0.81-2.35; OR = 2.34; 95% CI, 1.05-5.21; P(trend) = 0.02)].
A significant association was found among TCE-exposed subjects with at least
one intact GSTT1 allele (active genotype; OR = 1.88; 95% CI, 1.06-3.33) but not
among subjects with two deleted alleles (null genotype; OR = 0.93; 95% CI, 0.35-
2.44; P(interaction) = 0.18). Similar associations for all exposure metrics
including average intensity were observed among GSTT1-active subjects (OR =
1.56; 95% CI, 0.79-3.10; OR = 2.77; 95% CI, 1.01-7.58; P(trend) = 0.02) but not
among GSTT1 nulls (OR = 0.81; 95% CI, 0.24-2.72; OR =1.16; 95% CI, 0.27-
5.04; P(trend) = 1.00; P(interaction) = 0.34). Further evidence of heterogeneity
was seen among TCE-exposed subjects with >or=l minor allele of several
CCBLl-tagging single nucleotide polymorphisms: rs2293968, rs2280841,
rs2259043, and rs941960. These findings provide the strongest evidence to date
that TCE exposure is associated with increased renal cancer risk, particularly
among individuals carrying polymorphisms in genes that are important in the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-399 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
reductive metabolism of this chemical, and provides biological plausibility of the
association in humans.
B.3.2.13.1.2. Study description and comment.
The hospital case-control study of kidney cancer in men and women who were residents in areas
of the sevens tudy centers evaluated nonoccupational and occupational risk factors and included
a detailed exposure assessment for chloroinated organic solvents, including TCE.
Histologically-confirmed incident cases of renal cell carcinoma (ICD-O-2, Code C.64) between
20 and 79 years of age and diagnosed between 1999 and 2003 at Severn participating hospitals
were eligible as cases, with hospital in-patient or out-patient controls admitted to the same
hospital centers but with non-tobacco-related conditions, excluding genitourinary cancers, and
frequency matched to cases by sex and age, and by study center. The final study population
ieluded 1,097 cases and 1,476 controls for a participation rate, depending on study center of 90 -
98% and 90 - 96% for cases and controls, respectively. As part of the study, blood samples
obtained from 925 cases and 1,192 controls were assayed for deletion of the GSTT1
polymorphism and genetic variation across the renal cyctein P-lyase (CCBL1) gene.
Face-to-face interviews were conducted using standard questionnaires which asked about
life-style habits and personal, familial medical history, and for each job held >1 year. For
specific jobs or industries with likely exposure to know or suspected occupational carcinogens of
interest, a specialized occupation questionnaires were used to gather more detailed information.
For every job in a subject's work history, an exposure assessment team from each center, with
extensive knowledge of industries in the region and blinded to case or control status, evalauted
the frequency and intensity of exposure to organic and chlorinated solvents based on the general
and job-specific questionnaires. The general category of aliphatic chlorinate organic solvents
included PCE, methylene chloride, carbon tetrachloride, 1,1, 1-trichloroethane, and TCE.
Subjects identifed as exposed to organic solvents were reevaluated by the team at a later date to
confirm assignment as an attempt to reduce exposure misclassification. The reevaluation was
performed blinded to case and controls status. For each exposed job, the frequency, intensity and
confidence of exposure to TCE, organic solvents, and chlorinated solvents. While TCE exposure
was correlated with both chlorinated solvents and organic solvents exposure, it was not
associated with other coexposures. Exposure frequency was coded into three categories,
represeinting the average percentage of a working day exposure was likely (1-4.9%, 5-30%,
>30%>), with midpoint weights for cumulative exposure calculations of 0.-025, 0.175, and 0.50,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-400 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
respectively, and assuming a log-normal exposure distribution. TCE intensity was also coded
into three categories (0-<5 ppm), 5-50 ppm, >50 ppm) with midpoint weights for cumulative
exposure calculations of 2.5, 25, and 75 ppm, respectively. Exposure surrogates developed
included cumulative exposure, the product of the midpoints for intensity and frequency and
multiplied by duration. Average exposure intensity was a second exposure surrogate and defined
as the quotient of cumulative exposure and duration. Last, confidence of exposure that
represtend the expected percentage of workers that would be exposed in that job was categorized
as possible (<40%), probale (40-89%), or definite (>90%). Among subjects with probable
exposure (high confidence TCE exposure), the median intensity score was 0.076 ppm [25th and
75th percentile range among cases, 0.83 - 7.25 ppm] and median cumulative exposure scores
were 1.58 (25th and 75th percentiles, 0.77 - 2.87 ppm-year) and 1.95 ppm-years (25th and 75th
percentiles, 0.83 - 7.25 ppm-year) among cases and controls, respectively.
Assocation between renal cell carcinoma and organic solvents, chlorinated solvents and
TCE exposure for jobs with any confidence level and for holding a job with probable or definite
exposure was assessed using unconditional logistic regression to estimate ORs and 95%
confidence intervals. All statistical models included covarates for sex, age, and study center.
Analyses were also modeled to account for a 20-year lag. Almost all TCE exposure occurred at
least 20 years before renal cell carcinoma onset and Moore et al. (2010) did not report these
findings as odds ratio estimates were similar to those from the models using untagged exposure
surrogate.
The strong expsosure approach in Moore et al. (2010) and examination of exposure
probability or confidence are strengths of the study. TCE used did not appear widespread as
exposure prevalence was low, 6 % of cases had held a job of any exposure probability, compared
to 29% of cases identified with any exposure to organic solvents. The percentage of cases was
even lower, 4%, for higher confidence TCE exposure. Additionally, evaluation of GST
polymorphisms provides assessment of susceptibility factors.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-401 DRAFT—DO NOT CITE OR QUOTE
-------
1	Moore LE, Buffetta P, Karami S, Brennan P, Stewart PS, Hung R, et al. (2010). Occupational trichloroethylene exposure and
2	renal carcinoma risk: Evidence of genetic susceptibility by reductive metabolism gene variants. Cancer Res 20:6527-6536.
3
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-402 DRAFT—DO NOT CITE OR QUOTE
-------
2.
Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or hypothesis
Study hypotheses of investigating risk association with occupation TCE exposure and kidney
(excluding pelvis) cancers through analysis of job histories and use of detailed exposure
assessment method.
Selection and characterization in cohort studies of
exposure and control groups and of cases and
controls in case-control studies is adequate
Cases: 1,097 histologically-confirmed RCC cases in males and females, 20-79 years of age,
1999-2003, identified through 7 hospital centers in 4 countries (Czech Republic, Polant,
Romania, Russia).
Controls: 1,476 in-patient or out-patient hospital controls admistted to same hospital as case
with nontobacco-related conditions and frequency matched to cases by sex and age, and by
study center.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Renal cell carcinoma incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's lymphoma
ICD-0-2 [Codes C.54],
CATEGORY C: TCE-EXPOSURE CRITERIA
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-403 DRAFT—DO NOT CITE OR QUOTE
-------
Exposure assessment approach, including adoption
of JEM and quantitative exposure estimates
Job-specific questionnaire for job > year. Exposure assessment team from each
center with knowledge of region's industries to assess frequency, intensity and
confidence of exposure to TCE and organic solvent group (PCE, methylene chloride,
carbon tetrachloride, and 1,1, 1-trichloroethane). Exposure surrogates of frequency
(3 categories based on percentage of day), intensity (3 groups), cumulative exposure
(product of intensity, duration, frequency), and average exposure intensity
(cumulative exposure score divided by the number of years exposed). Exposure
confidence score (possible, probably, definite) defined as percentage of workers
exposed at a job.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
In-person interview using questionnaire.
Blinded interviewers
No information in published paper if interviewers were blinded Exposure assessment
assigned blinded.
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
No proxy interviews.
CATEGORY G: SAMPLE SIZE
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-404 DRAFT—DO NOT CITE OR QUOTE
-------
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies; numbers
of exposed cases and prevalence of exposure in
case-control studies
Cases: 90 - 99% participation rate; Controls: 90 - 96% participation rate.
Exposure prevalence, ever exposed to TCE (6% of cases holding TCE job, anv
confidence level; 4% of cases with probable or definite exposure).

CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and center Place of residence, tobacco smoking,
body mass index, and hypertension also examined but did not
alder odds ratio estimate by >10%, and thus, were not included
in final models.
Statistical methods
Unconditional logistic regression.
Exposure-response
analysis presented in
published paper
Test for trend reported for years, hours, cumulative and average
intensity fo exposure.
Documentation of results
Yes, study was well documented with supplemental material
available on publisher's webpage.
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
B-405 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
B.3.2.13.2. Parent et al. (2000a), Siemiatycki (1991).
B.3.2.13.2.1. Author's abstract.
BACKGROUND: Little is known about the role of workplace exposures on the
risk of renal cell cancer. METHODS: A population-based case-control study was
undertaken in Montreal to assess the association between hundreds of
occupational circumstances and several cancer sites, including the kidney. A total
of 142 male patients with pathologically confirmed renal cell carcinoma, 1900
controls with cancer at other sites and 533 population-based controls were
interviewed. Detailed job histories and relevant data on potential confounders
were obtained. A group of chemists-hygienists evaluated each job reported and
translated them into a history of occupational exposures using a checklist of 294
substances. Multivariate logistic regression models using either population, cancer
controls, or a pool of both groups were used to estimate odds ratios. RESULTS:
There were some indications of excess risks among printers, nursery workers
(gardening), aircraft mechanics, farmers, and horticulturists, as well as in the
following industries: printing-related services, defense services, wholesale trade,
and retail trade. Notwithstanding the low precision of many of the odds ratio
estimates, the following workplace exposures showed some evidence of excess
risk: chromium compounds, chromium (VI) compounds, inorganic acid solutions,
styrene-butadiene rubber, ozone, hydrogen sulphide, ultraviolet radiation, hair
dust, felt dust, jet fuel engine emissions, jet fuel, aviation gasoline, phosphoric
acid and inks. CONCLUSIONS: For most of these associations there exist no, or
very little, previous data. Some associations provide suggestive evidence for
further studies.
B.3.2.13.2.2. Study description and comment. This population case-control study of
histologically-confirmed kidney cancer among males who resided in the Montreal
Metropolitan area relies on the use of expert assessment of occupational information on a
detailed questionnaire and face-to-face interview and was part of a larger study of 10 other
site-specific cancers and occupational exposures (Parent et al., 2000a; Siemiatycki, 1991).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-406 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Interviewers were unblinded, although exposure assignment was carried out blinded as to
case and control status. The questionnaire sought information on the subject's complete
job history and included questions about the specific job of the employee and work
environment. Occupations considered with possible TCE exposure included machinists,
aircraft mechanics, and industrial equipment mechanics. An additional specialized
questionnaire was developed for certain job title of a prior interest that sought more
detailed information on tasks and possible exposures. For example, the supplemental
questionnaire for machinists included a question on TCE usage. A team of industrial
hygienists and chemicals assigned exposures blinded based on job title and other
information obtained by questionnaire. A semiquantitative scale was developed for
300 exposures and included TCE (any, substantial). Parent et al. (2000a) presents
observations of analyses examining job title, occupation, and some chemical-specific
exposures, but not TCE. Observations on TCE are found in the original report of
Siemiatycki (1991). Any exposure to TCE was 3% among cases but <1% for substantial
TCE exposure; "substantial" is defined as >10 years of exposure for the period up to 5
years before diagnosis. The TCE exposure frequencies in this study are lower than those in
Briining et al. (2003) and Charbotel et al. (2006), studies conducted in geographical areas
with a high prevalence of industries using TCE. The expert assessment method is
considered a valid and reliable approach for assessing occupational exposure in
community-base studies and likely less biased from exposure misclassification than
exposure assessment based solely on self-reported information (Fritschi et al., 2003; IOM,
2003; Siemiatycki et al., 1987). For example, Dewar et al. (1994) examine sensitivity of
JEM of Siemiatycki et al. (1987) to exposure assessment by chemists and industrial
hygienists using interview information and evaluation of job histories. Specific solvents are
not examined, although, a sensitive 84% and specificity of 97% was found for the JEM for
general solvent exposure.
This population study of several cancer sites included histologically-confirmed cases of
kidney cancer (ICD-0 189, malignant neoplasm of kidney and other and unspecified urinary
organs) ascertained from 16 Montreal-area hospitals between 1979 and 1985. A total of
227 eligible kidney cancer cases were identified were identified from 19 Montreal-area hospitals;
177 cases participated in the study (78% response). One control group (n = 1,295) consisted of
patients with other forms of cancer (excluding lung cancer and other intestinal cancers) recruited
through the same study procedures and time period as the rectal cancer cases. A
population-based control group (n = 533), frequency matched by age strata, was drawn using
electoral lists and random digit dialing. All controls were interviewed using face-to-face
methods; however, 20 % of the all cancer cases in the larger study were either too ill to interview
or had died and, for these cases, occupational information was provided by a proxy respondent.
The quality of interview conducted with proxy respondents was much lower, increasing the
potential for misclassification bias, than that with the subject. The direction of this bias would
diminish observed risk towards the null.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-407 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Statistical analysis are considered valid; logistic regression model which included terms
for respondent status, age, smoking and body mass index in Parent et al. (2000a) and
Mantel-Haenszel % stratified on age, family income, cigarette smoking, and ethic origin in
Siemiatycki (1991). Odds ratios are presented with 90% confidence intervals in Siemiatycki
(1991) and 95% confidence intervals in Parent et al. (2000a).
Overall, exposure assessment in this study adopted a superior approach, using expert
knowledge and use of a job-exposure matrix. However, examination of NHL and TCE exposure
is limited by statistical power considerations related to low exposure prevalence, particularly for
"substantial" exposure. For the exposure prevalence found in this study to TCE and for kidney
cancer, the minimum detectable odds ratio was 3.0 when p = 0.02 and a = 0.05 (one-sided). The
low statistical power to detect a doubling of risk and an increased possibility of misclassification
bias associated with case occupational histories resulting from proxy respondents suggests a
decreased sensitivity in this study for examining kidney cancer and TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-408 DRAFT—DO NOT CITE OR QUOTE
-------
Parent M-E, Hua Y, Siemiatycki J. (2000a). Occupational risk factors for renal cell carcinoma in Montreal. Am J Ind Med
38:609-618.
Siemiatycki J. (1991). Risk Factors for Cancer in the Workplace. Baca Raton: CRC Press.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical
exposures.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
277 kidney cancer cases were identified among male Montreal residents between
1979 and 1985 of which 177 (147 renal cell carcinomas) were interviewed.
740 male population controls were identified from the same source population using
random digit dialing; 533 were interviewed. A second control series consisted of all
other cancer controls excluding lung and bladder cancer cases.
Participation rate: cases, 78%; population controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD 189 (Malignant neoplasm of the kidney and other and unspecified urinary
organs) (Siemiatycki, 1991).
ICD 189.0, renal cell carcinoma (Parent et al., 2000a).
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Unblinded interview using questionnaire sought information on complete job history
with supplemental questionnaire for jobs of a priori interest (e.g., machinists,
painters). Team of chemist and industrial hygienist assigned exposure using job title
with a semiquantitative scale developed for 300 exposures, including TCE. For each
exposure, a 3-level ranking was used for concentration (low or background, medium,
high) and frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%;
and high, >30%).
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
100%o of cases and controls were interviewed face-to-face by a trained interviewer.
Cases interviews were conducted either at home or in the hospital; all population
control interviews were conducted at home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case
and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 16%o of cases, 13%> of population controls, and 22% of cancer controls had
proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
177 cases (78% response), 533 population controls (72%).
Exposure prevalence: Any TCE exposure, 2% cases; Substantial TCE exposure
(Exposure for >10 yrs and up to 5 yrs before disease onset), 1% cases.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, income, index for cigarette smoking (Siemiatycki, 1991).
Age, smoking, body mass index, and proxy status (Parent et al., 2000b).
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
Logistic regression (Parent et al., 2000a).
-------

§•
to
s
a, 5
o a
TO 3
s
o
o
s
Co
I?
f? Ng
^ 3
>9
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
8
TO
Co
0
S
1
a.
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
B.3.2.13.3. Dosemeci et al. (1999).
B.3.2.13.3.1. Author's abstract.
BACKGROUND: Organic solvents have been associated with renal cell cancer;
however, the risk by gender and type of solvents is nuclear. METHODS: We
evaluated the risk of renal cell carcinoma among men and women exposed to all
organic solvents-combined, all chlorinated aliphatic hydrocarbons (CAHC)-
combined, and nine individual CAHC using a priori job exposure matrices
developed by NCI in a population-based case-control study in Minnesota, U.S.
We interviewed 438 renal cell cancer cases (273 men and 165 women) and 687
controls (462 men and 225 women). RESULTS: Overall, 34% of male cases and
21% of female cases were exposed to organic solvents in general. The risk of
renal cell carcinoma was significantly elevated among women exposed to all
organic solvents combined (OR = 2.3; 95% CI = 1.3-4.2), to CAHC combined
(OR = 2.1; 95% CI = 1.1 -3.9), and to trichloroethylene (TCE) (OR = 2.0; 95% CI
= 1.0-4.0). Among men, no significant excess risk was observed among men
exposed to any of these nine individual CAHCs, all CAHCs-combined, or all
organic solvents-combined. DISCUSSION: These observed gender differences in
risk of renal cell carcinoma in relation to exposure to organic solvents may be
explained by chance based on small numbers, or by the differences in body fat
content, metabolic activity, the rate of elimination of xenobiotics from the body,
or by differences in the level of exposure between men and women, even though
they have the same job title.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-413 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
B.3.2.13.3.2. Study description and comment. Dosemeci et al. (1999) reported data from a
population-based case-control study of the association between occupation exposures and
renal cancer risk. The investigators identified newly diagnosed patients with histologically
confirmed renal cell carcinoma from the Minnesota Cancer Surveillance System from July
1,1988 to December 31,1990. The study was limited to white cases, and age and gender-
stratified controls were ascertained using random digit dialing (for subjects ages 20-64)
and from Medicare records (for subjects 65-85 years). Of the 796 cases and 796 controls
initially identified, 438 cases (273 men, 165 women) and 687 controls (462 men, 225 women)
with complete personal interviews were included in the occupational analysis.
Data were obtained using in-person interviews that included demographic variables,
residential history, diet, smoking habits, medical history, and drug use. The occupational history
included information about the most recent and usual industry and occupation (coded using the
standard industrial and occupation codes, Department of Commerce), job activities, hire and
termination dates, and full/part time status. A job exposure matrix developed by the National
Cancer Institute (Gomez et al., 1994) was used with the coded job data assign occupational
exposure potential for 10 chlorinated aromatic hydrocarbons and organic solvents, and includes
tri chl oroethyl ene.
Dosemeci et al. (1999) adopted logistic regression methods to evaluate renal cancer and
occupational exposures. Odds ratios were adjusted for age, smoking, hypertension, and use of
drugs for hypertension, and body mass index.
Strengths of this study include the use of incident cases of renal cancer from a defined
population area, with confirmation of the diagnosis using histology reports. The occupation
history was based on usual and most recent job, in combination with a relatively focused job
exposure matrix. In contrast to the type of exposure assessment that can be conducted in cohort
studies within a specific workplace, however, exposure measurements, based on personal or
workplace measurement, were not used, and a full lifetime job history was not obtained.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-414 DRAFT—DO NOT CITE OR QUOTE
-------
Dosemeci M, Cocco P, Chow W-H. (1999). Gender differences in risk of renal cell carcinoma and occupational exposures to
chlorinated aliphatic hydrocarbons. Am J Ind Med 36:54-59.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes. From abstract—study aim was to evaluate effect of organic solvents on RCC
risk using a priori job exposure matrices.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
796 white males and females identified through the Minnesota Cancer Surveillance
System with histological confirmed RCC between July 1, 1988 and December 31,
1990. Interviews were obtained for 690 subjects of which 241 were with next-of-kin
and excluded; 438 cases (273 males and 165 females) were included in analysis.
707 white population controls identified through random digit dialing, and matched to
cases, aged 20-65 yrs old, by age and sex using a stratified random sample or, for
cases aged 65-85, from Health Care Financing Administration list. 687 controls
(462 males and 225 females) are included in the analysis.
Participation rate: cases, 87%; controls, 86%.
Occupational analysis: cases, 55%, controls 83%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
N/A
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
s^
Sf
<§>
>S
TO
*
^s
«
s
TO
>;
0
s
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
A trained interviewer blinded to case and control status interviewed subjects at home
using a questionnaire which covered occupational, residential, and medical histories;
demographic information; and personal information. Occupational history included
self-reporting of the most recent job and usual occupation and industry, employment
dates, and focused on 13 specific occupations or industries.
Occupation and industry were coded according to a standard occupational
classification or standard industrial classification with potential chemical-specific
exposures to TCE and eight other chlorinated hydrocarbons identified using the job
exposure matrix of Dosemeci et al. (1999) and Gomez et al. (1994).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
All cases and controls had face-to-face interviews.
Blinded interviewers
Yes, interviewers were blinded as to case and control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
No, subjects with next-of-kin interviews were excluded from the analysis.
CATEGORY G: SAMPLE SIZE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
s^
Sf
<§>
>S
TO
*
^s
«
s
TO
1
#
I
a.
VO
to
o
Number of deaths in cohort mortality studies;
numbers of total cancers in incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
55 cases with TCE exposure (13% exposure prevalence among cases).
69 controls cases with TCE exposure (10% exposure prevalence among controls).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, smoking, body mass index, and hypertension/ use of diuretics/use of
anti-hypertension drugs.
Statistical methods
Logistic regression.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
B.3.2.14. Other Cancer Site Case-Control Studies
B.3 .2.14.1. Siemiatycki (1991), Siemiatycki et al. (1987).
B.3.2.14.1.1. Author's abstract.
A multi-cancer site, multi-factor, case-referent study was undertaken to generate
hypotheses about possible occupational carcinogens. About 20 types of cancer
were included. Incident cases among men aged 35-70 years and diagnosed in any
of the major Montreal hospitals were eligible. Probing interviews were carried out
for 3,726 eligible cases. The interview was designed to obtain detailed lifetime
job histories and information on potential confounders. Each job history was
reviewed by a team of chemists who translated it into a history of occupational
exposures. These occupational exposures were then analyzed as potential risk
factors in relation to the sites of cancer included. For each site of cancer
analyzed, referents were selected from among the other sites in the study. The
analysis was carried out in stages. First a Mantel-Haenszel analysis was
undertaken of all cancer-sub stance associations, stratifying on a limited number of
covariates, and, then, for those associations which were noteworthy in the initial
analysis, a logistic regression analysis was made taking into account all potential
confounders. This report describes the fieldwork and analytical methods.
B.3.2.14.1.2. Study description and comment. Siemiatycki (1991) reported data from a
case-control study of occupational exposures and several site-specific cancers, including
lung and pancreas, conducted in Montreal, Quebec (Canada). Other cases included in this
study were cancers of the bladder, colon, rectum, esophagus prostate, and lymphatic
system (NHL); a description of the other case series are found in other sections in this
appendix. The investigators identified 1,082 newly diagnosed cases of lung cancer (ICD-O,
162) and 165 newly diagnosed cases of pancreatic cancer (ICD-O, 157), confirmed on the
basis of histology reports, between 1979 and 1985; 857 lung cancer (79.2% ) and 117
pancreatic cancer cases (70.7%) participated in the study interview. One control group
consisted of patients with other forms of cancer recruited through the same study
procedures and time period as the melanoma cancer cases. The control series for lung
cancer cases excluded other lung cancer cases; the control series for pancreatic cancer
cases excluded all lung cancer cases. Additionally, a population-based control group
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-418 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
(n = 533, 72% response), frequency matched by age strata, was drawn using electoral lists
and random digit dialing. Face-to-face interviews were carried out with 82% of all cancer
cases with telephone interview (10%) or mailed questionnaire (8%) for the remaining
cases. Twenty percent of all case interviews were provided by proxy respondents. The
occupational assessment consisted of a detailed description of each job held during the
working lifetime, including the company, products, nature of work at site, job activities,
and any additional information that could furnish clues about exposure from the
interviews.
A team of industrial hygienists and chemists blinded to subject's disease status translated
jobs into potential exposure to 294 substances with three dimensions (degree of confidence that
exposure occurred, frequency of exposure, and concentration of exposure). Each of these
exposure dimensions was categorized into none, any, or substantial exposure. Any exposure to
TCE was 2% among cases (n = 21 lung cancer cases, 2 pancreatic cancer cases) and 1% for
substantial TCE exposure (n = 9 lung cancer cases); "substantial" is defined as >10 years of
exposure for the period up to 5 years before diagnosis. None of the pancreatic cancer cases was
identified with "substantial" exposure to TCE.
Mantel-Haenszel x analyses examined occupation exposures and lung cancer stratified
on age, family income, cigarette smoking, ethnic origin, alcohol consumption, and respondent
status or pancreatic cancer stratified on age, income, cigarette smoking, and respondent status
(Siemiatycki, 1991). Odds ratios for TCE exposure in Siemiatycki (1991) are presented with
90% confidence intervals.
The strengths of this study were the large number of incident cases, specific information
about job duties for all jobs held, and a definitive diagnosis of cancer. However, the use of the
general population (rather than a known cohort of exposed workers) reduced the likelihood that
subjects were exposed to TCE, resulting in relatively low statistical power for the analysis. The
job exposure matrix, applied to the job information, was very broad since it was used to evaluate
294 chemicals.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-419 DRAFT—DO NOT CITE OR QUOTE
-------
Siemitycki J. (1991). Risk Factors for Cancer in the Workplace. J Siemiatycki, Ed. Boca Raton: CRC Press.
Siemiatycki J, Wacholder S, Richardson L, Dewar R, Gerin M. (1987). Discovering carcinogens in the occupational
environment. Scand J Work Environ Health 13:486-492.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This population case-control study was designed to generate hypotheses on possible
association between 11 site-specific cancers and occupational title or chemical exposures.
Selection and characterization in cohort
studies of exposure and control groups and
of cases and controls in case-control
studies is adequate
1,082 lung cases were identified among male Montreal residents between 1979 and 1985
of which 857 were interviewed; 165 cases were identified among male Montreal residents
between 1979 and 1985 of which 117 were interviewed.
740 eligible male controls identified from the same source population using random digit
dialing or electoral lists; 533 were interviewed. A second control series consisted of other
cancer cases identified in the larger study.
Participation rate: lung cancer cases, 79.2 %, pancreatic cancer cases, 70.7%; population
controls, 72%.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-O, 122 (Malignant neoplasm of trachea, bronchus and lung).
ICD-O, 157 Malignant neoplasm of pancreas.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including Unblinded interview using questionnaire sought information on complete job history with
adoption of JEM and quantitative exposure supplemental questionnaire for jobs of a priori interest (e.g., machinists, painters). Team
estimates	of chemist and industrial hygienist assigned exposure using job title with a
semiquantitative scale developed for 294 exposures, including TCE. For each exposure, a
3-level ranking was used for concentration (low or background, medium, high) and
frequency (percent of working time: low, 1 to 5%; medium, >5 to 30%; and high, >30%).
-------

>3
o
o
s
Co
I?
"tS
3'
§•
to
£
a
to
s
Co
>T
<§>
>s
>S
TO
TO'
*
^S
S
s
TO
>;
0
s
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
82%o of all cancer cases interviewed face-to-face by a trained interviewer, 10%> telephone
interview, and 8%> mailed questionnaire. Cases interviews were conducted either at home
or in the hospital; all population control interviews were conducted at home.
Blinded interviewers
Interviews were unblinded but exposure coding was carried out blinded as to case and
control status.
CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents
Yes, 20%o of all cancer cases had proxy respondents.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality
studies; numbers of total cancer incidence
studies; numbers of exposed cases and
prevalence of exposure in case-control
studies
857 lung cancer cases (79.2% response), 117 pancreatic cancer cases (70.7% response);
533 population controls (72% response).
Exposure prevalence: Any TCE exposure, 2% cancer cases (n = 21 lung cancer cases and
2 pancreatic cancer cases); substantial TCE exposure (exposure for >10 yrs and up to
5 yrs before disease onset), 1% lung cancer cases (n = 9), no pancreatic cancer cases
assigned "substantial" TCE exposure.
CATEGORY H: ANALYSIS
Control for potential confounders in
statistical analysis
Lung cancer—age, family income, cigarette smoking, ethnic origin, alcohol consumption,
and respondent status.
Pancreatic cancer —age, income, cigarette smoking, and respondent status.
Statistical methods
Mantel-Haenszel (Siemiatycki, 1991).
-------

5S
>3
o
o
s
Co
I?
"tS
§*
3'
Co
a
>r
<§>
>S
>S
TO
to'
*
^S
S
s
TO
>;
0
s
1
a.
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
B.3.3. Geographic-Based Studies
B.3.3.1. Coyle et al. (2005)
B.3.3.1.1. Author's abstract.
Purpose. To investigate the role of environment in breast cancer development, we
conducted an ecological study to examine the association of releases for selected
industrial chemicals with breast cancer incidence in Texas.
Methods. During 1995-2000, 54,487 invasive breast cancer cases were reported
in Texas. We identified 12 toxicants released into the environment by industry
that: (1) were positively associated with breast cancer in epidemiological studies,
(2) were Environmental Protection Agency (EPA) Toxics Release Inventory
(TRI) chemicals designated as carcinogens or had estrogenic effects associated
with breast cancer risk, and (3) had releases consistently reported to EPA TRI for
multiple Texas counties during 1988-2000. We performed univariate, and
multivariate analyses adjusted for race and ethnicity to examine the association of
releases for these toxicants during 1988-2000 with the average annual age-
adjusted breast cancer rate at the county level.
Results. Univariate analysis indicated that formaldehyde, methylene chloride,
styrene, tetrachloroethylene, trichloroethylene, chromium, cobalt, copper, and
nickel were positively associated with the breast cancer rate. Multivariate
analyses indicated that styrene was positively associated with the breast cancer
rate in women and men (b = 0.219, p =0.004), women (b = 0.191, p=0.002), and
women J 50 years old (b = 0.187, p=0.002).
Conclusion. Styrene was the most important environmental toxicant positively
associated with invasive breast cancer incidence in Texas, likely involving
women and men of all ages. Styrene may be an important breast carcinogen due
to its widespread use for food storage and preparation, and its release from
building materials, tobacco smoke, and industry.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-424 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B.3.3.1.2. Study description and comment. Residential address in 254 Texas counties at
time of cancer diagnosis was the exposure surrogate in this ecologic study of invasive breast
cancer in over a 5-year period (1995-2000). Incident breast cancer cases in males and
females were identified from Texas Cancer Registry. During the 5-year period, 54,487
cases were diagnosed, of which 53,910 were in females (99%). Median average annual age-
adjusted breast cancer rates for women and men, women, women <50 years old, and
women >50 years old and 12 hazardous air pollutants identified as exposures of interested
were examined using nonparametric tests (Mann-Whitney U test) and linear regression
analyses. The 12 hazardous air pollutants (HAPs) were: carbon tetrachloride,
formaldehyde, methylene chloride, styrene, perchloroethylene, TCE, arsenic, cadmium,
chromium, cobalt, copper, and nickel. On-site atmospheric release data on individual
HAPs was identified from EPA's Toxics Release Inventory (TRI) for a 13-year period, 1998
to 2000 with an exposure surrogate as the annual total release in pounds/year for the 12
HAPs.
Coyle et al. (2005) compared average annual age-adjusted breast cancer rate for counties
reporting a release to that rate for non-reporting counties using Mann-Whitney U test.
Additionally, multiple linear regression analyses was used to determine the association of the
average annual age-adjusted breast cancer rates with the 12 HAPs, adjusting for race and
ethnicity when associated with the study's outcome variable.
While this study provides insight on cancer rates in studied population, TCE and other
hazardous air pollutant exposures are poorly defined and the exposure surrogate unable to
distinguish subjects more with higher exposure potential from those with low or minimal
exposure potential. Some information may be provided through examination of inter-county
release rates; however, no information is provided by Coyle et al. (2005). Furthermore, the
ecologic design of the study does not address residential history or other information on an
individual-subject level and is subject to bias from "ecologic fallacy" or improper inference
about individual-level associations based on aggregate-level analysis. Overall, this study is not
able to identify risk factors (etiologic exposures), has low sensitivity for examining TCE, and
provides little weight in an overall weight of evidence evaluation of TCE and cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-425 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Coyle YM, Hynan LS, Euhus DM, Minhajuddin ATM. (2005). An ecological study of the association of environmental
chemicals on breast cancer incidence in Texas. Breast Cancer Res Treat.92:107-114.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Hypothesis of this study was to evaluate breast risks in Texas counties and hazardous
air pollutants (HAPs).
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cases are incident breast cancers in males and females over a 5-yr period
(1995-2000) in subjects residing in Texas and reported to the Texas Cancer Registry.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Residence in Texas county as time of diagnosis is exposure surrogate. Annual
release by county of 12 HAPs (carbon tetrachloride, formaldehyde, methylene
chloride, styrene, perchloroethylene, TCE, arsenic, cadmium, chromium, cobalt,
copper, and nickel) are obtained from EPA's TRI database.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
>50% cohort with full latency
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
54,487 incident breast cancer cases in males and females.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and race/ethnicity.
Statistical methods
Mann-Whitney U test (nonparametric) to compared average annual age-adjusted
breast cancer rate between counties reported HAP release to that for non-reporting
counties.
Linear logistic regression
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
EPA = Environmental Protection Agency. HAP = hazardous air pollutant. TRI = Toxic Release Inventory.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
B.3.3.2. Morgan and Cassady (2002)
B.3.3.2.1. Author's abstract.
In response to concerns about cancer stemming from drinking water contaminated
with ammonium perchlorate and trichloroethylene, we assessed observed and
expected numbers of new cancer cases for all sites combined and 16 cancer types
in a California community (1988 to 1998). The numbers of observed cancer cases
divided by expected numbers defined standardized incidence ratios (SIRs) and
99% confidence intervals (CI). No significant differences between observed and
expected numbers were found for all cancers (SIR, 0.97; 99% CI, 0.93 to 1.02),
thyroid cancer (SIR, 1.00; 99% CI, 0.63 to 1.47), or 11 other cancer types.
Significantly fewer cases were observed than expected for cancer of the lung and
bronchus (SIR, 0.71; 99% CI, 0.61 to 0.81) and the colon and rectum (SIR, 0.86;
0.74 to 0.99), whereas more cases were observed for uterine cancer (SIR, 1.35;
99% CI, 1.06 to 1.70) and skin melanoma (SIR, 1.42; 99% CI, 1.13 to 1.77).
These findings did not identify a generalized cancer excess or thyroid cancer
excess in this community.
B.3.3.2.2. Study description and comment. Residential address in 13 census tracts in
Redlands (San Bernardino County, CA) at time of cancer diagnosis was the exposure
surrogate in this ecologic study of cancer incidence over a 10-year period (1988-1998).
Seventeen cancers in adults (all cancers, bladder, brain and other nervous system, breast
[females only], cervix, colon and rectum, Hodgkin lymphoma, kidney and renal pelvis,
leukemia [all], liver and bile duct, lung and bronchus, NHL, melanoma, ovary, prostate,
thyroid and uterus) and 3 site-specific incident cancers in children under 15 years of age
(leukemia [all], brain/CNS, and thyroid) were identified from the Desert Sierra Cancer
Surveillance Program, a regional cancer registry reporting to the California Cancer
Registry, with expected numbers of site-specific cancer using age-race annual site-specific
cancer incidence rates between 1988 and 1992 to 1990 census-reported information on
population size and demographics. The use of the Desert Sierra Cancer Surveillance
Program rates which include the studied population would inflate the number of site-
specific cancer expected; however, the potential magnitude of bias is likely minimal given
the Redlands populations was estimated as 2% of the total population of the regional
cancer registries ascertainment area (Morgan and Cassady, 2002). This is a record-based
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-429 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
study and information on personal habits and potential risk factors other than race, sex,
and age are lacking for individual subjects.
Morgan and Cassidy (2002) identified TCE and perchlorate from drinking water as
exposures of interest. Limited monitoring data from the 1,980 identified TCE concentrations in
Redlands wells as between 0.09 and 97 ppb TCE and drinking water concentrations as below the
maximum contaminant level (MCL; 5 ppb) since 1991. The paper lacks information if water
monitoring represented wells in the 13-census tract study area. Furthermore, the paper does not
include information on water treatment and distribution networks to provide an estimate of TCE
concentration in finished tap water to individual homes. These authors noted their inability to
identify higher or lower exposed subjects, as well, as minimally exposed subjects as a source of
uncertainty. No data are presented on perchlorate concentrations in well or drinking water. The
assumption of residence in 13 census tracts is insufficient as a surrogate of potential exposure to
TCE and perchlorate in the absence of exposure modeling and data on water distribution
patterns. Exposure misclassification bias is highly likely and of a nondifferential nature which
would dampen observed associations.
While this study provides insight on cancer rates in studied population, TCE exposure is
poorly defined and the exposure surrogate unable to distinguish subjects more with higher
exposure potential from those with low or minimal exposure potential. Furthermore, the
ecologic design of the study does not address residential history or other information on an
individual-subject level and is subject to bias from "ecologic fallacy" or improper inference
about individual-level associations based on aggregate-level analysis. Morgan and Cassidy
(2002) furthermore discuss the relatively high education and income levels in the Redlands
population compared with the average for the referent population may lead to lower tobacco use
and higher than average access to health care, biases that would dampen risks for lung and other
tobacco-related cancers, but may also increase risks for colon and cervical cancers. Overall, this
study is not able to identify risk factors (etiologic exposures), has low sensitivity for examining
TCE, and provides little weight in an overall weight of evidence evaluation of TCE and cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-430 DRAFT—DO NOT CITE OR QUOTE
-------
Morgan JW, Cassady RE. (2002). Community cancer assessment in response to long-time exposure to perchlorate and
trichloroethylene in drinking water. J Occup Environ Med 44:616-621.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Hypothesis of this study was to evaluate cancer risks in a California community, not
to evaluate TCE and cancer explicitly.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cases are incident cancers over a 10-yr period (1988-1989) in subjects residing in
13 Redlands (CA) census tracts at time of diagnosis. 17 site-specific cancers are
identified in adults and 3 site-specific cancers in children less than 15 yrs old.
Cancer cases identified from Desert Sierra Cancer Surveillance Program (DSCSP), a
regional cancer registry.
Annual age-race-site specific cancer rates from DSCSP for 1988 and 1992 and
age-race-sex specific population estimates for 1990.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Residence in a 13-census tract area of Redlands, CA is exposure surrogate. No data
are presented on TCE or perchlorate concentrations in treated drinking water supplied
to residents.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
3,098 incident cancers, the largest number from 536 breast cancer and fewest number
from Hodgkin disease.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and race/ethnicity.
Statistical methods
SIR with indirect standardization of estimated expected numbers of site-specific
cancers adjusted for population growth; 90% confidence intervals presented in tables.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
SIR = standardized incidence ratio.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3.3.3. Cohn et al. (1994a)
B.3.3.3.1. Author's abstract.
A study of drinking water contamination and leukemia and non-Hodgkin's
lymphoma (NHL) incidence (1979-1987) was conducted in a 75-town study area.
Comparing incidence in towns in the highest trichloroethylene (TCE) stratum (>5
microg/L) to towns without detectable TCE yielded an age-adjusted rate ratio
(RR) for total leukemia among females of 1.43 (95% CI 1.07-1.90). For females
under 20 years old, the RR for acute lymphocytic leukemia was 3.26 (95% CI
1.27-8.15). Elevated RRs were observed for chronic myelogenous leukemia
among females and for chronic lymphocytic leukemia among males and females.
NHL incidence among women was also associated with the highest TCE stratum
(RR = 1.36; 95% CI 1.08-1.70). For diffuse large cell NHL and non-Burkitt's
high-grade NHL among females, the RRs were 1.66 (95% CI 1.07-2.59) and 3.17
(95% CI 1.23-8.18), respectively, and 1.59 (95% CI 1.04-2.43) and 1.92 (95% CI
0.54-6.81), respectively, among males. Perchloroethylene (PCE) was associated
with incidence of non-Burkitt's high-grade NHL among females, but collinearity
with TCE made it difficult to assess relative influences. The results suggest a link
between TCE/PCE and leukemia/NHL incidence. However, the conclusions are
limited by potential misclassification of exposure due to lack of individual
information on long-term residence, water consumption, and inhalation of
volatilized compounds.
B.3.3.3.2. Study description and comment. This expanded study of a previous analysis of
TCE and perchloroethylene in drinking water in a 27-town study area (Fagliano et al.,
1990)examined leukemia and NHL incidence from 1979 to 1987 in residents and TCE and
other VOCs in drinking water delivered to 75 municipalities. Exposure estimates were
developed from data generated by a mandatory monitoring program for four
trihalomethane chemicals and 14 other volatile organic chemicals in 1984-1985 for public
water supplies and from historical monitoring data conducted in 1978-1984 by the New
Jersey Department of Environmental Protection and Energy and the New Jersey
Department of Health, which was the mean of monthly averages for this period. The
average and maximum concentration of TCE and other chemicals were estimated by
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-434 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
considering together, for the period prior to 1985, details of the distribution system size,
well or surface water use, patterns of water purchases among systems, and significant
changes in water supply, and for years after 1985, samples of finished water from the plant
and samples taken from the distribution system under the assumption of homogeneous
mixing. The number of distribution system samples for each supply varied from 2 to 50.
Additionally, a dilution factor assuming complete mixing was used to adjust for water
purchased from another source. A single summary average and maximum concentration
for each contaminate for a municipality was assigned to all cases residing in that
municipality at the time of cancer diagnosis. Concentrations of TCE and
perchloroethylene were highly correlated (r = 0.63). A ranking of municipalities was the
same when using average or maximum concentration and the maximum concentration of
TCE or perchloroethylene used in statistical analyses was grouped into three strata: <0.1
ppb (referent group), 0.1-5 ppb, >5-20 ppb, and >20 ppb.
Incident cases of NHL and forms of leukemia reported to the New Jersey State Cancer
Registry were identified from 1979 and 1987. Incidence rate ratios were estimated using Poisson
regression models fitted to age- and sex-specific numbers of cases by exposure strata and the
stratum-specific population. Statistical treatment considered exposure to other drinking water
contaminants, atmospheric emissions of hazardous air pollutants as reported to U.S. EPA's
Toxics Release Inventory (TRI) by municipality and two socioeconomic variables measured as
municipal—average annual household income and percentage of high school graduates. None of
the water trihalomethane or volatile organic contaminants other than perchloroethylene was
shown to be associated with childhood leukemia or adult lymphomas. Furthermore, neither
average income, education, nor TRI release data were associated with NHL or leukemia except
in one exception, TRI release was shown to modify the effects of TCE and high-grade
non-Burkett's lymphoma in females.
This ecological study is subject to known biases and confounding as introduced through
its study design (NRC, 1997). Exposure estimates are crude (averages), do not consider
individual differences in drinking water patterns, and assigns group exposure levels to all
subjects without consideration of residential history. Potential for misclassification bias is likely
great in this study as is the potential for bias. This study does attempt to examine three possible
confounding exposures, although these are crudely defined, and some potential for residual
confounding is possible given the study's use of aggregated data.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-43 5 DRAFT—DO NOT CITE OR QUOTE
-------
Cohn P, Klotz J, Bove F, Berkowitz M, Fagliano J. (1994a). Drinking water contamination and the incidence of leukemia and
non-Hodgkin's lymphoma. Environ Health Perspect 102:556-561.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This study was designed to further examine drinking water contaminates and
lymphoma; a previous study of TCE and perchloroethylene in drinking water found a
statistically significant association with leukemia among females residing in a
27-town study area (Fagliano et al., 1990).
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Incident cases of various forms of leukemia (all leukemia, acute lymphocytic, chronic
lymphocytic, acute myelogenous, chronic myelogenous, other specified and
unspecified leukemia) and NHL (total, low-grade, intermediate-grade [total and
diffuse large cell a B-cell lymphoma], high-grade including non-Burkett's
lymphoma) from 1979-1987 are identified from New Jersey State Cancer Registry.
Subjects grouped in lowest exposure category are referents.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Average and maximum concentration of TCE and other chemicals were estimated by
considering together, for the period prior to 1985, details of the distribution system
size, well or surface water use, patterns of water purchases among systems, and
significant changes in water supply, and for years after 1985, samples of finished
water from the plant and samples taken from the distribution system under the
assumption of homogeneous mixing. No difference in municipality ranking by
average or maximum concentration.
Three grouped categories of maximum concentration in statistical analysis are
<0.1 ppb (referent), 0.1-5 ppb, >5 ppb (U.S. EPA Maximum Contaminant Level for
TCE and perchloroethylene).
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
1,190 leukemia cases (663 males, 527 females), 119 cases assigned >5.0 ppb TCE.
1,658 NHL cases (841 males, 817 females), 165 cases assigned >5.0 ppb TCE.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and sex.
Statistical methods
Poisson regression fitted to the age-and sex-specific count of cases in towns grouped
by exposure strata and weighted by the logarithm of the strata-specific population.
Exposure-response analysis presented in
published paper
Yes.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
B.3.3.4. Vartiainen et al. (1993)
B.3.3.4.1. Author's abstract.
Concentrations up to 212 j_ig/l of trichloroethene (TCE) and 180 [j.g/1 of
tetrachloroethene (TeCE) were found in the drinking water from two villages in
Finland. To evaluate a possible exposure, urine sample fro m95 and 21
inhabitants in these villages and from two control groups of 45 and 15 volunteers
were collected. Dichloroacetic acid (DCA) and trichloroacetic acid (TCA), the
metabolites of TCE and TeCE, were also analyzed. The individuals using
contaminated water in one of the villages excreted TCE an average 19 jag/d (<1 -
110 (J,g/d) and in the other 7.9 j_ig/d (<1 - 50 j_ig/d), while the controls excreted an
average 2.0 jj,g/d (<1 - 6.4 jag/d) or 4.0 jag/d (<1 - 13 (J,g/d). No increased
incidence rates were found in the municipalities in question for total cancer, liver
cancer, non-Hodgkin's lymphomas, Hodgkin's disease, multiple myeloma, or
leukemia.
B.3 .3 .4.2. Study description and comment. This published study of two separate analyses,
(1) urinary biomonitoring of 106 subjects from two Finish municipalities, Hausjarvi and
Hattula, and, (2) calculation of total cancer and site-specific cancer incidence between 1953
and 1991 in Hausjarvi and Hattula residents. Limited exposure monitoring data are
presented in the paper. TCE concentrations in drinking water from Oitti are lacking other
than noting TCE and perchloroethylene were 100-200 jig/L in 1992. TCE concentrations
in drinking water from Hattula were below 10 jig/L in December 1991; however, samples
(number unknown) taken 6 months later contained 212 jig/L and 66 jig/L TCE. These two
municipalities discontinued use of these sources for drinking water in August 1992.
Cancer incidence for 6 sites (all cancers, liver cancer, NHL, Hodgkin's lymphoma,
multiple myeloma, and leukemia) between 1953-1991 in Hausjarvi and Hattula residents was
obtained from the Finnish Cancer Registry. A total of 1,934 cancers were observed during the
study period. Standardized incidence ratios for each municipality were calculated using
site-specific cancer incidence rates from the Finnish population for the entire time period and for
3 shorter periods, 1953-1971, 1972-1981, and 1982-1991. The paper does not identity the
source for or size of Hausjarvi and Hattula population estimates and if temporal changes in
population estimates were considered in the statistical analysis. This study using record systems
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-439 DRAFT—DO NOT CITE OR QUOTE
-------
1	did not include information obtained directly from subjects and lacks information on personal
2	and lifestyle factors that may introduce bias or confounding.
3	This study provides little information in an overall weight-of-evidence analysis on cancer
4	risks and TCE exposure. A major limitation is its lack of exposure assessment to TCE and
5	perchloroethylene. While this study provides some information on cancer incidence in the two
6	towns over a 40-year period, this study is not able to identify potential risk factors and exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-440 DRAFT—DO NOT CITE OR QUOTE
-------
Vartiainen T, Pukkala E, Rienoja T, Strandman T, Kaksonen K. (1993). Population exposure to tri- and tetrachloroethene
and cancer risk: two cases of drinking water pollution. Chemosphere 27:1171-1181.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Study aim was (1) to determine if residents of two villages in Finland had exposure
to TCE and perchloroethylene as indicated from urinary biomonitoring, (2) identify
biomarker for low-level exposure, and (3) to determine cancer incidence in Hausjarvi
and Hattula, two municipalities in Finland. This study could not identify potential
risk factors.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cancer incidence cases identified from Finnish Cancer Registry.
Site-specific cancer rates for the Finnish population was used a referent.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Residence in two municipalities is the exposure surrogate in this ecologic study. The
paper lacks exposure assessment to TCE and perchloroethylene in drinking water in
Hausjarvi and Hattula.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
3,846 cancer cases; 1,942 from Hausjarvi and 1,904 from Hattula.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and sex.
Statistical methods
SIR with cancer incidence rates in Finnish population as referent.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Cancer incidence analysis is not well documented.
SIR = standardized incidence ratio.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
B.3.3.5. Mallin (1990)
B.3.3.5.1. Author's abstract.
Cancer maps from 1950 through 1979 revealed areas of high mortality from
bladder cancer for both males and females in several northwestern Illinois
counties. In order to further explore this excess, a bladder cancer incidence study
was conducted in the eight counties comprising this region. Eligible cases were
those first diagnosed with bladder cancer between 1978 and 1985. Age adjusted
standardized incidence ratios were calculated for each county and for 97 zip codes
within these counties. County results revealed no excesses. Zip code results
indicated elevated risks in a few areas, but only two zip codes had significantly
elevated results. One of these zip codes had a significant excess in males
(standardized incidence ratio =1.5) and females (standardized incidence ratio =
1.9). This excess was primarily confined to one town in this zip code, in which
standardized incidence ratios were significantly elevated in males (1.7) and
females (2.6). Further investigation revealed that one of four public drinking
water wells in this town had been closed due to contamination; two wells were
within a half mile (0.8 km) of a landfill site that had ceased operating in 1972.
Tests of these two wells revealed traces of trichloroethylene, tetrachloroethylene,
and other solvents. Further investigation of this cluster is discussed.
B.3.3.5.2. Study description and comment. This ecologic study of bladder cancer incidence
and mortality among white residents in nine Illinois counties between 1978-1985 was
carried out to further investigate a previous finding of elevated bladder cancer mortality
rates in some counties. The study lacks exposure assessment to subjects and potential
sources of exposure was examined in a post hoc manner in one case only, for a community
with an observed elevated bladder cancer incidence. The limited exposure examination
focused on groundwater contamination and proximity of Superfund sites to the
community, lacked assignment of exposure surrogates to individual study subjects, and
findings are difficult to interpret given the lack of exposure assessment for the other eight
counties.
Histologically-confirmed incident bladder cancer cases were identified from hospital
records in eight of the nine counties. Since the 9-county area bordered on neighboring states of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-443 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Wisconsin and Iowa, incident bladder cancer cases were also ascertained from the Wisconsin
Cancer Reporting System and Iowa's State Health Registry. No information is provided in the
paper on completeness of ascertainment of bladder cancer cases among residents or on the source
for identifying bladder cancer deaths. Expected numbers of incident cancers calculated using
age-specific rates for white males and females from the SEER program (incidence) or the United
States population [mortality], and the census data on population estimates for the nine-county
area. Statistical analyses adopt indirect standardization methods to calculate SMR and
standardized incidence ratios (SIRs) for a community and SIRs for individual postal zip codes.
The use of records and absence of information collected from subject personal interviews
precluded examination of possible confounders other than age and race.
This ecological study is subject to known biases and confounding as introduced through
its study design (NRC, 1997). Ecological studies like this study are subject to bias known as
"ecological fallacy" since variables of exposure and outcome measured on an aggregate level
may not represent association at the individual level. Consideration of this bias is important for
diseases with more than one risk factor, such as the site-specific cancers evaluated in this
assessment. Lack of information on smoking is another uncertainty. While this study provides
insight on bladder cancer rates in the studied communities, it does not provide any evidence on
cancer and TCE exposure. For this reason, this study provides little weight in an overall
weight-of-evidence analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-444 DRAFT—DO NOT CITE OR QUOTE
-------
Mallin K. (1990). Investigation of a bladder cancer cluster in Northwestern Illinois. Amer J Epidemiol 132:S96-S106.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
The hypothesis of study was to "further exposure a previous finding of bladder
cancer excess in several northwestern Illinois counties." (from abstract).
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Incident cancer cases diagnosed between 1978-1985 were identified in residents in
9 northwestern Illinois counties from the Illinois Cancer Registry, the Wisconsin
Cancer Reporting System or the Iowa State Health Registry. Source for deaths in
subjects residing at the time of death in the 9 counties was not identified in the
published paper.
Expected number of bladder cancer derived using (1) SEER age-race-sex specific
incidence rates and (2) age-race-sex specific mortality rates of the U.S. population for
1978-1981 and for 1982-1985 and census estimates of population for each county or
postal zip code area.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence and mortality.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
This is a health survey and lacks exposure assessment to communities and to
individual subjects. Monitoring of volatile organic chemicals including
trichloroethylene in two municipal drinking water wells for 1982-1988 in a
community with elevated bladder cancer rates was identified in paper; TCE
concentrations were less than 15 ppb. It is not know whether monitoring data are
representative of exposure to study subjects.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
712 bladder cancer incident cases and 222 bladder cancer deaths among white males
and female residents in nine northwestern Illinois counties.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and sex .
Statistical methods
SIR with cancer incidence rates from Surveillance, Epidemiology and End Results
program and mortality rates of U.S. population as referents.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
B.3.3.6. Isacson et al. (1985)
B.3.3.6.1. Author's abstract.
With data from the Iowa Cancer Registry, age-adjusted sex-specific cancer
incidence rates for the years 1969-1981 were determined for towns with a
population of 1,000-10,000 and a public water supply from a single stable ground
source. These rates were related to levels of volatile organic compounds and
metals found in the finished drinking water of these towns in the spring of 1979.
Results showed association between 1,2 dichloroethane and cancers of the colon
and rectum and between nickel and cancers of the bladder and lung. The effects
were most clearly seen in males. These associations were independent of other
water quality and treatment variables and were not explained by occupational or
other sociodemographic features including smoking. Because of the low levels of
the metals and organics, the authors suggest that they are not causal factors, but
rather indicators of possible anthropogenic contamination of other types. The data
suggest that water quality variables other than chlorination and trihalomethanes
deserve further consideration as to their role in the development of human cancer.
B.3 .3 .6.2. Study description and comment. This ecologic study of cancer incidence at six
sites [bladder, breast, colon, lung, prostate, rectum] and chlorinated drinking water uses
monitoring data from finished public drinking water supplies to infer exposure to residents
of Iowa towns of 1,000-10,000 population sizes. Towns were included if they received
water from a single major source (surface water, wells of <150 feet depth, or wells >50 feet
depth) prior to 1965. Water monitoring for VOCs, trace elements and heavy metals was
carried in Spring, 1979, as part of a larger nation-wide collaborative study of bladder
cancer and artificial sweeteners (Hoover and Strasser, 1980), and samples analyzed using
proton-induced x-ray emission for trihalomethanes, TCE, perchloroethylene, 1,2-
dichloroethane, 1,1,1-trichloroethane, carbon tetrachloride, 1,2-dichloroethylene, and 43
inorganic elements. 1,1,1-trichloroethane was the most frequently detected VOC in both
surface and groundwater; TCE, perchloroethylene, and 1,2-dichloroethane were more
frequently detected in shallow wells than in deep (>150 feet) wells.
Cancer incidence was obtained for the period 1969 and 1981 with age-adjusted
site-specific cancer incidence rates for males and females calculated separately for four VOCs
(1,2-dichloroethane, TCE, perchloroethylene, and 1,1,1-trichloroethane) in finished groundwater
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-448 DRAFT—DO NOT CITE OR QUOTE
-------
1	supplies using the direct standardization method. Using the address at the time of diagnosis,
2	each cancer patient was classified into one of two groups: (1) residing within the city limits and,
3	thus, drinking the municipality's water, or (2) residing outside the city limits and consuming
4	water from a private source. Age-adjusted incidence rates are reported by group study town into
5	two TCE water concentrations categories of <0.15 [j,g/L and >0.15 (J,g/L.
6	This ecological study on drinking water exposure and cancer provides little information
7	in a weight-of-evidence analysis of TCE and cancer. Exposure estimates are crude (averages),
8	do not consider individual differences in drinking water patterns or other sources of exposure,
9	and assigns group exposure levels to all subjects. Potential for misclassification bias is likely
10	great in this study, likely of a nondifferential nature, and dampen observations.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-449 DRAFT—DO NOT CITE OR QUOTE
-------
Isacson P, Bean JA, Splinter R, Olson DB, Kohler J. (1985). Drinking water and cancer incidence in Iowa. III. Association
of cancer with indices of contamination. Amer J Epidemiol 121:856-869.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This ecological study was designed to examine consistency with the hypothesis of an
association between cancer and chlorinated water through examination of other water
contaminants besides water chlorination by-products and trihalomethanes.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Subjects are incident cases of cancer of the bladder, breast, prostate, lung rectum, and
stomach reported to the Iowa Cancer Registry between 1969 and 1981 and, who
resided in towns with a 1970 population of 1,000-10,000 and a public drinking water
supply coming solely from a single major source (wells) prior to 1965.
Age-adjusted site-specific incidence rates are calculated using the direct method and
the 1970 Iowa population.
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Not identified in paper.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
As part of another epidemiologic study on water chlorination and bladder cancer,
finished drinking water samples from treatment plant were collected in Iowa
municipalities with populations of 1,000 or larger in Spring 1979 and analyzed using
proton induced x-ray emission for 4 trihalomethanes (chloroform,
chlorodibromomethane, bromoform, dibromochloromethane), 7 VOCs (TCE,
perchloroethylene, 1,1,1-trichloroethane, carbon tetrachloride, 1,2-dichloroethane,
and cis- and trans-1,2-dichloroethylene) and 43 inorganic elements, including metals.
The predominant contaminant was 1,1,1-trichloroethane; detectable levels of TCE
were found in approximately 20% of sampled municipalities.
Study towns were ranked into two categories of TCE in finished water, <0.15 [j,g/L
and >0.15 [j,g/L in the statistical analysis.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face

Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
?
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
11,091 cancer cases of which -20% of cases resided in municipality with finished
water TCE concentration of >0.15 (J,g/L.
Bladder, 852 cases
Breast (female), 1,866 cases
Colon, 2,032 cases
Lung 1,828 cases
Prostate, 1,823 cases
Rectum, 824 cases
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and sex.
Statistical methods
Age-adjusted site-specific mortality rates calculated using direct standardization
method and 1970 Iowa population.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
B.3 .3 .7. Studies in the Endicott Area of New York
A series of health statistics reviews and exposure studies have been conducted in an area
with a history of VOCs, including trichloroethylene, detected in municipal wells used to supply
drinking water to residents of Endicott, Broome County, NY. These studies were carried out by
staff the New York State Department of Health (NYS DOH) with support from the ATSDR.
Early health surveys examined cancer incidence among Broome County residents between
1976-1980 or 1981-1990, with focused analyses of cancer incidence among residents of
Endicott Village and other nearby towns, childhood leukemia in the Town of Union and possible
etiologic factors, and adult leukemia deaths and employment in the shoe and boot manufacturing
industry (Forand, 2004; Nysdoh, 2005). Two recent studies focused on cancer incidence or birth
outcomes among Village of Endicott residents living in a geographically defined area with VOC
exposure potential as documented from indoor and soil vapor monitoring (ATSDR, 2006b,
2008).
The Village of Endicott is a mixed residential, commercial, and industrial community
with a rich industrial heritage and a number of VOCs were used at industrial locations in and
around Endicott, as well as, having been disposed at area landfills (ATSDR, 2006b). Three wells
provide drinking water to the Village of Endicott: Ranney, which supplied most of the water
used by the Endicott Municipal Water Works since it was first placed in service in 1950; and,
South Street, where two wells resided. The Endicott Municipal Water Supply operates on a
grid-water system, neighborhoods closest to the wells are usually supplied at a greater rate from
nearby wells as compared to wells farther away (ATSDR, 2006b).
Routine monitoring of the Ranney well in the early 1980s detected VOCs at levels above
New York State drinking water guidelines (ATSDR, 2006b). A groundwater contaminate plume
northwest of the Ranney Well was found in a lower aquifer from which the municipal drinking
supply is drawn. Several sources were initially recognized as contributing to contamination of
the wellfield with a supplemental remedial investigation concluding that the Endicott Village
Landfill was the source of the VOCs in the Endicott Wellfield water supply (ATSDR, 2006a).
Groundwater water samples collected from monitoring wells installed during previous
investigations, wells install as part of the supplemental remedial investigation, the Purge well,
and the Ranney well contained many VOCs. Remediation efforts starting in the 1980s have
reduced contamination in this well to current MCLs. Water monitoring of the South Street wells
(wells 5 and 28) has been carried out for VOCs since 1980 and 1981, respectively (ATSDR,
2006b). Detection limits for VOCs from the South Street wells varied from 0.5-1.0 (J,g/L;
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-453 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
1,1-dichloroethane had the highest detection frequency, in 44% of all samples and TCE was
detected in 3 of 116 samples obtained between 1980 and 2004 (ATSDR, 2006b).
An upper aquifer with a contaminant plume containing VOCs was also identified and
sampling data indicated there were multiple sources of vapor contamination including a former
IBM facility located in the Village (NYSDEC, 2007; U.S. EPA, 2005a). This groundwater
contaminant plume flows directly beneath the center of the Village of Endicott and serves as a
source of soil vapor contamination. Findings of a 2002 investigation indicated vapor migration
had resulted in detectable levels of contaminants in indoor air structures, including locations in
the Village of Endicott and Town of Union. Of soil gas and indoor air monitoring at more than
300 properties in an area south of the IBM Endicott facility, TCE was the most commonly found
contaminant in indoor air, at levels ranging from 0.18 to 140 (NYSDEC, 2007). This area is
identified as the Eastern study area in the health statistics review of ATSDR (2006b, 2008).
Other contaminants besides TCE detected in soil gas and indoor air less frequently and at lower
levels included tetrachloroethylene, cis-l,2-dichloroethene, 1,1,1-trichloroethane,
1,1-dichloroethylene, 1,1-dichloroethane, and Freon 113. Vapor-intrusion contamination was
also identified in a neighborhood adjacent to the Eastern area, call the Western study in the
health statistic review, and perchloroethylene and its degradation by-products were detected by
"3
vapor monitoring. Perchloroethylene levels generally ranged from 0.1 to 3.5 (J,g/m of air
(ATSDR, 2006a).
B.3 .3 .7.1. Agency for Toxic Substances and Disease Registry (2006a, 2008).
B.3 .3 .7.1.1. Agency for Toxic Substances and Disease Registry (2006a) executive
summary.
Background The New York State Department of Health (NYS DOH) conducted
this Health Statistics Review because of concerns about health issues associated
with environmental contamination in the Endicott area. Residents in the Endicott
area may have been exposed to volatile organic compounds (VOCs) through a
pathway known as soil vapor intrusion. Groundwater in the Endicott area is
contaminated with VOCs as a result of leaks and spills associated with local
industry and commercial businesses. In some areas of Endicott, VOC
contamination from the groundwater has contaminated the adjacent soil vapor
which has migrated through the soil into structures through cracks in building
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-454 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
foundations (soil vapor intrusion). Trichloroethene (TCE), tetrachloroethene
(PCE) and several other VOCs have been found in the soil vapor and in the indoor
air of some structures.
Conclusions This health statistics review was conducted because of concerns that
exposure to VOCs through vapor intrusion may lead to adverse health effects.
Although this type of study cannot prove whether there is a causal relationship
between VOC exposure in the study area and the increased risk of several health
outcomes observed, it does serve as a first step in providing guidance for further
health studies and interventions. The elevated rates of several cancers and birth
outcomes observed will be evaluated further to try to identify additional risk
factors which may have contributed to these adverse health outcomes.
Limitations in the current study included limited information about the levels
of VOCs in individual homes, the duration of the exposure, the amount of time
residents spent in the home each day and the multiple exposures and exposure
pathways that likely existed among long term residents of the Endicott area. In
addition, personal information such as medical history; dietary and lifestyle
choices such as smoking and drinking; and occupational exposures to chemicals
were not examined. Future evaluations of cancer and birth defects and VOC
exposures in the area should take these factors into account. The small population
size of the study area also limited the ability to detect meaningful elevations or
deficits in disease rates, especially for certain rare cancers and birth outcomes.
This study represents the first step in a step-wise approach to addressing
health outcome concerns related to environmental contamination in Endicott, NY.
Follow-up will consist of further reviewing of the cancer and birth outcome data
already collected. Additional efforts will include reviewing individual case
records of kidney and testicular cancers, heart defects, Down syndrome and term
low birth weight births. In addition, we will review spontaneous fetal deaths
among residents of the area. The information gained, along with the results of this
Health Statistics Review, will be used to assess if a follow up epidemiologic study
is feasible. Any follow-up study should be capable of accomplishing one of two
goals: either to advance the scientific knowledge about the relationship between
VOC exposure and health outcomes; or as part of a response plan to address
community concerns. While not mutually exclusive, the distinction between these
goals must be considered when developing a follow-up approach. Any plans for
additional study will need to address other risk factors for these health outcomes
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-455 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
such as smoking, occupation and additional information on environmental
exposures. As in the past, NYS DOH will solicit input from the community.
B.3.3.7.1.2. Agency for Toxic Substances and Disease Registry (2008) executive
summary.
This follow-up investigation was conducted to address concerns and to provide
more information related to elevated cancers and adverse birth outcomes
identified in the initial health statistics review entitled "Health Statistics Review:
Cancer and Birth Outcome Analysis, Endicott Area, Town of Union, Broome
County, New York" (2006a).
The initial health statistics review was carried out to address concerns about
health issues among residents in the Endicott area who may have been exposed to
volatile organic compounds (VOCs) through a pathway known as soil vapor
intrusion. The initial health statistics review reported a significantly elevated
incidence of kidney and testicular cancer among residents in the Endicott area. In
addition, elevated rates of heart defects and low birth weight births were
observed. The number of term low birth weight births, a subset of low birth
weight births, and the number of small for gestational age (SGA) births were also
significantly higher than expected.
The purpose of this follow-up investigation was to gather more information
and conduct a qualitative examination of medical and other records of individuals
identified with adverse birth outcomes and cancers found to be significantly
elevated. Quantitative analyses were also carried out for two additional birth
outcomes, conotruncal heart defects (specific defects of the heart's outflow
region), and spontaneous fetal deaths (stillbirths), and for cancer incidence
accounting for race.
Cancer Incidence Adjusting for Race: Because a higher percentage of the
population in the study area was white compared to the comparison population,
we examined the incidence of cancer among whites in the study area compared to
the incidence in the white population of New York State, excluding New York
City. Cancer incidence among whites was evaluated for the years 1980-2001.
Results: Limiting the analysis of cancer to only white individuals had little effect
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-456 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
on overall cancer rates or standardized incidence ratios compared to those of the
entire study area population analyzed previously. The only difference was the
lung cancer which had been borderline non-significantly elevated was not
borderline significantly elevated.
Cancer Case Record Review: We reviewed medical and other records of
individuals with kidney and testicular cancers to try to determine smoking,
occupational and residential histories. A number of preexisting data sources were
used including: hospital medical records; cancer registry records; death
certificates; newspaper obituaries; Motor Vehicle records; and city and telephone
directories. Results: The case record review did not reveal any unusual patterns in
terms of age, gender, year of diagnosis, cell type, or mortality rate among
individuals with kidney or testicular cancer. There was some evidence of an
increased prevalence of smoking among those with kidney cancer and some
indication that several individuals diagnosed with testicular and kidney cancer
may have been recent arrivals to the study area.
Conclusions/Recommendations: The purpose of the additional analyses
reported in the draft for public comment follow-up report was to provide
information on certain cancers and reproductive outcomes which were elevated in
the initial health statistics review. Although these additional analyses could not
determine whether there was a causal relationship between VOC exposures in the
study area and the increased risk of several health outcomes that were observed,
they did provide more information to help guide additional follow-up. The March
2007 public comment report provided a list of follow-up options for consideration
and stated, "Although an analytical (case-control) epidemiologic study of cancer
or birth defects within this community is not recommended at this time, we
describe several follow up options for discussion with the Endicott community. A
case-control study would be the preferable method for progressing with this type
of investigation, but the potentially exposed population in the Endicott area is too
small for conducting a study that would be likely to be able to draw strong
conclusions about potential health risks.
Alternative follow-up options were discussed at meetings with Endicott
stakeholders and were the subject of responses to comments on the draft report.
From these discussions and written responses, NYS DOH has noted community
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-457 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
interest in two possible options for future activities: a health statistics review
based on historic outdoor air emissions modeling, and a multi-site epidemiologic
study examining cancer outcomes in communities across the state with VOC
exposures similar to Endicott. NYS DOH has considered these comments and
examined whether these options would be able to accomplish one of two goals:
either to advance the scientific knowledge about the relationship between VOC
exposure and health outcomes or to be part of a response plan to address
community concerns.
An additional health statistics review using historic outdoor air emission
modeling results to identify and study a larger population of residents potentially
exposed to TCE is not likely to meet either of these goals at this time. Because of
the limitations of the health statistics review for drawing conclusions about cause
and effect, conducting an additional health statistics review is not likely to
increase our understanding of whether exposures in the Endicott area are linked to
health outcomes. Limitations with the available historic outdoor air data also
would make it difficult to accurately define the appropriate boundaries for the
exposure area. ATSDR historic outdoor air emissions modeling activity was
unable to model TCE due to a lack of available records.
A multi-site epidemiologic study of health outcomes in communities across
the state with VOC exposures similar to Endicott offers some promise of meeting
the goal of advancing the scientific knowledge about the relationship between
VOC exposures and health outcomes. The community has indicated its preference
that such a study focus on cancer outcomes. Given the complex issues involved in
conducting such a study (e.g., tracking down cases or their next of kin after many
years, participants' difficulty in accurately remembering possible risk factors from
many years ago, and the long time period between exposure to a carcinogen and
the onset of cancer), we do not consider a multisite case-control study of cancer as
the best option at this time. An occupational cancer study is a better option than a
community-based study because it can better incorporate information about past
workplace exposures and could use corporate records to assist in finding
individual employees many years after exposure.
Heart defects have been associated with TCE exposure in other studies. Given
the shorter latency period, and thus the shorter time period in which other risk
factors could come into play, a multi-site study of heart defects has some merit as
a possible option. Currently, NYS DEC and NYS DOH are investigating many
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-458 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
communities around New York State which could have VOC exposure patterns
similar to Endicott, and thus could be included in such a multi-site epidemiologic
study. However, in most of these communities exposure information sufficient to
identify a study population is not yet available. NYS DOH will continue to
evaluate these areas as additional exposure information becomes available, with
the goal of identifying other communities for possible inclusion in a multi-site
epidemiologic study of heart defects.
NYS DOH will continue to keep the Endicott community and stakeholders
informed about additional information regarding other communities with
exposures similar to those that occurred in the Endicott area. NYS DOH staff will
be available as needed to keep interested Endicott area residents up-to-date on the
feasibility of conducting a multi-site study that includes the Endicott area.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-459 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
B.3 .3 .7.1.3. Study description and comment. Health statistics review conducted by NYS
DOH because of concerns about possible exposures to VOCs in Endicott area groundwater
and vapor intrusion into residences examined cancer incidence between 1980 and 2001 and
birth outcomes among residents living in a study area defined by soil vapor sampling and
exposure modeling. The reviews were supported by ATSDR and conclusions presented in
final reports (ATSDR, 2006a, 2008) have received external comment, but the studies have
not been published in the open peer-reviewed literature. Testing of soil gas and indoor air
of more than 300 properties, including 176 residences [location not identified] for VOCs
detected TCE levels ranging from 0.18-140 jig/m3 ; other VOCs less commonly detected
included perchloroethylene, 1,1-dichloroethane, 1,1-dichloroethylene, 1,2-dichloroethylene,
vinyl chloride, 1,1,1-trichloroethane, methylene chloride, and Freon 113. A model was
developed to predict VOC presence in soil vapor based on measured results ("Groupwater
Vapor Project, Endicott, New York: Summary of findings, working draft. Cited in
ATSDR," 2006). Subsequent sampling and data collection verified this model. Initial
study area boundaries were determined based on the extent of the probable soil vapor
contamination greater than 10 jig/m3 of VOCs as defined by the model. Contour lines of
modeled VOC soil vapor contamination levels, known as isopleths, were mapped using a
geographic information system. This study area is referred to as the Eastern study area in
ATSDR (2006a, 2008). Additional sampling west of the initial study area identified further
contamination with the contaminant in this area primarily identified as perchloroethylene
at levels ranging from 0.1-3.5 jig/m3 in an area referred to as the Western study area
(ATSDR, 2006a, 2008). The source of perchloroethylene contamination was not known. A
digital map of the 2000 Census block boundaries was overlaid on these areas of
contamination. The study areas were then composed of a series of blocks combined to
conform as closely to the areas of soil vapor contamination as possible.
Incident cancer cases for 18 sites, including cancer in children 19 years or younger,
between 1980 and 2001 and obtained from the New York State Cancer Registry and addresses
were geocoded to identify cases residing in the study area. The observed numbers of site-
specific cancers were compared to that expected calculated using age-sex-year specific cancer
incidence rates for New York State exclusive of New York City and population estimates 1980,
1990 and 2000 Censuses. Expected numbers of site-specific cancer did not include adjustment
for race in (ATSDR, 2006a); however, race was examined in the 2008 follow-up study which
compared cancer incidence among the white residents in the study area to that of whites in New
York State (ATSDR, 2008). Over the 22-year period, a total of 347 incident cancers were
observed among residents in the study area, 339 of these were in white residents. Less than
6 cases of cancers in children 19 years of age or younger were identified and ATSDR (2006a)
did not present a SIR for this grouping, similar to their treatment of other site-specific cancers
with less than six observed cases.
The follow-up analysis by ATDR (2008) reviewed medical records of kidney and
testicular cancer cases for smoking, occupational and residential histories, and restricted the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-460 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
statistical analysis to white residents, given the few numbers of observed cancers in the small
population of nonwhite residents. Limiting the analysis to only white individuals in the study
area had little effect on overall cancer rates or SIR estimates (ATSDR, 2006a). As observed in
ATSDR (2006a), statistically significant excess risks were observed for kidney cancer in both
sexes and testicular cancer in males. In addition, lung cancer estimate risks in males and in
males and females were of the same magnitude in both analyses, but confidence intervals
excluded a risk of 1.0 in the ATSDR (2008) analyses which adjusted for race. Review of
medical records for the 15 kidney cancer and six testicular cancer cases provided limited
information about personal exposures and potential risk factors because of incomplete reporting
in records. The record review did not reveal any unusually patterns in either kidney cancer or
testicular cancer in terms of age, year of diagnosis, anatomical site, cell type, or mortality rate.
Occupational history suggested possible workplace chemical exposure for roughly half of the
13 kidney cancer cases and none of the testicular cancer cases whose medical records included
occupational history. For smoking, half of the 9 kidney cancer cases and some (number not
identified) of the 3 testicular cancer cases with such information in medical records were current
or former smokers; smoking habits were not reported for the other cases. Last, examination of
city and phone directories revealed while half the kidney cancer cases as long term Endicott
residents, several cases of testicular cancer were among residents who recently moved into the
Endicott area.
These health surveys are descriptive; they provide evidence of cancer rates in a
geographical area with some documented exposures to several VOCs including trichloroethylene
but are unable to identify possible etiologic factors for the observed elevations in kidney,
testicular, or lung cancers. The largest deficiency is the lack of exposure assessment, notably
historical exposure, to individual subjects. Review of city and phone directories suggests some
kidney and testicular cancer cases were among recently-arrived residents, a finding inconsistent
with a cancer latent period; however, of greater importance is the finding of cancers among
subjects with long residential history. On the other hand, the population in the study areas has
declined over the past 20 years (ATSDR, 2006a) and residents who may have moved from the
study area were not included, introducing potential bias if cancer risks differed in these
individuals. The medical history review suggests several risk factors including smoking and
occupational exposure as important to kidney and testicular cancer observations. Lacking
information for all subjects, there is uncertainty regarding the additive effect of other potential
risk factors such as smoking to residential exposures. For this reason, while excesses in several
incident cancers are observed in these reports, potential etiological risk factors are ill-defined,
and the weight these studies contribute in the overall weight-of-evidence analysis is limited.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-461 DRAFT—DO NOT CITE OR QUOTE
-------
ATSDR (Agency for Toxic Substances and Disease Registry). (2006a). Health Consultation. Cancer and Birth Outcome
Analysis, Endicott Area, Town of Union, Broome County, New York. Health Statistics Review. Atlanta, GA: U.S.
Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry. May
26, 2006.
ATSDR (Agency for Toxic Substances and Disease Registry). (2008). Health Consultation. Cancer and Birth Outcome
Analysis, Endicott Area, Town of Union, Broome County, New York. Health Statistics Review Follow-Up. Atlanta, GA: U.S.
Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry. May
15, 2008.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
This health statistics review examined incidence for 18 types of cancer in residents
living in the Village of Endicott at the time of diagnosis. This study was not
designed to identify possible etiologic factors.
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Subjects are incident cases of cancer of the 18 types of cancers including childhood
cancer (all cancers in children <19 yrs of age) reported to the New York Cancer
Registry between 1980 and 2001 among residents in two areas of the Village of
Endicott, NY.
The expected number of cancer cases for the period was calculated using cancer
incidence rates for New York State exclusion of New York City and population
estimates from 1980, 1990, and 2000 Censuses.
CATEGORY B: ENDPOINT MEASURED
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD 9th Revision.
-------
CATEGORY C: TCE-EXPOSURE CRITERIA
-------
Exposure assessment approach, including
This geographic-based study does not develop quantitative estimates of exposure,
adoption of JEM and quantitative exposure
rather study boundaries are defined using soil gas and indoor air monitoring data and
estimates
computer modeling.

Testing of soil gas and indoor air of more than 300 properties, including

176 residences (location not identified) in the Eastern study area for VOCs detected

-3
TCE levels ranging from 0.18-140 (J,g/m ; other VOCs less commonly detected

included perchloroethylene, 1,1-dichloroethane, 1,1-dichloroethylene,

1,2-dichloroethylene, vinyl chloride, 1,1,1-trichloroethane, methylene chloride, and

Freon 113. A model was developed to predict VOC presence in soil vapor based on

measured results ("Groupwater Vapor Project, Endicott, New York: Summary of

findings, working draft. Cited in AT SDR," 2006). Subsequent sampling and data

collection verified this model. Initial study area boundaries were determined based

"3
on the extent of the probable soil vapor contamination greater than 10 (j,g/m of

VOCs as defined by the model.

Additional sampling west of the initial study area identified further contamination

with the contaminant in this area primarily identified as perchloroethylene at levels

"3
ranging from 0.1-3.5 (J,g/m in an area referred to as the Western study area.

The study areas were then composed of a series of blocks combined to conform as

closely to the areas of soil vapor contamination as possible.

Cancer incident cases in residents at the time of diagnosis in the two areas were

included in the study.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up
No information.
>50% cohort with full latency
No information.
CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Record study.
Blinded interviewers

-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents
Record study.
CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
347 total cancers in males and females among an estimated population size of
3,540 (1980)-3,002 (2000).
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age and sex (AT SDR, 2006a).
Age, sex, race (ATSDR, 2008).
Medical record review of 15 kidney and 6 testicular cancer cases provided limited
information on smoking, work history, and residential history for a small percentage
of these cases (ATSDR, 2008).
Statistical methods

Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
B.3.3.8. Studies in Arizona
B.3 .3 .8.1. Studies of West Central Phoenix Area, Maricopa County, AZ.
B.3.3.8.1.1. Aickin et al. (1992), Aickin (2004).
B.3.3.8.1.1.1. Aickin et al. (1992) author's abstract.
Reports of a suspected cluster of childhood leukemia cases in West Central
Phoenix have led to a number of epidemiological studies in the geographical area.
We report here on a death certificate-based mortality study, which indicated an
elevated rate ratio of 1.95 during 1966-1986, using the remainder of the Phoenix
standard metropolitan statistical area (SMSA) as a comparison region. In the
process of analyzing the data from this study, a methodology for dealing with
denominator variability in a standardized mortality ratio was developed using a
simple linear Poisson model. This new approach is seen as being of general use in
the analysis of standardized rate ratios (SRR), as well as being particularly
appropriate for cluster investigations.
B.3.3.8.1.1.2. Aickin (2004) author's abstract.
BACKGROUND AND OBJECTIVES: Classical statistical inference has attained
a dominant position in the expression and interpretation of empirical results in
biomedicine. Although there have been critics of the methods of hypothesis
testing, significance testing (P-values), and confidence intervals, these methods
are used to the exclusion of all others. METHODS: An alternative metaphor and
inferential computation based on credibility is offered here. RESULTS: It is
illustrated in three datasets involving incidence rates, and its advantages over both
classical frequentist inference and Bayesian inference, are detailed.
CONCLUSION: The message is that for those who are unsatisfied with classical
methods but cannot make the transition to Bayesianism, there is an alternative
path.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-468 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
B.3.3.8.1.1.3. Study description and comment. This study by staff of Arizona Department of
Health Sendees of leukemia mortality or incidence rates among children <19 years old living
at the time a death in West Central Phoenix in Maricopa County assume residence in the
defined geographical area as a surrogate of undefined exposures. Aickin et al. (1994) adopted
a classical statistical approach, linear Poisson regression, to estimate age-, sex- and calendar
year adjusted relative risks for leukemia mortality between 1966 and 1986 among children 19
years of younger living in the study area at the time of death. Leukemia mortality rates for the
rest of Maricopa County, excluding the study area and three additional geographic areas
previously identified with hazardous waste contamination, were selected as the referent
(Aickin et al, 1992). Aickin (2004) adopt inferential or Bayesian approaches to test whether
childhood leukemia incidence between 1966 and 1986 would confirm the mortality analysis
observation.
Both studies use residence at time of diagnosis or death in the study area, West Central
Phoenix, AZ, as the exposure surrogate; specific exposures such as drinking water contaminates
are not examined nor is information on parental factors considered in the analysis. Some
information on potential exposures in the community-at-large may be obtained from reports
prepared by the AZ DHS of epidemiologic investigations of cancer mortality rates among
residents of this area. Aickin et al. (1992) is the published finding on childhood leukemia. Past
exposure to the population of West Central Phoenix to environmental contaminants has been
difficult to quantify because of a paucity of environmental monitoring data (ADHS, 1990).
Community concerns about the environment focused on TCE found in drinking water in the late
1981, air pollution, from benzene emission from a nearby major gasoline storage and distribution
facility, and pesticide residues. Two wells that occasionally supplemented the water supply in
West Central Phoenix were closed after TCE was detected at the wellhead. The levels of TCE
measured at the time contamination was detected were 8.9 ppb and 29.0 ppb (report does not
identify the number of samples nor concentration ranges). The period over which contaminant
water had been supplied from these wells was not known nor whether significant exposure to the
population occurred after mixing with surface water. Other compounds identified in the
contaminated plume besides TCE included 1,1-dichloroethylene, trans-1,2-dichloroethylene,
chloroform, and chromium. The exposure assessment in the AZ DHS reports is inadequate to
describe exposure potential to TCE to subjects of Aickin et al. (1992) and Aickin (2004).
Moreover, potential etiologic factors for the observed elevated estimated relative risk for
childhood leukemia bases are not examined. While these studies support an inference of
elevated childhood leukemia rates in residents of West Central Phoenix, these studies provide
little information on childhood leukemia and TCE exposure and contribute little weight in the
overall weight-of-evidence analysis of cancer and TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-469 DRAFT—DO NOT CITE OR QUOTE
-------
Aickin M, Chapin CA, Flood TJ, Englender SJ, Caldwell GG. (1992). Assessment of the spatial occurrence of childhood
leukemia mortality using standardized rate ratios with a simple linear Poisson model. Int J Epidemiol 21:649-655.
Aickin M. (2004). Bayes without priors. J Clin Epidemiol 57:4-13.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Aickin et al. (1992) illustrated a methodologic approach to reduce variability in rate
ratios from small-sized populations. Childhood leukemia mortality in a
geographically-defined area in central Phoenix, AZ, was the case study adopted to
illustrate methodologic approach. The analysis was not designed to examine possible
etiologic factors.
The purpose of Aickin (2004) "was to determine whether a 1.95 standardized
mortality ratio [19] for leukemia in West Central Phoenix (compared to the
remainder of Maricopa County) would be confirmed in an incidence study" [p. 8],
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Leukemia deaths among children <19 yrs of age between the years 1966 and 1986
and with addresses on death certificates in the geographically-defined study area
were identified from Arizona death tapes.
Referent group is childhood leukemia mortality rate of all other Maricopa residents
excluding the study area and 3 other areas with identified hazardous waste
contamination (Aickin et al., 1992).
Incident cases of childhood leukemia (<19 yrs) among residents living in study area
were identified from the Arizona Cancer Registry and from cancer registry and
medical record reviews at 13 area hospitals (ADHS, 1990).
CATEGORY B: ENDPOINT MEASURED
Levels of health outcome assessed
Cancer mortality (Aickin et al., 1992).
Cancer incidence (Aickin, 2004).
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
Mortality—ICD 7, ICDA 8, ICD 9 (Flood, 1988).
Incidence—ICD-O.
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Residence in geographical area is a surrogate of undefined exposures; possible
exposures are not identified in the paper.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
<90% face-to-face
Record study.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10% proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
38 childhood leukemia deaths over a period of 21 yrs.
49 childhood leukemia incident cases over a period of 21 yrs.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and year (1966-1969, 1979-1981, 1982-1986).
Statistical methods
Poisson regression using 1970, 1980, and 1985 population estimates from U.S.
Bureau of the Census.
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
B.3.3.8.2. Studies in Tucson, Pima County, AZ.
B.3.3.8.2.1. Arizona Department of Health Services (1990,1995).
B.3.3.8.2.1.1. Arizona Department of Health Services (1990) author's summary.
In 1986, responding to community concerns about possible past exposure to low
levels of trichloroethylene in drinking water, a committee appointed by the
Director of the Arizona Department of health Services recommended that the
incidence of childhood leukemia and testicular cancer be studied in the population
residing in the Tucson Airport Area (TAA). The study reported here was
designed to count all cancer cases occurring in 0-19 year-old Pima County
residents, and all testicular cancer cases in Pima County residents of all ages,
during the 1970-1986 time period. Based on the incidence rates in the remainder
of Pima County, approximately seven cases of childhood leukemia and
approximately eight cases of testicular cancer would have been expected in the
TAA. Eleven cases of leukemia (SIR = 1.50, 95% C.I. 0.76-2.70) and six cases of
testicular cancer (SIR = 0.78, 95% C.I. 0.32-1.59) were observed. Statistical
analyses showed that the incidence rates of these cancers were not significantly
elevated. Additionally, it was determined that the rates of other childhood cancers
in the TAA, grouped as lymphoma, brain/CNS and other, were not significantly
elevated. The childhood leukemia, childhood cancer, and testicular cancer rates
in Pima County were comparable to rates in other states and cities participating in
the National Cancer Institute's Surveillance Epidemiology and End Results
Program.
B.3.3.8.2.1.2. Arizona Department of Health Services (1995) author's summary.
In 1986, responding to community concerns about possible past exposure to low
levels of trichloroethylene in drinking water, a committee appointed by the
Director of the Arizona Department of health Services recommended that the
incidence of childhood leukemia and testicular cancer be studied in the population
residing in the Tucson Airport Area (TAA). The study reported here was
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-473 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
designed to count all cancer cases occurring in 0-19 year-old Pima County
residents, and all testicular cancer cases in Pima County residents of all ages,
during the 1986-1991 time period. Based on the incidence rates in the remainder
of Pima County, approximately 3 cases of childhood leukemia and 4 cases of
testicular cancer would have been expected in the TAA. Three cases of leukemia
(SIR= .80; 95% C.I. 0.31-2.05) and 4 cases of testicular cancer (SIR = .93; 95%
C.I. 0.37-2.35) were observed. Statistical analyses showed that the incidence
rates of these cancers were not significantly elevated. Additionally, results
indicate no statistically elevated incidence rates of childhood lymphoma,
brain/CNS, and other childhood cancers, for ages 0-19, in the TAA. No
consistent pattern of disease occurrence was observed when comparing the past
incidence and mortality studies conducted by ADHS in the TAA with this present
study regarding disease categories.
B . 3 .3 .8 .2 .1.3. Study description and comment. These reports by staff of AZ DHS of cancer
incidence among children <19 years old and of testicular cancer incidence among males living
at the time a diagnosis in 1970-1986 or 1987-1991 in the Tucson International Airport Area
(TAA) of southwest Tucson (ADHS, 1990,1995) compared to incidence rates for the rest of
Pima County were conducted in response to community concerns about cancer and possible
past exposure to low levels of TCE in drinking water. In contrast to studies in West Central
Phoenix, findings from the 1990 and 1995 AZ DHS studies in Tuscon have not been published
in the peer-reviewed literature. Childhood cancers included were leukemia, brain/CSN,
lymphoma, and a broad category of all other cancers diagnosed in children <19 years old.
The Arizona Cancer Registry and reviews of medical records of 10 Pima county hospitals
served as sources for identifying incident cases. The study area was defined as a geographical
area overlaying a plume of contaminated groundwater and was comprised offive census
tracts. The approximate areas boundaries areAjo Way (north), Los Reales Road (south),
Country Club Road (east), and the Santa Cruz River (west). Adjacent census tracts in Pima
County were aggregated into four separate study areas and incident cancer rates during the
1970-1986 time period (ADHS, 1990) or 1987-1991 (ADHS, 1995) of the aggregated 4-area
census tract, excluding the TAA area., were used to calculate expected numbers of cancers
using the indirect standardization method and population estimates from 1960,1970,1975,
1980, and 1985 (ADHS, 1990) or 1990 (ADHS, 1995) of the U.S. Bureau of Census. A
secondary analysis of AZ DHS (1990) compared the incidence rate of childhood leukemia and
testicular cancer among Pima County residents to that reported to the SEER for a similar time
period.
These studies assume residence in the defined geographical area as a surrogate of
undefined exposures. The reports do not identify specific exposures for the individual subjects
and some information on exposures in the community-at-large may be obtained from Public
Health Assessments of the Tucson International Airport Area Superfund Site prepared by the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-474 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
AZ DHS for the ATSDR (2000, 2001). The TAA site includes one main contaminated
groundwater plume with smaller areas of groundwater contamination located east of the main
plume. Insufficient data existed to evaluate groundwater contamination prior to 1981. Studies
conducted by AZ DHS in 1981-1982 showed TCE concentrations of above 5 ppb, the maximum
contaminate level, in the main groundwater plume with TCE detected in some municipal
drinking water wells at concentrations of up to 239 ppb. An ATSDR health assessment
conducted in 1988 indicated that soil and groundwater in the Main Plume had been contaminated
by chromium and volatile organic compounds such as TCE and dichloroethylene (DCE)
(ATSDR, 2000). Sampling of private wells from 1981 through 1994 identified both drinking and
irrigation private wells in and near the TIAA with TCE concentrations ranging from nondetect to
120 ppb. Concentrations of other VOCs and chromium from the 1980s are not presented in the
ATSDR reports. Besides groundwater, areas of contaminated soil and sediment have also been
identified as part of the site. The "Three Hangars" area of the airport was found to contain
polychlorinated biphenyls in drainage areas with migration off-site into residential
neighborhoods (ATSDR, 2001). The exposure assessment in these studies is inadequate to
describe exposure to TCE. The studies provide little information on cancer risks and TCE
exposure and carry little weight in the overall weight-of-evidence analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-475 DRAFT—DO NOT CITE OR QUOTE
-------
AZ DHS (Arizona Department of Health Services). (1990). The incidence of childhood leukemia and testicular cancer in Pima
County: 1970-1986. Prepared by the Arizona Department of Health Services, Division of Disease Prevention, Office of Risk
Assessment and Investigation, Office of Chronic Disease Epidemiology. September 17,1990.
AZ DHS (Arizona Department of Health Services). (1995). Update of the incidence of childhood leukemia and testicular
cancer in Southwest Tucson, 1987-1991. Prepared by the Arizona Department of Health Services, Office of Risk Assessment
and Investigation, Disease Prevention Services. June 6,1995.

Description
CATEGORY A: STUDY DESIGN
Clear articulation of study objectives or
hypothesis
Yes, from ADHS (1990), "1) To determine whether there was an elevated incidence
of leukemia or other cancers among children residing in the Tucson Airport Area
(TAA) and 2) To determine whether there was an elevated incidence of testicular
cancer in males in the TAA."
From ADHS (1995), "The objective of this study is to determine whether the
incidence rates of childhood leukemia (ages 0-19) and testicular cancer in males of
all ages were significantly elevated in the TAA when compared to the rest of Pima
County for the years 1987 through 1991."
Selection and characterization in cohort
studies of exposure and control groups and of
cases and controls in case-control studies is
adequate
Cases are identified from the Arizona Cancer Registry and review of medical records
at 10 Pima County hospitals. The referent is incidence rates for the remaining
population of Pima County, excluding the study area.
CATEGORY B: ENDPOINT MEASURED
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Levels of health outcome assessed
Cancer incidence.
Changes in diagnostic coding systems for
lymphoma, particularly non-Hodgkin's
lymphoma
ICD-0 and ICD-9 or equivalent codes from ICDA-8, ICD-7, HICDA, or SNODO.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
CATEGORY C: TCE-EXPOSURE CRITERIA
Exposure assessment approach, including
adoption of JEM and quantitative exposure
estimates
Residence in geographical area is a surrogate of undefined exposures; possible
exposures are not identified in the paper.
CATEGORY D: FOLLOW-UP (COHORT)
More than 10% loss to follow-up

>50% cohort with full latency

CATEGORY E: INTERVIEW TYPE
<90% face-to-face
Record study.
Blinded interviewers

CATEGORY F: PROXY RESPONDENTS
>10%) proxy respondents

CATEGORY G: SAMPLE SIZE
Number of deaths in cohort mortality studies;
numbers of total cancer incidence studies;
numbers of exposed cases and prevalence of
exposure in case-control studies
ADHS (1990), 31 childhood cancers—11 leukemia cases, 2 lymphoma,
3 CNS/Brain, and 15 other, and 6 testicular cancers.
ADHS (1995), 11 childhood cancers—3 leukemia, 1 lymphoma, 2 CNS/Brain, and 5
other, and 4 testicular cancers.
CATEGORY H: ANALYSIS
Control for potential confounders in statistical
analysis
Age, sex, and year.
Statistical methods
SIRs calculated using indirect standardization.
-------

§•
to
s
a, 5
o a
TO 3
s
o
o
s
Co
I?
f? Ng
^ 3
>9
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
8
TO
Co
0
S
1
a.
VO
to
o
Exposure-response analysis presented in
published paper
No.
Documentation of results
Yes.
-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	B-480 DRAFT—DO NOT CITE OR QUOTE
-------
APPENDIX C
Meta-Analysis of Cancer Results from
Epidemiological Studies
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-i DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
C.l. METHODOLOGY
An initial review of the epidemiological studies indicated some evidence for associations
between trichloroethylene (TCE) exposure and non-Hodgkin lymphoma (NHL) and cancers of
the kidney and liver (see Section 4.1). To investigate further these possible associations, we
performed meta-analyses of the epidemiological study results for these three cancer types. There
was suggestive evidence for some other cancer types, as well; however, fewer TCE studies
reported relative risk (RR) estimates for these other site-specific cancers, and meta-analysis was
not attempted for these cancer types (see Section 4.1). In addition, at the request of our Science
Advisory Board (SAB, 2011), we conducted a meta-analysis of lung cancer in the TCE cohort
studies to address the issue of smoking as a possible confounder in the kidney cancer studies (see
Section 4.4.2.3).
Meta-analysis provides a systematic way to combine study results for a given effect
across multiple (sufficiently similar) studies. The resulting summary (weighted average)
estimate is a quantitatively objective way of reflecting results from multiple studies, rather than
relying on a single study, for instance. Combining the results of smaller studies to obtain a
summary estimate also increases the statistical power to observe an effect, if one exists.
Furthermore, meta-analyses typically are accompanied by other analyses of the epidemiological
studies, including analyses of publication bias and investigations of possible factors responsible
for any heterogeneity across studies.
Given the diverse nature of the epidemiological studies for TCE, random-effects models
were used for the primary analyses, and fixed-effect analyses were conducted for comparison.
Both approaches combine study results (in this case, RR estimates) weighted by the inverse
variance; however, they differ in their underlying assumptions about what the study results
represent and how the variances are calculated. For a random-effects model, it is assumed that
there is true heterogeneity across studies and that both between-study and within-study
components of variation need to be taken into account; this was done using the methodology of
DerSimonian and Laird (1986). For a fixed-effect model, it is assumed that the studies are all
essentially measuring the same thing and all the variance is within-study variance; thus, for the
fixed-effect model, the RR estimate from each study is simply weighted by the inverse of the
(within-study) variance of the estimate.
Studies for the meta-analyses were selected as described in Appendix B, Section II-9.
Because each of the cancer types being evaluated is considered rare in the populations being
studied (all have lifetime risks < 10%, and all but lung cancer have lifetime risks < 3%), the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
different measures of relative risk (e.g., odds ratios, risk ratios, and rate ratios) are good
approximations of each other (Rothman and Greenland, 1998) and are included together as RR
estimates in the meta-analyses. (In addition, the meta-analyses of lung cancer and liver cancer
comprised only cohort studies and, thus, no odds ratios were included in those analyses.) The
general approach for selecting RR estimates was to select the reported RR estimate that best
reflected an RR for TCE exposure vs. no TCE exposure (overall effect). When multiple
estimates were available for the same study based on different subcohorts with different
inclusion criteria, the preference for overall exposure was to select the RR estimate that
represented the largest population in the study, while trying to minimize the likelihood of TCE
exposure misclassification. A subcohort with more restrictive inclusion criteria was selected if
the basis was to reduce exposure misclassification (e.g., including only subjects with more
probable TCE exposure) but not if the basis was to reflect subjects with greater exposure (e.g.,
routine versus any exposure).
When available, RR estimates from internal analyses were selected over standardized
incidence or mortality ratios (SIRs, SMRs) and adjusted RR estimates were generally selected
over crude estimates. Incidence estimates would normally be preferred to mortality estimates;
however, for the two studies providing both incidence and mortality results, incidence
ascertainment was for a substantially shorter period of time than mortality follow-up, so the
endpoint with the greater number of cases was used to reflect the results that had better case
ascertainment. Furthermore, RR estimates based on exposure estimates that discounted an
appropriate lag time prior to disease onset were typically preferred over estimates based on
untagged exposures, although few studies reported lagged results.
For separate analyses, an RR estimate for the highest exposure group was selected from
studies that presented results for different exposure groups. Exposure groups based on some
measure of cumulative exposure were preferred, if available; however, often duration was the
sole exposure metric used.
Sensitivity analyses were generally done to investigate the impact of alternate selection
choices, as well as to estimate the impact of study findings that were not reported. Specific
selection choices are described in the following subsections detailing the actual analyses.
The meta-analysis calculations are based on (natural) logarithm-transformed values.
Thus, each RR estimate was transformed to its natural logarithm (referred to here as "log RR,"
the conventional terminology in epidemiology), and either an estimate of the standard error (SE)
of the log RR was obtained, from which to estimate the variance for the weights, or an estimate
of the variance of the log RR was calculated directly. If the reported 95% confidence interval
limits were proportionally symmetric about the observed RR estimate (i.e., upper confidence
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
limit/RR ~ RR/lower confidence limit), then an estimate of the SE of the log RR estimate was
obtained using the formula
\log(UCL)-log(LCL)\
(Eq. C-l)
where UCL is the upper confidence limit and LCL is the lower confidence limit (for 90%
confidence intervals [CIs], the divisor is 3.29) (Rothman and Greenland, 1998). In all the TCE
cohort studies reporting SMRs or SIRs as the overall RR estimates, reported CIs were calculated
assuming the number of deaths (or cases) is approximately Poisson distributed. In such cases,
the CIs are not proportionally symmetric about the RR estimate (unless the number of deaths is
fairly large), and the SE of the log RR estimate was estimated as the inverse of the square root of
the observed number of deaths (or cases) (Breslow and Day, 1987). In some case-control
studies, no overall odds ratio (OR) was reported, so a crude OR estimate was calculated as
OR = (a/b)/(c/d), where a, b, c, and d are the cell frequencies in a 2 x 2 table of cancer cases vs.
TCE exposure, and the variance of the log OR was estimated using the formula
in accordance with the method proposed by Woolf (1955), as described by Breslow and Day
(1980).

(Eq. C-2)
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
9	C-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The analyses that were performed for this assessment include
•	meta-analyses to obtain overall summary estimates of RR (denoted RRm)
•	heterogeneity analyses
•	analyses of the influence of single studies on the summary estimates
•	analyses of the sensitivity of the summary estimates to alternate study inclusion
selections or to alternate selections of RR estimates from a study
•	publication bias analyses
•	meta-analyses to obtain summary estimates for the highest exposure groups in studies
that provide data by exposure group, and
•	consideration of some potential sources of heterogeneity across studies.
The analyses were conducted using Microsoft Excel spreadsheets and the software package
Comprehensive Meta-Analysis, Version 2 (© 2006, Biostat, Inc.). Funnel plots and cumulative
analyses plots were generated using the Comprehensive Meta-Analysis software, and forest plots
were created using SAS, Version 9.2 (© 2002-2008, SAS Institute Inc.).
The heterogeneity (or homogeneity) analysis tests the hypothesis that the study results are
homogeneous, i.e., that all the RR estimates are estimating the same population RR and the total
variance is no more than would be expected from within-study variance. Heterogeneity was
assessed using the statistic Q described by DerSimonian and Laird (1986). The g-statistic
represents the sum of the weighted squared differences between the summary RR estimate
(obtained under the null hypothesis, i.e., using a fixed-effect model) and the RR estimate from
each study, and, under the null hypothesis, Q approximately follows a % distribution with
degrees of freedom equal to the number of studies minus one. However, this test can be under-
powered when the number of studies is small, and it is only a significance test, i.e., it is not very
informative about the extent of any heterogeneity. Therefore, the / value (Higgins et al., 2003)
was also considered. / = 100% x (Q- df)IQ, where Q is the ^-statistic and df is the degrees of
freedom, as described above. This value estimates the percentage of variation that is due to
study heterogeneity. Typically, / values of 25%, 50%, and 75% are considered low, moderate,
and high amounts of heterogeneity, respectively. For a negative value of (Q - df), I2 is set to 0%,
indicating no observable heterogeneity.
Subgroup analyses were sometimes conducted to examine whether or not the combined
RR estimate varied significantly between different types of studies (e.g., case-control vs. cohort
studies). In such subgroup analyses of categorical variables (e.g., study design), analysis of
variance was used to determine if there was significant heterogeneity between the subgroups.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Applying analysis of variance to meta-analyses with two subgroups (df = 1), (^between subgroups =
Qoverall ~ (Ssubgroupi + 0subgrouP2) = 2-value2, where ^overall is the (^-statistic calculated across all the
studies and <2subgrouPi and 0SubgrouP2 are the O-statisties calculated within each subgroup.
Publication bias is a systematic error that occurs if statistically significant studies are
more likely to be submitted and published than non-significant studies. Studies are more likely
to be statistically significant if they have large effect sizes (in this case, RR estimates); thus, an
upward bias would result in a meta-analysis if the available published studies have higher effect
sizes than the full set of studies that were actually conducted. One feature of publication bias is
that smaller studies tend to have larger effect sizes than larger studies, since smaller studies need
larger effect sizes in order to be statistically significant. Thus, many of the techniques used to
analyze publication bias examine whether or not effect size is associated with study size.
Methods used to investigate potential publication bias for this assessment included funnel plots,
which plot effect size vs. study size (actually, SE vs. log RR here); the "trim and fill" procedure
of Duval and Tweedie (2000), which imputes the "missing" studies in a funnel plot (i.e., the
studies needed to counterbalance an asymmetry in the funnel plot resulting from an ostensible
publication bias) and recalculates a summary effect size with these studies present; forest plots
(arrays of RRs and CIs by study) sorted by precision (i.e., SE) to see if effect size shifts with
study size; Begg and Mazumdar rank correlation test (Begg and Mazumdar, 1994), which
examines the correlation between effect size estimates and their variances after standardizing the
effect sizes to stabilize the variances; Egger's linear regression test (Egger et al., 1997), which
tests the significance of the bias reflected in the intercept of a regression of effect size/SE on
1/SE; and cumulative meta-analyses after sorting by precision to assess the impact on the
summary effect size estimate of progressively adding the smaller studies.
C.2. META-ANALYSIS FOR NON-HODGKIN LYMPHOMA (NHL)
C.2.1. Overall Effect of TCE Exposure
C.2.1.1. Selection of RR Estimates
The selected RR estimates for NHL associated with TCE exposure from the selected
epidemiological studies are presented in Table C-l for cohort studies and in Table C-2 for case-
control studies. Some of the more recent case-control studies classified NHLs along the lines of
the recent WHO/REAL classification system (World Health Organization/Revised European-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
American Classification of Lymphoid Neoplasms) (Harris et al., 2000), which recognizes
lymphocytic leukemias and multiple myelomas (plasma cell myelomas) as (non-Hodgkin)
lymphomas; however, most of the available TCE studies reported NHL results according to the
International Classification of Diseases (ICD), Revisions 7, 8, and 9, using a traditional
definition of NHL that excluded lymphocytic leukemias and multiple myelomas and focused on
ICD-7, -8, -9 codes 200 + 202. For consistency of endpoint in the NHL meta-analyses, RR
estimates for ICD 200 + 202 were selected, wherever possible; otherwise, estimates for the
classification(s) best approximating this traditional definition of NHL were selected. In addition,
many of the studies provided RR estimates only for males and females combined, and we are not
aware of any basis for a sex difference in the effects of TCE on NHL risk; thus, wherever
possible, RR estimates for males and females combined were used. The only study of much size
(in terms of number of NHL cancer cases) that provided results separately by sex was
Raaschou-Nielsen (2003). This study reports an insignificantly higher SIR for females (1.4,
95% CI: 0.73, 2.34) than for males (1.2, 95% CI: 0.98, 1.52).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-6 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-l. Selected RR estimates for NHL associated with TCE exposure (overall effect) from cohort studies
S"4
§•
to
s
Study
RR
95%
LCL
95%
UCL
RR type
log RR
SE(log RR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)
1.81
0.78
3.56
SIR
0.593
0.354
None
ICD-7 200 + 202.
Axelson et
al. (1994)
1.52
0.49
3.53
SIR
0.419
0.447
1.36 (0.44, 3.18)
with estimated
female
contribution to SIR
added (see text)
ICD-7 200 and 202. Results reported
separately; combined assuming Poisson
distribution. Results reported for males only,
but there was a small female component to the
cohort.
Boice et
al.(1999)
1.19
0.83
1.65
SMR
0.174
0.267
1.19 (0.65, 1.99)
for routine
potential exposure
ICD-9 200 + 202. For any potential exposure.
Greenland
et al. (1994)
0.76
0.24
2.42
OR
-0.274
0.590
None
ICD-8 200-202. Nested case-control study.
Hansen et
al. (2001)
3.1
1.3
6.1
SIR
1.13
0.354
None
ICD-7 200 + 202. Male and female results
reported separately; combined assuming
Poisson distribution.
Morgan et
al. (1998)
1.01
0.46
1.92
SMR
0.00995
0.333
1.36 (0.35, 5.21)
unpublished RR
for ICD 200 (see
text)
ICD 200 + 202. Results reported by Mandel et
al. (2006). ICD Revision 7, 8, or 9, depending
on year of death.
Raaschou-
Nielsen et
al. (2003)
1.24
1.01
1.52
SIR
0.215
0.104
1.5(1.2, 2.0) for
subcohort with
expected higher
exposures
ICD-7 200 + 202.
Radican et
al. (2008)
1.36
0.77
2.39
Mortality
HR
0.307
0.289
None
ICD-8,-9 200 + 202; ICD-10 C82-C85. Time
variable = age; covariates = sex and race.
Referent group is workers with no chemical
exposures.
o
2 »
s s
a, Co'
TO Sj-
§ ^
>S
>S
TO
TO'
*
o
VO
to
o
o
-------
is*
Table C-l. Selected RR estimates for NHL associated with TCE exposure (overall effect) from cohort studies
(continued)
Study
RR
95%
LCL
95%
UCL
RR type
log RR
SE(log RR)
Alternate RR
estimates
Comments
Zhao et al.
(2005)
1.44
0.90
2.30
Mortality
RR
0.363
0.239
Incidence RR:
0.77 (0.42, 1.39)
Boice 2006 SMR
for ICD-9 200 +
202: 0.21 (0.01,
1.18)
All lymphohematopoietic cancer (ICD-9 200-
208), not just 200 + 202. Males only; adjusted
for age, SES, time since first employment.
Mortality results reflect more exposed cases
(33) than do incidence results (17). Overall RR
estimated by combining across exposure
groups (see text). Boice (2006b) cohort
overlaps Zhao et al. (2005) cohort; just
1 exposed death for ICD 200 + 202; 9 for
200-208 (vs. 33 in Zhao et al.).
o
vo
to
o
HR: hazard ratio; SES: socioeconomic status
O
oo
-------
K
s
TO
>3
Table C-2. Selected RR estimates for NHL associated with TCE exposure from case-control studies"
S"4
§•
to
s
Study
RR
95%
LCL
95%
UCL
log RR
SE(log
RR)
NHL type
Comments
Cocco et al.
(2010)
0.8
0.5
1.1
-0.223
0.201
NHL
Grouping consistent with traditional NHL definition provided
by author (see text). High-confidence subgroup. Adjusted
for age, sex, center, and education.
Hardell et al.
(1994)
7.2
1.3
42
1.97
0.887
NHL
Rappaport classification system. Males only; controls
matched for age, place of residence, vital status.
Miligi et al.
(2006)
0.93
b
b
-0.0726
0.168
NHL + CLL
NCI Working Formulation. Crude OR; overall adjusted OR
not presented.
Nordstrom
et al. (1998)
1.5
0.7
3.3
0.405
0.396
HCL
HCL specifically. Males only; controls matched for age and
county; analysis controlled for age.
Persson and
Frederikson
(1999)
1.2
0.5
2.4
0.182
0.400
NHL
Classification system not specified. Controls selected from
same geographic areas; ORs stratified on age and sex.
Purdue et al.
(2011)
1.4
0.8
2.4
0.336
0.280
NHL
ICD-O-3 codes 967-972. Probable-exposure subgroup.
Adjusted for age, sex, SEER center, race, and education.
Siemiatycki
(1991)
1.1
0.5
2.5
0.0953
0.424
NHL
ICD-9 200 + 202. SE and 95% CI calculated from reported
90% CIs; males only; adjusted for age, income, and
cigarette smoking index.
Wang et al.
(2009)
1.2
0.9
1.8
0.182
0.177
NHL
ICD-O M-9590-9595, 9670-9688, 9690-9698, 9700-9723.
Females only; adjusted for age, family history of
lymphohematopoietic cancers, alcohol consumption, and
race.
o
2 »
s s
to
O
S
>S
TO
TO'
*
o
VO
to
o
aThe RR estimates are all ORs for incident cases.
bNot calculated.
NHL: non-Hodgkin lymphoma; CLL: chronic lymphocytic leukemia; HCL: hairy cell leukemia (a subgroup of NHL).
O
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Most of the selections in Tables C-l and C-2 should be self-evident, but some are
discussed in more detail here, in the order the studies are presented in the tables. For Axelson et
al. (1994), in which a small subcohort of females was studied but only results for the larger male
subcohort were reported, the reported male-only results were used in the primary analysis;
however, an attempt was made to estimate the female contribution to an overall RR estimate for
both sexes and its impact on the meta-analysis. Axelson et al. (1994) reported that there were no
cases of NHL observed in females, but the expected number was not presented. To estimate the
expected number, the expected number for males was multiplied by the ratio of female-to-male
person-years in the study and by the ratio of female-to-male age-adjusted incidence rates for
NHL.4 The male results and the estimated female contribution were then combined into an RR
estimate for both sexes assuming a Poisson distribution, and this alternate RR estimate for the
Axelson et al. (1994) study was used in a sensitivity analysis.
For Boice et al. (1999), results for "any potential exposure" were selected for the primary
analysis, because this exposure category was considered to best represent overall TCE exposure,
and results for "potential routine exposure", which was characterized as reflecting workers
assumed to have received more cumulative exposure, were used in a sensitivity analysis.
The Greenland et al. (1994) study is a case-control study nested within a worker cohort,
and we treat it here as a cohort study (see Appendix B, Section II-9.1). Greenland et al. (1994)
report results only for all lymphomas, including Hodgkin lymphoma (ICD-8 201).
For Morgan et al. (1998), the reported results did not allow for the combination of
ICD 200 and 202, so the SMR estimate for the combined 200 + 202 grouping was taken from the
meta-analysis paper of Mandel et al. (2006), who included one of the investigators from the
Morgan et al. (1998) study. RR estimates for overall TCE exposure from internal analyses of the
Morgan et al. (1998) cohort data were available from an unpublished report (EHS, 1997) (the
published paper only presented the internal analyses results for exposure subgroups), but only for
ICD 200; from these, the RR estimate from the Cox model which included age and sex was
selected, because those are the variables deemed to be important in the published paper (Morgan
et al., 1998). Although the results from internal analyses are generally preferred, in this case the
SMR estimate was used in the primary analysis and the internal analysis RR estimate was used in
4Person-years for men and women <79 years were obtained from Axelson et al. (1994): 23516.5 and 3691.5,
respectively. Lifetime age-adjusted incidence rates for NHL for men and women were obtained from the National
Cancer Institute's 2000-2004 SEER-17 (Surveillance Epidemiology and End Results from 17 geographical areas)
database (http://seer.cancer.gov/statfacts/html/nhl. html): 23.2/100,000 and 16.3/100,000, respectively. The
calculation for estimating the expected number of cases in females in the cohort assumes that the males and females
have similar TCE exposures and that the relative distributions of age-related incidence risk for the males and
females in the Swedish cohort are adequately represented by the ratios of person-years and U.S. lifetime incidence
rates used in the calculation.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
a sensitivity analysis because the latter estimate represented an appreciably smaller number of
deaths (3, based on ICD 200 only) than the SMR estimate (9, based on ICD 200 + 202).
For Raaschou-Nielsen et al. (2003), results for the full cohort were used for the primary
analysis and results for the subcohort with expected higher exposure levels (> 1-year duration of
employment and year of 1st employment before 1980) were used in a sensitivity analysis.
Raaschou-Nielsen et al. (2003), in their Table 3, also present overall results for NHL with a lag
time of 20 years; however, they use a definition of lag that is different from a lagged exposure in
which exposures prior to disease onset are discounted and it is not clear what their lag time
actually represents5, thus these results were not used in any of the meta-analyses for NHL.
For Radican et al. (2008), the Cox model hazard ratio (HR) from the 2000 follow-up was
used. In the Radican et al. (2008) Cox regressions, age was the time variable, and sex and race
were covariates. It should also be noted that the referent group is composed of workers with no
chemical exposures, not just no exposure to TCE.
For Zhao et al. (2005), RR estimates were only reported for ICD-9 200-208 (all
lymphohematopoietic cancers), and not for 200 + 202 alone. Given that other studies have not
reported associations between leukemias and TCE exposure, combining all lymphohematopoietic
cancers would dilute any NHL effect, and the Zhao results are expected to be an underestimate
of any TCE effect on NHL alone. Another complication with the Zhao et al. (2005) study is that
no results for an overall TCE effect are reported. We were unable to obtain any overall estimates
from the study authors, so, as a best estimate, the results across the "medium" and "high"
exposure groups were combined, under assumptions of group independence, even though the
exposure groups are not independent (the "low" exposure group was the referent group in both
cases). Zhao et al. (2005) present RR estimates for both incidence and mortality; however, the
time frame for the incidence accrual is smaller than the time frame for mortality accrual and
fewer exposed incident cases (17) were obtained than deaths (33). Thus, because better case
ascertainment occurred for mortality than for incidence, the mortality results were used for the
primary analysis, and the incidence results were used in a sensitivity analysis. A sensitivity
analysis was also done using results from Boice et al. (2006b) in place of the Zhao et al. (2005)
RR estimate. The cohorts for these studies overlap, so they are not independent studies and
should not be included in the meta-analysis concurrently. Boice et al. (2006b) report an RR
estimate for an overall TCE effect for NHL alone; however, it is based on far fewer cases
(1 death in ICD-9 200 + 202; 9 deaths for 200-208) and is an SMR rather than an internal
analysis RR estimate, so the Zhao et al. (2005) estimates are preferred for the primary analysis.
5 In their Methods section, Raaschou-Nielsen et al. (2003) define their lag period as the period "from the date of
first employment to the start of follow-up for cancer".
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-l 1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
For the case-control studies, the main issue was the NHL classifications. Cocco et al.
(2010) present results for NHLs classified according to the WHO/REAL classification system,
i.e., including lymphocytic leukemias and multiple myelomas. For this meta-analysis, we were
able to obtain results for a grouping of lymphomas generally consistent with the traditional
definition of NHL (T-cell lymphomas and B-cell lymphomas, excluding Hodgkin lymphomas,
chronic lymphocytic leukemias [CLLs], multiple myelomas, and unspecified lymphomas) from
Dr. Cocco (personal communication from Pierluigi Cocco, University of Cagliari, Italy, to
Cheryl Scott, U.S. EPA, 19 March 2011; see Section 4.6.1.2). The results used in the meta-
analyses are for the high-confidence subgroup, which included workers with jobs with a
"certain" probability of exposure and > 90% of workers exposed (5.5% of cases).
Hardell et al. (1994) used the Rappaport classification system, which, according to
Weisenburger (1992) is consistent with the traditional definition of NHL.
Miligi et al. (2006) include CLLs in their NHL results, consistent with the current
WHO/REAL classification. Also, Miligi et al. (2006) do not report an overall adjusted RR
estimate, so a crude estimate of the OR was calculated for the two TCE exposure categories
together vs. no TCE exposure.
The Nordstrom et al. (1998) study was a case-control study of hairy cell leukemias
(HCLs), so only results for HCL were reported. HCLs are a subgroup of NHLs under current
classification systems, but they were not included in the traditional definition of NHL.
Persson and Frederikson (1999) did not report the classification system used.
According to Schenk et al.(2009), Purdue et al. (2011) used ICD-O-3 codes 967-972,
which are generally consistent with the traditional definition of NHL. The results used in the
meta-analyses are for the probable-exposure subgroup, which includes workers with at least one
job assigned an exposure probability of > 50% (3.8% of cases).
According to Zhang et al. (2004), Wang et al. (2009) used ICD-O-2 codes M-9590-9595,
9670-9688, 9690-9698, 9700-9723, which are consistent with the traditional definition of NHL
(i.e., ICD-7, -8, -9 codes 200 + 202).
No alternate RR estimates were considered for any of the case-control studies of NHL.
For the Cocco et al. (2010) and Purdue et al. (2011) studies, the RR estimates used are for a
higher confidence subgroup. No overall results for the full studies were presented to use as
alternative estimates. Results for lower confidence subgroups were presented separately, but no
attempt was made to combine the results across confidence groups because these results were not
independent, as they relied on the same referent groups.
An alternate analysis was done including only the studies for which RR estimates for the
traditional definition of NHL were available. In this analysis, Miligi et al. (2006), Nordstrom et
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
al.(1998), Persson and Frederikson (1999), and Greenland et al. (1994) were omitted and the
Boice et al. (2006b) cohort study was used instead of Zhao et al. (2005).
C.2.1.2. Results of Meta-Analyses
Results from some of the meta-analyses that were conducted on the epidemiological
studies of TCE and NHL are summarized in Table C-3. The summary estimate (RRm) from the
primary random-effects meta-analysis of the 17 studies was 1.23 (95% CI: 1.07, 1.42) (see
Figure C-l). No single study was overly influential; removal of individual studies resulted in
RRm estimates that ranged from 1.18 (with the removal of Hansen (2001)) to 1.27 (with the
removal of Miligi et al. (2006) or Cocco et al. (2010)) and were all statistically significant (all
with p < 0.02). Removal of Hardell (1994) whose RR estimate is a relative outlier (see
Figure C-l), only decreased the RRm estimate to 1.21 (1.07, 1.38), since this study does not
contribute a lot of weight to the meta-analysis. Removal of studies other than Hansen (2001)
resulted in RRm estimates that were all greater than 1.20.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-13 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-3. Summary of some meta-analysis results for TCE (overall) and NHL
S"4
>}'
§•
to
s
Analysis
# of
studies
Model
Summary
RR
estimate
(RRm)
95%
LCL
95%
UCL
Heterogeneity
Comments
All studies
17
Random
1.23
1.07
1.42
Not significant
(p = 0.16)
P = 26%
Statistical significance of RRm not dependent on
individual studies.
Fixed
1.21
1.08
1.35

Cohort
9
Random
1.33
1.13
1.58
Not significant
(p = 0.34)
P = 12%
Not significant difference between CC and cohort
studies (p = 0.19).
Fixed
1.31
1.14
1.51
Not significant difference between CC and cohort
studies (p = 0.08).
Case-control
8
Random
1.11
0.89
1.38
Not significant
(P = 0.22)
P = 27%

Fixed
1.07
0.90
1.28

Alternate
RR
selections3
17
Random
1.20
1.03
1.39
Not significant
(P = 0.11)
P = 27%
With estimated Zhao (2005) overall RR for incidence
rather than mortality.
17
Random
1.22
1.03
1.43
Not significant
(p = 0.09)
r = 27%
With Boice (2006b) study rather than Zhao (2005).
17
Random
1.23
1.07
1.42
Not significant
(p = 0.16)
P = 27%
With estimated female contribution to Axelson
(1994).
17
Random
1.24
1.07
1.44
Not significant
(P = 0.10)
P = 27%
With Boice (1999) potential routine exposure SMR.
17
Random
1.25
1.08
1.44
Not significant
(P = 0.17)
P = 27%
With Morgan et al. (1998) unpublished RR.
17
Random
1.28
1.09
1.49
Not significant
(p = 0.09)
r = 27%
With Raaschou-Nielsen (2003) subgroup expected to
have higher exposures
Alternate
analysis;
13
Random
1.27
1.05
1.55
Not significant
(p = 0.054)
Omitting Miligi (2006), Nordstrom (1998), Persson
and Frederikson (1999), and Greenland (1994), and
5
o
s
, ^ 1
§ I 5
5 a, co-
c §- a
5 S S-
?• i
r: §
TO ^
s >•
S to
^ 1
5J-
TO
VO
to
o
O
-------
K
s
TO
>1
O
K. S
§ I
r;
5
ri'
r;
S"4
>!'
§•
Si
a
>3*
s
s*
<§>
>!
>{
rS*
*
traditional
definition of
NHL only





? = 42%
including Boice (2006b) instead of Zhao (2005).
Highest
exposure
groups
13
Random
1.43
1.13
1.82
Not significant
(p = 0.30)
r = 14%
Statistical significance not dependent on single study.
See Table C-5 for results with alternate RR
selections.
Fixed
1.43
1.16
1.75


§•
to
¦? >>
a
o
^k
TO
O
a
>;
"Changing the primary analysis by one alternate RR each time; more details on alternate RR estimates in text.
o
vo
o
to
o
O
-------
TCE Exposure 
-------
model (1.21 vs. 1.23) and had a slightly narrower CI (1.08-1.35 vs. 1.07-1.42). In addition, non-
significant heterogeneity was apparent in each of the meta-analyses with alternate RR selections
— /"-values ranged from 0.09 to 0.17 and / -values ranged from 25% to 36%.
To investigate the heterogeneity, subgroup analyses were done examining the cohort and
case-control studies separately. With the random-effects model (and tau-squared not pooled
across subgroups), the resulting RRm estimates were 1.33 (95% CI: 1.13, 1.58) for the cohort
studies and 1.11 (0.89, 1.38) for the case-control studies. There was residual heterogeneity in
each of the subgroups, but in neither case was it statistically significant. / -values were 12% for
the cohort studies, suggesting low heterogeneity, and 27% for the case-control studies,
suggesting low-to-moderate heterogeneity. The difference between the RRm estimates for the
cohort and case-control subgroups was not statistically significant. Some thought was given to
further analyses to investigate the source(s) of the heterogeneity, such as qualitative tiering or
subgroups based on likelihood for correct exposure classification or on likelihood for higher vs.
lower exposures across the studies. Ultimately, these approaches were rejected because in many
of the studies it was difficult to judge (and weight) the extent of exposure misclassification or the
degree of TCE exposure with any precision. In other words, there was inadequate information to
reliably assess either the extent to which each study accurately classified exposure status or the
relative TCE exposure levels and prevalences of exposure to different levels across studies. See
Section C.2.3 below for a qualitative discussion of some potential sources of heterogeneity.
As discussed in Section C.l, publication bias was examined in several different ways.
The funnel plot in Figure C-2 suggests some relationship between RR estimate and study size (if
there were no relationship, the studies would be symmetrically distributed around the summary
RR estimate rather than veering towards higher RR estimates with increasing SEs), although the
observed asymmetry is highly influenced by the Hardell (1994) study, which is a relative outlier
and which contributes little weight to the overall meta-analysis, as discussed above. The Begg
and Mazumdar (1994) rank correlation test and Egger et al.'s (1997) linear regression test were
not statistically significant (the one-tailed />values were 0.18 and 0.07, respectively); it should be
noted, however, that both of these tests have low power. Duval and Tweedie et al. (2000)'s trim-
and-fill procedure yielded a summary RR estimate (under the random-effects model) of 1.15
(95%) CI: 0.97, 1.36) when the 4 studies deemed missing from the funnel plot were filled into the
meta-analysis (these studies are filled in so as to counter-balance the apparent asymmetry of the
more extreme values in the funnel plot). Eliminating the Hardell (1994) study made little
difference to the results of the publication bias analyses. The results of a cumulative
meta-analysis, incorporating studies with increasing SE one at a time, are depicted in Figure C-3.
This procedure is a transparent way of examining the effects of including studies with increasing
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-17 DRAFT—DO NOT CITE OR QUOTE
-------
SE. The figure shows that the summary RR estimate is 1.16 after inclusion of the 4 largest (i.e.,
most precise) studies, which constitute about 50% of the weight. The RRm estimate decreases to
1.10 with the inclusion of the next most precise study, which contributes another 9% of the total
weight. The RRm estimate increases to 1.22 with inclusion of the 6 next most precise studies;
this summary estimate represents 11 of the 17 studies and about 87% of the weight. Adding in
the 6 least precise studies (13% of the weight) barely increases the RRm estimate further. In
summary, there is some evidence of potential publication bias in this data set. It is uncertain,
however, that this reflects actual publication bias rather than an association between effect size
and SE resulting for some other reason, e.g., a difference in study populations or protocols in the
smaller studies. Furthermore, if there is publication bias in this data set, it does not appear to
account completely for the findings of an increased NHL risk.
Funnel Plot of Standard Error by Log rate ratio
0.0
0.2
04
o
LU
¦o
0.6
w
0.8
1.0
-2.0	-1.5	-1.0	-0.5	0.0	0.5	1.0	1.5	2.0
Log rate ratio
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-18 DRAFT—DO NOT CITE OR QUOTE
-------
Figure C-2. Funnel plot of SE by log RR estimate for TCE and NHL studies.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-19 DRAFT—DO NOT CITE OR QUOTE
-------
TCE and Non-Hodgkin Lymphoma
Study name
Cumulative statistics
Cumulative rate ratio (95% CI)


Lower
Upper


Point
limit
limit
p-Value
Raaschou-Nielsen 2003
1.240
1.0152
1.5146
0.035
Miligi 2006
1.108
0.8412
1.4588
0.466
Wang 2009
1.153
0.9785
1.3577
0.089
Boice 1999 any
1.163
1.0103
1.3383
0.036
Cocco 2010
1.096
0.9355
1.2846
0.256
Zhao 2005 mort
1.124
0.9682
1.3054
0.125
Purdue 2011
1.142
0.9959
1.3100
0.057
Radican 2008
1.156
1.0212
1.3082
0.022
Morgan 1998
1.152
1.0223
1.2976
0.020
Arrttila 1995
1.167
1.0374
1.3123
0.010
Hansen 2001
1.221
1.0378
1.4368
0.016
Nordstrom 1998
1.227
1.0508
1.4330
0.010
Persson&Fredrikson 1999
1.222
1.0555
1.4148
0.007
Siemiatycki 1991
1.215
1.0575
1.3958
0.006
Axelson 1994
1.218
1.0667
1.3910
0.004
Greenland 1994
1.210
1.0627
1.3767
0.004
Hardell 1994
1.233
1.0676
1.4247
0.004

1.233
1.0676
1.4247
0.004
0.5
random effects model; cumulative analysis, sorted by SE
Figure C-3. Cumulative meta-analysis of TCE and lymphomaNHL studies,
progressively including studies with increasing SEs.
C.2.2. LymphomaNHL Effect in the Highest Exposure Groups
C.2.2.1. Selection of RR Estimates
The selected RR estimates for NHL in the highest TCE exposure categories, for studies
that provided such estimates, are presented in Table C-4. All 8 cohort studies (but not the nested
case-control study of Greenland et al. (1994) and 5 of the 8 case-control studies did report NHL
risk estimates categorized by exposure level. As in Section C.2.1.1 for the overall risk estimates,
estimates to best correspond to NHL as represented by ICD-7, -8, and -9 200 and 202 were
selected, and, wherever possible, RR estimates for males and females combined were used.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-20 DRAFT—DO NOT CITE OR QUOTE
-------
As above for the overall TCE effect, for Axelson et al. (1994), in which a small subcohort
of females was studied but only results for the larger male subcohort were reported, the reported
male-only high-exposure group results were used in the primary analysis; however, an attempt
was made to estimate the female contribution to a high-exposure group RR estimate for both
sexes and its impact on the meta-analysis. To estimate the expected number in the highest
exposure group for females, the expected number in the highest exposure group for males was
multiplied by the ratio of total female-to-male person-years in the study and by the ratio of
female-to-male age-adjusted incidence rates for NHL. The RR estimate for both sexes was used
as an alternate RR estimate for the Axelson et al. (1994) study in a sensitivity analysis.
For Boice et al. (1999), only results for workers with "any potential exposure" (rather
than "potential routine exposure") were presented by exposure category, and the referent group is
workers not exposed to any solvent.
For Hansen et al. (2001), exposure group data were presented only for males. To
estimate the female contribution to a highest exposure group RR estimate for both sexes, it was
assumed that the expected number of cases in females had the same overall-to-highest-exposure-
group ratio as in males. The RR estimate for both sexes was then calculated assuming a Poisson
distribution, and this estimate was used in the primary analysis. Hansen et al. (2001) present
results for three exposure metrics; the cumulative exposure metric was preferred for the primary
analysis, and results for the other two metrics were used in sensitivity analyses.
For Morgan et al. (1998), results did not allow for the combination of ICD 200 and 202,
so the highest exposure group RR estimate for ICD 200 only was used. The primary analysis
used results for the cumulative exposure metric, and a sensitivity analysis was done with the
results for the peak exposure metric.
For Radican et al. (2008), it should be noted that the referent group is composed of
workers with no chemical exposures, not just no exposure to TCE. In addition, results for
exposure groups (based on cumulative exposure scores) were reported separately for males and
females and were combined for this assessment using inverse-variance weighting, as in a fixed-
effect meta-analysis. Radican et al. (2008) present only mortality HR estimates by exposure
group; however, in an earlier follow-up of this same cohort, Blair et al. (1998) present both
incidence and mortality RR estimates by exposure group. The mortality RR estimate based on
the more recent follow-up of Radican et al. (Blair et al., 1998; 2008) (17 deaths in the highest
exposure group) was used in the primary analysis, while the incidence RR estimate based on
similarly combined results from Blair et al. (1998) (9 cases) was used as an alternate estimate in
a sensitivity analysis. Radican et al. (2008) also present results for categories based on
frequency and pattern of exposure; however, subjects weren't distributed uniquely across the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-21 DRAFT—DO NOT CITE OR QUOTE
-------
categories (the numbers of cases across categories exceeded the total number of cases), thus it
was difficult to interpret these results and they were not used in a sensitivity analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-22 DRAFT—DO NOT CITE OR QUOTE
-------
Table C-4. Selected RR estimates for NHL risk in highest TCE exposure groups
S"4
s
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE(log
RR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)
1.4
0.17
5.04
100+ |jmol/L
U-TCAa
0.336
0.707
none
SIR. ICD 200 + 202.
Axelson et
al. (1994)
6.25
0.16
34.83
>2-year
exposure
and 100+
mg/L U-TCA
1.83
1.00
5.62 (0.14, 31.3)
with estimated
female
contribution
added (see text)
SIR. ICD 200 + 202. Results reported for
males only, but there was a small female
component to the cohort.
Boice et al.
(1999)
1.62
0.82
3.22
>5-year
exposure
0.482
0.349
None
Mortality RR. ICD 200 + 202. For potential
routine or intermittent exposure. Adjusted for
date of birth, dates 1st and last employed,
race, and sex. Referent group is workers not
exposed to any solvent.
Hansen et
al. (2001)
2.7
0.56
8.0
>1080
months *
mg/m3
0.993
0.577
3.7 (1.0, 9.5) for
>75 months
exposure
duration
2.9 (0.79, 7.5) for
>19 mg/m3 mean
exposure
SIR. ICD 200 + 202. Exposure-group
results presented only for males. Female
results estimated and combined with male
results assuming Poisson distribution (see
text).
Morgan et
al. (1998)
0.81
0.1
6.49
High
cumulative
exposure
score
-0.211
1.06
1.31 (0.28, 6.08)
for med/high
peak vs. low/no
Mortality RR. ICD 200 only. Adjusted for
age and sex.
Raaschou-
Nielsen et
al. (2003)
1.6
1.1
2.2
>5 years in
subcohort
with
expected
higher
exposure,
levels
0.470
0.183
1.45 (0.99, 2.05)
for > 5 years in
full cohort, both
sexes combined
SIR. ICD 200+ 202.
s
o
i»"t-1
s § _
5 a, co-
c §- a
5 S S-
?• i
r: §
s >•
§ 3
*¦>1.
to
o
o
to
LtJ
-------
Table C-4. Selected RR estimates for NHL risk in highest TCE exposure groups (continued)
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE(log
RR)
Alternate RR
estimates
Comments
Radican et
al. (2008)
1.41
0.71
2.81
>25 unit-
years
0.337
0.350
Blair et al.
(1998) 0.97
(0.42, 2.2)
incidence RR
Mortality HR. ICD 200 + 202. Male and female
results presented separately and combined (see
text). Cox regression time variable = age;
covariate = race. Referent group is workers with
no chemical exposures.
Zhao et al.
(2005)
1.30
0.52
3.23
High
exposure
score
0.262
0.466
Incidence RR:
0.20 (0.03,
1.46)
Mortality RR. Results for all
lymphohematopoietic cancer (ICD-9 200-208),
not just 200 + 202. Males only; adjusted for age,
SES, time since first employment. Mortality
results reflect more exposed cases (6 in high-
exposure group) than do incidence results (1 in
high-exposure group).
Cocco et al.
(2010)
0.7
0.4
1.3
High
cumulative
exposure
-0.357
0.301
None
Incidence OR. Grouping consistent with
traditional NHL definition provided by author (see
text). High-confidence subgroup. Adjusted for
age, sex, center, and education.
Miligi et al.,
(2006)
1.2
0.7
2.0
Med/high
exposure
intensity
0.182
0.268
1.0 (0.5, 2.6)
for med/high
intensity and
>15-years
Incidence OR. NHL + CLL (see Section C.2.1.1).
Adjusted forage, sex, education, and area.
Purdue et
al. (2011)
3.3
1.1
10.1
Cumulative
exposure >
234,000
ppmxhours
1.194
0.566
2.3 (1.0, 5.0)
for highest
exposure fertile
(>112,320
ppmxhours)
ICD-O-3 codes 967-972. Probable-exposure
subgroup. Adjusted for age, sex, SEER center,
race, and education.
Siemiatycki
(1991)
0.8
0.2
3.3
Substantial
-0.223
0.719
None
Incidence OR. NHL. SE and 95% CI calculated
from reported 90% CIs. Males only; adjusted for
age, income, and cigarette smoking index.
Wang et al.
(2009)
2.2
0.9
5.4
Medium-
high
intensity
0.788
0.457
None
Incidence OR. NHL. Females only; adjusted for
age, family history of lymphohematopoietic
cancers, alcohol consumption, and race.
aMean personal trichloroacetic acid in urine. 1 |imol/L = 0.1634 mg/L.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
For Zhao et al. (2005), RR estimates were only reported for ICD-9 200-208 (all
lymphohematopoietic cancers), and not for 200 + 202 alone. Given that other studies have not
reported associations between leukemias and TCE exposure, combining all lymphohematopoietic
cancers would dilute any NHL effect, and the Zhao results are expected to be an underestimate
of any TCE effect on NHL alone. Zhao et al. (2005) present RR estimates for both incidence and
mortality in the highest exposure group; however, the time frame for the incidence accrual is
smaller than the time frame for mortality accrual and fewer incident cases (1) were obtained than
deaths (6), so the mortality results were used for the primary analysis to reflect the better case
ascertainment in the mortality data, and the incidence results were used in a sensitivity analysis.
Cocco et al. (2010) present exposure group results only for their high-confidence
subgroup, which included workers with jobs with a "certain" probability of exposure and > 90%
of workers exposed (5.5% of cases). Results for a grouping of lymphomas generally consistent
with the traditional definition of NHL (T-cell lymphomas and B-cell lymphomas, excluding
Hodgkin lymphomas, CLLs, multiple myelomas, and unspecified lymphomas) were kindly
provided by Dr. Cocco (personal communication from Pierluigi Cocco, University of Cagliari,
Italy, to Cheryl Scott, U.S. EPA, 19 March 2011; see Section 4.6.1.2).
Miligi et al. (2006) include CLLs in their NHL results, consistent with the current
WHO/REAL classifications. Miligi et al. (2006) report RR estimates for medium and high
exposure intensity overall and by duration of exposure; however, there was incomplete
information for the duration breakdowns (e.g., a case missing), so the RR estimate for med/high
exposure intensity overall was used in the primary analysis, and the RR estimate for med/high
exposure for >15 years was used in a sensitivity analysis.
Purdue et al. (2011) used ICD-O-3 codes 967-972, generally consistent with a traditional
definition of NHL. These investigators present exposure group results only for their probable-
exposure subgroup, which included workers with jobs with an assigned probability of exposure
of > 50% (3.8%) of cases). The exposure groups are cumulative exposure tertiles, with cutpoints
determined from the exposure distribution in the probably exposed controls. The highest
exposure tertile was further subdivided using the intra-category median. The highest exposure
group from the subdivided highest exposure tertile was used for the primary analysis (4 cases),
and the results for the complete highest tertile were used in a sensitivity analysis (9 cases).
Wang et al. (2009) used ICD-O-2 codes (M-9590-9595, 9670-9688, 9690-9698, 9700-
9723), consistent with the traditional definition of NHL (i.e., ICD-7, -8, -9 codes 200 + 202).
Wang et al. (2009) present exposure-group (low or medium/high intensity) results cross-
categorized by exposure probability (low and medium/high). The medium and high exposure-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
intensity category was used as the highest exposure group, although all of the subjects with
medium and high exposure intensity were in the low exposure-probability category.
C.2.2.2. Results of Meta-Analyses
Results from the meta-analyses that were conducted for NHL in the highest exposure
groups are summarized at the bottom of Table C-3 and reported in more detail in Table C-5. The
summary RR estimate from the primary random-effects meta-analysis of the 13 studies with
results presented for exposure groups was 1.43 (95% CI: 1.13, 1.82) (see Figure C-4). No single
study was overly influential; removal of individual studies resulted in RRm estimates that were all
statistically significant (all withp < 0.025) and that ranged from 1.38 (with the removal of
Purdue et al. (2011) ) to 1.57 (with the removal of Cocco (2010)) . In addition, the RRm estimate
was not highly sensitive to alternate RR estimate selections. Use of the 9 alternate selections,
individually, resulted in RRm estimates that were all statistically significant (all withp < 0.025)
and all in the narrow range from 1.40 to 1.49 (see Table C-5).
There was some heterogeneity apparent across the 13 studies, although it was not
statistically significant (p = 0.30). The / -value was 14%, suggesting low heterogeneity. This
small amount of heterogeneity is also indicated by the finding that the RRm from the fixed-effect
analysis had a slightly narrower CI (1.16-1.75 vs. 1.13-1.82), although the RRm estimates
themselves were essentially identical. In addition, non-significant heterogeneity was apparent in
each of the meta-analyses with alternate RR selections —^-values ranged from 0.12 to 0.37 and
I -values ranged from 9% to 37%.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-26 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-5. Summary of some meta-analysis results for TCE (highest exposure groups) and NHL
S"4
§•
to
s
o
* v	§
§ I	5
a,	Co'
§-	a
TO	Sj-
§	^
O
a
>S
>S
TO
'S
TO
*
O
VO
to
o
o
to
^1
Analysis
Model
Combined
RR
estimate
95% LCL
95% UCL
Heterogeneity
Comments
All studies (13)
Random
1.43
1.13
1.82
Not significant
(p = 0.30)
r = 14%
Statistical significance not dependent on single
study.
Fixed
1.43
1.16
1.75

Cohort Studies
(8)
Random
1.60
1.24
2.08
None
observable
(random =
fixed)
Not significant difference between CC and cohort
studies (p = 0.47).
Fixed
1.60
1.24
2.08
Not significant difference between CC and cohort
studies (p = 0.15).
Case-Control
Studies (5)
Random
1.29
0.76
2.20
NS (p = 0.08)
I2 = 53%

Fixed
1.18
0.84
1.64

Alternate RR
selections3
(all studies)
Random
1.40
1.11
1.75
NS (p = 0.33)
l2= 11%
With Raaschou-Nielsen (2003) full cohort instead
of subgroup expected to have higher exposures.
Random
1.40
1.09
1.80
NS (p = 0.25)
I2 =19%
With Blair et al. (1998) incidence RR instead of
Radican mortality HR.
Random
1.41
1.05
1.88
NS (p = 0.12)
I2 = 33%
With Zhao (2005) incidence.
Random
1.43
1.13
1.80
NS (p = 0.32)
I2 = 13%
With estimated female contribution for Axelson
(1994).
Random
1.43
1.15
1.78
NS (p = 0.37)
I2 = 9%
With Purdue (2011) highest cumulative exposure
fertile
Random
1.44
1.12
1.85
NS (p = 0.29)
I2 = 16%
With Miligi (2006) with >15 years.
Random
1.44
1.14
1.83
NS (p = 0.32)
I2 = 13%
With Morgan (1998) peak.
Random
1.45
1.14
1.86
NS (p = 0.25)
I2 =19%
With Hansen (2001) mean exposure.
Random
1.49
1.14
1.93
NS (p = 0.17)
I2 = 27%
With Hansen (2001) duration.
aChanging the primary analysis by one alternate RR estimate each time.
NS: not statistically significant; CC: case-control
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
TCE Exposme 
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
study with the next largest RR estimate (Purdue et al., 2011), whose removal results in the lowest
RRm estimate in the analyses of study influence (see above) does not eliminate the
heterogeneity. On the other hand, removal of the study with the lowest RR estimate (Cocco et
al., 2010), which also has a substantial amount of weight and whose removal results in the
highest RRm estimate in the analyses of study influence (see above), eliminates all of the
heterogeneity. This suggests that the result from Cocco et al. (2010) for the highest exposure
group might be an outlier, but it is unclear what about the study might account for this result
being inordinately low.
C.2.3. Discussion of NHL Meta-Analysis Results
The meta-analyses of the overall effect of TCE exposure on NHL suggest a small,
statistically significant increase in risk. The summary estimate from the primary random-effects
meta-analysis of the 17 studies was 1.23 (95% CI: 1.07, 1.42). This result was not overly
influenced by any single study, nor was it overly sensitive to individual RR estimate selections or
to restricting the analysis to only those studies for which RR estimates based on the traditional
definition of NHL were available, and in all of the influence and sensitivity analyses, the RRm
estimate was statistically significantly increased. Thus, the finding of an increased risk of NHL
associated with TCE exposure, though the increased risk is not large in magnitude, is robust.
There is some evidence of potential publication bias in this data set; however, it is
uncertain that this is actually publication bias rather than an association between SE and effect
size resulting for some other reason, e.g., a difference in study populations or protocols in the
smaller studies. Furthermore, if there is publication bias in this data set, it does not appear to
account completely for the finding of an increased NHL risk. For example, using Duval and
Tweedie et al. (2000)'s trim-and-fill procedure to impute the values from the 4 'missing' studies
that would balance the funnel plot yields an RRm estimate of 1.15 (95% CI: 0.97, 1.36).
Although there was some heterogeneity across the 17 studies, it was not statistically
significant (p = 0.16). The / -value was 26%, suggesting low-to-moderate heterogeneity.
Similarly, when subgroup analyses were done of cohort and case-control studies separately, there
was some observable heterogeneity in each of the subgroups, but it was not statistically
significant in either case. / -values were 12% for the cohort studies, suggesting low
heterogeneity, and 27% for the case-control studies, suggesting low-to-moderate heterogeneity.
In the subgroup analyses, the increased risk of NHL was strengthened in the cohort study
analysis and nearly eliminated in the case-control study analysis, although the subgroup RRm
estimates were not statistically significantly different. Study design itself is unlikely to be an
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
underlying cause of heterogeneity and, to the extent that it may explain some of the differences
across studies, is more probably a surrogate for some other difference(s) across studies that may
be associated with study design. Furthermore, other potential sources of heterogeneity may be
masked by the broad study design subgroupings. The true source(s) of heterogeneity across
these studies is an uncertainty. As discussed above, further quantitative investigations of
heterogeneity were ruled out because of database limitations. A qualitative discussion of some
potential sources of heterogeneity follows.
Study differences in exposure assessment approach, exposure prevalence, average
exposure intensity, and NHL classification are possible sources of heterogeneity. Many studies
included TCE assignment from information on job and task exposures, e.g., a job-exposure
matrix (JEM) (Boice et al., 1999; Boice et al., 2006b; Miligi et al., 2006; Morgan et al., 1998;
Radican et al., 2008; Siemiatycki, 1991; Zhao et al., 2005)(Cocco et al., 2010; Purdue et al.,
2011; Wang et al., 2009), or from an exposure biomarker in either breath or urine (Anttila et al.,
1995; Axelson et al., 1994; Hansen et al., 2001). Three case-control studies relied on self-
reported exposure to TCE (Hardell et al., 1994; Nordstrom et al., 1998; Persson and Fredrikson,
1999). Misclassification is possible with all exposure assessment approaches. No information is
available to judge the degree of possible misclassification bias associated with a particular
exposure assessment approach; it is quite possible that in some cohort studies, in which past
exposure is inferred from various data sources, exposure misclassification may be as great as in
population-based or hospital-based case-control studies. Approaches based upon JEMs can
provide order-of-magnitude estimates that are useful for distinguishing groups of workers with
large differences in exposure; however, smaller differences usually cannot be reliably
distinguished (NRC, 2006). Biomonitoring can provide information on potential TCE exposure
in an individual, but the biomarkers used aren't necessarily specific for TCE and they reflect only
recent exposures.
General population studies have special problems in evaluating exposure, because the
subjects could have worked in any job or setting that is present within the population ('t
Mannetje et al., 2002; Copeland et al., 1977; Mcguire et al., 1998; Nelson et al., 1994; NRC,
2006). Low exposure prevalence in the case-control studies may be another source of
heterogeneity. Prevalence of TCE exposure among cases in the case-control studies was low,
ranging from 3% in Siemiatycki (1991) to 13% in Wang et al. (2009). However, prevalence of
high TCE exposure in these case-control studies was even rarer—3% of all cases in Miligi et al.
(2006), 2% in Wang et al. (2009) and Cocco et al. (2010); high-confidence assessments; personal
communication from Pierluigi Cocco, University of Cagliari, Italy, to Cheryl Scott, U.S. EPA, 19
March 2011; see Section 4.6.1.2), 1% (with probable exposure) in Purdue (2011), and less than
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
1% in Siemiatycki (1991). Low exposure prevalence may be one of the underlying
characteristics differentiating the case-control and cohort studies and explaining some of the
heterogeneity across the studies.
Study differences in NHL groupings and in NHL classification schemes are another
potential source of heterogeneity in the meta-analysis, although restricting the meta-analysis to
only those studies for which RR estimates based on the traditional NHL definition were available
did not eliminate all heterogeneity. All studies included a broad but sometimes slightly different
group of lymphosarcoma, reticulum-cell sarcoma, and other lymphoid tissue neoplasms, with the
exception of the Nordstrom et al. (1998) case-control study, which examined hairy cell leukemia,
now considered a (non-Hodgkin) lymphoma, and the Zhao et al. (2005) cohort study, which
reported only results for all lymphohematopoietic cancers, including nonlymphoid types.
Persson and Fredrikson (1999) do not identify the classification system for defining NHL, and
Hardell et al. (1994) define NHL using the Rappaport classification system. Miligi et al.
(2006)vused the NCI Working Formulation and also considered CLLs as (non-Hodgkin)
lymphomas. Cocco et al. (2010) used the WHO/REAL classification system, which reclassifies
lymphocytic leukemias and NHLs as lymphomas of B-cell or T-cell origin and considers CLLs
and multiple myelomas as (non-Hodgkin) lymphomas; however, we were able to obtain results
generally consistent with the traditional NHL definition from Dr. Cocco, although lymphomas
not otherwise specified were excluded. Wang et al. (2009) defined NHL using ICD-O-2 codes
(M-9590-9595, 9670-9688, 9690-9698, 9700-9723), which is consistent with the traditional
definition of NHL (i.e., ICD-7, -8, -9 codes 200 + 202). Purdue et al. (2011) used ICD-O-3
codes 967-972, which is generally consistent with the traditional definition of NHL, although
this grouping doesn't include the malignant lymphomas of unspecified type coded as M-9590-
9599. The cohort studies (except for Zhao et al., 2005) and the case-control study of Siemiatycki
(1991) have some consistency in coding NHL, with NHL defined as lymphosarcoma and
reticulum-cell sarcoma (ICD code 200) and other lymphoid tissue neoplasms (ICD 202) using
the ICD Revisions 7, 8, or 9. Revisions 7 and 8 are essentially the same with respect to NHL;
under Revision 9, the definition of NHL was broadened to include some neoplasms previously
classified as Hodgkin lymphomas (Banks, 1992).
Thirteen of the 17 studies categorized results by exposure level. Different exposure
metrics were used, and the purpose of combining results across the different highest exposure
groups was not to estimate an RRm associated with some level of exposure, but rather to see the
impacts of combining RR estimates that should be less affected by exposure misclassification.
In other words, the highest exposure category is more likely to represent a greater differential
TCE exposure compared to people in the referent group than the exposure differential for the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
overall (typically any vs. none) exposure comparison. Thus, if TCE exposure increases the risk
of NHL, the effects should be more apparent in the highest exposure groups. Indeed, the RRm
estimate from the primary meta-analysis of the highest exposure group results was 1.43 (95% CI:
1.13, 1.82), which is greater than the RRm estimate of 1.23 (95% CI: 1.07, 1.42) from the overall
exposure analysis. The statistical significance of the increased RR estimate for the highest
exposure groups was not dependent on any single study, nor was it sensitive to individual RR
estimate selections. The robustness of this finding lends substantial support to a conclusion that
TCE exposure increases the risk of NHL.
Although there was some heterogeneity apparent across the 13 highest-exposure-group
studies, it was not statistically significant (p = 0.30). The / -value was 14%, suggesting low
heterogeneity. When subgroup analyses were done examining the cohort and case-control
studies separately, there was no residual heterogeneity in the cohort subgroup (/ = 0%).
Heterogeneity remained in the case-control subgroup, but it was not statistically significant (p =
0.08) — the / -value was 53%, suggesting moderate heterogeneity. In the subgroup analyses, the
increased risk of NHL was strengthened in the cohort study analysis and reduced in the case-
control study analysis, although the subgroup RRm estimates were not statistically significantly
different. As with the meta-analysis for overall TCE exposure discussed above, no further
attempt was made to quantitatively investigate potential sources of heterogeneity. It is, however,
noted that removal of the Cocco et al. (2010) study, whose removal had the greatest impact in the
analyses of study influence (RRm = 1.57, 95% CI: 1.27, 1.95), eliminates all of the
heterogeneity, suggesting that the RR estimate for the highest exposure group from that study is
a relative outlier.
C.3. META-ANALYSIS FOR KIDNEY CANCER
C.3 .1. Overall Effect of TCE Exposure
C.3 .1.1. Selection of RR Estimates
The selected RR estimates for kidney cancer associated with TCE exposure from the
epidemiological studies are presented in Table C-6 for cohort studies and in Table C-7 for
case-control studies. The majority of the cohort studies reported results for all kidney cancers,
including cancers of the renal pelvis and ureter (i.e., ICD-7 180; ICD-8 and -9 189.0-189.2;
ICD-10 C64-C66); whereas the majority of the case-control studies focused on renal cell
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
carcinoma (RCC), which comprises roughly 85% of kidney cancers. Where both all kidney
cancer and RCC were reported, the primary analysis used the results for RCC, because RCC and
the other forms of kidney cancer are very different cancer types and it seemed preferable not to
combine them; the results for all kidney cancers were then used in a sensitivity analysis. The
preference for the RRC results alone is supported by the results in rodent cancer bioassays,
where TCE-associated rat kidney tumors are observed in the renal tubular cells (Section 4.3.5),
and in metabolism studies, where the focus of studies for the GSH conjugation pathway
(considered the primary metabolic pathway for kidney toxicity) is in renal cortical and tubular
cells (Sections 3.3.3.2 and 4.3.6).
As for NHL, many of the studies provided RR estimates only for males and females
combined, and we are not aware of any basis for a sex difference in the effects of TCE on kidney
cancer risk; thus, wherever possible, RR estimates for males and females combined were used.
Of the three larger (in terms of number of cases) studies that did provide results separately by
sex, Dosemeci et al. (1999) suggest that there may be a sex difference for TCE exposure and
RCC (OR= 1.04 [95% CI: 0.6, 1.7] in males and 1.96 [1.0, 4.0] in females), while
Raaschou-Nielsen et al. (2003) report the same SIR (1.2) for both sexes and crude ORs
calculated from data from the Pesch et al. (2000b) study (provided in a personal communication
from Baeta Pesch, Forschungsinstitut fur Arbeitsmedizin (BGFA), to Cheryl Scott, U.S. EPA,
21 February 2008) are 1.28 for males and 1.23 for females. Radican et al. (2008) and Hansen et
al. (2001) also present some results by sex, but both of these studies have too few cases to be
informative about a sex difference for kidney cancer.
Most of the selections in Tables C-6 and C-7 should be self-evident, but some are
discussed in more detail here, in the order the studies are presented in the tables. For Axelson et
al. (1994), in which a small subcohort of females was studied but only results for the larger male
subcohort were reported, the reported male-only results were used in the primary analysis;
however, as for NHL, an attempt was made to estimate the female contribution to an overall RR
estimate for both sexes and its impact on the meta-analysis. Axelson et al. (1994) reported
neither the observed nor the expected number of kidney cancer cases for females. It was
assumed that none was observed. To estimate the expected number, the expected number for
males was multiplied by the ratio of female-to-male person-years in the study and by the ratio of
female-to-male age-adjusted incidence rates for kidney cancer.6 The male results and the
6Person-years for men and women <79 years were obtained from Axelson et al. (1994): 23516.5 and 3691.5,
respectively. Lifetime age-adjusted incidence rates for cancer of the kidney and renal pelvis for men and women
were obtained from the National Cancer Institute's 2000-2004 SEER-17 (Surveillance Epidemiology and End
Results from 17 geographical locations) database (http://seer.cancer.gov/statfacts/html/kidrp.html): 17.8/100,000
and 8.8/100,000, respectively. The calculation for estimating the expected number of cases in females in the cohort
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-33 DRAFT—DO NOT CITE OR QUOTE
-------
1	estimated female contribution were then combined into an RR estimate for both sexes assuming
2	a Poisson distribution, and this alternate RR estimate for the Axelson et al. (1994) study was
3	used in a sensitivity analysis.
assumes that the males and females have similar TCE exposures and that the relative distributions of age-related
incidence risk for the males and females in the Swedish cohort are adequately represented by the ratios of person-
years and U.S. lifetime incidence rates used in the calculation.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-34 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
s
Table C-6. Selected RR estimates for kidney cancer associated with TCE exposure (overall effect) from
cohort studies
Study
RR
95%
LCL
95%
UCL
RR type
log RR
SE(log RR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)
0.87
0.32
1.89
SIR
-0.139
0.408
None
ICD-7 180.
Axelson et al.
(1994)
1.16
0.42
2.52
SIR
0.148
0.408
1.07 (0.39, 2.33)
with estimated
female contribution
to SIR added (see
text)
ICD-7 180. Results reported for males only, but
there was a small female component to the
cohort.
Boice et al.
(1999)
0.99
0.4
2.04
SMR
-0.010
0.378
None
ICD-9 189.0-189.2. For potential routine
exposure. Results for any potential exposure not
reported.
Greenland et
al. (1994)
0.99
0.30
3.32
OR
-0.010
0.613
None
Nested case-control study. ICD-8 codes not
specified, presumably all of 189.
Hansen et al.
(2001)
1.1
0.3
2.8
SIR
0.095
0.500
None
ICD-7 180. Male and female results reported
separately; combined assuming Poisson
distribution.
Morgan et al.
(1998)
1.14
0.51
2.58
Mortality
RR
0.134
0.415
Published SMR
1.32 (0.57, 2.6)
ICD-9 189.0-189.2. Unpublished RR, adjusted
for age and sex (see text).
Raaschou-
Nielsen et al.
(2003)
1.20
0.94
1.50
SIR
0.182
0.115
1.20 (0.98, 1.46)
for ICD-7 180
1.4 (1.0, 1.8) for
subcohort with
expected higher
exposures
RCC.
Radican et
al. (2008)
1.18
0.47
2.94
Mortality
HR
0.166
0.468
None
ICD-8, -9 189.0, ICD-10 C64. Time variable =
age; covariates = sex and race. Referent
group is workers with no chemical exposures.
Zhao et al.
(2005)
1.7
0.38
7.9
Mortality
RR
0.542
0.775
Incidence RR: 2.0
(0.47, 8.2)
Mortality RR no
lag: 0.89 (0.22,
3.6)
Incidence RR no
lag : 2.1 (0.56, 8.1)
Boice (2006b)
SMR: 2.22 (0.89,
4.57)
ICD-9 189. Males only. Adjusted for age, SES,
time since first employment, exposure to other
carcinogens. 20-yr lag. Mortality results reflect
same number exposed cases (10 with no lag) as
do incidence results, so no reason to prefer
mortality results, but they are used in primary
analysis to avoid appearance of "cherry-picking."
Overall RR estimated by combining across
exposure groups (see text). Boice (2006b)
cohort overlaps Zhao cohort; just 7 exposed
o
* V §
§ I 5
5 a, co-
c §- a
S. 5 &•
^ 
>{
>{
*
to
o
-------
K
s
TO
O
s
S"
Si
^ 1
TO Si
§
1. § S-
i <3
§ ^

TO
O
a
>;
>S
>S
TO
TO'
*
VO
to
o
O
LtJ
On
deaths.
-------
S"4
>}'
§•
to
s
Table C-7. Selected RR estimates for renal cell carcinoma associated with TCE exposure from case-control
studies"
Study
RR
estimate
95%
LCL
95%
UCL
log RR
SE(logRR)
Alternate RR
estimates
Comments
Bruning et
al. (2003)
2.47
1.36
4.49
0.904
0.305
1.80 (1.01,
3.20) for
longest job
held in
industry with
TCE exposure
Self-assessed exposure. Adjusted for age, sex,
and smoking.
Charbotel et
al. (2006)
1.88
0.89
3.98
0.631
0.382
1.64 (0.95,
2.84) for full
study
1.68 (0.97,
2.91) for full
study with 10-
year lag
Subgroup with good level of confidence about
exposure assessment. Matched on sex, age.
Adjusted for smoking, body mass index.
Dosemeci et
al. (1999)
1.30
0.9
1.9
0.262
0.191
None
Adjusted for age, sex, smoking, hypertension
and/or use of diuretics and/or anti-hypertension
drugs, body mass index.
Moore et al.
(2010)
2.05
1.13
3.73
0.718
0.305
1.63 (1.04,
2.54) for all
subjects
Subgroup with high-confidence assessments.
Adjusted for age, sex, and center.
Pesch et al.
(2000b)
1.24
b
b
0.215
0.094
1.13 with
German JEM
With JTEM (job task exposure matrix). Crude
OR calculated from data provided in personal
communication (see text).
Siemiatycki
(1991)
0.8
0.3
2.2
-0.223
0.524
None
"Kidney cancer." SE and 95% CI calculated
from reported 90% CIs. Males only; adjusted for
age, income, and cigarette smoking index.
o
* V §
§ I 5
5 a, co-
c §- a
S_ 5 &-
to ^
o
a
>;
>S
>S
TO
'S
TO
*
to
o
aThe RR estimates are all ORs for incident cases.
bNot calculated.
O
LtJ
^1
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
For Boice et al. (1999), only results for "potential routine exposure" were reported for
kidney cancer. Boice et al. (1999) report in general that the SMRs for workers with any potential
exposure "were similar to those for workers with daily potential exposure."
In their published paper, Morgan et al. (1998) present only SMRs for overall TCE
exposure, although the results from internal analyses are presented for exposure subgroups. RR
estimates for overall TCE exposure from the internal analyses of the Morgan et al. (1998) cohort
data were available from an unpublished report (EHS, 1997); from these, the RR estimate from
the Cox model which included age and sex was selected, because those are the variables deemed
to be important in the published paper. The internal analysis RR estimate was preferred for the
primary analysis, and the published SMR result was used in a sensitivity analysis.
Raaschou-Nielsen et al. (2003) reported results for RCC and renal pelvis/ureter
separately. As discussed above, RCC estimates were used in the primary analysis, and the
results for both kidney cancer categories were combined (across sexes as well), assuming a
Poisson distribution, and used in a sensitivity analysis. In another sensitivity analysis, results for
RCC from the subcohort with expected higher exposure levels (> 1-year duration of employment
and year of 1st employment before 1980) were used. Raaschou-Nielsen et al. (2003), in their
Table 3, also present overall results for RCC and for renal pelvis/ureter cancer with a lag time of
20 years; however, they use a definition of lag that is different from a lagged exposure in which
exposures prior to disease onset are discounted and it is not clear what their lag time actually
represents?, thus, as for NHL, these results were not used in any of the meta-analyses for kidney
cancer.
For Radican et al. (2008), the Cox model hazard ratio (HR) from the 2000 follow-up was
used. In the Radican et al. (2008) Cox regressions, age was the time variable, and sex and race
were covariates. It should also be noted that the referent group is composed of workers with no
chemical exposures, not just no exposure to TCE.
For Zhao et al. (2005), no results for an overall TCE effect are reported. We were unable
to obtain any overall estimates from the study authors, so, as a best estimate, as was done for
NHL, the results across the "medium" and "high" exposure groups were combined, under
assumptions of group independence, even though the exposure groups are not independent (the
"low" exposure group was the referent group in both cases). Unlike for NHL, adjustment for
exposure to other carcinogens made a considerable difference, so Zhao et al. (2005) also present
kidney results with this additional adjustment, with and without a 20-year lag. Estimates of RR
with this additional adjustment were selected over those without. In addition, a 20-year lag
7 In their Methods section, Raaschou-Nielsen et al. (2003) define their lag period as the period "from the date of
first employment to the start of follow-up for cancer".
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-38 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
seemed reasonable for kidney cancer, so the lagged estimates were preferred to the unlagged;
unlagged estimates were used in sensitivity analyses. Zhao et al. (2005) present RR estimates for
both incidence and mortality. Unlike for NHL, the number of exposed incident cases (10 with no
lag) was identical to the number of deaths, so there was no reason to prefer the mortality results
over the incidence results. (In fact, there were more exposed incident cases [10 vs. 7] after
lagging.) However, the mortality results, which yield a lower RR estimate, were selected for the
primary analysis to avoid any appearance of "cherry-picking," and incidence RR estimates were
used in sensitivity analyses. A sensitivity analysis was also done using results from Boice et al.
(2006b) in place of the Zhao et al. (2005) RR estimate. The cohorts for these studies overlap, so
they are not independent studies and should not be included in the meta-analysis concurrently.
Boice et al. (2006b) report results for an overall TCE effect for kidney cancer; however, the
results are SMR estimates rather than internal comparisons and are based on fewer exposed
deaths (7), so either Zhao et al. (2005) estimate is preferred over the Boice et al. (2006b)
estimate.
Regarding the case-control studies, for Briining et al. (2003), the results based on
self-assessed exposure were preferred because, although TCE exposure was probably under
ascertained with this measure, there were greater concerns about the result based on the alternate
measure reported—longest-held job in an industry with TCE exposure. Even though this study
was conducted in the Arnsberg region of Germany, an area with high prevalence of exposure to
TCE, the exposure prevalence in both cases (87%) and controls (79%) seemed inordinately high,
and this for not just any job in an industry with TCE exposure, but for the longest-held job.
Furthermore, Table V of Briining et al. (2003), which presents this result, states that the result is
for longest-held job in industries with TCE or tetrachloroethylene exposure. Additionally, some
of the industries with exposure to TCE presented in Table V have many jobs that would not
entail TCE exposure (e.g., white-collar workers), so the assessment based on industry alone
likely has substantial misclassification. Both of these—inclusion of tetrachloroethylene and
exposure assessment by industry—could result in overstating TCE exposure prevalence. Results
based on the longest-held-job measure were used in a sensitivity analysis.
For Charbotel et al. (2006), results from the analysis that considered "only job periods
with a good level of confidence for TCE exposure assessment" (Table 7 of Charbotel et al.,
2006) were preferred, as these estimates would presumably be less influenced by exposure
misclassification. Estimates from the full study analysis were used in a sensitivity analysis.
Results for exposure with a 10-year lag are also provided in an unpublished report (Charbotel et
al., 2005); however, lagged results are presented only for the full study and, thus, were similarly
used in a sensitivity analysis.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Likewise, for Moore et al. (2010), results from the analysis that considered high-
confidence assessments only were preferred. Here the definition of TCE exposure was restricted
to jobs classified as having probable or certain exposure (i.e., at least 40% of workers with that
job were expected to be exposed), so these estimates should be less influenced by exposure
misclassification. The RR estimate from the analysis of all subjects was used in a sensitivity
analysis.
For Pesch et al. (2000b), TCE results were presented for 2 different exposure
assessments. Estimates using the job-task-exposure-matrix (JTEM) approach were preferred
because they seemed to represent a more comprehensive exposure assessment (see Appendix B,
Section II-4); estimates based on the JEM approach were used in a sensitivity analysis.
Furthermore, results were presented only by exposure category, with no overall RR estimate
reported. Case and control numbers for the different exposure categories were kindly provided
by Dr. Pesch (personal communication from Baete Pesch, BGFA, to Cheryl Scott, U.S. EPA,
21 February 2008), and we calculated crude overall ORs for males and females combined for
each exposure assessment approach.
C.3.1.2. Results of Meta-Analyses
Results from some of the meta-analyses that were conducted on the epidemiological
studies of TCE and kidney cancer are summarized in Table C-8. The summary estimate from the
primary random-effects meta-analysis of the 15 studies was 1.27 (95% CI: 1.13, 1.43) (see
Figure C-5). As shown in Figure C-5, the analysis was dominated by 2 (contributing over 65%
of the weight) or 3 (about 75% of the weight) large studies. No single study was overly
influential; removal of individual studies resulted in RRm estimates that were all statistically
significant (all withp < 0.005) and that ranged from 1.24 (with the removal of (Briining et al.,
2003) to 1.30 (with the removal of Raaschou-Nielsen (2003).
Similarly, the RRm estimate was not highly sensitive to alternate RR estimate selections.
Use of the 13 alternate selections, individually, resulted in RRm estimates that were all
statistically significant (all withp < 0.0005) and that ranged from 1.21 to 1.32 (see Table C-8).
In fact, as can be seen in Table C-8, all but two of the alternates had negligible impact. The
Zhao (2005), Axelson (1994), Morgan (1998), Briining (2003), Charbotel (2006), and Moore
(2010) original values and alternate selections were associated with very little weight and, thus,
had little influence in the RRm. The Raaschou-Nielsen (2003) all-kidney-cancer value carried
more weight, but the alternate RR estimate was identical to the original, although with a
narrower CI, and so did not alter the RRm. Only the Raaschou-Nielsen high-exposure-subcohort
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
alternate and the Pesch (2000b) alternate (with the JEM exposure assessment approach instead of
the JTEM approach) had much impact, resulting in RRm estimates of 1.32 (95% CI: 1.17, 1.49)
and 1.21 (95% CI: 1.09, 1.34), respectively. As noted above, the JTEM approach is preferred,
thus the lower RRm estimate obtained with the JEM alternate is considered clearly inferior. The
JEM approach takes jobs into account but not tasks; thus, it is expected to have greater potential
for exposure misclassification. Indeed, a comparison of exposure prevalences for the
two approaches suggests that the JEM approach is less discriminating about exposure; 42% of
cases were defined as TCE-exposed under the JEM approach, but only 18% of cases were
exposed under the JTEM approach. On the other hand, the higher RRm estimate obtained with
the Raaschou-Nielsen (2003) high-exposure-subcohort alternate is consistent with an expectation
that the subgroup has higher exposures and less exposure misclassification.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-41 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-8. Summary of some meta-analysis results for TCE (overall) and kidney cancer
S"4
>}'
§•
to
s
Analysis
# of
studies
Model
Combined
RR estimate
95% LCL
95% UCL
Heterogeneity
Comments
All studies
15
Random
1.27
1.13
1.43
None obs
(fixed = random)
Statistical significance not dependent on
single study. No apparent publication
bias.


Fixed
1.27
1.13
1.43


Cohort
9
Random
1.16
0.96
1.40
None obs
Not significant difference between CC
and cohort studies (p = 0.12).


Fixed
1.16
0.96
1.40

Not significant difference between CC
and cohort studies (p = 0.19).
Case-control
6
Random
1.48
1.15
1.91
Not significant
(p = 0.14)



Fixed
1.36
1.17
1.39


Alternate RR
selections3
15
Random
1.27-1.28
1.13-1.14
1.42-1.43
None obs
With 3 different alternates from Zhao
(see Table C-6).

15
Random
1.29
1.15
1.45
None obs
With Boice (2006b) study rather than
Zhao (2005)

15
Random
1.27
1.13
1.43
None obs
With estimated female contribution to
Axelson (1994).

15
Random
1.28
1.14
1.43
None obs
With Morgan (1998) published SMR.

15
Random
1.27
1.13
1.42
None obs
With Raaschou-Nielsen (2003) all
kidney cancer.

15
Random
1.32
1.17
1.49
None obs
With Raaschou-Nielsen (2003) high-
exposure subcohort.

15
Random
1.26
1.12
1.41
None obs
With Bruning (2003) longest job held in
industry with TCE.

15
Random
1.28
1.14
1.43
None obs
With Charbotel full study (2006), with
and without 10-year lag.

15
Random
1.27
1.13
1.43
None obs
With Moore full study (2010).

15
Random
1.21
1.09
1.34
None obs
With Pesch JEM (2000b).
Highest
10
Random
1.64
1.31
2.04
None obs

exposure
13
Random
1.58
1.28
1.96
None obs
Usina RR = 1 for Anttila (1995), Axelson
o
* v §
§ I 5
5 a, co-
c 8- a
5 S S-
?• i <§
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
O
VO
to
o
o
-k
to
-------
K
s
TO
0
s
1
a,
' >3
o
o
s
S?
r?
S"4
>1'
§•
Si
a
>3*
S5
S*
<§>
*s
*s
'
*
groups






(1994), and Hansen (2001) (see text).
13
Random
1.47-1.60
1.20-1.29
1.79-1.98
See Table C-10
Using RR = 1 for Anttila (1995), Axelson
(1994), and Hansen (2001) and various
alternate RR selection results (see
Table C-10)a.
aChanging the primary analysis by one alternate RR each time,
obs = observable.
O
-k
u>
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
TCE Exposure and Kidney Cancer
Study	Relative Risk arid 95* CI	RR LCL UCL
Anttila (1905)
	—	Hf———
0.87
0.32
1.80
Anelson (1994)
	H	
lie
0.42
2.52
Boies (1900) '
_	j	
' 0.09
0.40
2.04
Greenland (1994) '
-—	f-		
: 0.09
0.30
3.32
Hansen (2001)
1 1
; i.io
0.30
2.80
!
Morgan (1808)
	H	
1.14
0.51
2.58
Raasohou-Nielsen (2003)
...
1.20
0.94
1.50
Radiean (2008) !
	H—		
; 1.18
0.47
2.94
Zhao (2005)
! 1
, , i ?n
0.38
7.90
	1		1 ~ '
>: I J u
Bruning (2003)
I 	0	
2.47
1.36
4.49
Charbotel (2006) ¦
1 ,
1.88
0.89
3.98

Dosemeci (1909)
_|__Q_	
1.30
0.90
1.90
Moore (2010) •
[ 	B_——
2.05
1.13
3.73
Pesch (2000) '

1.24
1.00
1.50
Siemiat/oki (1901) '
	1—j	
0.80
0.30
2.20
OVERALL

1.27
1.13
1.43
|	1	1	5	1—(—i i i |	?	s	1	1—i—i—r~r~|
0.1	1	10
Figure C-5. Meta-analysis of kidney cancer and overall TCE exposure.
Random-effects model; fixed-effect model same. The summary estimate is in
the bottom row, represented by the diamond. Symbol sizes reflect relative
weights of the studies.
There was no apparent heterogeneity across the 15 studies, i.e., the random-effects model
and the fixed-effect model gave the same results (phetero = 0.67; I2 = 0%). Nonetheless, subgroup
analyses were done examining the cohort and case-control studies separately. With the random-
effects model (and tau-squared not pooled across subgroups), the resulting RRm estimates were
1.16 (95% CI: 0.96, 1.40) for the cohort studies and 1.48 (1.15, 1.91) for the case-control studies.
There was no heterogeneity in the cohort subgroup (p = 0.998; I2 = 0%). There was
heterogeneity in the case-control subgroup, but it was not statistically significant (p = 0.14) and
the / -value of 41% suggests that the extent of the heterogeneity in this subgroup was low-to-
moderate. Nor was the difference between the RRm estimates for the cohort and case-control
subgroups statistically significant under either the random-effects model or the fixed-effect
model. Further quantitative investigations of heterogeneity were not pursued because of
database limitations and, in any event, there is no evidence for heterogeneity of study results in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
this database. A qualitative discussion of some potential sources of heterogeneity across studies
is nonetheless included in Section C.3.3.
As discussed in Section C.l, publication bias was examined in several different ways.
The funnel plot in Figure C-6 shows little relationship between RR estimate and study size, and,
indeed, none of the other tests performed found any evidence of publication bias. Duval and
Tweedie et al. (2000)'s trim-and-fill procedure, for example, determined that no studies were
missing from the funnel plot, i.e., there was no asymmetry to counterbalance. Similarly, the
results of a cumulative meta-analysis, incorporating studies with increasing SE one at a time,
shows no evidence of a trend of increasing effect size with addition of the less precise studies.
Including the 3 most precise studies, reflecting 75% of the weight, the RRm goes from 1.24 to
1.22 to 1.23. The addition of the Moore (2010) study brings the RRm to 1.26 and the weight to
79% and further addition of the Briining (2003) study increases the RRm to 1.38 and the weight
to 83%). After the addition of the next 6 studies, the RRm stabilizes at about 1.28, and further
addition of the 4 least precise studies has little impact.
C.3.2. Kidney Cancer Effect in the Highest Exposure Groups
C.3 .2.1. Selection of RR Estimates
The selected RR estimates for kidney cancer in the highest TCE exposure categories, for
studies that provided such estimates, are presented in Table C-9. Five of the 9 cohort studies and
5 of the 6 case-control studies reported kidney cancer risk estimates categorized by exposure
level. As in Section C.3.1.1 for the overall risk estimates, estimates for RCC were preferentially
selected when presented, and, wherever possible, RR estimates for males and females combined
were used.
Funnel Plot of Standard Error by Log risk ratio
0.0
0.2
Standard Eirror
0.4
This document is a draft
10/20/09
0.6
-------
1
2	Figure C-6. Funnel plot of SE by log RR estimate for TCE and kidney
3	cancer studies
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-46 DRAFT—DO NOT CITE OR QUOTE
-------
Table C-9. Selected RR estimates for kidney cancer risk in highest TCE exposure groups
S"4
to
s
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE
(logRR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)



100+ |jmol/L
U-TCAa


1.0 assumed
Reported high exposure group results for
some cancer sites but not kidney.
Axelson et
al. (1994)



>2 yr exposure
and 100+ mg/L
U-TCA


1.0 assumed
Reported high exposure group results for
some cancer sites but not kidney.
Boice et al.
(1999)
0.69
0.22
2.12
>5 years exp
-0.371
0.578
None
Mortality RR. ICD-9 189.0-189.2. For
potential routine or intermittent exposure,
adjusted for date of birth, dates 1s and
last employed, race, and sex. Referent
group is workers not exposed to any
solvent.
Hansen et
al. (2001)



>1080 months
x mg/m3


1.0 assumed
Reported high exposure group results for
some cancer sites but not kidney.
Morgan et
al. (1998)
1.59
0.68
3.71
High
cumulative
exposure
score
0.464
0.433
1.89 (0.85, 4.23)
for med/high
peak vs. low/no
Mortality RR. ICD-9 189.0-189.2.
Adjusted for age and sex.
Raaschou-
Nielsen et
al. (2003)
1.7
1.1
2.4
>5 yrs in
subcohort with
expected
higher
exposure
levels
0.531
0.183
1.6 (1.1, 2.2) for
RCC for >5 years
in total cohort
1.4 (0.99, 1.9)
ICD-7 180
>5 years in total
cohort
SIR. RCC.
Radican et
al. (2008)
1.11
0.35
3.49
>25 unit-years
0.104
0.582
Blair et al. (1998)
incidence RR
0.9 (0.3, 3.2)
Mortality HR. ICD-8, -9 189.0, ICD-10
C64. Male and female results presented
separately and combined (see text).
Referent group is workers with no
chemical exposures.
o
* v §
§ I 5
5 a, co-
c 8- a
5 S S-
?• i <§
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
to
o
O
-k
-------
Table C-9. Selected RR estimates for kidney cancer risk in highest TCE exposure groups (continued)

* v §
§ I 5
5 a, co-
c §- a
5 S S-
?• i
t: §
O a*
o
a
>;
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE
(logRR)
Alternate RR
estimates
Comments
>s
>S
TO
*
to
o
Zhao et al.
(2005)
7.40
0.47
116
High exposure
score
2.00
1.41
Mortality RR:
1.82 (0.09, 38.6)
Incidence RR no
lag: 7.71 (0.65,
91.4)
Mortality RR no
lag: 0.96 (0.09,
9.91)
Boice (2006b)
mortality RR:
2.12 (0.63, 7.11)
for >5 years as
test stand
mechanic; 3.13
(0.74,13.2) for
>4 test-year
engine flush
Incidence RR. ICD-9 189. Males only.
Adjusted for age, SES, time since first
employment, exposure to other
carcinogens. 20-yr lag. Incidence results
reflect more exposed cases (4 with no
lag) than do mortality results (3), so they
are used in primary analysis.
Bruning et
al. (2003)
2.69
0.84
8.66
>20 years
self-assessed
exposure
0.990
0.595
None
Incidence OR. RCC. Adjusted for age,
sex, and smoking.
Charbotel et
al. (2006)
3.34
1.27
8.74
High
cumulative
dose
1.21
0.492
3.80 (1.27, 11.40)
for high cum +
peaks.
Full study, high
cum: 2.16 (1.02,
4.60)
+ peaks: 2.73
(1.06, 7.07)
Full study with
10-year lag, high
cum: 2.16 (1.01,
4.65)
+ peaks: 3.15
(1.19, 8.38)
Incidence OR. RCC. In subgroup with
good level of confidence for TCE
exposure. Adjusted for smoking and
body mass index. Matched on sex and
age. Alternate full study estimates
(without lag) were additionally adjusted
for exposure to cutting fluids and other
petroleum oils.
-------
K
s
TO
>3

§•
to
s
0
* v §
§ I 5
a, Co'
§- a
TO Sj-
1	<2
§ ^
O
a
>;
r
t:
C
?:'
r:
>S
>S
TO
'S
TO
*
VO
to
o
O
-k
VO
Full study, addl
adj, high cum:
1.96 (0.71, 5.37)
+ peaks:
2.63 (0.79, 8.83)
-------
K
s
TO
>3
Table C-9. Selected RR estimates for kidney cancer risk in highest TCE exposure groups (continued)
S"4
§•
TO
S
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE(logRR)
Alternate RR
estimates
Comments
Moore et al.
(2010)
2.23
1.07
4.64
> 1.58 ppm
x years
0.802
0.374
2.02 (1.14,
3.59) for all
subjects
Subgroup with high-confidence
assessments. Adjusted for age, sex, and
center.
Pesch et al.
(2000b)
1.4
0.9
2.1
Substantial
0.336
0.219
1.2 (0.9, 1.7)
for JEM
Incidence OR. RCC. JTEM approach.
Adjusted for age, study center, and
smoking. Sexes combined.
Siemiatycki
(1991)
0.8
0.2
3.4
Substantial
-0.233
0.736
none
Incidence OR. Kidney cancer. SE and
95% CI calculated from reported 90%
CIs. Males only; adjusted for age,
income, and cigarette smoking index.
o
* v	§
§ I	5
a,	Co'
8-	a
TO	Sj-
§	^
O
a
>S
>S
TO
'S
TO
*
O
VO
to
o
"Mean personal trichloroacetic acid in urine. 1 |imol/L = 0.1634 mg/L.
O
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Three of the 9 cohort studies (Anttila et al., 1995; Axelson et al., 1994; Hansen et al.,
2001) did not report kidney cancer risk estimates categorized by exposure level even though
these same studies reported such estimates for selected other cancer sites. To address this
reporting bias, attempts were made to obtain the results from the primary investigators, and,
failing that, an alternate analysis was performed in which null estimates (RR = 1.0) were
included for all 3 studies. This alternate analysis was then used as the main analysis, e.g., the
basis of comparison for the sensitivity analyses. For the SE (of the logRR) estimates for these
null estimates, SE estimates from other sites for which highest-exposure-group results were
available were used. For Anttila et al.(1995), the SE estimate for liver cancer in the highest
exposure group was used, because liver cancer and kidney cancer had similar numbers of cases
in the overall study (5 and 6, respectively). For Axelson et al. (1994), the SE estimate for NHL
in the highest exposure group was used, because NHL and kidney cancer had similar numbers of
cases in the overall study (5 and 6, respectively). For Hansen et al. (2001), the SE estimate for
NHL in the highest exposure group was used, because NHL was the only cancer site of interest
in this assessment for which highest-exposure-group results were available.
For Boice et al. (1999), only results for workers with "any potential exposure" (rather
than "potential routine exposure") were presented by exposure category, and the referent group is
workers not exposed to any solvent.
For Morgan et al. (1998), the primary analysis used results for the cumulative exposure
metric, and a sensitivity analysis was done with the results for the peak exposure metric.
For Raaschou-Nielsen et al. (2003), results for RCC in the highest duration subgroup
from the subcohort with expected higher exposure levels (> 1-year duration of employment and
year of 1st employment before 1980) were preferred for the highest-exposure-group analyses.
Results for RCC in the highest duration subgroup from the whole cohort were combined across
sexes, assuming a Poisson distribution, and used in a sensitivity analysis. Also, for the whole
cohort, results for RCC and renal pelvis/ureter cancers in the highest duration group were
combined (across sexes as well), assuming a Poisson distribution, and used in an additional
sensitivity analysis.
For Radican et al. (2008), it should be noted that the referent group is workers with no
chemical exposures, not just no TCE exposure. In addition, results for exposure groups (based
on cumulative exposure scores) were reported separately for males and females and were
combined for this assessment using inverse-variance weighting, as in a fixed-effect meta-
analysis. Radican et al. (2008) present only mortality HR estimates by exposure group; however,
in an earlier follow-up of this same cohort, Blair et al. (1998) present both incidence and
mortality RR estimates by exposure group. The mortality RR estimate based on the more recent
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
follow-up of Radican et al. (2008) (6 deaths in the highest exposure group) was used in the
primary analysis, while the incidence RR estimate based on similarly combined results from
Blair et al. (1998) (4 cases) was used as an alternate estimate in a sensitivity analysis. Radican et
al. (2008) also present results for categories based on frequency and pattern of exposure;
however, subjects weren't distributed uniquely across the categories (the numbers of cases across
categories exceeded the total number of cases), thus it was difficult to interpret these results and
they were not used in a sensitivity analysis.
Zhao et al. (2005) present kidney cancer RR estimates adjusted for exposure to other
carcinogens, because, unlike for NHL, this adjustment made a considerable difference.
Estimates of RR with this additional adjustment were selected over those without. Furthermore,
the kidney results were presented with and without a 20-year lag. A 20-year lag seemed
reasonable for kidney cancer, so the lagged estimates were preferred to the untagged; untagged
estimates were used in sensitivity analyses. In addition, the incidence results reflect more cases
(4 with no lag) in the highest exposure group than do the mortality results (3), so the incidence
result (with the 20-year lag) was used for the primary analysis, and the untagged incidence result
and the mortality results were used in a sensitivity analysis. Sensitivity analyses were also done
using results from Boice et al. (2006b) in place of the Zhao et al. (2005) RR estimate. The
cohorts for these studies overlap, so they are not independent studies. Boice et al. (2006b) report
mortality RR estimates for kidney cancer by years worked as a test stand mechanic, a job with
potential TCE exposure, and by a measure that weighted years with potential exposure from
engine flushing by the number of flushes each year. No results were presented for a third metric,
years worked with potential exposure to any TCE, because the Cox proportional hazards model
did not converge. The Boice et al. (2006b) estimates are adjusted for years of birth and hire and
for hydrazine exposure.
For Charbotel et al. (2006), results from the analysis that considered "only job periods
with a good level of confidence for TCE exposure assessment" (Table 7 of Charbotel et al.,
2006) were preferred, as these estimates would presumably be less influenced by exposure
misclassification. Additionally, the high cumulative dose results were preferred, but the results
for high cumulative dose + peaks were included in a sensitivity analysis. Exposure group results
with a 10-year lag are provided in an unpublished report (Charbotel et al., 2005); however,
lagged results are presented only for the full study and, thus, were used in sensitivity analyses.
Estimates from the full study analysis (without the lag) that were further adjusted for exposure to
cutting fluids and other petroleum oils were also used in sensitivity analyses.
Similarly, for Moore et al. (2010), results from the analysis that considered high-
confidence assessments only were preferred. Here the definition of TCE exposure was restricted
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
to jobs classified as having probable or certain exposure (i.e., at least 40% of workers with that
job were expected to be exposed), so these estimates should be less influenced by exposure
misclassification. Estimates from the analysis of all subjects were used in a sensitivity analysis.
The highest exposure group was reported as > 1.58 ppm x years; however, this value is not based
on continuous exposure estimates but rather calculated from midpoints of estimated ranges
corresponding to categorical groups, i.e, cumulative exposure = categorical intensity weight
(ppm) xcategorical frequency weight x duration (years).
For Pesch et al. (2000b), TCE results were presented for two different exposure
assessments. As discussed above, estimates using the JTEM approach were preferred because
they seemed to represent a more comprehensive exposure assessment; estimates based on the
JEM approach were used in a sensitivity analysis.
C.3.2.2. Results of Meta-Analyses
Results from the meta-analyses that were conducted for kidney cancer in the highest
exposure groups are summarized at the bottom of Table C-8 and reported in more detail in
Table C-10. The RRm estimate from the random-effects meta-analysis of the 10 studies with
results presented for exposure groups was 1.64 (95% CI: 1.31, 2.04). The RRm estimate from
the primary random-effects meta-analysis with null RR estimates (i.e., 1.0) included for Anttila
et al.(1995), Axelson (1994), and Hansen (2001) to address reporting bias (see above) was 1.58
(1.28, 1.96) (see Figure C-7). The inclusion of these 3 additional studies contributed just over
7%> of the total weight. As with the overall kidney cancer meta-analyses, the meta-analyses of
the highest exposure groups were dominated by 2 studies (Pesch et al., 2000b; Raaschou-Nielsen
et al., 2003), which provided about 60% of the weight. No single study was overly influential;
removal of individual studies resulted in RRm estimates that were all statistically significant (all
with p < 0.005) and that ranged from 1.52 (with the removal of Raaschou-Nielsen (2003) to 1.64
(with the removal of Pesch (2000b).
Similarly, the RRm estimate was not highly sensitive to alternate RR estimate selections.
Use of the 18 alternate selections, individually, resulted in RRm estimates that were all
statistically significant (all withp < 0.0005) and that ranged from 1.47 to 1.60, with all but 2 of
the alternate selections yielding RRm estimates in the narrow range of 1.54-1.60 (see
Table C-10). The lowest RRm estimates, 1.47 in both cases, were obtained when the alternate
selections involved the 2 large studies. One of the alternate selections was for Raaschou-Nielsen
(2003), with a highest-exposure-group estimate for all kidney cancer in the total cohort, rather
than RCC in the subcohort expected to have higher exposure levels. The latter value is strongly
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-53 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
preferred because, as discussed above, the subcohort is likely to have less exposure
misclassification. Furthermore, RCC is very different from other types of kidney cancer, and
TCE, if an etiological factor, may not be etiologically associated with all kidney cancers, so
using the broad category may dilute a true association with RCC, if one exists. The other
alternate selection with a considerable impact on the RRm estimate was for Pesch (2000b), with
the highest exposure group result based on the JEM exposure assessment approach, rather than
the JTEM approach. As discussed above, the JTEM approach is preferred because it seemed to
be a more comprehensive and discriminating approach, taking actual job tasks into account,
rather than just larger job categories. Thus, although results with these alternate selections are
presented for comprehensiveness and transparency, the primary analysis is believed to reflect
better the potential association between kidney cancer (in particular, RCC) and TCE exposure.
Other than a negligible amount of heterogeneity observed in the sensitivity analysis with
the Pesch (2000b) JEM alternate discussed above (/ = 0.64%), there was no observable
heterogeneity across the studies for any of the meta-analyses conducted with the highest
exposure groups, including those in which RR values for Anttila et al.(1995), Axelson (1994),
and Hansen (2001) were assumed. No subgroup analyses (e.g., cohort vs. case-control studies)
were done with the highest exposure group results.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-54 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-10. Summary of some meta-analysis results for TCE (highest exposure groups) and kidney cancer
S"4
§•
TO
S
Analysis
Model
Combined
RR estimate
95% LCL
95% UCL
Heterogeneity
Comments
Analysis
based on
reported
results
Random
1.64
1.31
2.04
None obs
(fixed =
random)

Primary
analysis
Random
1.58
1.28
1.96
None obs
Includes assumed values for Anttila (1995),
Axelson (1994), and Hansen (2001) (see text).
Statistical significance not dependent on single
study.
Alternate RR
selections3
Random
1.57
1.27
1.95
None obs
With Blair et al. (1998) incidence RR instead of
Radican (2008) mortality HR.

Random
1.60
1.29
1.98
None obs
With Morgan (1998) peak metric.

Random
1.47, 1.55
1.20, 1.25
1.80, 1.91
None obs
With Raaschou-Nielsen (2003) >5 years in total
cohort for all kidney cancer and for RCC,
respectively.

Random
1.56-1.58
1.26-1.28
1.93-1.96
None obs
With Zhao (2005) incidence unlagged and
mortality with and without lag.

Random
1.58-1.59
1.28-1.29
1.95-1.96
None obs
With Boice (2006b) alternates for Zhao (2005).

Random
1.59
1.29
1.95
None obs
With Moore full study.

Random
1.54-1.58
1.24-1.27
1.90-1.95
None obs
With Charbotel (2006) high cumulative dose +
peaks in subgroup; and high cumulative dose
and high cumulative dose + peaks in full study
with and without 10-year lag and with and
without additional adjustment for exposure to
cutting fluids and other petroleum oils.

Random
1.47
1.20
1.79
Not significant
(p = 0.44)
With Pesch (2000b) JEM.
o
* v §
§ I 5
5 a, co-
c §- a
5 S S-
?• £ <§
r: §
TO -
o
a
>;
>S
>S
TO
'S
TO
*
O
VO
to
o
"Changing the primary analysis by one alternate RR each time.
O
dfi obs = observable.
i^r>
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
TCE Exposure and Kidney Cancel
Study
Relative Risk and 95% CI
RR
LCL
UCL
Boies (1999)
	J	:	
0.69
0.22
n
Morgan (1998)
0
1.59
0.68
3.71
Raaschou-Nielsen (2003)
	cm	
1.70
1.10
2.40
Radican (2008)
	t	
1.11
0.35
3.49
Zhao (2005)
	i	
7.40
0.47
116.00
Burning (2003)
i
2.69
0.34
8.86
Charbotel (2006)
i 	1	
3.34
1.27
8.74
Moore (2010)
:	1	

1.07
4.64
Pesch (2000)
—~—
1.40
0.90
2.10
Siemiatycki (1991)
!
0.80
0.20
3.40
Anttila (1995)
	}	
1.00
0.25
4.00
Axelson (1994) 	
	)	
1.00
0.14
7.10
Hansen (2001)
	1	
1.00
0.32
3.10
OVERALL

1.58
1.28
1.80
i	1	1	¦«	1	1		""i	'	'	'	¦«	'	1	1	n
0.1	1	10
Figure C-7. Meta-analysis of kidney cancer and TCE exposure—highest
exposure groups, with assumed null RR estimates for Anttila, Axelson, and
Hansen (see text). Random-effects model; fixed-effect model same. The
summary estimate is in the bottom row, represented by the diamond. Symbol
sizes reflect relative weights of the studies.
C.3.3. Discussion of Kidney Cancer Meta-Analysis Results
For the most part, the meta-analyses of the overall effect of TCE exposure on kidney
cancer suggest a small, statistically significant increase in risk. The summary estimate from the
primary random-effects meta-analysis of the 15 studies was 1.27 (95% CI: 1.13, 1.43). Although
the analysis was dominated by 2-3 large studies that contribute 65-75% of the weight, the
summary estimate was not overly influenced by any single study, nor was it overly sensitive to
individual RR estimate selections. The largest downward impacts were from the removal of the
Briining (2003) study, resulting in an RRm estimate of 1.24 (95% CI: 1.10, 1.40), and from the
substitution of the Pesch (2000b) JTEM RR estimate with the RR estimate based on the JEM
approach, resulting in an RRm estimate of 1.21 (1.09, 1.34). Thus, the finding of an increased
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-56 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
risk of kidney cancer associated with TCE exposure is robust. Furthermore, there is no evidence
of publication bias in this data set.
In addition, there was no heterogeneity observed across the results of the 15 studies.
When subgroup analyses were done of cohort and case-control studies separately, there was
some observable heterogeneity among the case-control studies, but it was not statistically
significant {p = 0.14) and the / -value of 41% suggested the extent of the heterogeneity was low-
to-moderate. The increased risk of kidney cancer was strengthened in the case-control study
analysis and weakened in the cohort study analysis, but the difference between the 2 RRm
estimates was not statistically significant. One difference between the case-control and cohort
studies is that the case-control studies were of RCC and almost all of the cohort studies were of
all kidney cancers, including renal pelvis. As discussed above, RCC is very different from other
types of kidney cancer, and TCE, if an etiological factor, may not be etiologically associated
with all kidney cancers, so using the broad category may dilute a true association with RCC, if
one exists.
With respect to the non-significant heterogeneity in the 6 case-control studies, these
studies differ in TCE exposure potential to the underlying population from which case and
control subjects were identified, and this may be a source of some heterogeneity. Prevalence of
exposure to TCE among cases in these studies was 27% in Charbotel et al. (2006) (for
high-level-of-confidence jobs), 18% in Briining et al. (2003) (for self-assessed exposure), 18% in
Pesch et al. (2000b), 13% in Dosemeci et al. (1999), 3.6% in Moore et al. (2010) (for high-
confidence jobs), and 1% in Siemiatycki (1991). Both Briining et al. (2003) and Charbotel et al.
(2006) are studies designed specifically to assess RCC and TCE exposure. These studies were
carried out in geographical areas with both a high prevalence and a high degree of TCE
exposure. Some information is provided in these and accompanying papers to describe the
nature of exposure, making it possible to estimate the order of magnitude of exposure, even
though there were no direct measurements (Briining et al., 2003; Cherrie et al., 2001; Fevotte et
al., 2006). The Charbotel et al. (2006) study was carried out in the Arve Valley region in France,
where TCE exposure was through metal-degreasing activity in small shops involved in the
manufacturing of screws and precision metal parts (Fevotte et al., 2006). Industrial hygiene data
from shops in this area indicated high intensity TCE exposures of 100 ppm or higher, particularly
from exposures from hot degreasing processes. Considering exposure only from the jobs with a
high level of confidence about exposure, 18% of exposed cases were identified with high
cumulative exposure to TCE. The source population in the Briining et al. (2003) study includes
the Arnsberg region in Germany, which also has a high prevalence of TCE exposure. A large
number of small companies used TCE in metal degreasing in small workrooms. Subjects in this
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-57 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
study also described neurological symptoms previously associated with higher TCE intensities.
While subjects in the Briining et al. (2003) study had potential high TCE exposure intensity,
average TCE exposure in this study is considered lower than that in the Charbotel et al. (2006)
study because the base population was enlarged beyond the Arnsberg region to areas which did
not have the same focus of industry.
Siemiatycki (1991), Dosemeci et al. (1999), and Pesch et al. (2000b) are
population-based studies. Sources of exposure to TCE and other chlorinated solvents are much
less well defined in these studies, and most subjects identified with TCE exposure probably had
minimal contact; estimated average concentrations to exposed subjects were of about 10 ppm or
less (NRC, 2006). Pesch et al. (2000b) includes the Arnsberg area and 4 other regions. Neither
Dosemeci et al. (1999) nor Siemiatycki (1991) describe the nature of the TCE exposure. TCE
exposure potential in these two studies is likely lower than in the other studies and closer to
background. Furthermore, the use of generic job-exposure-matrices for exposure assessment in
these studies may result in a greater potential for exposure misclassification bias.
Moore et al. (2010) is a hospital-based study which identified subjects from 4 Eastern and
Central European countries with high kidney cancer rates (Czech Republic, Poland, Russia, and
Romania). In their exposure assessment, Moore et al. (2010) accounted for the likelihood of
TCE exposure, defined as possible, probable, or definite exposure. This likely increased
exposure potential in their subgroup of high-confidence TCE assessments, which was restricted
to subjects with probable or definite exposure. Although their semi-quantitative exposure
assessment most probably improved exposure rankings, TCE exposure potential is likely lower
in their study than in Briining et al. (2003) and Charbotel et al. (2006), given the many jobs and
industries included.
Ten of the 15 studies categorized results by exposure level. Three other studies reported
results for other cancer sites by exposure level, but not kidney cancer; thus, to address this
reporting bias, null values (i.e., RR estimates of 1.0) were used for these studies. Different
exposure metrics were used in the various studies, and the purpose of combining results across
the different highest exposure groups was not to estimate an RRm associated with some level of
exposure, but rather to see the impacts of combining RR estimates that should be less affected by
exposure misclassification. In other words, the highest exposure category is more likely to
represent a greater differential TCE exposure compared to people in the referent group than the
exposure differential for the overall (typically any vs. none) exposure comparison. Thus, if TCE
exposure increases the risk of kidney cancer, the effects should be more apparent in the highest
exposure groups. Indeed, the RRm estimate from the primary meta-analysis of the highest
exposure group results was 1.58 (95% CI: 1.28, 1.96), which is greater than the RRm estimate of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
1.27 (95% CI: 1.13, 1.43) from the overall exposure analysis. This result for the highest
exposure groups was not overly influenced by any single study, nor was it overly sensitive to
individual RR estimate selections. Heterogeneity was not observed in any of the analyses, with
the exception of some negligible heterogeneity (/ = 0.64%) in one sensitivity analysis. The
robustness of this finding lends substantial support to a conclusion that TCE exposure increases
the risk of kidney cancer.
C.4. META-ANALYSIS FOR LIVER CANCER
C.4.1. Overall Effect of TCE Exposure
CAl.l. Selection of RR Estimates
The selected RR estimates for liver cancer associated with TCE exposure from the
epidemiological studies are presented in Table C-l 1. There were no case-control studies for
liver cancer and TCE exposure that were selected for inclusion in the meta-analysis (see
Appendix B, Section II-9), so all of the relevant studies are cohort studies. All of the studies
reported results for liver cancers plus cancers of the gall bladder and extrahepatic biliary
passages (i.e., ICD-7 155.0 + 155.2; ICD-8 and -9 155 + 156). Three of the studies also report
results for liver cancer alone (ICD-7 155.0; ICD-8 and -9 155). For the primary analysis, results
for cancers of the liver, gall bladder, and biliary passages combined were selected, for the sake of
consistency, since these were reported in all the studies. An alternate analysis was also done
using results for liver cancer alone for the 3 studies that reported them and the combined liver
cancer results for the remainder of the studies.
As for NHL and kidney cancer, many of the studies provided RR estimates only for
males and females combined, and we are not aware of any basis for a sex difference in the
effects of TCE on liver cancer risk; thus, wherever possible, RR estimates for males and females
combined were used. The only study of much size (in terms of number of liver cancer cases)
that provided results separately by sex was Raaschou-Nielsen (2003). The results of this study
suggest that liver cancer risk in females might be slightly higher than the risk in males, but the
number of female cases is small (primary liver cancer SIR: males 1.1 [95% CI: 0.74, 1.64;
27 cases], females 2.8 [1.13, 5.80; 7 cases]; gallbladder and biliary passage cancers SIR:
males 1.1 [0.61, 1.87; 14 cases]; females 2.8 [1.28, 5.34; 9 cases]). Radican et al. (2008) report
HRs for liver/biliary passage cancers combined of 1.36 (95% CI: 0.59, 3.11; 28 deaths) for males
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-59 DRAFT—DO NOT CITE OR QUOTE
-------
1	and 0.74 (95% CI: 0.18, 2.97; 3 deaths) for females, but these results are based on fewer cases,
2	especially in females.
3
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-60 DRAFT—DO NOT CITE OR QUOTE
-------
is*
Table C-ll. Selected RR estimates for liver cancer associated with TCE exposure (overall effect) from cohort
studies
Study
RR
95%
LCL
95%
UCL
RR type
log RR
SE(log
RR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)
1.89
0.86
3.59
SIR
0.637
0.333
2.27 (0.74, 5.29)
for 155.0 alone
ICD-7 155.0 + 155.1; combined assuming Poisson
distribution.
Axelson et al.
(1994)
1.41
0.38
3.60
SIR
0.344
0.5
1.34 (0.36, 3.42)
with estimated
female
contribution to
SIR added (see
text)
ICD-7 155. Results reported for males only, but
there was a small female component to the cohort.
Boice et al.
(1999)
0.81
0.45
1.33
SMR
-0.616
0.5
0.54 (0.15, 1.38)
for potential
routine exposure
ICD-9 155 + 156. For any potential exposure.
Greenland et
al. (1994)
0.54
0.11
2.63
OR
-0.616
0.810
None
ICD-8 155 + 156. Nested case-control study.
Hansen et al.
(2001)
2.1
0.7
5.0
SIR
0.742
0.447
None
ICD-7 155. Male and female results reported
separately; combined assuming Poisson
distribution.
Morgan et al.
(1998)
1.48
0.56
3.91
SMR
0.393
0.495
Published SMR
0.98 (0.36, 2.13)
ICD-9 155 + 156. Unpublished RR, adjusted for
age and sex (see text).
Raaschou-
Nielsen et al.
(2003)
1.35
1.03
1.77
SIR
0.300
0.132
1.28 (0.89, 1.80)
for ICD-7 155.0
ICD-7 155.0 + 155.1. Results for males and
females and different liver cancer types reported
separately; combined assuming Poisson
distribution.
Radican et al.
(2008)
1.12
0.57
2.19
Mortality
HR
0.113
0.343
1.25 (0.31, 4.97)
for ICD-8, -9
155.0
ICD-8, -9 155 + 156, ICD-10 C22-C24. Time
variable = age; covariates = sex, race. Referent
group is workers with no chemical exposures.
Boice et al.
(2006a)
1.28
0.35
3.27
SMR
0.247
0.5
1.0 assumed for
Zhao et al.
(2005)
ICD-9 155 + 156. Boice et al. (2006a) used in lieu
of Zhao et al. (2005) because Zhao et al. (2005)
do not report liver cancer results. Boice et al.
(2006b) cohort overlaps Zhao cohort.
o
On
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Most of the selections in Table C-l 1 should be self-evident, but some are discussed in
more detail here, in the order the studies are presented in the table. For Axelson et al. (1994), in
which a small subcohort of females was studied but only results for the larger male subcohort
were reported, the reported male-only results were used in the primary analysis; however, as for
NHL and kidney cancer, an attempt was made to estimate the female contribution to an overall
RR estimate for both sexes and its impact on the meta-analysis. Axelson et al. (1994) reported
that there were no cases of liver cancer observed in females, but the expected number was not
presented. To estimate the expected number, the expected number for males was multiplied by
the ratio of female-to-male person-years in the study and by the ratio of female-to-male
age-adjusted incidence rates for liver cancer.8 The male results and the estimated female
contribution were then combined into an RR estimate for both sexes assuming a Poisson
distribution, and this alternate RR estimate for the Axelson et al. (1994) study was used in a
sensitivity analysis.
For Boice et al. (1999), results for "any potential exposure" were selected for the primary
analysis, because this exposure category was considered to best represent overall TCE exposure,
and results for "potential routine exposure", which was characterized as reflecting workers
assumed to have received more cumulative exposure, were used in a sensitivity analysis. To
estimate the SE(logRR) for the primary RR selection, it was assumed that the number of exposed
cases (deaths) was 15. The actual number was not presented, but 15 was the number that
allowed us to reproduce the reported CIs. The number suggested by exposure level in Boice et
al. (1999) Table 9 is 13; however, it may be that exposure level data were not available for all the
cases.
In their published paper, Morgan et al. (1998) present only SMRs for overall TCE
exposure, although the results from internal analyses are presented for exposure subgroups. RR
estimates for overall TCE exposure from the internal analyses of the Morgan et al. (1998) cohort
data were available from an unpublished report (EHS, 1997); from these, the RR estimate from
the Cox model which included age and sex was selected, because those are the variables deemed
to be important in the published paper. The internal analysis RR estimate was preferred for the
primary analysis, and the published SMR result was used in a sensitivity analysis.
8 Person-years for men and women <79 years were obtained from Axelson et al. (1994): 23516.5 and 3691.5,
respectively. Lifetime age-adjusted incidence rates for liver cancer for men and women were obtained from the
National Cancer Institute's 2000-2004 SEER-17 (Surveillance Epidemiology and End Results from 17 geographical
areas) database (http://seer.cancer.gov/statfacts/html/livibd.html): 9.5/100,000 and 3.4/100,000, respectively. The
calculation for estimating the expected number of cases in females in the cohort assumes that the males and females
have similar TCE exposures and that the relative distributions of age-related incidence risk for the males and
females in the Swedish cohort are adequately represented by the ratios of person-years and lifetime U.S. incidence
rates used in the calculation.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-62 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Raaschou-Nielsen et al. (2003) reported results for primary liver cancer (ICD-7 155.0),
gallbladder and biliary passage cancers (ICD-7 155.1), and unspecified liver cancers (ICD-7 156)
separately. As discussed above, RR estimates for cancers of the liver, gall bladder, and biliary
passages combined were preferred for the primary analysis; thus, the results for primary liver
cancer and gallbladder/biliary passage cancers were combined (across sexes as well), assuming a
Poisson distribution. The results for primary liver cancer only (similarly combined across sexes)
were used in an alternate analysis. The results for unspecified liver cancers (ICD-7 156) were
not included in any analyses because, under the ICD-7 coding, 156 can include secondary liver
cancers. Raaschou-Nielsen et al. (2003), in their Table 3, also present overall results for primary
liver cancer and gallbladder/biliary passage cancers with a lag time of 20 years; however, they
use a definition of lag that is different from a lagged exposure in which exposures prior to
disease onset are discounted and it is not clear what their lag time actually represents9, thus, as
for NHL and kidney cancer, these results were not used in any of the meta-analyses for liver
cancer. In addition, results for the subcohort with expected higher exposure levels were not
provided for liver cancer, so no alternate analysis was done based on the subcohort.
For Radican et al. (2008), the Cox model HR from the 2000 follow-up was used. In the
Radican et al. (2008) Cox regressions, age was the time variable, and sex and race were
covariates. It should also be noted that the referent group is composed of workers with no
chemical exposures, not just no exposure to TCE.
Zhao et al. (2005) did not present RR estimates for liver cancer; thus, results from Boice
et al. (2006b) were used in the primary analysis. The cohorts for these studies overlap, so they
are not independent studies. Zhao et al. (2005), however, was our preferred study for NHL and
kidney cancer results; thus, in a sensitivity analysis, a null value (RR = 1.0) was assumed for
Zhao et al. (2005) to address the potential reporting bias. The SE estimate for kidney cancer
(incidence with 0 lag) was used as the SE for the liver cancer. (It is not certain that there was a
reporting bias in this case. In the "Methods" section of their paper, Zhao et al. (2005) list the
cancer sites examined in the cohort, and liver was not listed; it is not clear if the list of sites was
determined a priori or post hoc)
Also, on the issue of potential reporting bias, the Siemiatycki (1991) study should be
mentioned. This study was a case-control study for multiple cancer sites, but only the more
common sites, in order to have greater statistical power. Thus, NHL and kidney cancer results
were available, but not liver cancer results. Because no liver results were presented for any of
the chemicals, this is not a case of reporting bias.
9 In their Methods section, Raaschou-Nielsen et al. (2003) define their lag period as the period "from the date of
first employment to the start of follow-up for cancer".
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
C.4.1.2. Results of Meta-Analyses
Results from some of the meta-analyses that were conducted on the epidemiological
studies of TCE and liver cancer are summarized in Table C-12. The RRm from the primary
random-effects meta-analysis of the 9 studies was 1.29 (95% CI: 1.07, 1.56) (see Figure C-8).
As shown in Figure C-8, the analysis was dominated by one large study (contributing about 53%
of the weight). That large study was critical in terms of the statistical significance of the RRm
estimate. Without the large Raaschou-Nielsen study, the RRm estimate decreases somewhat and
is no longer statistically significant (RRm = 1.22; 95% CI: 0.93, 1.61). No other single study
was overly influential; removal of any of the other individual studies resulted in RRm estimates
that were all statistically significant (all withp <0.03) and that ranged from 1.24 (with the
removal of Anttila) to 1.39 (with the removal of Boice (1999).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-64 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-12. Summary of some meta-analysis results for TCE and liver cancer
S"4
>}'
§•
to
s
Analysis
# of
studies
Model
Combined
RR estimate
95% LCL
95% UCL
Heterogeneity
Comments
All studies
(all cohort
studies)
9
Random
1.29
1.07
1.56
None obs
(fixed =
random)
Statistical significance not dependent
on single study, except for
Raaschou-Nielsen, without which
p = 0.15. No apparent publication
bias.

Fixed
1.29
1.07
1.56


All studies;
liver cancer
only, when
available
9
Random
1.25
0.99
1.57
None obs
Used RR estimates for liver cancer
alone for the 3 studies that
presented these; remaining RR
estimates are for liver and gall
bladder/biliary passage cancers.
Alternate RR
selections3
9
Random
1.28
1.06
1.55
None obs
With RR = 1 assumed for Zhao in
lieu of Boice (2006b) (see text).
9
Random
1.34
1.09
1.63
None obs
With Boice (1999) potential routine
exposure rather than any potential
exposure.
9
Random
1.29
1.07
1.55
None obs
With estimated female contribution to
Axelson.
9
Random
1.26
1.05
1.52
None obs
With Morgan published SMR.
Highest
exposure
groups
6
Random
1.32
0.93
1.86
None obs

8
Random
1.28
0.93
1.77
None obs
Primary analysis. Using RR = 1 for
Hansen and Zhao (see text).
7-8
Random
1.24-1.26
0.88-0.91
1.73-1.82
None obs
Using alternate selections for
Morgan and Raaschou-Nielsen and
excluding Axelson (see text).3
o
* v	§
§ I	5
a,	Co'
§-	a
TO	Sj-
§	^
O
a
>S
>S
TO
'S
TO
*
O
VO
to
o
o
On
aChanging the primary analysis by one alternate RR each time,
obs = observable.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Study
Anttila (1905)
A*elson (1994)
Bo ice (10&Q)
Boice (2006)
Greenland (1954)
Hansen (2001)
Morgan (1998)
R a asc h o u- N i a Ise n (2003)
Radiean (2008)
OVERALL
0.1
Figure C-8. Meta-analysis of liver cancer and TCE exposure. Random-effects
model; fixed-effect model same. The summary estimate is in the bottom row,
represented by the diamond. Symbol sizes reflect relative weights of the studies.
As discussed in Section C.4.1.1, all of the 9 studies presented results for liver and gall
bladder/biliary passage cancers combined, and these results were the basis for the primary
analysis discussed above. An alternate analysis was performed substituting, simultaneously,
results for liver cancer alone for the 3 studies for which these were available. The RRm estimate
from this analysis was slightly lower than the one based entirely on results from the combined
cancer categories and was just short of statistical significance (1.25; 95% CI: 0.99, 1.57). This
result was driven by the fact that the RR estimate from the large Raaschou-Nielsen et al. (2003)
study decreased from 1.35 for liver and gall bladder/biliary passage cancers combined to 1.28 for
liver cancer alone.
Similarly, the RRm estimate was not highly sensitive to other alternate RR estimate
selections. Use of the 4 other alternate selections, individually, resulted in RRm estimates that
were all statistically significant (all withp < 0.02) and that ranged from 1.26 to 1.34 (see
Table C-12). In fact, as can be seen in Table C-12, only one of the alternates had notable impact.
The Boice (2006b), Morgan, and Axelson original values and alternate selections were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-66 DRAFT—DO NOT CITE OR QUOTE
TCE Exposure and Liver Cancer
Relative Risk and 05% CI

UZH
RR
1.39
1.41
0.31
1.28
0.54
2.10
1.48
1.35
1.12
1.20
LyL
0.86
0.38
0.45
0.35
0.11
0.70
0.50
1.03
0.57
1.07
UCL
3.50
3.60
1.33
3.27
2.63
5.00
3.01
1.77
2.10
1.56
10
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
associated with very little weight and, thus, have little influence in the RRm. Using the Boice
(1999) alternate RR estimate based on potential routine exposure rather than any potential
exposure increased the RRm slightly from 1.29 to 1.34. The alternate Boice (1999) RR estimate
is actually smaller than the original value (0.54 vs. 0.81); however, use of the more restrictive
exposure metric captures fewer liver cancer deaths, causing the weight of that study to decrease
from almost 14% to about 4.1%.
There was no apparent heterogeneity across the nine studies, i.e., the random-effects
model and the fixed-effect model gave the same results (I2 = 0%). Furthermore, all of the liver
cancer studies were cohort studies, so no subgroup analyses examining cohort and case-control
studies separately, as was done for NHL and kidney cancer, were conducted. No alternate
quantitative investigations of heterogeneity were pursued because of database limitations and, in
any event, there is no evidence of heterogeneity of study results in this database.
As discussed in Section C.l, publication bias was examined in several different ways.
The funnel plot in Figure C-9 shows little relationship between RR estimate and study size, and,
indeed, none of the other tests performed found any evidence of publication bias. Duval and
Tweedie's trim-and-fill procedure, for example, suggested that no studies were missing from the
funnel plot, i.e., there was no asymmetry to counterbalance. Similarly, the results of a
cumulative meta-analysis, incorporating studies with increasing SE one at a time, shows no
evidence of a trend of increasing effect size with addition of the less precise studies. The
Raaschou-Nielsen study contributes about 53% of the weight. Including the 2 next most precise
studies, the RRm goes from 1.35 to 1.10 to 1.25 and the weight to 75%. With the addition of the
next 2 most precise studies, the RRm estimate goes to 1.23 and then 1.29. Further addition of the
4 least precise studies leaves the RRm essentially unchanged.
C.4.2. Liver Cancer Effect in the Highest Exposure Groups
C.4.2.1. Selection of RR Estimates
The selected RR estimates for liver cancer in the highest TCE exposure categories, for
studies that provided such estimates, are presented in Table C-13. Six of the 9 cohort studies
reported liver cancer risk estimates categorized by exposure level. As in Section C.4.1.1 for the
overall risk estimates, estimates for cancers of the liver and gall bladder/biliary passages
combined were preferentially selected, when presented, for the sake of consistency, and,
wherever possible, RR estimates for males and females combined were used.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-67 DRAFT—DO NOT CITE OR QUOTE
-------
Funnel Plot of Standard Error by Log risk ratio
0.0
0.2
0.4
Standard Eirror
O CD
0.6
0.8
-2.0	-1.5	-1.0	-0.5	0.0	0.5	1.0	1.5	2.0
Log risk ratio
Figure C-9. Funnel plot of SE by log RR estimate for TCE and liver cancer
studies.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-68 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-13. Selected RR estimates for liver cancer risk in highest TCE exposure groups
S"4
>}'
§•
to
s
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE(log
RR)
Alternate RR
estimates
Comments
Anttila et
al. (1995)
2.74
0.33
9.88
100+ |jmol/L
U-TCAa
1.008
0.707
None
SIR. ICD-7 155.0 (liver only).
Axelson et
al. (1994)
3.7
0.09
21
100+ mg/L
U-TCA
1.308
1.000
Exclude study
SIR. ICD-7 155. 0 cases observed in
highest exposure group (i.e., >2 years and
100+ U-TCA), so combined with <2 years
and 100+ subgroup and females, estimating
the expected numbers (see text).
Boice et al.
(1999)
0.94
0.36
2.46
> 5 years
exposure
-0.062
0.490
None
Mortality RR. ICD-9 155 + 156. For
potential routine or intermittent exposure.
Adjusted for date of birth, dates 1st and last
employed, race, and sex. Referent group is
workers not exposed to any solvent.
Hansen et
al. (2001)



> 1080 months
x mg/m3


1.0 assumed
Reported high exposure group results for
some cancer sites but not liver.
Morgan et
al. (1998)
1.19
0.34
4.16
High
cumulative
exposure score
0.174
0.639
0.98 (0.29,
3.35) for
med/high peak
vs. low/no
Mortality RR. ICD-9 155 + 156. Adjusted
for age and sex.
Raaschou-
Nielsen et
al. (2003)
1.2
0.7
1.9
> 5 years
0.182
0.243
1.1 (0.5, 2.1)
ICD-7 155.0
(liver only)
SIR. ICD-7 155.0 + 155.1. Male and
female results presented separately and
combined assuming a Poisson distribution.
Radican et
al. (2008)
1.49
0.67
3.34
> 25 unit-years
0.399
0.411
None (see text)
Mortality HR. ICD-8, -9 155 + 156, ICD-10
C22-C24. Male and female results
presented separately and combined (see
text). Time variable = age, covariate = race.
Referent group is workers with no chemical
exposures.
Zhao et al.
(2005)



High exposure
score


1.0 assumed
No liver results reported.
o
* v §
§ I 5
5 a, co-
c §- a
5 S S-
?• i
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
O
VO
to
o
^ aMean personal trichloroacetic acid in urine. 1 |imol/L = 0.1634 mg/L.
On
VO
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Two of the 9 cohort studies (Hansen et al., 2001; Zhao et al., 2005) did not report liver
cancer risk estimates categorized by exposure level even though these same studies reported such
estimates for selected other cancer sites. To address this reporting bias (as discussed above,
Zhao et al. (2005) did not present any liver results, and it is not clear if this was actual reporting
bias or an a priori decision not to examine liver cancer in the cohort), attempts were made to
obtain the results from the primary investigators, and, failing that, alternate analyses were
performed in which null estimates (RR = 1.0) were included for both studies. This alternate
analysis was then used as the main analysis, e.g., the basis of comparison for the sensitivity
analyses. For the SE (of the logRR) estimates for the null estimates, SE estimates from other
sites for which highest-exposure-group results were available were used. For Hansen et al.
(2001), the SE estimate for NHL in the highest exposure group was used, because NHL was the
only cancer site of interest in this assessment for which highest-exposure-group results were
available. For Zhao et al. (2005), the SE estimate for kidney cancer in the highest exposure
group (incidence with 0 lag) was used. (Note that Boice et al. (2006b), who studied a cohort that
overlapped that of Zhao et al. (2005), also did not present liver cancer results by exposure level.)
For Axel son et al. (1994), there were no liver cancer cases in the highest exposure group
(>2 years and 100+ mean urinary-trichloroacetic acid [U-TCA] level), so no log RR and
SE(log RR) estimates were available for the meta-analysis. Instead, the <2 years and >2 years
results were combined, assuming expected numbers of cases were proportional to person-years,
and 100+ U-TCA (with any exposure duration) was used as the highest exposure category. The
female contribution to the expected number was also estimated, again assuming proportionality
to person-years, and adjusting for the difference between female and male age-adjusted liver
cancer incidence rates. The estimated RR and SE values for the combined exposure times and
sexes were used in the primary analysis. In an alternate analysis, the Axelson et al. (1994) study
was excluded altogether, because we estimated that less than 0.2 cases were expected in the
highest exposure category, suggesting that the study had low power to detect an effect in the
highest exposure group and would contribute little weight to the meta-analysis.
For Boice et al. (1999), only results for workers with "any potential exposure" were
presented by exposure category, and the referent group is workers not exposed to any solvent.
For Morgan et al. (1998), the primary analysis used results for the cumulative exposure metric,
and a sensitivity analysis was done with the results for the peak exposure metric. For Raaschou-
Nielsen et al. (2003), unlike for NHL and RCC, liver cancer results for the subcohort with
expected higher exposure levels were not presented, so the only highest-exposure-group results
were for duration of employment in the total cohort. Results for cancers of the liver and gall
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
bladder/biliary passages combined were used for the primary analysis and results for liver cancer
alone in a sensitivity analysis.
For Radican et al. (2008), it should be noted that the referent group is workers with no
chemical exposures, not just no TCE exposure. Furthermore, results for exposure groups (based
on cumulative exposure scores) were reported separately for males and females and were
combined for this assessment using inverse-variance weighting, as in a fixed-effect meta-
analysis. In addition to results for biliary passage and liver cancer combined, Radican et al.
(2008) present results for liver only by exposure group; however, there were no liver cancer
deaths in females and the number expected was not reported, so no alternate analysis for the
highest exposure groups with an RR estimate from Radican et al. (2008) for liver cancer only
was conducted. Radican et al. (2008) present only mortality HR estimates by exposure group;
however, in an earlier follow-up of this same cohort, Blair et al. (1998) present both incidence
and mortality RR estimates by exposure group. As with the Radican et al. (2008) liver cancer
only results, however, there were no incident cases for females in the highest exposure group in
Blair et al. (1998) (and the expected number was not reported). Additionally, there were more
biliary passage/liver cancer deaths (31) in Radican et al. (2008) than incident cases (13) in Blair
et al. (1998) overall and in the highest exposure group (14 vs. 4). Thus, we elected to use only
the Radican et al. (2008) mortality results from this cohort and not to include an alternate
analysis based on incidence results from the earlier follow-up. Radican et al. (2008) also present
results for categories based on frequency and pattern of exposure; however, subjects weren't
distributed uniquely across the categories (the numbers of cases across categories exceeded the
total number of cases), thus it was difficult to interpret these results and they were not used in a
sensitivity analysis.
C.4.2.2. Results of Meta-Analyses
Results from the meta-analyses that were conducted for liver cancer in the highest
exposure groups are summarized at the bottom of Table C-12. The RRm estimate from the
random-effects meta-analysis of the 6 studies with results presented for exposure groups was
1.32 (95% CI: 0.93, 1.86). As with the overall liver cancer meta-analyses, the meta-analyses of
the highest exposure groups were dominated by one study (Raaschou-Nielsen), which provided
about 52% of the weight. The RRm estimate from the primary random-effects meta-analysis
with null RR estimates (i.e., 1.0) included for Hansen and Zhao to address (potential) reporting
bias (see above) was 1.28 (95% CI: 0.93, 1.77) (see Figure C-10). The inclusion of these
2 additional studies contributed about 10% of the total weight. No single study was overly
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-71 DRAFT—DO NOT CITE OR QUOTE
-------
1	influential (removal of individual studies resulted in non-significant RRm estimates that ranged
2	from 1.23 to 1.36), and the RRm estimate was not highly sensitive to alternate RR estimate
3	selections (RRm estimates with alternate selections ranged from 1.24 to 1.26, all non-significant;
4	see Table C-12). In addition, there was no observable heterogeneity across the studies for any of
5	the meta-analyses conducted with the highest exposure groups (/ = 0%). However, none of the
6	RRm estimates was statistically significant.
7
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Stud-
TCE Exposure and Livei Cancel - hitjhest exposure (jioups
Relative Risk and 95* CI	RR LCL UCL
Anttila (1095)
An.elson f1994)est
Bo ice (1699)
Morgan (1988)
Raaschou-Nielsen (2003)
Radioan (2008)
Hansen (2001)riull
Zhao (200o)nuli
OVERALL
0.1
2,74
3,70
0.94
1.10
1.20
1.49
1.00
1.00
1.28
0.33
0.09
0.36
0.34
0.70
0.67
0.32
0.03
0.93
9.88
21.00
2.43
4.16
1.90
3.34
3.10
11.00
1.77
10
Figure C-10. Meta-analysis of liver cancer and TCE exposure—highest
exposure groups, with assumed null RR estimates for Hansen and Zhao (see
text). Random-effects model; fixed-effect model same. The summary estimate is
in the bottom row, represented by the diamond. Symbol sizes reflect relative
weights of the studies.
Furthermore, most of the RRm estimates for the highest exposure groups were less than
the significant RRm estimate for an overall effect on liver cancer (1.29; 95% CI: 1.07, 1.56; see
Section C.4.2.2 and Table C-12). This contradictory result is driven by the fact that the RR
estimate for the highest exposure group was less than the overall RR estimate for Raaschou-
Nielsen, which contributes the majority of the weight to the meta-analyses. The liver cancer
results are relatively underpowered with respect to numbers of studies and number of cases, and
the Raaschou-Nielsen study, which dominates the analysis, uses duration of employment as an
exposure-level surrogate for liver cancer, and duration of employment is a notoriously weak
exposure metriclO. Thus, the contradictory finding that most of the RRm estimates for the
10 Moreover, this study is prone to misclassifying some of the subjects with longer durations of employment as
having lesser durations of employment due to the fact that employment information prior to 1964 was not available
and, thus, employment prior to 1964 was not included in the calculations of duration of employment. For example,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-73 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
highest exposure groups were less than the RRm estimate for an overall effect does not rule out
an effect of TCE on liver cancer; however, it certainly does not provide additional support for
such an effect.
C.4.3. Discussion of Liver Cancer Meta-Analysis Results
For the most part, the meta-analyses of the overall effect of TCE exposure on liver (and
gall bladder/biliary passages) cancer suggest a small, statistically significant increase in risk.
The summary estimate from the primary random-effects meta-analysis of the 9 (all cohort)
studies was 1.29 (95% CI: 1.07, 1.56). The analysis was dominated by one large study that
contributed about 53% of the weight. When this study was removed, the RRm estimate
decreased somewhat and was no longer statistically significant (RRm = 1.22; 95% CI: 0.93,
1.61). The summary estimate was not overly influenced by any other single study, nor was it
overly sensitive to individual RR estimate selections. The next largest downward impacts were
from the removal of the Anttila study, resulting in an RRm estimate of 1.24 (95% CI: 1.02, 1.51),
and from the substitution of the Morgan unpublished RR estimate with the published SMR
estimate, resulting in an RRm estimate of 1.26 (1.05, 1.52). Substituting the RR estimates for
liver/gall bladder/biliary passage cancers with those of liver cancer alone for the 3 studies that
provided these results yielded an RRm estimate of 1.25 (0.99, 1.57). There was no evidence of
publication bias in this data set, and there was no observable heterogeneity across the study
results.
Six of the 9 studies provided liver cancer results by exposure level. Two other studies
reported results for other cancer sites by exposure level, but not liver cancer; thus, to address this
reporting bias, null values (i.e., RR estimates of 1.0) were used for these studies. Different
exposure metrics were used in the various studies, and the purpose of combining results across
the different highest exposure groups was not to estimate an RRm associated with some level of
exposure, but rather to see the impacts of combining RR estimates that should be less affected by
exposure misclassification. In other words, the highest exposure category is more likely to
represent a greater differential TCE exposure compared to people in the referent group than the
exposure differential for the overall (typically any vs. none) exposure comparison. Thus, if TCE
exposure increases the risk of liver cancer, the effects should be more apparent in the highest
17 of the 27 primary liver cancer cases in men were observed in men first employed before 1970 and some of these
might have occurred in men first employed before 1964. Thus, some of the 18 cases with durations of employment
reported as < 5 years may actually have had durations > 5 years and hence may have belonged in the highest
exposure group.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-74 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
exposure groups. However, the RRm estimate from the primary meta-analysis of the highest
exposure group results (and most of the RRm estimates from the sensitivity analyses) was less
than the RRm estimate from the overall exposure analysis. This anomalous result is driven by
the fact that for Raaschou-Nielsen, which contributes the majority of the weight to the meta-
analyses, the RR estimate for the highest exposure group, although greater than 1, was less than
the overall RR estimate.
Thus, while there is the suggestion of an increased risk for liver cancer associated with
TCE exposure, the statistical significance of the overall summary estimate is dependent on one
study, which provides the majority of the weight in the meta-analyses. Removal of this study
yields an RRm estimate that is decreased somewhat but is still greater than 1; however, it
becomes non-significant (p = 0.15). Furthermore, meta-analysis results for the highest exposure
groups yielded generally lower RRm estimates than for an overall effect. These results do not
rule out an effect of TCE on liver cancer, because the liver cancer results are relatively
underpowered with respect to numbers of studies and number of cases and the overwhelming
study in terms of weight uses the weak exposure surrogate of duration of employment for
categorizing exposure level; however, at present, there is only modest support for such an effect.
C.5. META-ANALYSIS FOR LUNG CANCER
C.5 .1. Overall Effect of TCE Exposure
C. 5.1.1. Selection of RR Estimates
Although there was no general indication of an increased risk of lung cancer associated
with TCE exposure in the epidemiologic literature, the Science Advisory Board recommended a
meta-analysis for lung cancer to more exhaustively examine the issue of smoking as a possible
confounder in the kidney cancer studies (SAB, 2011). Only the cohort studies were considered
for the meta-analysis because these provide a consistent group of studies to compare RRm
estimates for kidney cancer to those for lung cancer and the cohort studies are the studies of
concern for potential confounding since the kidney cancer results from these studies were not
adjusted for smoking. The selected RR estimates for lung cancer from the 9 cohort studies are
presented in Table C-14. All of the studies, with the possible exception of Greenland et al.
(1994), reported cancers of the lung and bronchus combined. Some also included cancer of the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-75 DRAFT—DO NOT CITE OR QUOTE
-------
1	trachea; however, this is a rare tumor (<0.1% of tumors) (Macchiarini, 2006) and so its inclusion
2	is negligible.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-76 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table C-14. Selected RR estimates for lung (& bronchus) cancer associated with TCE exposure (overall effect)
2	from cohort studies
3
Study
RR
95%
LCL
95%
UCL
RR type
log RR
SE(log
RR)
Alternate RR
estimates
Comments
Anttila et al.
(1995)
0.92
0.59
1.35
SIR
-0.0834
0.2
None

Axelson et al.
(1994)
0.69
0.31
1.30
SIR
-0.371
0.333
None
Results reported for males only, but there was a
small female component to the cohort.
Boice et al.
(1999)
0.76
0.66
0.87
SMR
-0.274
0.0705
0.76 (0.60, 0.95)
for potential
routine exposure
For any potential exposure.
Greenland et
al. (1994)
1.01
0.69
1.47
OR
0.00995
0.193
None
Nested case-control study.
Hansen et al.
(2001)
0.8
0.5
1.3
SIR
-0.223
0.243
None
Male and female results reported separately;
combined assuming Poisson distribution.
Morgan et al.
(1998)
1.14
0.90
1.44
SMR
0.133
0.119
Published SMR
1.10 (0.89, 1.34)
Unpublished RR, adjusted for age and sex (see
text).
Raaschou-
Nielsen et al.
(2003)
1.43
1.32
1.55
SIR
0.358
0.0398
None

Radican et al.
(2008)
0.83
0.63
1.08
Mortality
HR
-0.186
0.138
None
Time variable = age; covariates = sex, race.
Referent group is workers with no chemical
exposures.
Zhao et al.
(2005)
1.04
0.81
1.34
RR
0.0392
0.128
1.27 (0.88, 1.83)
for incidence.
1.24 (0.92, 1.63)
for Boice
(2006b)
mortality.
mortality
4
10/20/09
This document is a draft for review purposes only and does not constitute Agency policy.
DRAFT^EX) NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
As for NHL and kidney and liver cancer, many of the studies provided RR estimates only
for males and females combined, and we are not aware of any basis for a sex difference in the
effects of TCE on lung cancer risk; thus, wherever possible, RR estimates for males and females
combined were used. The only two studies of much size (in terms of number of lung cancer
cases) that provided results separately by sex were Raaschou-Nielsen (2003) and Radican et al.
(2008). The results from Raaschou-Nielsen (2003) suggest that lung cancer relative risk in
females might be slightly higher than the relative risk in males (SIR: males 1.4 [95% CI: 1.3, 1.5;
559 cases], females 1.9 [1.5, 2.4; 73 cases]), but the difference narrows when a 20-year lag is
taken into account (males 1.4 [1.2, 1.6; 202 cases], females 1.6 [1.0, 2.3; 26 cases]). Radican et
al. (2008) report HRs for lung cancer of 0.91 (95% CI: 0.67, 1.24; 155 deaths) for males and 0.53
(0.27, 1.07; 11 deaths) for females, but these results are based on fewer cases, especially in
females.
Most of the selections in Table C-14 should be self-evident, but some are discussed in
more detail here, in the order the studies are presented in the table. For Axelson et al. (1994), in
which a small subcohort of females was studied but only results for the larger male subcohort
were reported, only the reported male results were used. Unlike for NHL and kidney and liver
cancer, no attempt was made to estimate the female contribution to an overall RR estimate for
both sexes and its impact on the meta-analysis because, unlike for those other cancer types, the
meta-analysis for lung cancer was not done to test a null hypothesis of no effect but, rather, to
investigate whether or not smoking might be confounding the kidney cancer results. An
association of TCE exposure and lung cancer might indicate a confounding effect of smoking (or
a causal association with lung cancer), but a finding of no association would essentially rule out
a confounding effect of smoking, since smoking is such a strong risk factor for lung cancer.
Axelson et al. (1994) reported neither the number of lung cancers observed in females nor the
number expected. To test a null hypothesis of no effect, one might conservatively assume none
was observed and estimate the number expected, as was done for kidney cancer; however, since
that is not the hypothesis here, we chose not to make any assumptions or estimates for the female
component of the cohort.
For Boice et al. (1999), results for "any potential exposure" were selected for the primary
analysis, because this exposure category was considered to best represent overall TCE exposure,
and results for "potential routine exposure", which was characterized as reflecting workers
assumed to have received more cumulative exposure, were used in a sensitivity analysis. The
number of cases (deaths) with "any potential exposure" was not presented, but a value of 200
allowed us to reproduce the reported CIs. The number suggested by exposure level in Boice et
al. (1999) Table 9 is 173; however, it may be that exposure level data were not available for all
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-78 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
the cases. Because the exact number is unknown but is a large number, consistent with CIs that
are proportionally symmetric, the SE(logRR) was calculated as from symmetric CIs (see Section
c.i).
In their published paper, Morgan et al. (1998) present only SMRs for overall TCE
exposure, although the results from internal analyses are presented for exposure subgroups. RR
estimates for overall TCE exposure from the internal analyses of the Morgan et al. (1998) cohort
data were available from an unpublished report (EHS, 1997); from these, the RR estimate from
the Cox model which included age and sex was selected, because those are the variables deemed
to be important in the published paper. The internal analysis RR estimate was preferred for the
primary analysis, and the published SMR result was used in a sensitivity analysis.
Raaschou-Nielsen et al. (2003) reported results for lung cancer for both sexes combined
in the text. In their Table 3, Raaschou-Nielsen et al. (2003) also present overall results for lung
cancer with a lag time of 20 years; however, they use a definition of lag that is different from a
lagged exposure in which exposures prior to disease onset are discounted and it is not clear what
their lag time actually represents 11, thus, these results were not used in any of the meta-analyses
for lung cancer. In addition, results for the subcohort with expected higher exposure levels were
not provided for lung cancer, so no alternate analysis was done based on the subcohort.
For Radican et al. (2008), the Cox model HR from the 2000 follow-up was used. In the
Radican et al. (2008) Cox regressions, age was the time variable, and sex and race were
covariates. It should also be noted that the referent group is composed of workers with no
chemical exposures, not just no exposure to TCE.
Zhao et al. (2005) do not report results for an overall TCE effect. Therefore, as for NHL
and kidney cancer, the results across the "medium" and "high" exposure groups were combined,
under assumptions of group independence, even though the exposure groups are not independent
(the "low" exposure group was the referent group in both cases). Zhao et al. (2005) present RR
estimates for both incidence and mortality; however, the time frame for the incidence accrual is
smaller than the time frame for mortality accrual and fewer exposed incident cases (49) were
obtained than deaths (95). Thus, because better case ascertainment occurred for mortality than
for incidence, the mortality results were used for the primary analysis, and the incidence results
were used in a sensitivity analysis. A sensitivity analysis was also done using results from Boice
et al. (2006b) in place of the Zhao et al. (2005) RR estimate. The cohorts for these studies
overlap, so they are not independent studies and should not be included in the meta-analysis
concurrently. Boice et al. (2006b) report an RR estimate for an overall TCE effect for lung
11 In their Methods section, Raaschou-Nielsen et al. (2003) define their lag period as the period "from the date of
first employment to the start of follow-up for cancer".
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-79 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
cancer mortality; however, it is based on fewer deaths (51) and is an SMR rather than an internal
analysis RR estimate, so the Zhao et al. (2005) mortality estimate is preferred for the primary
analysis.
C.5.1.2. Results of Meta-Analyses
Results from some of the meta-analyses that were conducted on the epidemiological
studies of TCE and lung cancer are summarized in Table C-15. The RRm from the fixed-effect
meta-analysis of the 9 studies was 1.16 (95% CI: 1.09, 1.23) (see Figure C-ll). As shown in
Figure C-ll, the analysis was dominated by one large study (Raaschou-Nielsen, contributing
about 58% of the weight). The RR estimate from that large study was higher than the RR
estimates from all the other studies and, with its relatively narrow confidence interval, was
largely inconsistent with the results of the other studies, in particular that of the next largest
study (Boice (1999), contributing about 18% of the weight). While the RR estimate of
Raaschou-Nielsen was statistically significantly elevated, that of Boice (1999) was statistically
significantly decreased. This heterogeneity of study results is corroborated by a statistically
8	7
significant p-value for the test of heterogeneity (p < 10") and an I -value of 90%, indicating a
high amount of heterogeneity. Because of this heterogeneity, the appropriateness of conducting
any meta-analysis without attempting to explain the heterogeneity is arguable, but a fixed-effect
meta-analysis is clearly improper (see Section C.l).
The RRm from the primary random-effects meta-analysis of the 9 studies was 0.96
(95% CI: 0.76, 1.21) (see Figure C-12). As shown in Figure C-12, because the random-effects
model takes both between-study and within-study variation into account in the study weight, and
because the between-study variation is fairly substantial for these studies, study size has minimal
impact on study weight. The relative weights for the 9 studies range from 6.7% to 13.9% in the
random-effects meta-analysis, thus no single study dominates the analysis in terms of weight.
The most influential single study is nonetheless the largest study, Raaschou-Nielsen, because it
also has an RR estimate well above the others, and its removal from the analysis reduces the
RRm estimate to 0.90 (0.79, 1.04). In contrast, removal of Boice (1999), the study with the
lowest RR estimate, increases the RRm estimate to 1.01 (0.82, 1.24). Removal of any of the
other individual studies resulted in RRm estimates that were all non-significantly decreased and
that ranged from 0.93 (with the removal of Morgan) to 0.98 (with the removal of Axelson,
Hansen, or Radican). Use of the 4 alternate selections, individually, resulted in RRm estimates
that were all non-significant and that fell in a narrower range — 0.96 to 0.98 (see Table C-15).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-80 DRAFT—DO NOT CITE OR QUOTE
-------
Study
Anttila (1905)
A*elson (1994)
Bo ice (10&Q)
Greenland (1904)
Hansen (2001)
Morgan (1908)
Raaschou-Nielsen (2003)
Radican (2003)
Zhao (2005)
OVERALL
TCE Exposure <111(1 Liinij Cancer
Relative Risk and 05% CI

RR
LyL
UCL
0.92
0.59
1.35
0.89
0.31
1.30
0.70
0.68
0.87
1.01
0.09
1.47
0.30
0.50
1.30
1.14
0.90
1.44
1.43
1.32
1.55
0.83
0.03
1.08
1.04
0.81
1.34
1.10
1.09
1.23
0.1
10
Figure C-ll. Meta-analysis of lung cancer and TCE exposure - fixed-effect
model. The summary estimate is in the bottom row, represented by the diamond.
Symbol sizes reflect relative weights of the studies.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-81 DRAFT—DO NOT CITE OR QUOTE
-------

TCE Exposure and Lung Cancer



Study
Relative Risk and 05% CI
RR
LyL
UCL
Anttila (1905)
		~
_
0.92
0.59
1.35
A*elson (1994)
	D	
.
0.89
0.31
1.30
Bo ice (10&Q)
-Q

0.70
0.68
0.87
Greenland (1904)
—C

1.01
0.09
1.47
Hansen (2001)
	
_
0.30
0.50
1.30
Morgan (1908)


1.14
0.90
1.44
Raaschou-Nielsen (2003)

~
1.43
1.32
1.55
Radican (2003)


0.83
0.03
1.08
Zhao (2005)
-fa-
1.04
0.81
1.34
OVERALL


0.96
0.76
1.21
1	
0.1
1
10


2
3	Figure C-12. Meta-analysis of lung cancer and TCE exposure - random-
4	effects model. The summary estimate is in the bottom row, represented by the
5	diamond. Symbol sizes reflect relative weights of the studies.
6
7
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-82 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-15. Summary of some meta-analysis results for TCE and lung cancer
S"4
>}'
§•
to
s
Analysis
# of
studies
Model
Combined
RR estimate
95% LCL
95% UCL
Heterogeneity
Comments
All studies
(all cohort
studies)
9
Random
0.96
0.76
1.21
Significant
(p< 10"8)
r = 90%
Non-significance of RRm not
dependent on any single study.
No apparent publication bias.

Fixed
1.16
1.09
1.23
Because of
significant
heterogeneity,
fixed-effect
model not
appropriate.
Significant elevation in RRm
dependent on single study,
Raaschou-Nielsen, without which the
RRm would be significantly
decreased (RRm = 0.87, p = 0.004).
Alternate RR
selections3
9
Random
0.98
0.78
1.25
Significant
(p< 10"8)
r = 90%
With Zhao incidence instead of
mortality.
9
Random
0.98
0.77
1.24
Significant
(p< 10"8)
r = 90%
With Boice (2006b) instead of Zhao.
9
Random
0.97
0.78
1.20
Significant
(p< 10"7)
r = 85%
With Boice (1999) potential routine
exposure rather than any potential
exposure.
9
Random
0.96
0.76
1.20
Significant
(p< 10"8)
r = 90%
With Morgan published SMR.
Highest
exposure
groups
6
Random
0.96
0.72
1.27
Significant
See Table C-17 for details.
6
Random
0.92-0.98
0.67-0.75
1.25-1.30

Using alternate selections (see
text).3
o
* v §
§ I 5
5 a, co-
c 8- a
5 S S-
?• i <§
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
O
VO
to
o
aChanging the primary analysis by one alternate RR each time.
O
oo
LtJ
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
As discussed above, there was significant heterogeneity across the nine studies. All of
the lung cancer studies were cohort studies, so no subgroup analyses examining cohort and case-
control studies separately, as was done for NHL and kidney cancer, were conducted. In addition,
no alternate quantitative investigations of heterogeneity were pursued because our goal here was
to investigate lung cancer risks as an indication of possible confounding of the kidney cancer
results by smoking, not to do an all-encompassing meta-analysis of lung cancer. The majority of
the studies have non-significant RR estimates for lung cancer that fall near or below 1. The
relative outliers are the significantly increased RR estimate from Raaschou-Nielsen and the
significantly decreased RR estimate from Boice (1999). The Raaschou-Nielsen et al. (2003)
study considered a lot of different job titles and the RR estimate could reflect a TCE effect or
exposure to other chemicals that are lung carcinogens. Alternatively, because the study is an
SMR study of largely blue-collar workers and the comparison population is the general Danish
population, the elevated RR estimate could reflect small differences in smoking rates between
those two populations. However, if the observed increase is attributable to smoking, it's not
enough of an effect to explain the increased RR estimate for RCC in the same study because
smoking is a much stronger risk factor for lung cancer than for RCC, whereas the increased RR
estimate for lung cancer in the study was relatively small (Raaschou-Nielsen et al., 2003); see
also Section 4.4.2.3). It is unclear why the Boice et al. (1999) study reports a significantly
decreased RR estimate. In any event, there's no increase in the RRm estimate for all 9 studies
from the random-effects model, suggesting that there's no confounding of the overall RRm for
kidney cancer by smoking, in particular for the cohort studies.
As discussed in Section C.l, publication bias was examined in several different ways, and
there's no indication of publication bias for these lung cancer studies (results not shown). If
anything, the relationship between study size and RR estimate is the opposite of what would be
expected if publication bias were occurring because the one large study is the only study with a
significantly increased RR estimate and incorporating studies with increasing SE one at a time,
generally shows a decrease in effect size with addition of the less precise studies.
C.5.2. Lung Cancer Effect in the Highest Exposure Groups
C.5 .2.1. Selection of RR Estimates
The selected RR estimates for lung cancer in the highest TCE exposure categories, for
studies that provided such estimates, are presented in Table C-16. Six of the 9 cohort studies
reported lung cancer risk estimates categorized by exposure level. As in Section C.5.1.1 for the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-84 DRAFT—DO NOT CITE OR QUOTE
-------
1	overall risk estimates, RR estimates for males and females combined were used, wherever
2	possible.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-85 DRAFT—DO NOT CITE OR QUOTE
-------
Table C-16. Selected RR estimates for lung cancer risk in highest TCE exposure groups
S"4
to
s
Study
RR
95%
LCL
95%
UCL
Exposure
category
log RR
SE
(log RR)
Alternate RR
estimates
Comments
Anttila et
al. (1995)
0.83
0.33
1.71
100+ |jmol/L
U-TCAa
-0.186
0.378
None
SIR.
Boice et al.
(1999)
0.64
0.46
0.89
> 5 years
exposure
-0.446
0.168
None
Mortality RR. For any potential exposure.
Adjusted for date of birth, dates 1st and last
employed, race, and sex. Referent group is
workers not exposed to any solvent.
Morgan et
al. (1998)
0.96
0.72
1.29
High
cumulative
exposure
score
-0.041
0.149
1.07 (0.82, 1.40)
for med/high peak
vs. low/no
Mortality RR. Adjusted for age and sex.
Raaschou-
Nielsen et
al. (2003)
1.4
1.2
1.6
> 5 years
0.336
0.070
None
SIR. Male and female results presented
separately and combined assuming a
Poisson distribution.
Radican et
al. (2008)
0.90
0.63
1.27
> 25 unit-
years
-0.105
0.179
0.8 (0.4, 1.7) for
Blair incidence
Mortality HR. Male and female results
presented separately and combined (see
text). Time variable = age, covariate = race.
Referent group is workers with no chemical
exposures.
Zhao et al.
(2005)
1.0
0.68
1.53
High
exposure
score
0.020
0.207
1.1 (0.60, 2.06) for
Zhao incidence.
Boice (2006b):
0.80 (0.46, 1.41)
for > 4 years with
any potential exp;
0.86 (0.56, 1.33)
for > 5 years test
stand mechanic,
0.76 (0.42, 1.36)
for> 4 test-years.
Mortality RR. Males only. Adjusted for time
since 1st employment, SES, age.
o
* v §
§ I 5
5 a, co-
c 8- a
5 S S-
?• i <§
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
to
o
aMean personal trichloroacetic acid in urine. 1 |imol/L = 0.1634 mg/L.
O SES: socio-economic status.
oo
On
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Three of the 9 cohort studies (Axelson et al., 1994); (Hansen et al., 2001); (Zhao et al.,
2005) did not report lung cancer risk estimates categorized by exposure level even though these
same studies reported such estimates for selected other cancer sites. Unlike for the other cancer
types, we did not attempt to address the issue of unreported results by including RR estimates of
1 for the missing estimates. This is because, as discussed in Section 5.1.1 above with respect to
estimate a female contribution to the Axelson study, unlike for the other cancer types, we are not
testing a null hypothesis of no effect for lung cancer but rather investigating whether smoking
might be a confounder in the kidney cancer studies. Thus, we would not want to bias the RRm
estimate toward 1 in this case by including estimates of 1 for missing RR values.
For Boice et al. (1999), only results for workers with "any potential exposure" were
presented by exposure category, and the referent group is workers not exposed to any solvent.
For Morgan et al. (1998), the primary analysis used results for the cumulative exposure
metric, and a sensitivity analysis was done with the results for the peak exposure metric.
For Raaschou-Nielsen et al. (2003), unlike for NHL and RCC, lung cancer results for the
subcohort with expected higher exposure levels were not presented, so the only highest-
exposure-group results were for duration of employment in the total cohort. Results for males
and females combined were estimated assuming a Poisson distribution.
For Radican et al. (2008), it should be noted that the referent group is workers with no
chemical exposures, not just no TCE exposure. Furthermore, results for exposure groups (based
on cumulative exposure scores) were reported separately for males and females and were
combined for this assessment using inverse-variance weighting, as in a fixed-effect meta-
analysis. Radican et al. (2008) present only mortality HR estimates by exposure group; however,
in an earlier follow-up of this same cohort, Blair et al. (1998) present both incidence and
mortality RR estimates by exposure group. There were no incident cases for females in the
highest exposure group in Blair et al. (1998) (and the expected number was not reported), thus,
for the same reasons we didn't use RR estimates of 1 for unreported RR estimates in the Axelson
et al. (1994), Hansen et al. (2001), and Zhao et al. (2005) studies discussed above, the male-only
results were used for the RR estimate without attempting to approximate a contribution to the RR
estimate from the females in the cohort. Radican et al. (2008) also present results for categories
based on frequency and pattern of exposure; however, subjects weren't distributed uniquely
across the categories (the numbers of cases across categories exceeded the total number of
cases), thus it was difficult to interpret these results and they were not used in a sensitivity
analysis.
Unlike for kidney cancer, Zhao et al. (2005) present lung cancer RR estimates only for
untagged exposures. The mortality results reflect more cases (33) in the highest exposure group
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
than do the incidence results (14), so the mortality RR estimate was used for the primary
analysis, and the incidence estimate was used in a sensitivity analysis. Sensitivity analyses were
also done using results from Boice et al. (2006b) in place of the Zhao et al. (2005) RR estimate.
The cohorts for these studies overlap, so they are not independent studies. Boice et al. (2006b)
report mortality RR estimates for lung cancer by years worked with any potential exposure, years
worked as a test stand mechanic, a job with potential TCE exposure, and by a measure that
weighted years with potential exposure from engine flushing by the number of flushes each year.
The Boice et al. (2006b) estimates are adjusted for years of birth and hire and for hydrazine
exposure.
C.5.2.2. Results of Meta-Analyses
Results from the meta-analyses that were conducted for lung cancer in the highest
exposure groups are summarized at the bottom of Table C-15 and reported in more detail in
Table C-17. The RRm estimate from the random-effects meta-analysis of the 6 studies with
results presented for exposure groups was 0.96 (95% CI: 0.72, 1.27). As with the overall results
for lung cancer, the highest-exposure-group results exhibited significant heterogeneity, with the
largest study (Raaschou-Nielsen) having a statistically significantly increased RR estimate and
the next largest (Boice et al., 1999) having a statistically significantly decreased RR estimate (see
Figure C-13). The remaining 4 studies all had non-significant RR estimates closer to 1. Non-
significance of the RRm estimate was not dependent on any single study; although removing
Raaschou-Nielsen decreased the RRm estimate to 0.86 and removing Boice (1999) increased the
RRm estimate to 1.07. The RRm estimate was not highly sensitive to alternate RR estimate
selections. Use of the 6 alternate selections, individually, resulted in RRm estimates that were all
non-significant and that ranged from 0.92 to 0.98 (see Table C-17). As with the primary
analysis, significant heterogeneity was observed for all the meta-analyses with alternate
selections (see Table C-17).
The RRm estimate from the primary analysis of the highest exposure groups was the
same as that for the overall TCE analysis (0.96), indicating no evidence of an exposure-response
relationship and confirming the absence of evidence of an increased risk of lung cancer
associated with TCE exposure from these studies as a whole.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-88 DRAFT—DO NOT CITE OR QUOTE
-------
K
s
TO
>3
Table C-17. Summary of some meta-analysis results for TCE (highest exposure groups) and lung cancer
S"4
§•
TO
S
Analysis
Model
Combined
RR estimate
95% LCL
95% UCL
Heterogeneity
Comments
Primary
analysis
Random
0.96
0.72
1.27
Significant
to < 0.0002)
? = 80%
Non-significance of RRm not dependent on any
single study.

Fixed
1.15
1.03
1.27
Because of
significant
heterogeneity,
fixed-effect
model not
appropriate.
Significant elevation in RRm dependent on
single study, Raaschou-Nielsen, without which
the RRm would be non-significantly decreased
(RRm = 0.86, p = 0.07).
Alternate RR
selections3
Random
0.95
0.70
1.29
Significant
to < 0.0003)
r = 79%
With Blair et al. (1998) incidence RR instead of
Radican mortality HR.

Random
0.98
0.75
1.29
Significant
to = 0.0003)
r = 79%
With Morgan peak metric.

Random
0.96
0.71
1.30
Significant
to = 0.0002)
r = 79%
With Zhao incidence.

Random
0.92-0.93
0.67-0.69
1.25
Significant
to < 0.0002)
r = 81%
With Boice (2006b) alternates for Zhao (see
text).
o
* v §
§ I 5
5 a, co-
c §- a
5 S S-
?• £ <§
r: §
to a*
o
a
>;
>S
>S
TO
'S
TO
*
O
VO
to
o
"Changing the primary analysis by one alternate RR each time.
O
oo
VO
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
TCE Exposure and Lung Cancel - hitjhest exposure yioups
Relative Risk and 05% CI
RR
LCL UCL
A.rrttila (1085)
Boies (1900)
Morgan (1008)
Raaschoy-Nielsen (2003)
Radio an (2003)
Zhao (2005)
OVERALL

~
~ •
0.83 0.33 1.71
0.64 0.46 0.89
0.06 o;
1.29
1.40 1.20 1.80
0.00 0 .S3 1.2
1.00 0.08 1.53
0.96 0.72
0.1
10
Figure C-13. Meta-analysis of lung cancer and TCE exposure—highest
exposure groups. Random-effects model. The summary estimate is in the
bottom row, represented by the diamond. Symbol sizes reflect relative weights of
the studies.
C.5 .3. Discussion of Lung Cancer Meta-Analysis Results
Significant heterogeneity was observed in the lung cancer results (for both overall TCE
exposure and for the highest exposure groups) from the different studies, and there was no clear
explanation for the source(s) of the heterogeneity, as discussed in Section C. 5.1.2. Nonetheless,
we conducted (random-effects) meta-analyses of the lung cancer results with the goal of
addressing the question of whether or not there was evidence of an association between TCE
exposure and lung cancer that might suggest that smoking could be confounding the kidney
cancer results, in particular in the cohort studies, which did not adjust for smoking.
Both the overall and highest-exposure-group analyses yielded non-significant RRm
estimates of 0.96 for lung cancer. Influence analyses and sensitivity analyses using alternate RR
estimate selection for various studies similarly found no evidence of an association between TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-90 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
exposure and lung cancer from these studies as a whole. This finding suggests that there is no
confounding of the overall RRm for kidney cancer by smoking, in particular from the cohort
studies (see Section 4.4.2.3 for a more comprehensive discussion of the issue of potential
confounding of the kidney cancer results by smoking).
C.6. DISCUSSION OF STRENGTHS, LIMITATIONS, AND UNCERTAINTIES IN THE
META-ANALYSES
Meta-analysis provides a systematic way of objectively and quantitatively combining the
results of multiple studies to obtain a summary effect estimate. Use of meta-analysis can help
risk assessors avoid some of the potential pitfalls in overly relying on a single study or in making
more subjective qualitative judgments about the apparent weight of evidence across studies.
Combining the results of smaller studies also increases the statistical power to observe an effect,
if one exists. In addition, meta-analysis techniques assist in systematically investigating issues
such as potential publication bias and heterogeneity in a database.
While meta-analysis can be a useful tool for analyzing a database of epidemiological
studies, the analysis is limited by the quality of the input data. If the individual studies are
deficient in their abilities to observe an effect (in ways other than low statistical power, which
meta-analysis can help ameliorate), the meta-analysis will be similarly deficient. A critical step
in the conduct of a meta-analysis is to establish eligibility criteria and clearly and transparently
identify all relevant studies for inclusion in the meta-analysis. For the TCE database, a
comprehensive qualitative review of available studies was conducted and eligible studies were
identified, as described in Appendix B, Section II-9.
Identifying all relevant studies may be hampered if publication bias has occurred.
Publication bias is a systematic error that can arise if statistically significant studies are more
likely to be published than non-significant studies. This can result in an upward bias on the
effect size measure, i.e., the relative risk estimate. To address this concern, potential publication
bias was investigated for the databases for which meta-analyses were undertaken. For the
studies of kidney cancer and liver cancer, there was no evidence of publication bias. For the
studies of NHL, there was some evidence of potential publication bias. It is uncertain whether
this reflects actual publication bias or rather an association between SE and effect size (as
discussed in Section C.l, a feature of publication bias is that smaller studies tend to have larger
effect sizes) resulting for some other reason, e.g., a difference in study populations or protocols
in the smaller studies. Furthermore, if there is publication bias in this data set, it may be creating
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-91 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
an upward bias on the relative risk estimate, but this bias does not appear to account completely
for the finding of an increased NHL risk (see Section C.2.1.2).
Another concern in meta-analyses is heterogeneity across studies. Random-effects
models were used for the primary meta-analyses in this assessment because of the diverse nature
of the individual studies. When there is no heterogeneity across the study results, the
random-effects model will give the same result as a fixed-effect model. When there is
heterogeneity, the random-effects model estimates the between-study variance. Thus, when
there is heterogeneity, the random-effects model will generate wider confidence intervals and be
more "conservative" than a fixed-effect model. However, if there is substantial heterogeneity, it
may be inappropriate to combine the studies at all. In cases of significant heterogeneity, it is
important to try to investigate the potential sources of the heterogeneity.
For the studies of kidney cancer and liver cancer, there was no apparent heterogeneity
across the study results, i.e., random- and fixed-effects models gave identical summary
estimates. For the NHL studies, there was heterogeneity, but it was not statistically significant
(p = 0.16). The / -value was 26%, suggesting low-to-moderate heterogeneity. When subgroup
analyses were done for the cohort and case-control studies separately, there was some
heterogeneity in both groups, but in neither case was it statistically significant. Further attempts
to quantitatively investigate the heterogeneity were not pursued because of limitations in the
database. The sources of heterogeneity are an uncertainty in the database of studies of TCE and
NHL. Some potential sources of heterogeneity, which are discussed qualitatively in
Section C.2.3, include differences in exposure assessment or in the intensity or prevalence of
TCE exposures in the study population and differences in NHL classification.
The joint occurrence of heterogeneity and potential publication bias in the database of
studies of TCE and NHL raises special concerns. Because of the heterogeneity, a random-effects
model should be used if these studies are to be combined; yet, the random-effects model gives
relatively large weight to small studies, which could exacerbate the potential impacts of
publication bias. For the NHL studies, the summary relative risk estimates from the random-
effects and fixed-effect models are not very different (RRm = 1.23 [95% CI: 1.07, 1.42] and 1.21
[1.08, 1.35], respectively); however, the confidence interval for the fixed-effect estimate does not
reflect the between-study variance and is, thus, overly narrow.
8	7
Heterogeneity was statistically significant for the lung cancer studies (p < 10") and the I -
value was 90%, indicating that the amount of heterogeneity was high. Nonetheless, (random-
effects) meta-analyses were conducted for the purpose of investigating the potential for smoking
to be confounding the kidney cancer results (see Sections C.5 and 4.4.2.3).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-92 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
C.7. CONCLUSIONS
The strongest finding from the meta-analyses was for TCE and kidney cancer. The
summary estimate from the primary random-effects meta-analysis of the 15 studies was
RRm = 1.27 (95% CI: 1.13, 1.43). There was no apparent heterogeneity across the study results
(i.e., fixed-effect model gave same summary estimate), and there was no evidence of potential
publication bias. The summary estimate was robust across influence and sensitivity analyses; the
estimate was not markedly influenced by any single study, nor was it overly sensitive to
individual RR estimate selections. The findings from the meta-analyses of the highest exposure
groups for the studies that provided kidney cancer results categorized by exposure level were
similarly robust. The summary estimate was RRm = 1.58 (95% CI: 1.28, 1.96) for the 13 studies
included in the analysis. There was no apparent heterogeneity in the highest-exposure-group
results, and the estimate was not markedly influenced by any single study, nor was it overly
sensitive to individual RR estimate selections. In sum, these robust results support a conclusion
that TCE exposure increases the risk of kidney cancer.
The meta-analyses of the overall effect of TCE exposure on NHL also suggest a small,
statistically significant increase in risk. The summary estimate from the primary random-effects
meta-analysis of the 17 studies was 1.23 (95% CI: 1.07, 1.42). This result was not overly
influenced by any single study, nor was it overly sensitive to individual RR estimate selections.
There is some evidence of potential publication bias in the NHL study data set; however, it is
uncertain that this is actually publication bias rather than an association between SE and effect
size resulting for some other reason, e.g., a difference in study populations or protocols in the
smaller studies. Furthermore, if there is publication bias, it does not appear to account
completely for the findings of an increased NHL risk. There was some heterogeneity across the
results of the 17 studies, but it was not statistically significant (p = 0.16). The / -value was 26%,
suggesting low-to-moderate heterogeneity. The source(s) of this heterogeneity remains an
uncertainty. The summary estimate from the meta-analysis of the highest exposure groups for
the 13 studies which provided NHL results categorized by exposure level was RRm = 1.43
(95% CI: 1.13, 1.82). The statistical significance of the increased RR estimate for the highest
exposure groups was not dependent on any single study, nor was it sensitive to individual RR
estimate selections. Although there was some heterogeneity across the 13 highest-exposure-
group studies, it was not statistically significant (p = 0.30) and the / -value was 14%, suggesting
that the amount of heterogeneity was low. Furthermore, the heterogeneity is dependent on a
single study, Cocco et al. (2010), suggesting that the RR estimate for the highest exposure group
from that study is a relative outlier. Overall, the robustness of the finding of an increased NHL
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
risk for the highest exposure groups strengthens the more moderate evidence from the meta-
analyses for overall effect.
The meta-analyses of the overall effect of TCE exposure on liver (and gall bladder/biliary
passages) cancer also suggest a small, statistically significant increase in risk, but the study
database is more limited. The summary estimate from the primary random-effects meta-analysis
of the 9 (all cohort) studies was 1.29 (95% CI: 1.07, 1.56). The analysis was dominated by one
large study that contributed about 53% of the weight. When this study was removed, the RRm
estimate decreased somewhat and was less precise (RRm = 1.22; 95% CI: 0.93, 1.61). The
summary estimate was not overly influenced by any other single study, nor was it overly
sensitive to individual RR estimate selections. There was no evidence of publication bias in this
data set, and there was no observable heterogeneity across the study results. However, the
findings from the meta-analyses of the highest exposure groups for the studies that provided liver
cancer results categorized by exposure level do not add support to the overall effect findings.
The summary estimate was RRm = 1.28 (95% CI: 0.93, 1.77) for the 8 studies included in the
analysis, which is slightly lower than the summary estimate for the overall effect. This
contradictory result is driven by the fact that the RR estimate for the highest exposure group in
the individual study which contributes the majority of the weight to the meta-analyses, although
greater than 1, was less than the overall RR estimate for the same study. In sum, these results do
not rule out an effect of TCE on liver cancer, because the liver cancer results are relatively
underpowered with respect to numbers of studies and number of cases and the overwhelming
study in terms of weight uses the weak exposure surrogate of duration of employment for
categorizing exposure level; however, at present, there is only modest support for an increased
risk of liver cancer.
Meta-analyses were also conducted for lung cancer with the goal of addressing the
question of whether or not there was evidence of an association between TCE exposure and lung
cancer that might suggest that smoking could be confounding the kidney cancer results, in
particular in the cohort studies, which did not adjust for smoking. Both the overall and highest-
exposure-group random-effects meta-analyses yielded non-significant RRm estimates of 0.96 for
lung cancer. Influence analyses and sensitivity analyses using alternate RR estimate selection for
various studies similarly found no evidence of an association between TCE exposure and lung
cancer from these studies as a whole. This finding suggests that there is no confounding of the
overall RRm for kidney cancer by smoking (see Section 4.4.2.3 for a more comprehensive
discussion of the issue of potential confounding of the kidney cancer results by smoking).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-94 DRAFT—DO NOT CITE OR QUOTE
-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	C-95 DRAFT—DO NOT CITE OR QUOTE
-------
APPENDIX D
Neurological Effects of Trichloroethylene
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-i DRAFT—DO NOT CITE OR QUOTE
-------
D.l. HUMAN STUDIES ON Till NEUROLOGICAL EFFECTS OF
TRICHLOROETHYLENE (TCE)
There is an extensive body of evidence in the literature on the neurological effects caused
by exposure to trichloroethylene (TCE) in humans. The primary functional domains that have
been studied and reported are trigeminal nerve function and nerve conductivity (latency),
psychomotor effects, including reaction times (simple and choice), visual and auditory effects,
cognition, memory, and subjective neurological symptoms, such as headache and dizziness. This
section discusses the primary studies presented for each of these effects. Summary tables for all
the human TCE studies are at the end of this section.
D. 1.1. Changes in Nerve Conduction
There is strong evidence in the literature that exposure to TCE results in impairment of
trigeminal nerve function in humans exposed occupationally, by inhalation, or environmentally,
by ingestion. Functional measures such as the blink reflex and masseter reflex tests were used to
determine if physiological functions mediated by the trigeminal nerve were significantly
impacted. Additionally, trigeminal somatosensory evoked potentials were also measured in
some studies to ascertain if nerve activity was directly affected by TCE exposure.
D.l. 1.1. Blink Reflex and Masseter Reflex Studies—Trigeminal Nerve
Barret et al. (1984) conducted a study on 188 workers exposed to TCE occupationally
from small and large factories in France (type of factories not disclosed). The average age of the
workers was 41 (standard deviation [SD] not provided, but authors noted 14% <30 years and
25% >50 years) and the average exposure duration was 7 hours/day for 7 years. The
188 workers were divided into high and low exposure groups for both TCE exposure measured
using detector tubes and trichloroacetic acid (TCA) levels measured in urine. There was no
unexposed control population, but responses in the high-exposure group were compared response
in the low-exposure group. TCE exposure groups were divided into a low exposure group
(<150 ppm; n = 134) and a high exposure group (>150 ppm; n = 54). The same workers
(n = 188) were also grouped by TCA urine measurements such that a high exposure was
>100 mg TCA/g creatinine. Personal factors including age, tobacco use and alcohol intake were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-1 DRAFT—DO NOT CITE OR QUOTE
-------
also analyzed. No mention was made regarding whether or not the examiners were blind to the
subjects' exposure status. Complete physical examination including testing visual performance
(acuity and color perception), evoked trigeminal potential latencies and audiometry, facial
sensitivity, reflexes, and motoricity of the masseter muscles. Chi squared analysis was used to
examine distribution of the different groups for comparing high and low exposed workers
followed by one way analysis of variance. Overall, 22 out of 188 workers (11.7%) experienced
trigeminal nerve impairment (p < 0.01) as measured by facial sensitivity, reflexes (e.g., jaw,
corneal, blink) and movement of the masseter muscles. When grouped by TCE exposure, 12 out
of 54 workers (22.2%) in the high exposure group (>150 ppm) and 10 out of 134 workers (7.4%)
in the low exposure group had impaired trigeminal nerve mediated responses. When grouped by
the presence of TCA in the urine, 41 workers were now in the high TCA group and 10 out of 41
workers (24.4%) experienced trigeminal nerve impairment in comparison to the 12 out of 147
(8.2%>) in the low TCA (<100 mg TCA/g creatinine) group. Statistically significant results were
also presented for the following symptoms based on TCE and TCA levels: trigeminal nerve
impairment (p < 0.01), asthenia (p < 0.01), optic nerve impairment (p < 0.001), and dizziness
(0.05 
-------
not available for the entire exposure period. TCE concentrations for the control population were
less than the maximum contaminant level (MCL) (5 ppb). The BR was used to measure the
neurotoxic effects of TCE. The BR was measured using an electrode to stimulate the
supraorbital nerve (above the eyelid) with a shock (0.05 ms in duration) resulting in a response
and the response was measured using a recording electrode over the orbicularis oculi muscle (the
muscle responsible for closing the eyelid and innervated by the trigeminal nerve). The BR
generated an R1 and an R2 component from each individual. BRs were recorded and the
supraorbital nerve was stimulated with single electrical shocks of increasing intensity until nearly
stable R1 and R2 ipsilateral and R2 contralateral responses were obtained. The student's t-test
was used for testing the difference between the group means for the blink reflex component
latencies. Because of the variability of R2 responses, this study focused primarily on the R1
response latencies. Highly significant differences in the conduction latency means of the BR
components for the TCE exposed population versus control population were observed when
comparing means for the right and left side R1 to the controls. The mean R1 BR component
latency for the exposed group was 11.35 ms, SD = 0.74 ms, 95% confidence interval (CI):
11.03-11.66. The mean for the controls was 10.21 ms, SD = 0.78 ms, 95% CI: 9.92-10.51;
(p < 0.001). The study was well conducted with consistency of methods, and statistically
significant findings for trigeminal nerve function impairment resulting from environmental
exposures to TCE. However, the presence of other solvents in the well water, self selection of
subjects involved in litigation, and incomplete characterization of exposure present problems in
drawing a clear conclusion of TCE causality or dose-response relationship.
Kilburn and Warshaw (1993) conducted an environmental study on 544 Arizona
residents exposed to TCE in well-water. TCE concentrations were from 6 to 500 ppb and
exposure ranged from 1 to 25 years. Subjects were recruited and categorized in 3 groups.
Exposed group 1 consisted of 196 family members with cancer or birth defects. Exposed group
2 consisted of 178 individuals from families without cancer or birth defects; and exposed group 3
included 170 parents whose children had birth defects and rheumatic disorders. Well-water was
measured from 1957 to 1981 by several governmental agencies and average annual TCE
exposures were calculated and then multiplied by each individual's years of residence for
170 subjects. A referent group of histology technicians (n = 113) was used as a comparison for
the BR test. For this test, recording electrodes were placed over the orbicularis oculi muscles
(upper and lower) and the BR was elicited by gently tapping the glabeela (located on the mid-
frontal bone at the space between the eyebrows and above the nose). A two-sided Student's
t-test and linear regression were used for statistical analysis. Significant increases in the R1
component of the BR response was observed in the exposed population as compared to the
referent group. The R1 component measured from the right eye appeared within 10.9 ms in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-3 DRAFT—DO NOT CITE OR QUOTE
-------
TCE-exposed subjects whereas in referents, this component appeared 10.2 ms after the stimulus
was elicited indicating a significant delay {p < 0.008) in the reflex response. Similarly, delays in
the latency of appearance for the R1 component were also noted for the left eye but the effect
was not statistically significant {p = 0.0754). This study shows statistically significant
differences in trigeminal nerve function between subjects environmentally exposed and
nonexposed to TCE. This is an ecological study with TCE exposure inferred to subjects by
residence in a geographic area. Estimates of TCE concentrations in drinking water to individual
subjects are lacking. Additionally, litigation is suggested and may introduce a bias, particularly
if no validity tests were used.
Kilburn (2002b) studied 236 residents (age range: 18-83 years old) lived nearby
manufacturing plants (e.g., microchip plants) in Phoenix, AZ. Analysis of the groundwater in
the residential area revealed contamination with many volatile organic compounds including
TCE. Concentrations of TCE in the well water ranged from 0.2 ppb to more than 10,000 ppb
and the exposure duration varied between 2 to 37 years. Additional associated solvents included
dichloroethane (DCE), perchloroethylene, and vinyl chloride. A group-match design was used to
compare the 236 TCE-exposed residents to 161 unexposed regional referents and 67 referents in
NE Phoenix in the BR test. The BR response was recorded from surface electrodes placed over
the location of the orbicularis oculi muscles. The reflex response was elicited by gently tapping
the left and right supraorbital notches with a small hammer. The R1 component of the BR
response was measured for both the left and right eye. Statistically significant increases in
latency time for the R1 component was observed for residents exposed to TCE in comparison to
the control groups. In unexposed individuals, the R1 component occurred within 13.4 ms from
the right eye and 13.5 ms from the left eye. In comparison, the residents near the manufacturing
plant had latency times of 14.2 ms (p < 0.0001) for the right eye and 13.9 ms (p < 0.008) for the
left eye. This study shows statistically significant differences between environmentally exposed
and unexposed populations for trigeminal nerve function, as a result of exposures to TCE. This
is an ecological study with TCE exposure potential to subjects inferred by residence in a
geographic area. Estimates of TCE concentrations in drinking water to individuals are lacking.
Additionally, litigation is suggested and may introduce a bias, particularly if no validity tests
were used.
Feldman et al. (1992) evaluated the BR reflex in 18 subjects occupationally exposed to
neurotoxic chemicals (e.g., degreasers, mechanics, and pesticide sprayers among many others).
Eight of the subjects were either extensively (n = 4) or occupationally (n = 4) exposed to TCE.
The remaining subjects (n= 10) were exposed to other neurotoxic chemicals, but not TCE.
Quantitative exposure concentration data were not reported in the study, but TCE exposure was
characterized as either "extensive" or "occupational." Subjects in the "extensive" exposure
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-4 DRAFT—DO NOT CITE OR QUOTE
-------
group were chronically exposed (>1 year) to TCE at least 5 days a week and for at greater than
50% of the workday (n = 3) or experienced a direct, acute exposure to TCE for greater than
15 minutes (n = 1). Subjects in the "occupational" group were chronically exposed (>1 year) to
TCE for 1-3 days/week and for greater than 50% of the workday. The BR responses from the
TCE-exposed subjects were compared to a control group consisting of 30 nonexposed subjects
with no noted neurological disorders. BR responses were measured using surface electrodes
over the lower lateral portion of the orbicularis oculi muscle. Electrical shocks with durations of
0.05 ms were applied to the supraorbital nerve to generate the R1 and R2 responses. All of the
subjects that were extensively exposed to TCE had significantly increased latency times in the
appearance of the R1 component (no /> value listed) and for 3 subjects this increased latency time
persisted for at least 1 month and up to 20 years postexposure. However, none of the subjects
occupationally exposed to TCE had changes in the BR response in comparison to the control
group. In comparing the remaining neurotoxicant exposed subjects to the TCE-exposed
individuals, the sensitivity, or the ability of a positive blink reflex test to identify correctly those
who had TCE exposure was 50%. However, in workers with no exposure to TCE, 90%
demonstrated a normal R1 latency.
Mixed results were obtained in a study by Ruijten et al. (1991) on 31 male printing
workers exposed to TCE. The mean age was 44; mean exposure duration was 16 years and had
at least 6 years of TCE exposure. The control group consisted of 28 workers with a mean age
45 years. Workers in the control group were employed at least 6 years in print factories (similar
to TCE-exposed), had no exposure to TCE, but were exposed to "turpentine-like organic
solvents." TCE exposure potential was inferred from historical monitoring of TCE at the plant
using gas detection tubes. These data indicated TCE concentrations in the 1960s of around
80 ppm, mean concentration of 70 ppm in the next decade, with measurements from 1976 and
1981 showing a mean concentration of 35 ppm. The most recent estimate of TCE concentrations
in the factory was 17 ppm (stable for 3 years) at the time of the report. The authors calculated
that mean cumulative TCE exposure would be 704 ppm x years worked in factory. The masseter
and blink reflexes were measured to evaluate trigeminal nerve function in TCE-exposed and
control workers. For measurement of the masseter reflex, surface electrodes were attached over
the right masseter muscle (over the cheek area). A gentle tap on a roller placed under the
subject's chin was used to elicit the masseter reflex. For measurement of the blink reflex,
surface electrodes were placed on the muscle near the upper eyelid. Electrical stimulation of the
right supraorbital nerve was used to generate the blink reflex. There was a significant increase in
the latency of the masseter reflex to appear for the TCE-exposed workers (p < 0.05). However,
there was no significant change in the blink reflex measure between TCE-exposed workers and
control. Although no change in the blink reflex measures were observed between the two
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-5 DRAFT—DO NOT CITE OR QUOTE
-------
groups, it should be noted that the control group was exposed to other volatile organic solvents
(not specified) and this volatile organic compound exposure could be a possible confounder for
determination of TCE-induced effects.
There are two studies that reported no effect of TCE exposure on trigeminal nerve
function (El Ghawabi et al., 1973; Rasmussen et al., 1993a). El Ghawabi et al. (1973) conducted
a study on 30 money printing shop workers occupationally exposed to TCE. Metabolites of total
trichloroacetic acid and trichloroethanol were found to be proportional to TCE concentrations up
to 100 ppm (550 mg/m ). Controls were 20 age- and socio-economic status (SES)-matched
nonexposed males and 10 control workers not exposed to TCE. Trigeminal nerve involvement
was not detected, but the authors failed to provide details as to how this assessment was made. It
is mentioned that each subject was clinically evaluated and trigeminal nerve involvement may
have been assessed through a clinical evaluation. As a result, the conclusions of this study are
tempered since the authors did not provide details as to how trigeminal nerve function was
evaluated in this study.
Rasmussen et al. (1993a) conducted an historical cohort study on 99 metal degreasers.
Subjects were selected from a population of 240 workers from 72 factories in Denmark. The
participants were divided into three groups based on solvent exposure durations where low
exposure was up to 0.5 years, medium was 2.1 years and high was 11.0 years (mean exposure
duration). Most of the workers (70 out of 99) were primarily exposed to TCE with an average
exposure duration of 7.1 years for 35 hours/week. TCA and trichloroethanol (TCOH) levels
were measured in the urine samples provided by the workers and mean TCA levels in the high
group was 7.7 mg/L and was as high as 26.1 mg/L. Experimental details of trigeminal nerve
evaluation were not provided by the authors. It was reported that 1 out of 21 people (5%) in the
low exposure, 2 out of 37 (5%) in the medium exposure and 4 out of 41 (10%) in the high
exposure group experienced abnormalities in trigeminal nerve sensory function. No linear
association was seen on trigeminal nerve function (Mantel-Haenzel test for linear association,
p = 0.42). However, the trigeminal nerve function findings were not compared to a control (no
TCE exposure) group and it should be noted that some of the workers (29 out of 99) were not
exposed to TCE.
D.l .1.2. Trigeminal Somatosensory Evoked Potential (TSEP) Studies—Trigeminal Nerve
In a preliminary study, Barret et al. (1982) measured trigeminal sensory evoked potentials
(TSEPs) in eleven workers that were chronically exposed to TCE. Nine of these workers were
suffering effects from TCE intoxication (changes in facial sensitivity and clinical changes in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-6 DRAFT—DO NOT CITE OR QUOTE
-------
trigeminal nerve reflexes), and two were TCE-exposed without exhibiting any clinical
manifestations from exposure. A control group of 20 nonexposed subjects of varying ages were
used to establish the normal response curve for the trigeminal nerve function. In order to
generate a TSEP, a surface electrode was placed over the lip and a voltage of 0.05 ms in duration
was applied. The area was stimulated 500 times at a rate of two times per second. TSEPs were
recorded from a subcutaneous electrode placed between the international CZ point (central
midline portion of the head) and the ear. In eight of the eleven workers, an increased voltage
ranging from a 25 to a 45 volt increase was needed to generate a normal TSEP. Two of the
11 workers had an increased latency of appearance for the TSEP and three workers had increases
in TSEP amplitudes. The preliminary findings indicate that TCE exposure results in
abnormalities in trigeminal nerve function. However, the study does not provide any exposure
data and lacks information with regards to the statistical treatment of the observations.
Barret et al. (1987) conducted a study on 104 degreaser machine operators in France
(average age = 41.6 years; range = 18-62 years) who were highly exposed to TCE with an
average exposure of 7 hours/day for 8.23 years. Although TCE exposure concentrations were
not available, urinary concentrations of TCOH and TCA were measured for each worker. A
control group consisting of 52 subjects without any previous solvent exposure and neurological
deficits was included in the study. Trigeminal nerve symptoms and TSEPs were collected for
each worker. Trigeminal nerve symptoms were clinically assessed by examining facial
sensitivity and reflexes dependent on this nerve such as the jaw and blink reflex. TSEPs were
elicited by electrical stimulation (70-75 V for 0.05 ms) of the nerve using an electrode on the lip
commissure. Eighteen out of 104 TCE-exposed machine operators (17.3%) had trigeminal nerve
symptoms. The subjects that experienced trigeminal nerve symptoms were significantly older
(47.8 years vs. 40.5; p < 0.001). Both groups had a similar duration of exposure with a mean of
9.2 years in the sensitive group and 7.8 years in the nonsensitive group. Urinary concentrations
of TCOH and TCA were also statistically similar although the levels were slightly higher in the
sensitive group (245 mg/g creatinine vs. 162 mg/g creatinine for TCOH; 131 mg/g creatinine vs.
93 mg/g creatinine for TCA). However, in the same group, 40 out of 104 subjects (38.4%) had
an abnormal TSEP. Abnormal TSEPs were characterized as potentials that exhibited changes in
latency and/or amplitude that were at least 2.5 times the standard deviation of the normal TSEPs
obtained from the control group. Individuals with abnormal TSEP were significantly older
(45 years vs. 40.1 years; p < 0.05) and were exposed to TCE longer (9.9 years vs. 5.6 years;
p < 0.01). Urinary concentrations TCOH and TCA were similar between the groups with
sensitive individuals having average metabolite levels of 195 mg TCOH/g creatinine and
98.3 mg TCA/g creatinine in comparison to 170 mg TCOH/g creatinine and 96 mg TCA/g
creatinine in nonsensitive individuals. When a comparison was made between workers that had
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-7 DRAFT—DO NOT CITE OR QUOTE
-------
normal TSEP and no trigeminal symptoms and workers that had an abnormal TSEP and
experienced trigeminal symptoms, it was found that in the sensitive individuals (abnormal TSEP
and trigeminal symptoms) there was a significant increase in age (48.5 vs. 39.5 years old,
p < 0.01), duration of exposure (11 vs. 7.5 years, p < 0.05) and an increase in urinary TCA (313
vs. 181 mg TCA/g creatinine). No significant changes were noted in urinary TCOH, but the
levels were slightly higher in sensitive individuals (167 vs. 109 mg TCOH/g creatinine).
Overall, it was concluded that abnormal TSEPs were recorded in workers who were exposed to
TCE for a longer period (average duration 9.9 years). This appears to be a well designed study
with statistically significant results reported for abnormal trigeminal nerve response in TCE
exposed workers. Exposure assessment to TCE is by exposure duration and mean urinary TCOH
and TCA concentrations. TCE concentrations to exposed subjects as measured by atmospheric
or personal monitoring are lacking.
Mhiri et al. (2004) measured TSEPs from 23 phosphate industry workers exposed to TCE
for 6 hours/day for at least two years while cleaning tanks. Exposure assessment was based on
measurement of urinary metabolites of TCE, which were performed 3 times/worker, and air
measurements. Blood tests and hepatic enzymes were also collected. The mean exposure
duration was 12.4 ± 8.3 years (exposure duration range = 2-27 years). Although TCE exposures
were not provided, mean urinary concentrations of TCOH, TCA, and total trichlorides were
79.3 ± 42, 32.6 ± 22, and 111.9 ± 55 mg/g urinary creatinine, respectively. The control group
consisted of 23 unexposed workers who worked in the same factory without being exposed to
any solvents. TSEPs were generated from a square wave pulses (0.1 ms in duration) delivered
through a surface electrode that was placed 1 cm under the corner of the mouth. The responses
to the stimuli (TSEPs) were recorded from another surface electrode that was placed over the
contralateral parietal area of the brain. The measured TSEP was divided into several
components and labeled according to whether it was (1) a positive (P) or negative (N) potential
and (2) the placement of the potential in reference to the entire TSEP (e.g., PI is the first positive
potential in the TSEP). TSEPs generated from the phosphate workers that were ±2.5 times the
standard deviation from the TSEPs obtained from the control group were considered abnormal.
Abnormal TSEP were observed in 6 workers with clinical evidence of trigeminal involvement
and in 9 asymptomatic workers. Significant increases in latency were noted for all TSEP
potentials (Nl, PI, N2, P2, N3, p < 0.01) measured from the phosphate workers. Additionally,
significant decreases in the PI (p < 0.02) and N2 (p < 0.05) amplitudes were observed. A
significant positive correlation was demonstrated between duration of exposure and the N2
latency (p < 0.01) and P2 latency (p < 0.02). Only one subject had urinary TCE metabolite
levels over tolerated limits. TCE air contents were over tolerated levels, ranging from
"3
50-150 ppm (275-825 mg/m ). The study is well presented with statistically significant results
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-8 DRAFT—DO NOT CITE OR QUOTE
-------
for trigeminal nerve impairment resulting from occupational exposures to TCE. Exposure
potential to TCE is defined by urinary biomarkers, TCA, total trichloro-compounds, and TCOH.
The study lacks information on atmospheric monitoring of TCE in this occupational setting.
D.l .1.3. Nerve Conduction Velocity Studies
Nerve conduction latencies were also studied in two occupational studies by Triebig et al.
(1983; 1982) using methods for measurement of nerve conduction which differ from most
published studies, but the results indicate a potential impact on nerve conduction following
occupational TCE exposure. There was no impact seen on latencies in the 1982 study, but a
statistically significant response was observed in the latter study. The latter study, however, is
confounded by multiple solvent exposures.
In Triebig et al. (1982), 24 healthy workers (20 males, 4 females) were exposed to TCE
occupationally at three different plants. The ages ranged from 17-56, and length of exposure
ranged from 1 month to 258 months (mean 83 months). TCE concentrations measured in air at
work places ranged from 5-70 ppm (27-385 mg/m ). A control group of 144 healthy,
complaint-free individuals were used to establish 'normal' responses on the nerve conduction
studies. The matched control group consisted of twenty-four healthy nonexposed individuals
(20 males, 4 females), chosen to match the subjects for age and sex. TCA, TCE, and
trichloroethanol were measured in blood, and TCE and TCA were measured in urine. Nerve
conduction velocities were measured for sensory and motor nerve fibers using the following
tests: MCVmax (U): Maximum NLG of the motor fibers of the N. ulnaris between the wrist joint
and the elbow; dSCV (U): Distal NLG of mixed fibers of the N. ulnaris between finger V and the
wrist joint; pSCV (U): Proximal NLG of sensory fibers of the N. medianus between finger V and
Sulcus ulnaris; and dSCV (M): Distal NLG of sensory fibers of the N. medianus between finger
III and the wrist joint. Data were analyzed using parametric and nonparametric tests, rank
correlation, linear regression, with 5% error probability. Results show no statistically significant
difference in nerve conduction velocities between the exposed and unexposed groups. This
study has measured exposure data, but exposures/responses are not reported by dose levels.
Triebig et al. (1983) has a similar study design to the previous study (Triebig et al., 1982)
in the tests used for measurement of nerve conduction velocities, and in the analysis of blood and
urinary metabolites of TCE. However, in this study, subjects were exposed to a mixture of
solvents, including TCE, specifically "ethanol, ethyl acetate, aliphatic hydrocarbons (gasoline),
methyl ethyl ketone (MEK), toluene, and trichloroethene." The exposed group consists of
66 healthy workers selected from a population of 112 workers. Workers were excluded based on
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-9 DRAFT—DO NOT CITE OR QUOTE
-------
polyneuropathy (n = 46) and alcohol consumption (n = 28). The control group consisted of
66 healthy workers with no exposures to solvents. Subjects were divided into three exposure
groups based on length of exposure, as follows: 20 employees with "short-term exposure"
(7-24 months); 24 employees with "medium-term exposure" (25-60 months); 22 employees
with "long-term exposure" (over 60 months). TCA, TCE, and trichloroethanol were measured in
blood, and TCE and TCA were measured in urine. Subjects were divided into exposure groups
based on length of exposures, and results were compared for each exposure group to the control
group. In this study, there was a dose-response relationship observed between length of
exposure to mixed solvents and statistically significant reduction in nerve conduction velocities
observed for the medium and long-term exposure groups for the ulnar nerve (NCV).
Interpretation of this study is limited by the mixture of solvent exposure, with no results reported
for TCE alone.
D. 1.2. Auditory Effects
There are three large environmental studies reported which assessed the potential impact
of TCE exposures through groundwater ingestion on auditory functioning. They present mixed
results. All three studies were conducted on the population in the TCE Subregistry from the
National Exposure Registry (NER) developed by the Agency for Toxic Substances Disease
Registry (ATSDR). The two studies conducted by Burg et al. (1999; 1995) report an increase in
auditory effects associated with TCE exposure, but the auditory endpoints were self reported by
the population, as opposed to testing of measurable auditory effects in the subject population.
The third of these studies, reported by ATSDR (2003b) conducted measurements of auditory
function on the subject population, but failed to demonstrate a positive relationship between TCE
exposure and auditory effects. Results from these studies strongly suggest that children <9 years
are more susceptible to hearing impairments from TCE exposure than the rest of the general
population. These studies are described below.
Burg et al. (1995) conducted a study on registrants in the National Health Interview
Survey (NHIS) TCE subregistry of 4,281 (4,041 living and 240 deceased) residents
environmentally exposed to TCE via well water in Indiana, Illinois, and Michigan. Morbidity
baseline data were examined from the TCE Subregistry from the NER developed by the ATSDR.
Participants were interviewed in the NHIS, which consists of 25 questions about health
conditions. Data were self reported via face-to-face interviews. Neurological endpoints were
hearing and speech impairments. This study assessed the long-term health consequences of
long-term, low-level exposures to TCE in the environment. The collected data were compared to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-10 DRAFT—DO NOT CITE OR QUOTE
-------
the NHIS, and the National Household Survey on Drug Abuse. Poisson Regression analysis
model was used for registrants 19 and older. The statistical analyses performed treated the NHIS
population as a standard population and applied the age- and sex-specific period prevalence and
prevalence rates obtained from the NHIS data to the corresponding age- and sex-specific
denominators in the TCE Subregistry. This one-sample approach ignored sampling variability in
the NHIS data because of the large size of the NHIS database when compared to the TCE
Subregistry data file. A binomial distribution was assumed in estimating standard errors for the
TCE Subregistry data. Weighted age- and sex-specific period prevalence and prevalence rates
by using the person-weights were derived for the TCE subregistry. These "standard" rates were
applied to the corresponding TCE Subregistry denominators to obtain expected counts in each
age and sex combination. In the NHIS sample, 18% of the subjects were nonwhite. In the TCE
Subregistry sample, 3% of the subjects were nonwhite. Given this discrepancy in the proportion
of nonwhites and the diversity of races reported among the nonwhites in the TCE Subregistry,
the statistical analyses included 3,914 exposed white TCE registrants who were alive at baseline.
TCE registrants that were 9 years old or younger had a statistically significant increase in hearing
impairment as reported by the subjects. The relative risk (RR) in this age group for hearing
impairments was 2.13. The RR decreased to 1.12 for registrants aged 10-17 years and to 0.32 or
less for all other age groups. As a result, the effect magnitude was lower for children
10-17 years and for all other age groups. The study reports a dose-response relationship, but the
hearing effects are self-reported, and exposure data are modeled estimates.
Burg and Gist (1999) reported a study conducted on the same subregistry population
described for Burg et al. (1995). It investigated intrasubregistry differences among 3,915 living
members of the National Exposure Registry's Trichloroethylene Subregistry (4,041 total living
members). The participants' mean age was 34 years (SD = 19.9 years), and included children in
the registry. All registrants had been exposed to TCE through domestic use of contaminated well
water. All were Caucasian. All registrants had been exposed to TCE though domestic use of
contaminated well water; there were four exposure Subgroups, each divided into quartiles:
(1) Maximum TCE measured in well water, exposure subgroups include 2-12 ppb; 12-60 ppb;
60-800 ppb; (2) Cumulative TCE exposure subgroups include <50 ppb, 50-500 ppb,
500-5,000 ppb, >5,000 ppb; (3) Cumulative chemical exposure subgroups include TCA, DCE,
dichloroacetic acid (DCA), in conjunction with TCE, with the same exposure Categories as in #
2; and (4) Duration of exposure subgroups include <2 years, 2-5 years, 5-10 years, >10 years;
2,867 had TCE exposure of <50 ppb; 870 had TCE exposure of 51-500 ppb; 190 had TCE
exposure of 501-5,000 ppb; 35 had TCE exposure >5,000 ppb. The lowest quartile was used as
a control group. Interviews included occupational, environmental, demographic, and health
information. A large number of health outcomes were analyzed, including speech impairment
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-11 DRAFT—DO NOT CITE OR QUOTE
-------
and hearing impairment. Statistical methods used include Logistic Regression and Odds Ratios.
The primary purpose was to evaluate the rate of reporting health-outcome variables across
exposure categories. The data were evaluated for an elevation of the risk estimates across the
highest exposure categories or for a dose-response effect, while controlling for potential
confounders. Estimated prevalence odds ratios for the health outcomes, adjusted for the
potential confounders, were calculated by exponentiating the (3-coefficients from the exposure
variables in the regression equations. The standard error of the estimate was used to calculate
95% confidence intervals (CIs). The referent group used in the logistic regression models was
the lowest exposure group. The results variables were modeled as dichotomous, binary
dependent variables in the regression models. Nominal, independent variables were modeled,
using dummy variables. The covariables used were sex, age, occupational exposure, education
level, smoking history, and the sets of environmental subgroups. The analyses were restricted to
persons 19 years of age or older when the variables of occupational history, smoking history, and
education level were included. When the registrants were grouped by duration of exposure to
TCE, a statistically significant association (adjusted for age and sex) between duration of
exposure and reported hearing impairment was found. The prevalence odds ratios were 2.32
(95% CI: 1.18, 4.56) (>2 to <5 years); 1.17 (95% CI: 0.55, 2.49) (>5 to <10 years); and 2.46
(95%) CI = 1.30, 5.02) (>10 years). Higher rates of speech impairment (although not statistically
significant) were associated with maximum and cumulative TCE exposure, and duration of
exposure. The study reports dose-response relationships, but the effects are self reported, and
exposure data are estimates. No information was reported on presence or absence of additional
solvents in drinking water.
ATSDR (2003b) conducted a follow-up study to the TCE subregistry findings (Burg and
Gist, 1999; Burg et al., 1995) and focused on the subregistry children. Of the 390 subregistry
children (<10 years at time of original study), 116 agreed to participate. TCE exposure ranged
from 0.4 to 5,000 ppb from the drinking water. The median TCE exposure for this subgroup was
estimated to be 23 ppb per year of exposure. To further the hearing impairments reported in
Burg et al. (1999; 1995), comprehensive auditory tests were conducted with the 116 children and
compared to a control group of 182 children that was age-matched. The auditory tests consisted
of a hearing screening (typanometry, pure tone and distortion product otoacoustic emissions
[DPOAE]) and a more in-depth hearing evaluation for children that failed the initial screening.
Ninety percent of the TCE-exposed children passed the typanometry and pure tone tests, and
there were no significant differences between control and TCE-exposed groups. Central auditory
processing tests were also conducted and consisted of a test for acoustic reflexes and a screening
test for auditory processing disorders (SCAN). The acoustic reflex tested the ipsilateral and
contralateral auditory pathway at 1,000 Hz for each ear. In this test, each subject hears the sound
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-12 DRAFT—DO NOT CITE OR QUOTE
-------
frequency and determines if the sound causes the stapedius muscle to tighten the stapes (normal
reflex to noise). Approximately 20% of the children in the TCE subregistry and 5-7% in the
controls exhibited an abnormal acoustic reflex, and this increased abnormality in the test was a
significant effect (p = 0.003). No significant effects were noted in the SCAN tests. The authors
concluded that the significant decrease in the acoustic reflex for the TCE subregistry children is
reflective of potential abnormalities in the middle ear, which may reflect abnormalities in lower
brainstem auditory pathway function. Lack of effects with the pure tone and typanometry tests
suggests that the cochlea is not affected by TCE exposure.
Although auditory function was not directly measured, Rasmussen et al. (1993c) used a
psychometric test to measure potential auditory effects of TCE exposure in an environmental
study. Results from 96 workers exposed to TCE and other solvents were presented in this study.
The workers were divided into three exposure groups: low, medium, and high. Details of the
exposure groups and exposure levels are provided in Table 4-21 (under study description of
Rasmussen et al., 1993c). Three auditory-containing tasks were included in this study, but only
the acoustic motor function test could be used for evaluation of auditory function. In the
acoustic motor function test, high and low frequency tones were generated and heard through a
set of earphones. Each individual then had to imitate the tones by knocking on the table using
the flat hand for a low frequency and using a fist for a high frequency. A maximal score of 8
could be achieved through this test. The tones were provided in either a set of 1 or 3 groups. In
the one group acoustic motor function test, the average score for the low exposure group was 4.8
in comparison to 2.3 in the high exposure group. Similar decrements were noted in the 3 group
acoustic motor function test. A significant association was reported for TCE exposure and
performance on the one group acoustic motor function test (p < 0.05) after controlling for
confounding variables.
D. 1.3. Vestibular Effects
The data linking acute TCE exposure with transient impairment of vestibular function are
quite strong based on human chamber studies, occupational exposure studies, and laboratory
animal investigations. It is clear from the human literature that these effects can be caused by
exposures to TCE, as they have been reported extensively in the literature.
The earliest reports of neurological effects resulting from TCE exposures focused on
subjective symptoms, such as headaches, dizziness, and nausea. These symptoms are subjective
and self-reported, and, therefore, offer no quantitative measurement of cause and effect.
However, there is little doubt that these effects can be caused by exposures to TCE, as they have
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 DRAFT—DO NOT CITE OR QUOTE
-------
been reported extensively in the literature, resulting from occupational exposures (Grandjean et
al., 1955; Liu et al., 1988; Rasmussen and Sabroe, 1986; Smith, 1970), environmental exposures
(Hirsch et al., 1996), and in chamber studies (Kylin et al., 1967; Stewart et al., 1970). These
studies are described below in more detail.
Grandjean et al. (1955) reported on 80 workers exposed to TCE from 10 different
factories of the Swiss mechanical engineering industry. TCE air concentrations varied from
"3
6-1,120 ppm (33-6,200 mg/m ) depending on time of day and proximity to tanks, but mainly
"3
averaged between 20-40 ppm (100-200 mg/m ). Urinalysis (TCA) varied from 30 mg/L to
300 mg/L. This study does not include an unexposed referent group, although prevalences of
self-reported symptoms or neurological changes among the higher-exposure group are compared
to the lower-exposure group. Workers were classified based on their exposures to TCE and there
were significant differences (p = 0.05) in the incidence of neurological disorder between
Groups I (10-20 ppm), II (20-40 ppm; 110-220 mg/m3) and III (>40 ppm; 220 mg/m3).
Thirty-four percent of the workers had slight or moderate psycho-organic syndrome; 28% had
neurological changes. Approximately 50% of the workers reported incidences of vertigo and
30%) reported headaches (primarily an occasional and/or minimal disorder). Based on TCA
eliminated in the urine, results show that subjective, vegetative, and neurological disorders were
more frequent in Groups II (40-100 mg/L) and III (101-250 mg/L) than in Group I
(10-39 mg/L). Statistics do support a dose-effect relationship between neurological effects and
TCE exposure, but exposure data are questionable.
Liu et al. (1988) evaluated the effects of occupational TCE exposure on 103 factory
workers in Northern China. The workers (79 men, 24 women) were exposed to TCE during
vapor degreasing production or operation. An unexposed control group of 85 men and
26 women was included for comparison. Average TCE exposure was mostly at less than 50 ppm
"3
(275 mg/m ). The concentration of breathing zone air during entire shift was measured by
diffusive samplers placed on the chest of each worker. Subjects were divided into three exposure
groups; 1-10 ppm (5.5-55 mg/m3), 11-50 ppm (60-275 mg/m3) and 51-100 ppm
(280-550 mg/m ). Results were based on a self-reported subjective symptom questionnaire.
The frequency of subjective symptoms, such as nausea, drunken feeling, light-headedness,
floating sensation, heavy feeling of the head, forgetfulness, tremors and/or cramps in extremities,
body weight loss, changes in perspiration pattern, joint pain, and dry mouth (all >3 times more
common in exposed workers); reported as 'prevalence of affirmative answers', was significantly
greater in exposed workers than in unexposed (p < 0.01). "Bloody strawberry jam-like feces "
was borderline significant in the exposed group and ''frequent flatus" was statistically
significant. Dose-response relationships were established (but not statistically significant) for
symptoms. Most workers were exposed below 10 ppm, and some at 11-50 ppm. The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-14 DRAFT—DO NOT CITE OR QUOTE
-------
differences in exposure intensity between men and women was of borderline significance
(0.05 
-------
TCE between 0 to 2,441 ppb. The distance of residence from contaminated well was used to
estimate exposure level. Sixty-six subjects (62%) complained of headaches at the time of
evaluation. Diagnosis of TCE-induced cephalagia was considered credible for 57 patients
(54%). Forty-seven of these had a family history of headaches. Retrospective TCE level of well
water or well's distance from the industrial site analysis did not correlate with the occurrence of
possibly-TCE induced headaches. This study shows a general association between headaches
and exposure to TCE in drinking water wells. There were no statistics to support a
dose-response relationship. All subjects were involved in litigation.
Stewart et al. (1970) evaluated vestibular effects in 13 subjects who were exposed to TCE
3	3
vapor 100 ppm (550 mg/m ) and 200 ppm (1,100 mg/m ) for periods of 1 hour to a 5-day work
week. Experiments 1-7 were for a duration of 7 hours with a mean TCE concentration of
198-200 ppm (1,090-1,100 mg/m3). Experiments 8 and 9 exposed subjects to 190-202 ppm
(1,045-1,110 mg/m ) TCE for a duration of 3.5 and 1 hour, respectively. Experiment 10
"3
exposed subjects to 100 ppm (550 mg/m ) TCE for 4 hours. Experiments 2-6 were carried out
with the same subjects over 5 consecutive days. Gas chromatography of expired air was
measured. There were no self controls. Subjects reported symptoms of lightheadedness,
headache, eye, nose, and throat irritation. Prominent fatigue and sleepiness by all were reported
"3
above 200 ppm (1,100 mg/m ). There were no quantitative data or statistics presented regarding
dose and effects of neurological symptoms.
Kylin et al. (1967) exposed 12 volunteers to 1,000 ppm (5,500 mg/m3) TCE for 2 hours
in a 1.5 x 2 x 2 meters chamber. Volunteers served as their own controls since 7 of the 12 were
pretested prior to exposure and the remaining 5 were post-tested days after exposure. Subjects
were tested for optokinetic nystagmus, which was recorded by electronystogmography, that is,
"the potential difference produced by eye movements between electrodes placed in lateral angles
between the eyes." Venous blood was also taken from the volunteers to measure blood TCE
levels during the vestibular task. The authors concluded that there was an overall reduction in
the limit ("fusion limit") to reach optokinetic nystagmus when individuals were exposed to TCE.
Reduction of the "fusion limit" persisted for up to 2 hours after the TCE exposure was stopped
and the blood TCE concentration was 0.2 mg/100 mL.
D.1.4. Visual Effects
Kilburn (2002b) conducted an environmental study on 236 people exposed to TCE in
groundwater in Phoenix, AZ. Details of the TCE exposure and population are described earlier
in Section D. 1.1.1 (see Kilburn, 2002b). Among other neurological tests, the population and 161
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-16 DRAFT—DO NOT CITE OR QUOTE
-------
nonexposed controls was tested for color discrimination using the desaturated Lanthony 15-hue
test, which can detect subtle changes in color vision deficiencies. Color discrimination errors
were significantly increased in the TCE exposed population (p < 0.05) with errors scores
averaging 12.6 in the TCE exposed in comparison to 11.9 in the control group. This study shows
statistically significant differences in visual response between exposed and nonexposed subjects
exposed environmentally. Estimates of TCE concentrations in drinking water to individual
subjects are lacking.
Reif et al. (2003) conducted a cross sectional environmental study on 143 residents of the
Rocky Mountain Arsenal community of Denver whose water was contaminated with TCE and
related chemicals from nearby hazardous waste sites between 1981 and 1986. The residents
were divided into three groups based on TCE exposure with the lowest exposure group at
<5 ppb, the medium exposure group at 5 tol5 ppb and the high exposure group defined as
>15 ppb TCE. Visual performance was measured by two different contrast sensitivity tests
(C and D) and the Benton visual retention test. In the two contrast sensitivity tests, there was a
20 to 22% decrease in performance between the low and high TCE exposure groups and
approached statistical significance (p = 0.06 or 0.07). In the Benton visual retention test, which
measures visual perception and visual memory, scores, dropped by 10% from the lowest
exposure to the highest TCE exposure group and was not statistically significant. It should be
noted that the residents were potentially exposed to multiple solvents including TCE and a
nonexposed TCE group was not included in the study. Additionally, modeled exposure data are
only a rough estimate of actual exposures, and possible misclassification bias associated with
exposure estimation may limit the sensitivity of the study.
Rasmussen et al. (1993c) conducted a cross-sectional study on 96 metal workers, working
in degreasing at various factories in Denmark (industries not specified) with chlorinated solvents.
These subjects were identified from a larger cohort of 240 workers. Details of the exposure
groups and TCE exposure levels are presented in Section D. 1.1.1 (under Rasmussen et al.,
1993a). Neuropsychological tests including the visual gestalts (test of visual perception and
retention) and the stone pictures test (test of visual learning and retention) were administered to
the metal workers. In the visual gestalts test, cards with a geometrical figure containing four
items were presented and workers had to redraw the figure from memory immediately (learning
phase) after presentation and after 1 hour (retention phase). In the learning phase, the figures
were redrawn until the worker correctly drew the figure. The number of total errors significantly
increased from the low group (3.4 errors) to the high exposure group (6.5 errors; p = 0.01) during
the learning phase (immediate presentation). Similarly, during the retention phase of this task
(measuring visual memory), errors significantly increased from an average of 3.2 in the low
group to 5.9 in the high group (p < 0.001). In the stone pictures test, slides of 10 stones
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-17 DRAFT—DO NOT CITE OR QUOTE
-------
(different shapes and sizes) were shown and the workers had to identify the 10 stones out of a
lineup of 25 stones. There were no significant changes in this task, but the errors increased from
4.6 in the low exposure group to 6.3 in the high exposure group during the learning phase of this
task. Although this study identifies visual performance deficits, a control group (no TCE
exposure) was not included in this study and the presented results may actually underestimate
visual deficits from TCE exposure.
Troster and Ruff (1990) presented case studies conducted on two occupationally exposed
workers to TCE and included a third case study on an individual exposed to
1,1,1-trichloroethane. Case #1 was exposed to TCE (concentration unknown) for 8 months and
Case #2 was exposed to TCE over a 3-month period. Each patient was presented with a
visual-spatial task (Ruff-Light Trail Learning test as referenced by the authors). Both of the
individuals exposed to TCE were unable to complete the visual-spatial task and took the
maximum number of trials (10) to attempt to complete the visual task. A control group of
30 individuals and the person exposed to 1,1,1-trichloroethane were able to complete this task
accordingly. The lack of quantitative exposure data and a small sample size severely limits the
study and does not allow for statistical comparisons.
Vernon and Ferguson (1969) exposed eight male volunteers (ages 21-30) to 0, 100, 300,
and 1,000-ppm TCE for 2 hours. Each individual was exposed to all TCE concentrations and a
span of at least 3 days was given between exposures. The volunteers were presented with six
visuo-motor tests during the exposure sessions. When the individuals were exposed to
-3
1,000-ppm TCE (5,500 mg/m ), significant abnormalities were noted in depth perception as
measured by the Howard-Dolman test (p < 0.01), but no effects on the flicker fusion frequency
test (threshold frequency at which the individual sees a flicker as a single beam of light) or on the
form perception illusion test (volunteers presented with an illusion diagram). This is one of the
earliest chamber studies of TCE. This study included only healthy young males, is of a small
size, limiting statistical power, and reports mixed results on visual testing following TCE
exposure.
D.1.5. Cognition
There is a single environmental study in the literature that presents evidence of a negative
impact on intelligence resulting from TCE exposure. Kilburn and Warshaw (1993) (study details
in Section D. 1.1.1) evaluated the effects on cognition for 544 Arizona residents exposed to TCE
in well-water. Subjects were recruited and categorized into three groups. Exposed Group 1
consisted of 196 family members with cancer or birth defects. Exposed Group 2 consisted of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-18 DRAFT—DO NOT CITE OR QUOTE
-------
178 individuals from families without cancer or birth defects; and exposed Group 3 included 170
parents whose children had birth defects and rheumatic disorders. Sixty-eight referents were
used as a comparison group for the clinical memory tests. Several cognitive tests were
administered to these residents in order to test memory recall skills and determine if TCE
exposure resulted in memory impairment. Working or short-term memory skills were tested by
asking each individual to recall two stories immediately after presentation (verbal recall) and
also draw three diagrams immediately after seeing the figures (visual recall). Additionally, a
digit span test where increasing numbers of digits were presented and then the subject had to
recall the digits was conducted to the extent of the short-term memory. Exposed subjects had
lower intelligence scores and there were significant impairments in verbal recall (p = 0.001),
visual recall (p = 0.03) and with the digit span test (p = 0.07). Significant impairment in
short-term memory as measured by three different cognitive test was correlated with TCE
exposure. Lower intelligence scores (p = 0.0001) as measured by the Culture Fair IQ test may be
a possible confounder in these findings. Additionally, the large range of TCE concentrations
(6-500 ppb) and exposure durations (1 to 25 years) and overall poor exposure characterization
precludes a no-observed-adverse-effect level (NOAEL)/lowest-observed-adverse-effect level
(LOAEL) from being estimated from this study on cognitive function.
Rasmussen et al. (1993 c; 1993 d) and Troster and Ruff (1990) present results of positive
findings in occupational studies for cognitive effects of TCE. Rasmussen et al. (1993d) reported
an historical cohort study conducted on 96 metal degreasers, identified 2 years previously and
were selected from a population of 240 workers from 72 factories in Denmark. They reported
psychoorganic syndrome, a mild syndrome of dementia characterized by cognitive impairment,
personality changes, and reduced motivation, vigilance, and initiative, was increased in the three
exposure groups. The medium and high exposure groups were compared with the low exposure
group. Neuropsychological tests included WAIS (original version, Vocabulary, Digit Symbol,
Digit Span), Simple Reaction Time, Acoustic-motor function (Luria), Discriminatory attention
(Luria), Sentence Repetition, Paced Auditory Serial Addition Test (PASAT), Text Repetition,
Rey's Auditory Verbal Learning, Visual Gestalts, Stone Pictures (developed for this study,
nonvalidated), revised Santa Ana, Luria motor function, and Mira. The prevalence of
psychoorganic syndrome was 10.5% in low exposure group; 38.9% in medium exposure group;
2	2
63.4%) in high exposure group, (x trend analysis: low vs. medium exposure x = 11.0,/? value
<0.001; low vs. high exposure x = 19.6,/?-value <0.001.) Psychoorganic syndrome increased
with age (p < 0.01). Age was strongly correlated with exposure.
Rasmussen et al. (1993c) used a series of cognitive tests to measure effects of
occupational TCE exposure. Short-term memory and retention following an latency period of
one hour was evaluated in several tests including a verbal recall (auditory verbal learning test),
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-19 DRAFT—DO NOT CITE OR QUOTE
-------
visual gestalts, visual recall (stone pictures), and the digit span test. Significant cognitive
performance decreases were noted in both short-term memory and memory retention. In the
verbal recall test immediate memory and learning were significantly decreased (p = 0.03 and
0.04, respectively). No significant effects were noted for retention following a one hour latency
period was noted. Significant increases in errors were noted in both the learning (p = 0.01) and
memory (p < 0.001) phases for the visual gestalts test. No significant effects were found in the
visual recall test in either the learning or memory phases or in the digit span test. As a result,
there were some cognitive deficits noted in TCE-exposed individuals as measured through
neuropsychological tests.
Troster and Ruff (1990) provides additional supporting evidence in an occupational study
for cognitive impairment, although the results reported in a qualitative fashion are limited in their
validity. In the two case studies that were exposed to TCE, there were decrements (no statistical
analysis performed) in cognitive performance as measured in verbal and visual recall tests that
were conducted immediately after presentation (learning phase) and one hour after original
presentation (retention/memory phase).
Triebig et al. (1977b) presents findings of no impairment of cognitive ability resulting
from TCE exposure in an occupational setting. This study was conducted on 8 subjects
occupationally exposed to TCE. Subjects were 7 men and 1 woman with an age range from
23-38 years. Measured TCE in air averaged 50 ppm (260 mg/m ). Length of occupational
exposure was not reported. There was no control group. Results were compared after exposure
periods, and compared to results obtained after periods removed from exposure. TCA and TCE
metabolites in urine and blood were measured. The testing consisted of the Syndrome Short
Test, which consists of nine subtests through which amnesic and simple perceptive and cognitive
functional deficits are detected; the "Attention Load Test" or "d2 Test" from Brickenkamp is a
procedure that measures attention, concentration, and stamina. Number recall test, letter recall
test, the "Letter Reading Test," "Word Reading Test." Data were assessed using Wilcoxon and
Willcox nonparametric tests. Due to the small sample size a significance level of 1% was used.
The concentrations of TCE, trichloroethanol, and TCA in the blood and total TCE and total TCA
elimination in the urine were used to assess exposure in each subject. The mean values observed
were 330 mg trichloroethanol and 319 mg TCA/g creatinine, respectively, at the end of a work
shift. The psychological tests showed no statistically significant difference in the results before
or after the exposure-free time period. The small sample size may limit the sensitivity of the
study.
Salvini et al. (1971), Gamberale et al. (1976), and Stewart et al. (1970) reported positive
findings for the impairment of cognitive function following TCE exposures in chamber studies.
Salvini et al. (1971) reported a controlled exposure study conducted on six male university
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-20 DRAFT—DO NOT CITE OR QUOTE
-------
students. TCE concentration was 110 ppm (550 mg/m ) for 4-hour intervals, twice per day.
Each subject was examined on two different days, once under TCE exposure, and once as self
controls, with no exposure. Two sets of tests were performed for each subject corresponding to
exposure and control conditions. The test battery included a perception test with tachistoscopic
presentation, the Wechsler memory scale test, a complex reaction time test, and a manual
dexterity test. Statistically significant results were observed for perception tests learning
(p < 0.001), mental fatigue (p < 0.01), subjects (p < 0.05); and CRT learning (p < 0.01), mental
fatigue (p < 0.01), subjects (p < 0.05). This is controlled exposure study with measured dose
(110 ppm; 600 mg/m ) and clear, statistically significant impact on neurological functional
domains. However, it only assesses acute exposures.
Gamberale et al. (1976) reported a controlled exposure study conducted on 15 healthy
men aged 20-31 yrs old, employed by the Department of Occupational Medicine in Stockholm,
Sweden. Controls were within subjects (15 self-controls), described above. Test used included
reaction time (RT) Addition and short term memory using an electronic panel. Subjects also
assessed their own conditions on a 7-point scale. Researchers used a repeated measures analysis
of variance (ANOVA) for the 4 performance tests based on a 3 x 3 Latin square design. In the
short-term memory test (version of the digit span test), a series of numbers lasting for one second
was presented to the subject. The volunteer then had to reproduce the numerical sequence after a
latency period (not specified). No significant effect on the short-term memory test was observed
with TCE exposure in comparison to air exposure. Potential confounders from this study include
repetition of the same task for all exposure conditions, volunteers served as their own controls,
and TCE exposure preceded air exposure in two of the three exposure experimental designs.
This is a well controlled study of short term exposures with measured TCE concentrations and
significant response observed for cognitive impairment.
Additional qualitative support for cognitive impairment is provided by Stewart et al.
(1970). This was a controlled exposure study conducted on 13 subjects in 10 experiments, which
-3
consisted of ten chamber exposures to TCE vapor of 100 ppm (550 mg/m ) and 200 ppm
"3
(1,100 mg/m ) for periods of 1 hour to a 5-day work week. Experiments 1-7 were for 7 hours
-3
with a mean TCE concentration of 198-200 ppm (1,090-1,100 mg/m ). Experiments 8 and
9 exposed subjects to 190-202 ppm (1,045-1,110 mg/m3) TCE for a duration of 3.5 and 1 hour,
"3
respectively. Experiment 10 exposed subjects to 100 ppm (550 mg/m ) TCE for 4 hours.
Experiments 2-6 were carried out with the same subjects over 5 consecutive days. Gas
chromatography of expired air was measured. There were no self controls. All had normal
neurological tests during exposure, but 50% reported greater mental effort was required to
perform a normal modified Romberg test on more than one occasion. There were no quantitative
data or statistics presented regarding dose and effects of neurological symptoms.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-21 DRAFT—DO NOT CITE OR QUOTE
-------
Two chamber studies conducted by Triebig et al. (1976; 1977a) report no impact of TCE
exposure on cognitive function. Triebig et al. (1976) was a controlled exposure study conducted
on 7 healthy male and female students (4 females, 3 males) exposed for 6 hours/day for 5 days to
100 ppm (550 mg/m TCE). The control group was 7 healthy students (4 females, 3 males)
exposed to hair care products. This was assumed as a zero exposure, but details of chemical
composition were not provided. Biochemical and psychological testing was conducted at the
beginning and end of each day. Biochemical tests included TCE, TCA, and trichloroethanol in
blood. Psychological tests included the d2 test, which was an attention load test; the short test
(as characterized in the translated version of Treibig, 1976) is used to record patient performance
with respect to memory and attention; daily Fluctuation Questionnaire measured the difference
between mental states at the start of exposure and after the end of exposure is recorded; The
MWT-A is a repeatable short intelligence test; Culture Fair Intelligence Test (CFT-3) is a
nonverbal intelligence test that records the rather "fluid" part of intelligence, that is, finding
solution strategies; Erlanger Depression Scale. Results were not randomly distributed. The
median was used to describe the mean value. Regression analyses were conducted. In this study
the TCE concentrations in blood reported ranged from 4 to 14 [j,g/mL. A range of 20 to
60 ng/mL was obtained for TCA in the blood. There was no correlation seen between exposed
and unexposed subjects for any measured psychological test results. The biochemical data did
demonstrate subjects' exposures. This is a well controlled study with excellent exposure data,
although the small sample size may have limited sensitivity.
Triebig et al. (1977a) is an additional report on the seven exposed subjects and seven
controls evaluated in Triebig et al. (1976). Additional psychological testing was reported. The
testing included the Syndrome Short Test, which consists of nine subtests, described above.
Statistics were conducted using Whitney Mann. Results indicated the anxiety values of the
placebo random sample group dropped significantly more during the course of testing (p < 0.05)
than those of the active random sample group. No significantly different changes were obtained
with any of the other variables. Both these studies were well controlled with excellent exposure
data, which may provide some good data for establishing a short term NOAEL. The small
sample size may have limited the sensitivity of the study.
Additional reports on the impairment of memory function as a result of TCE exposures
have been reported, and provide additional evidence of cognitive impairment. The studies by
Chalupa et al. (1960), Rasmussen et al. (1993c; 1986), and Troster and Ruff (1990) report
impairment of memory resulting from occupational exposures to TCE. Kilburn and Warshaw
(1993) and Kilburn (2002b) report impairment of memory following environmental exposures to
TCE. Salvini et al. (1971) reports impairment of memory in a chamber study, although Triebig
et al. (1976) reports no impact on memory following TCE exposure in a chamber study.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-22 DRAFT—DO NOT CITE OR QUOTE
-------
D. 1.6. Psychomotor Effects
There is evidence in the literature that TCE can have adverse psychomotor effects in
humans. The effects of TCE exposure on psychomotor response have been studied primarily as
the impact on RTs, which provide a quantitative measure of the impact TCE exposure has on
motor skills. Studies on motor dyscoordination resulting from TCE exposure are more
subjective, but provide additional evidence that TCE may cause adverse psychomotor effects.
These studies are described below.
D. 1.6.1. Reaction Time
There are several reports in the literature that report an increase in reaction times
following exposures to TCE. The best evidence for TCE exposures causing an increase in choice
reaction times comes from environmental studies by Kilburn (2002b), Kilburn and Warshaw
(1993), Reif et al. (2003), and Kilburn and Thornton (1996), which were all conducted on
populations which were exposed to TCE through groundwater contaminated as the result of
environmental spills. Kilburn (2002b) (study details described in Section D.l.l) evaluated
reaction times in a Phoenix, Arizona population exposed to TCE through groundwater.
Volunteers were tested for response rates in the simple reaction time (SRT) and 2 choice reaction
time (CRT) tests. Various descriptive statistics were used, as well as analysis of covariance
(ANCOVA) and a step-wise adjustment of demographics. The principal comparison, between
the 236 exposed persons and the 161 unexposed regional controls, revealed significant
differences (p < 0.05) indicating that SRTs and CRTs were delayed. Balance was also abnormal
with excessive sway speed (eyes closed), but this was not true when both eyes were open. This
study shows statistically significant differences in psychomotor responses between exposed and
nonexposed subjects exposed environmentally. However, it is limited by poor exposure
characterization.
Kilburn and Warshaw (1993) (study details described in Section D. 1.1.1) evaluated
reaction times in 170 Arizona residents exposed to TCE in well water. A referent group of
68 people was used for comparison. TCE concentration was from 6 to 500 ppb and exposure
ranged from 1 to 25 years. SRT was determined by presenting the subject a letter on a computer
screen and measuring the time (in milliseconds [msec]) it took for the person to type that letter.
SRT significantly increased from 281 ± 55 msec to 348 ± 96 msec in TCE-exposed individuals
(p < 0.0001). Similar increases were reported for CRT where subjects were presented with two
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-23 DRAFT—DO NOT CITE OR QUOTE
-------
different letters and required to make a decision as to which letter key to press. CRT of the
exposed subjects was 93 msec longer in the third trial (p < 0.0001) than referents. It was also
longer in all trials, and remained significantly different after age adjustment. This study shows
statistically significant differences for neurological test results between subjects environmentally
exposed and nonexposed to TCE, but is limited by poor exposure data on individual subjects
given the ecological design of this study. Additionally, litigation is suggested and may introduce
a bias, particularly if no validity tests were used.
Kilburn and Thornton (1996) conducted an environmental study that attempts to use
reference values from two control groups in assessing neurological responses for chemically
exposed subjects using neurophysiological and neuropsychological testing on three groups.
Group A included randomly selected registered voters from Arizona and Louisiana with no
exposure to TCE: n = 264 unexposed volunteers aged 18-83. Group B included volunteers from
California n = 29 (17 males and 12 females) that were used to validate the equations; Group C
included those exposed to TCE and other chemicals residentially for 5 years or more n = 237.
Group (A), was used to develop the regression equations for SRT and choice reaction time
(CRT). A similarly selected comparison group B was used to validate the equations. Group C,
the exposed population, was submitted to SRT and CRT tests (n = 237) and compared to the
control groups. All subjects were screened by a questionnaire. Reaction speeds were measured
using a timed computer visual-stimulus generator. No exposure data were presented. The Box-
Cox transformation was used for dependent variables and independent variables. They evaluated
graphical methods to study residual plots. Cook's distance statistic was used as a measure of
influence to exclude outliers with undue influence and none of the data were excluded. Lack-of-
fit test was performed on Final model and F statistic was used to compare estimated error to
lack-of-fit component of the model's residual sum of squared error. Final models were validated
using group B data and paired t-test to compare observed values for SRT and CRT. F statistic
was used to test the hypothesis that parameter estimates obtained with group B were equal to
those of Group A, the model. The results are as follows: Group A: SRT = 282 ms;
CRT = 532 ms. Group B: SRT = 269 ms; CRT = 531 ms. Group C: SRT = 334 ms;
CRT = 619 ms. TCE exposure produced a step increase in reaction times (SRT and CRT). The
coefficients from Group A were valid for group B. The predicted value for SRT and for CRT,
plus 1.5 SDs selected 8% of the model group as abnormal. The model produced consistent
measurement ranges with small numerical variation. This study is limited by lack of any
exposure data, and does not provide statistics to demonstrate dose-response effects.
Kilburn (2002b) conducted an environmental study on 236 residents chronically exposed
to TCE-associated solvents in the groundwater resulting from a spill from a microchip plant in
Phoenix, AZ. Details of the TCE exposure and population are described earlier in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-24 DRAFT—DO NOT CITE OR QUOTE
-------
Section D. 1.1.1 (see Kilburn, 2002b). The principal comparison, between the 236 exposed
persons and the 161 unexposed regional controls, revealed significant differences indicating that
SRTs and choice reaction times (CRTs) were increased. SRTs significantly increased from
283 ± 63 msec in controls to 334 ±118 msec in TCE exposed individuals (p < 0.0001).
Similarly, CRTs also increased from 510 ± 87 msec to 619 ± 153 msec with exposure to TCE
(p < 0.0001). This study shows statistically significant differences in psychomotor responses as
measured by reaction times between TCE-exposed and nonexposed subjects. Estimates of TCE
concentrations in drinking water to individual subjects were not reported in the paper. Since the
TCE exposure ranged from 0.2 to over 10,000 ppb in well water, it is not possible to determine a
NOAEL for increased reaction times through this study. Additionally, litigation is suggested and
may introduce a bias, particularly if no validity tests were used.
Reif et al. (2003) conducted a cross sectional study on 143 residents of the Rocky
Mountain Arsenal (RMA) community of Denver exposed environmentally to drinking water
contaminated with TCE and related chemicals from nearby hazardous waste sites between 1981
and 1986. The referent group was at the lowest estimated exposure concentration (<5 ppb). The
socioeconomic profile of the participants closely resembled those of the community in general.
"A total of 3393 persons was identified through the census, from which an age- and
gender-stratified sample of 1267 eligible individuals who had lived at their current residence for
at least 2 years was drawn. Random selection was then used to identify 585 persons from within
the age-gender strata, of whom 472 persons aged 2-86 provided samples for biomonitoring.
Neurobehavioral testing was conducted on 204 adults who lived in the RMA exposure area for a
minimum of 2 years. Among the 204 persons who were tested, 184 (90.2%) lived within the
boundaries of the LWD and were originally considered eligible for the current analysis.
Therefore, participants who reported moving into the LWD after 1985 were excluded from the
total of 184, leaving 143 persons available for study." An elaborate hydraulic simulation model
(not validated) was used in conjunction with a geographic information system (GIS) to model
estimates of residential exposures to TCE. The TCE concentration measured in community
wells exceeded the MCL of 5 ppb in 80% of cases. Approximately 14% of measured values
exceeded 15 ppb. Measured values were used to model actual exposure estimates based on
distance of residences from sampled wells. The estimated exposure for the high exposure group
was >15 ppb; the estimate for the low exposure referent group was <5 ppb. The medium
exposure group was estimated at exposures 5< x <15 ppb TCE. The test battery consisted of the
Neurobehavioral Core Test Battery (NCTB), which consists of 7 neurobehavioral tests including
simple reaction time. Results were assessed using the Multivariate Model. Results were
statistically significant (p < 0.04) for the simple reaction time tests. The results are confounded
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-25 DRAFT—DO NOT CITE OR QUOTE
-------
by exposures to additional solvents and modeled exposure data, which while highly technical,
are still only a rough estimate of actual exposures, and may limit the sensitivity of the study.
Gamberale et al. (1976) conducted a controlled exposure (chamber) study on 15 healthy
men aged 20-31 yrs old, employed by the Department of Occupational Medicine in Stockholm,
Sweden. Controls were within subjects (15 self-controls). Subjects were exposed to TCE for
3	3
70 minutes via a breathing valve to 540 mg/m (97 ppm), 1,080 mg/m (194 ppm), and to
ordinary atmospheric air (0 ppm). Sequence was counterbalanced between the 3 groups, days,
and exposure levels. Concentration was measured with a gas chromatographic technique every
third minute for the first 50 minutes, then between tests thereafter. Test used were RT addition,
simple RT, choice RT and short term memory using an electronic panel. Subjects also assessed
their own conditions on a 7 point scale. The researchers performed Friedman two-way analysis
by ranks to evaluate differences between the 3 conditions. The results were nonsignificant when
tested individually, but significant when tested on the basis of six variables. Nearly half of the
subjects could distinguish exposure/nonexposure. Researchers performed ANOVA for the four
performance tests based on a 3 x 3 Latin square design with repeated measures. In the RT-
Addition test the level of performance varied significantly between the different exposure
conditions (F[2.24] = 4.35; p < 0.05) and between successive measurement occasions
(F[2.24] = 19.25; p < 0.001). The level of performance declined with increased exposure to
TCE, whereas repetition of the testing led to a pronounced improvement in performance as a
result of the training effect. No significant interaction effects were observed between exposure
to TCE and training. This is a good study of short term exposures with measured TCE
concentrations and significant response observed for reaction time.
Gun et al. (1978) conducted an occupational study on 8 TCE-exposed workers who
operated degreasing baths in two different plants. Four female workers were exposed to TCE
only in one plant and four female workers were exposed to TCE and nonhalogenated
hydrocarbon solvents in the second plant. The control group (n = 8) consisted of 4 female
workers from each plant who did not work near TCE. Each worker worked 2 separate 4-hour
shifts daily, with one shift exposed to TCE and the second 4-hour shift not exposed. Personal air
samples were taken continuously over separate 10-minute sessions. Readings were taken every
30 seconds. Eight-choice reaction times were carried out in four sessions; at the beginning and
end of each exposure to TCE or TCE + solvents; a total of 40 reaction time trials were
completed. TCE concentrations in the TCE only plant 1 (148-418 ppm [800-2,300 mg/m3])
"3
were higher than in the TCE + solvent plant 2 (3-87 ppm (16-480 mg/m ). Changes in choice
reaction times (CRT) were compared to level of exposure. The TCE only group showed a mean
increase in reaction time, with a probable cumulative effect. In the TCE + solvent group, mean
reaction time shortened in Session 2, then increased to be greater than at the start. Both control
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-26 DRAFT—DO NOT CITE OR QUOTE
-------
groups showed a shortening in mean choice reaction time in Session 2, which was sustained in
Sessions 3 and 4 consistent with a practice effect. This is a study with well-defined exposures
and reports of cause and effect (TCE exposure on reaction time); however, no statistics were
presented to support the conclusions or the significance of the findings, and the small sample size
is a limitation of the study.
D. 1.6.2. Muscular Dyscoordination
Effects on motor dyscoordination resulting from TCE exposure have been reported in the
literature. These impacts are subjective, but may provide additional evidence that TCE can cause
adverse psychomotor effects. There are three reports summarized below which suggest that
muscular dyscoordination resulted from TCE exposure, although all three have significant
limitations due to confounding factors. Rasmussen et al. (1993a) presented findings on muscular
dyscoordination as it relates to TCE exposure. This was a historical cohort study conducted on
96 metal degreasers, identified 2 years previously. Subjects were selected from a population of
240 workers from 72 factories in Denmark. Although the papers report a population of
99 participants, tabulated results were presented for a total of only 96. No explanation was
provided for this discrepancy. These workers had chronic exposure to fluorocarbon (CFC 113)
(n = 25) and mostly TCE (n = 70; average duration: 7.1 years.). There were no external controls.
The range of working full-time degreasing was 1 month to 36 years. Researchers collected data
regarding the workers' occupational history, blood and urine tests, as well as biological
monitoring for TCE and TCE metabolites. A chronic exposure index (CEI) was calculated based
on number of hours per week worked with solvents multiplied by years of exposure multiplied
by 45 weeks per year. No TCE air concentrations were reported. Participants were categorized
into three groups: (1) "Low exposure:" n = 19, average full-time exposure = 0.5 years.
(2) "Medium exposure:" n = 36, average full-time exposure = 2.1 years. (3) "High exposure:"
n = 41, average full-time exposure =11 years. The mean TCA level in the "high" exposure
group was 7.7 mg/L (max = 26.1 mg/L). Time-weighted average (TWA) measurements of
CFC 113 levels were 260-420 ppm (U.S. and Danish TLV was 500 ppm). A significant trend of
dyscoordination from low to high solvent exposure was observed (p = 0.003). This study
provides evidence of causality for muscular dyscoordination resulting from exposure to TCE, but
no measured exposure data were reported.
Additional evidence of the psychomotor effects caused by exposure to TCE are presented
in Gash et al. (2008) and Troster and Ruff (1990). There are, however, significant limitations
with each of these studies. In Gash et al. (2008), the researchers evaluated the clinical features of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-27 DRAFT—DO NOT CITE OR QUOTE
-------
1 Parkinson's disease (PD) patient, identified in a Phase 1 clinical trial study, index case, and an
additional 29 coworkers of the patient, all with chronic occupational exposures to TCE. An
additional 2 subjects with Parkinson's Disease were included, making the total of 3 Parkinson's
disease patients, and 27 non-Parkinson's coworkers making up the study population. Coworkers
for the study were identified using a mailed questionnaire to 134 former coworkers. No details
are provided in the paper on selection criteria for the 134 former coworkers. Of the 134 former
workers sent questionnaires, 65 responded. Twenty-one self-reported no symptoms, 23 endorsed
1-2 symptoms, and 21 endorsed 3 or more signs of Parkinsonism. Fourteen of the 21 with 3 or
more signs and 13 of the 21 without any signs agreed to a clinical exam; this group comprises the
27 additional workers examined for Parkinsonian symptoms. No details were provided on
nonresponders. All subjects were involved in degreasing with long-term chronic exposure to
TCE through inhalation and dermal exposure (14 symptomatic: age range = 31-66, duration of
employment range: 11-35 yrs) (13 asymptomatic: age range = 46-63, duration of employment
range: 8-33 yrs). The data were compared between groups and with data from 110 age- matched
controls. Exposure to TCE is self-reported and based on job proximity to degreasing operations.
The paper lacks any description of degreasing processes including TCE usage and quantity.
Mapping of work areas indicated that workers with PD worked next to the TCE container, and
all symptomatic workers worked close to the TCE container. Subjects underwent a general
physical exam, neurological exam and Unified Parkinson's Disease Rating Scale (UPDRS),
timed motor tests, occupational history survey, and mitochondrial neurotoxicity. ANOVA
analysis was conducted, comparing symptomatic versus nonsymptomatic workers, and
comparing symptomatic workers to age-matched nonexposed controls. No description of the
control population (n = 110), nor how data were obtained for this group, was presented. The
symptomatic non-Parkinson's group was significantly slower in fine motor hand movements
than age-matched nonsymptomatic group (p < 0.001). The symptomatic group was significantly
slower (p < 0.0001) than age-matched unexposed controls as measured in fine motor hand
movements on the Movement Analysis Panel. All symptomatic workers had positive responses
to 1 or more questions on UPDRS Part II (diminished activities of daily life), and/or
deteriorization of motor functions on Part III. The fine motor hand movement times of the
asymptomatic TCE-exposed group were significantly slower (p < 0.0001) than age-matched
nonexposed controls. Also, in TCE-exposed individuals, the asymptomatic group's fine motor
hand movements were slightly faster (p < 0.01) than those of the symptomatic group. One
symptomatic worker had been tested 1 year prior and his UPDRS score had progressed from
9 to 23. Exposures are based on self-reported information, and no information on the control
group is presented. One of the PD patients predeceased the study and had a family history
of PD.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-28 DRAFT—DO NOT CITE OR QUOTE
-------
Troster and Ruff (1990) reported a case study conducted on two occupationally exposed
workers to TCE. Patients were exposed to low levels of TCE. There were 2 groups of n = 30
matched controls (all age and education matched) whose results were compared to the
performance of the exposed subjects. Exposure was described as "Unknown amount of TCE for
8 months." Assessment consisted of the San Diego Neuropsychological Test Battery (SDNTB)
and "1 or more of' Thematoc Apperception Test (TAT), Minnesota Multiphasic Personal
Inventory (MMPI), and Rorschach. Medical examinations were conducted, including
neurological, CT scan, and/or chemo-pathological tests, and occupational history was taken, but
not described. There were no statistical results reported. Results were reported for each test, but
no tests of significance were included, therefore, the authors presented their conclusions for each
"case" in qualitative terms, as such: Case 1: Intelligence "deemed" to drop from premorbid
function at 1 year 10 months after exposure. Impaired functions improved for all but reading
comprehension, visuospatial learning and categorization (abstraction). Case 2: Mild deficits in
motor speed, but symptoms subsided after removal from exposure.
D.1.7. Summary Tables
The following Tables (D-l through D-3) provide a detailed summary of all the
neurological studies conducted with TCE in humans. Tables D-l and D-2 summarize each
individual human study where there was TCE exposure. Table D-l consists of studies where
humans were primarily or soley exposed to TCE. Table D-2 contains human studies where there
was a mixed solvent exposure and TCE was one of the solvents in the mixture. For each study
summary, the study population, exposure assessment, methods, statistics, and results are
provided. Table D-3 indicates the neurological domains that were tested from selected
references (primarily from Table D-l).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-29 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>s
to
to'
*
^S
«
8
TO
Co
0
S
1
a.
Table D-l. Epidemiological studies: Neurological effects of trichloroethylene


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
§•
>3
o
o
s
Co
I?
£u
VO
to
o
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
>r
<§>
>S
TO
*
^S
«
s
TO
Co
0
S
1
a.
VO
to
o
Barret et al.
188 workers exposed
Review of medical
Complete physical
examined
Symptoms for which TCE role is statistically
(1984)
to TCE occupationally
records and analysis of
examination including
distribution of the
significant include the following: Trigeminal

from small and large
TCE atmospheric levels
testing visual
different groups
nerve impairment was reported in 22.2% (n = 12)

factories in France
(detector tubes) and
performance (acuity
for comparing high
of workers in the high-exposure group for TCE,

(type of factories not
level of urinary
and color perception),
and low exposed
7.4% (n = 10) in the low-exposure group for TCE,

disclosed); average
metabolites
evoked trigeminal
workers, one way
24.4% (n = 10) in the high-exposure group for

age = 41; 6 yrs
measurement (TCA).
potential latencies and
analysis of
TCA and 8.2% (n = 12) in the low-exposure

average exposure
TCE exposure groups
audiometry, facial
variance, Mann
group for TCA.



time.
included high exposure
sensitivity, reflexes,
Whitney U and






group (>150 ppm;
and motoricity of the
t-test for analyzing

High
Low


The workers were
n = 54) and low
masseter muscles.
personal factors.
TCE Results
dose%
dose%
P

divided into high and
exposure group


Trigeminal




low exposure groups
<150 ppm; n = 134).


nerve
22.2
7.4
<0.01

for both TCE and
urinary TCA. No
Personal factors
including age, tobacco


Impairment
asthenia
18.5
4.5
<0.01

control group was
mentioned.
use, and alcohol intake
were also analyzed;
Exposure duration =
7 h/d for 7 yrs; no
mention was made


Optic nerve
impairment
Headache
Dizziness
14.8
20.3
13
0.75
19.4
4.5
<0.001
NS
0.05 
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>s
to
to'
*
^S
«
8
TO
Co
0
S
1
a.
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
>3
o
o
s
Co
I?
£u
VO
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Barret et al.
104 occupationally
Urinary analysis
Evoked trigeminal
Student's t-test and
Dizziness (71.4%), headache (55.1%), asthenia
(1987)
exposed workers
determined TCE and
potentials were studied
one-way ANOVA
(46.9%), insomnia (24.4%), mood perturbation

highly exposed to TCE
TCA rates. The average
while eyes closed and
used as well as
(20.4%), and sexual problems (12.2%) were found.

during work as
of the last
fully relaxed. Also,
nonparametric tests
Symptomatic patients had significantly longer

degreaser machine
5 measurements were
physical exams with
Mann-Whitney U
exposure periods and were older than

operators in France.
considered indicative of
emphasis on nervous
test and Kruskal-
asymptomatic patients. 17.3% of patients had

Controls: 52 healthy,
the average level of past
system, a clinical study
Wallistest. Also
trigeminal nerve symptoms. Bilateral

nonexposed controls of
exposure. Mean
of facial sensitivity, and
decision matrix and
hypoesthesia with reflex alterations in 9 cases.

various ages who were
exposure 8.2 yrs,
of the reflexes
the analysis of the
Hypoesthesia was global and predominant in the

free from neurological
average daily exposure
depending on the
receiver operating
mandibular and maxillary nerve areas. Several

problems.
7 hrs/d. Mean age
trigeminal nerve were
curve to appreciate
reflex abolitions were found without facial palsy


41.6 yrs.
systematically
the accuracy of the
and without convincing hypoesthesia in 9 cases.



performed. Normal
TSEP method. The
Corneal reflexes were bilaterally abolished in



latency and amplitude
distribution of the
5 cases as were naso-palpebral reflexes in 6 cases;



values for TSEP
different
length of exposure positively correlated with



obtained from data from
populations was
functional manifestations (p < 0.01); correlation



control population.
compared by a chi
between symptoms and exposure levels were



Normal response
square test.
nonsignificant; 40 (38.4%) subjects had



characterized from

pathological response to TSEP with increased



4 main peaks,

latencies, amplitude or both; of these 28 had



alternating from

normal clinical trigeminal exam and 12 had



negative to positive,

abnormal exam; TSEP was positively correlated



respective latency of

with length of exposure (p < 0.01); and with age



12.8 ms (SD = 0.6),

(p < 0.05), but not with exposure concentration;



19.5 ms(SD= 1.3),

trigeminal nerve symptoms (n= 18) were



27.6 ms (SD = 1.6), and

positively correlated with older age (p < 0.001).



36.8 ms (SD = 2.2),





mean amplitude of





response is 2.5 ^v





(SD = 0.5 nv).





Pathological responses





ii ara mcnltc 0 1/.


-------
IS*
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)




Exposure assessment




a
Reference
Study population
and biomarkers
Tests used
Statistics
Results


Barret et al.
Eleven workers with
Selected following
Somatosensory evoked
SEP recordings
3 pathological abnormalities present in exposed
o
s*
(1982)
chronic TCE exposure;
clinical evaluations of
potential (SEP)
illustrated from
(TCE intoxicated) workers: (1) in 8 workers higher
o
S*

9 were suffering
their facial sensitivity
following stimulation of
trigeminal nerve
voltage required to obtain normal response, (2)
s
Co


effects of solvent
and trigeminal nerve
the trigeminal nerve
graphs.
excessive delay in response observed twice, (3)
1?
>{

intoxication; 2 were
reflexes; exposures
through the lip

excessive graph amplitude noted in 3 cases. One

—»

work place controls.
verified by urinalysis.
alternating right and left

subject exhibited all 3 abnormalities. Correlation
x


Control group was
Presence of TCE and
by a bipolar surface

was reported between clinical observation and test

t:
^3

20 unexposed subjects
TCA found. (Exposure
electrode utilizing

results. Most severe SEP alternations observed in
o
St
"3

of all ages.
rates not reported).
voltage, usually 75 to

subjects with the longest exposure to TCE
t
8
to



80 V, just below what is

(although exposure levels or exposure durations
Co
0
S
1



necessary to stimulate
the orbicularis oris
muscle. Duration was

are not reported). No statistics presented.
VO
a.



approx. 0.05 ms


o
to
o



stimulated 500 times
(2x/sec).


-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Burg et al.
From an NHIS TCE
Morbidity baseline data
Self report via face-to-
Poisson Regression
Speech impairments showed statistically
(1995)
subregistry of 4,281
were examined from the
face interviews—
analysis model used
significant variability in age-specific risk ratios

(4,041 living and
TCE Subregistry from
25 questions about
for registrants 19
with increased reporting for children <9 yrs

240 deceased)
the NER developed by
health conditions; were
and older.
(RR: 2.45, 99% CI: 1.31, 4.58) and for registrants

residents
the ATSDR; were
compared to data from
Maximum
>35 yrs (data broken down by 10-yr ranges).

environmentally
interviewed in the
the entire NHIS
likelihood
Analyses suggest a statistically significant increase

exposed to TCE via
NHIS.
population;
estimation and
in reported hearing impairments for children <9

well water in Indiana,

neurological endpoints
likelihood ratio
yrs (RR: 2.13, 99% CI: 1.12, 4.06). It was lower

Illinois, and Michigan;

were hearing and
statistics and Wald
for children 10-17 yrs (RR: 1.12, 99% CI: 0.52,

compared to NHIS

speech impairments.
CI; TCE
2.44) and <0.32 for all other age groups.

registrants.


subregistry





population was





compared to larger





NHIS registry





population.

>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Burg and
4,041 living members
All registrants exposed
Interviews
Logistic
When the registrants were grouped by duration of
Gist (1999)
of the National
to TCE though domestic
(occupational,
Regression, Odds
exposure to TCE, a statistically significant

Exposure Registry's
use of contaminated
environmental,
Ratios; lowest
association (adjusted for age and sex) between

T richloroethylene
well water; 4 exposure
demographic, and
quartile used as
duration of exposure and reported hearing

Subregistry; 97%
Subgroups, each divided
health information); A
reference
impairment was found. The prevalence odds ratios

white; mean age 34 yrs
into quartiles:
large number of health
population.
were 2.32 (95% CI: 1.18, 4.56) (>2 to <5 yrs); 1.17

(SD = 19.9 yrs.);
(1) Maximum TCE
outcomes analyzed,

(95% CI: 0.55, 2.49) (>5 to <10 yrs); and 2.46

divided in 4 groups
measured in well water,
including speech

(95% Cl:1.30, 5.02) (>10 yrs); Higher rates of

based on type and
exposure subgroups:
impairment and hearing

speech impairment (not statistically significant)

duration of exposure;
2-12 ppb; 12-60 ppb;
impairment.

associated with maximum and cumulative TCE

analysis reported only
60-800 ppb;


exposure, and duration of exposure.

for 3,915 white
(2) Cumulative TCE




registrants; lowest
exposure subgroups:




quartile used as control
<50 ppb, 50-500 ppb,




group.
500-5,000 ppb,





>5,000 ppb;





(3) Cumulative





chemical exposure





subgroups: include





TCA, DCE, DCA, in





conjunction with TCE,





with the same exposure





Categories as in # 2;





(4) Duration of exposure





subgroups: <2 yrs,





2-5 yrs, 5-10 yrs.,





>10 yrs.; 2,867 had TCE





exposure of <50 ppb;





870 had TCE exposure





of 51-500 ppb; 190 had





TCE exposure of





sni-'S nnn nnh- ^ ViqH



-------

§•
to
s
a
to
s
Co
a,
>r
>S
>S
to'
*
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Buxton and
Hayward
(1967)
This was a case study
on 4 workers exposed
to very high
concentrations of TCE,
which resulted from an
industrial accident. No
controls were
evaluated.
Case 1 was a 44-yr old
man exposed for
10 min; Case 2 was a
39-yr old man exposed
for 30 min; Case 3 was
a 43-yr old man exposed
for 2.5 h; Case 4 was a
39-yr old man exposed
for 4 h. TCE
concentrations were not
reported.
Clinical evaluations
were conducted by a
physician when patients
presented with
symptoms; numbness of
face, ocular pain,
enlarged right blind
spot, nausea, loss of
taste, headache,
dizziness, unsteadiness,
facial diplesia, loss of
gag and swallowing
reflex, absence of
corneal reflex, and
reduction of trigeminal
response.
There was no
statistical
assessment of
results presented.
Case 1 exhibited headaches and nausea for 48 h,
but had a full recovery. Case 2 exhibited nausea
and numbness of face, but had a full recovery.
Case 3 was seen and treated at a hospital with
numbness of face, insensitivity to pin prick over
the trigeminal distribution, ocular pain, enlarged
right blind spot, nausea, and loss of taste. No loss
of mental faculty was observed. Case 4 was seen
and treated for headache, nausea, dizziness,
unsteadiness, facial diplesia, loss of gag and
swallowing reflex, facial analgesia, absence of
corneal reflex, and reduction of trigeminal
response. The patient died and was examined
postmortem. There was demyelination of the 5th
cranial nerve evident.
Chalupa et
al. (1960)
This was a case study
conducted on 22
patients with acute
poisoning caused by
carbon monoxide and
industrial solvents.
Six subjects were
exposed to TCE (doses
not known). Average
age 38.
No exposure data were
reported.
Medical and
psychological exams
were given to all
subjects. These
included EEGs,
measuring middle
voltage theta activity of
5-6 sec duration.
Subjects were tested for
memory disturbances.
No statistics were
performed.
80% of those with pathological EEG displayed
memory loss; 30% of those with normal EEGs
displayed memory loss. Pathology and memory
loss were most pronounced in subjects exposed to
carbon monoxide.
-------
ko	Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
is*
5}'




Exposure assessment




a
Reference
Study population
and biomarkers
Tests used
Statistics
Results


El Ghawabi
30 money printing
Air samples on
Inquiries about
Descriptive
Most frequent symptoms: prenarcotic headache
o
s*
etal. (1973)
shop workers
30 workers. Mean TCE
occupational, past and
statistics and
(86% vs. 30% for controls), dizziness (67% vs.
o
S*

occupationally
air concentrations
present medical
central tendency
6.7% for controls), and sleepiness (53% vs. 6% for
s
Co


exposed to TCE;
ranged from 41 to
histories, and family
evaluation for
controls) main presenting symptoms in addition to
1?


Controls: 20 age and
163 ppm throughout the
histories in addition to
metabolites; no
suppression of libido. Trigeminal nerve



SES matched
Intalgio process
age and smoking habits.
stats reported for
involvement was not detected. The concentration
x


nonexposed males and
Colorimetric
EKGs were performed
neurological
of total trichloro-compounds increased toward

t:
^3

10 control workers not
determination of both
on 25 of the workers.
symptoms.
mid-week and was stationary during the last 2
o
si
"3

exposed to TCE but
TCA and total trichloro-
Lab investigations

working days. Metabolites of total trichloroacetic
t
s

exposed to inks used in
compounds in urine
included complete

acid and trichloroethanol are only proportional to
Co
O
S

printing.
withFujiware reaction.
blood and urine
analysis, and routine

TCE concentrations up to 100 ppm.

1
a.



liver function tests.


VO
Feldman et
21 Massachusetts
TCE in residential well
BR used as an objective
Student's t-test
Highly significant differences in the conduction
o
to
o
(—¦)
al. (1988)
residents with alleged
water was 30-80 times
indicator of neurotoxic
used for testing the
latency means of the BR components for the TCE


chronic exposure to
greater than U.S. EPA
effects of TCE; clinical
difference between
exposed population vs. control population, when



TCE in drinking water;
MCL; maximum
neurological exam,
the group means for
comparing means for the right and left side R1 to



27 laboratory controls.
reported concentration
was 267 ppb; other
solvents also present.
EMGs to evaluate blink
reflex, nerve conduction
studies, and extensive
neuropsychological
testing.
the Blink reflex
component
latencies.
the controls (p < 0.001). The mean R1 BR
component latency for the exposed group was
11.35 ms, SD = 0.74 ms, 95% CI: 11.03-11.66.
The mean for the controls was 10.21 ms, SD =
0.78 ms, 95% CI: 9.92-10.51; p < 0.001. Suggests
a subclinical alteration of the trigeminal nerve
function due to chronic, environmental exposure to
TCE.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Feldman et
18 workers
Reviewed exposure
Blink reflexes using
Non-Gaussian
The "extensive" group revealed latencies greater
al. (1992)
occupationally
histories of each worker
TECA 4 EMG.
distribution and
than 3 SD above the nonexposed group mean on

exposed to TCE; 30
(job type, length of

high coefficient of
R1 component of blink reflex; none of the

laboratory controls.
work) and audited

variance data were
"occasional" group exhibited such latencies,


medical records to

log-transformed
however, two of them demonstrated evidence of


categorize into three

and then compared
demyelinating neuropathy on conduction velocity


exposure categories:

to the log-
studies; the sensitivity, or the ability of a positive


"extensive,"

transformed control
blink reflex test to correctly identify those who had


"occasional," and

mean values. MRV
TCE exposure, was 50%. However, the specificity


"chemical other than

was calculated by
was 90%, which means that of those workers with


TCE".

subtracting the
no exposure to TCE, 90% demonstrated a normal




subjects value (x)
K1 latency. Subclinical alteration of the Vth




from the control
cranial nerve due to chronic occupational exposure




group mean (M),
to TCE is suggested.




and the difference





is divided by the





control group





standard deviation.

>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Gash et al.
30 Parkinson's Disease
Mapping of work areas.
General physical exam,
Workers' raw
Symptomatic non-Parkinson's group was
(2008)
patients and

neurological exam and
scores given;
significantly slower in fine motor hand movements

27 non-Parkinson

UPDRS, timed motor
ANOVA
than age-matched nonsymptomatic group

coworkers exposed to

tests, and occupational
comparing
(p < 0.001); All symptomatic workers had positive

TCE; No unexposed

history survey;
symptomatic vs.
responses to 1 or more questions on UPDRS Part I

controls.

mitochondrial
nonsymptomatic
and Part II, and/or had signs of parkinsonism on



neurotoxicity;
workers.
Part III; One symptomatic worker had been tested



Questionnaire mailed to

1 yr. prior and his UPDRS score had progressed



134 former

from 9 to 23.



non-Parkinson's





workers,





(14 symptomatic of





parkinsonism: age





range = 31-66, duration





of employment range:





11-35 yrs)





(13 asymptomatic: age





range = 46-63, duration





of employment range:





8-33 yrs);.


>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Grandjean et
80 workers employed
Vapors were collected
Medical exam,
Coefficient of
Men working all day with TCE showed on average
al. (1955)
in 10 different
in ethylic alcohol 95%.
including histories;
determination,
larger amounts of TCA than those who worked

factories of the Swiss
Volume of air was
Blood and biochem.
Regression
part time with TCE. Relatively high frequency of

mechanical
checked using a
tests, and psychiatric
coefficient.
subjective complaints, of alterations of the

engineering industry
flowmeter, and
exam. Psychological

vegetative nervous system, and of neurological

exposed to TCE, seven
quantitatively measured
exam; Meggendorf,

and psychiatric symptoms. 34% had slight or

of whom stopped
according to the method
Bourdon, Rorschach,

moderate psycho-organic syndrome; 28% had

working with TCE
of Truhaut (1951),
Jung, Knoepfel's

neurological changes; There is a relationship

from 3 wks to 6 yrs
which is based on a
"thirteen mistakes" test,

between the frequency of those alterations and the

prior; no unexposed
colored reaction
and Bleuler's test.

degree of exposure to TCE. There were

control group.
between TCE and the


significant differences (p = 0.05) in the incidence


pyridine in an alkaline


of neurological disorder between Groups I and III,


medium (with


while between Groups II and III there were


modifications). Urine


significant differences (p = 0.05) in vegetative and


analysis of TCA levels;


neurological disorders. Based on TCA eliminated


TCE air concentrations


in the urine, results show that subjective,


varied from


vegetative, and neurological disorders were more


6-1,120 ppm depending


frequent in Groups II and III than in Group I.


on time of day and


Statistical analysis revealed the following


proximity to tanks, but


significant differences (p < 0.01): subjective


mainly averaged


disorders between I and II; vegetative disorders


between 20-40 ppm.


between I and II and between I and III;


Urinalysis varied from


neurological disorders between I and (II and III).


30 mg/L to 300 mg/L;


Vegetative, neurological, and psychological


Could not establish a


symptoms increased with the length of exposure to


relationship between


TCE. The following definite differences were


TCE eliminated through


shown by statistical analysis (p < 0.03): vegetative


urine and TCE air


disorders between I and IV; neurological


levels. Four exposure


disorders between I and II and between I and IV;


groups estimated based


psychological disorders between I and III and


on air sampling data.


between I and IV.
-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
to
o
VO
Gun, el al.
(1978)
8 exposed: 4 female
workers from one
plant exposed to TCE
and 4 female workers
from another plant
exposed to TCE +
nonhalogenated
hydrocarbon solvent
used in degreasing;
control group (n = 8)
consisted of 4 female
workers from each
plant who did not work
near TCE.
Air sampled
continuously over
separate 10 min
durations drawn into a
Davis Halide Meter.
Readings taken every 30
sec.; ranged from 3-419
ppm.
Eight-Choice reaction
times carried out in four
sessions; 40 reaction
time trials completed.
Variations in RT by
level of exposure;
ambient air
exposure TCE
concentrations and
mean air TCE
values.
TCE only group had consistently high mean
ambient air TCE levels (which exceeded the 1978
TLV of 100 ppm) and showed a mean increase in
reaction time, with a probable cumulative effect.
In TCE + solvent group, ambient TCE was lower
(did not exceed 100 ppm) and mean reaction time
shortened in Session 2, then rose subsequently to
be greater than at the start. Both control groups
showed a shortening in mean choice reaction time
in Session 2 which was sustained in Sessions 3 and
4 consistent with a practice effect; No stats
provided.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Hirsch et al.
106 residents of
Random testing of the
Medical, neurologic,
Student t-test, Chi
66 subjects (62%) complained of headaches,
(1996)
Roscoe, a community
wells between 1983-84
and psychiatric exams
square analysis,
Diagnosis of TCE-induced cephalagia was

in Illinois on the Rock
revealed groundwater in
and histories. For those
nonparametric t-test
considered credible for 57 patients (54%).

River, in direct
wells to have levels of
who complained of
and ANOVA,
Retrospective TCE level of well water or well's

proximity to an
TCE between 0 to 2,441
headaches, a detailed
correlating all
distance from the industrial site analysis did not

industrial plant that
ppb; distance of
headache history was
history, physical
correlate with the occurrence of possibly-TCE

released an unknown
residence from well
taken, and an extensive
exam findings, test
induced headaches. Studies that were not

amount of TCE into
used to estimate
exam of nerve-threshold
data, TCE levels in
statistically significant with regard to possible

the River. All
exposure level.
measurements of toes,
wells, and distance
TCE-cephalalgia included P300, FFT, VER,

involved in litigation.

fingers, face, olfactory
from plant.
BAER, MMPI, MCMI, Beck Depression

Case series report; No

threshold tests for

Inventory, SSER, and nerve threshold

unexposed controls.

phenylethyl methylethyl

measurements. Headache might be associated



carbinol, brain map,

with exposure to TCE at lower levels than



Fast Fourier Transform

previously reported. Headaches mainly occurred



(FFT), P300 Cognitive

without sex predominance, gradual onset,



auditory evoked

bifrontal, throbbing, without associated features;



response, EEG, Visual

No quantitative data presented to support



Evoked Response

statement of headache in relation to TCE exposure



(VER), Somato sensory

levels, except for incidences of headache reporting



Evoked Potential

and measured TCE levels in wells.



(SSER), Brainstem





Auditory Evoked





Response (BAER),





MMPI-II, MCMI-II,





and Beck Depression





Inventory were also





given.


-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Kilburn and
Group A: Randomly
No exposure or
Reaction speed using a
Box-Cox
Group A: SRT = 282 ms CRT = 532 ms
Thornton
selected registered
groundwater analyses
timed computer visual-
transformation for
Group B: SRT = 269 ms CRT = 531 ms
(1996)
voters from Arizona
reported.
stimulus generator;
dependent and
Group C: SRT = 334 ms CRT = 619 ms

and Louisiana with no

Compared groups to
independent
Lg(SRT) = 5.620, SD = 0.198

exposure to TCE:

plotted measured SRT
variables.
Regression equation for Lg(CRT) = 6.094389 +

n = 264 unexposed

and CRT Questionnaire
Evaluated graphical
0.0037964 x age. TCE exposure produced a step

volunteers aged

to eliminate those
methods to study
increase in SRT and CRT, but no divergent lines.

18-83: Group B

exposed to possibly
residual plots.
Coefficients from Group A were valid for Group

volunteers from

confounding chemicals.
Cooks distance
B. Predicted value for SRT and for CRT, plus 1.5

California n = 29 17


statistic measured
SDs. selected 8% of the model group as abnormal.

males and 12 females


influence of outliers


to validate the


examined. Lack-


equations; Group C


of-fit test


exposed to TCE and


performed on Final


other chemicals


model and


residentially for 5 yrs


F statistic to


or more n = 231.


compare estimated





error to lack-of-fit





component of the





model's residual





sum of squared





error. Final models





were validated





using Group B data





and paired t-test to





compare observed





values for SRT and





CRT. F statistic to





test hypothesis that





parameter estimates





obtained with





Hmnn R u prp

-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Kilburn and
Well-water exposed
Well-water was
Neurobehavioral testing
Two sided student
Exposed subjects had lower intelligence scores and
Warshaw
subjects to 6 to 500
measured from 1957 to
- augmented NBT; Eye
t-test with a
more mood disorders.
(1993)
ppb of TCE for 1 to
1981 by several
Closure and Blink using
p < 0.05.


25 yrs; 544 recruited
governmental agencies,
EMG;

NPH: Significant impairments in sway speed with

test subjects; Group 1
and average annual TCE
neuropsychological
Linear regression
eyes open and closed, blink reflex latency (R-l),

= 196 exposed family
exposures were
(NPS) test - Portions of
coefficients to test
eye closure speed, and two choice visual reaction

members of subjects
calculated and then
Wechsler's Memory
how demographic
time.

with cancer or birth
multiplied by each
Scale, and WAIS and
variables or other


defects; Group 2 = 178
individual's years of
embedded figures test,
factors may
NPS: Significant impairments in Culture Fair

from exposed families
residence for
grooved pegboard, Trail
contribute.
(intelligence) scores, recall of stories, visual recall,

without cancer or birth
170 subjects.
Making A and B,

digit span, block design, recognition of fingertip

defects; Group 3

POMS, and Culture Fair

numbers, grooved pegboard, and Trail Making A

= 170 exposed parents

Test;

andB.

whose children had

neurophysiological



birth defects and

(NPH) testing - Simple

POMS: all subtests, but the fatigue, were elevated

rheumatic disorders;

visual reaction time,

Mean speeds of sway were greater with eyes open

Controls: 68 referents

body balance apparatus,

atp< 0.0001) and with eyes closedp < 0.05) in

and 113 histology

cerebellar function,

the exposed group compared to the combined

technicians (HTs)

proprioception, visual,

referents. The exposed group mean simple

without environmental

associative links and

reaction time was 67 msec longer than the referent

exposure to TCE.

motor effector function.

group p < 0.0001). Choice reaction time (CRT) of
the exposed subjects was 93 msec longer in the
third trial (p < 0.0001) than referents. It was also
longer in all trials, and remained significantly
different after age adjustment. Eye closure latency
was slower for both eyes in the exposed and
significantly different (p< 0.0014) on the right
compared to the HT referent group.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Kilburn
236 residents
Exposure estimate based
Simple reaction time,
Descriptive
The principal comparison, that was between the
(2002a)
chronically exposed to
on groundwater plume
choice reaction time,
statistics;
236 exposed persons and the 161 unexposed

TCE and associated
based on contour
Balance sway speed
ANCOVA; step-
regional controls, revealed 13 significant

solvents, including
mapping; concentrations
(with eyes open and
wise adjustment of
differences (p < 0.05). SRTs and CRTs were

DCE, PCE, and vinyl
between 0.2-10,000 ppb
eyes closed), color
demographics.
delayed. Balance was abnormal with excessive

chloride, in the
of TCE over a 64 km2
errors, blink reflex

sway speed (eyes closed), but this was not true

environment from a
area; additional
latency, Supra orbital

when both eyes were open. Color discrimination
Kilburn
nearby microchip
associated solvents,
tap (left and right),

errors were increased. Both right and left blink
(2002a)
plant, some involved
including DCE, PCE,
Culture Fair A,

reflex latencies (R-l) were prolonged. Scores on
(continued)
in litigation, prior to
and vinyl chloride, No
Vocabulary, Pegboard,

Culture Fair 2A, vocabulary, grooved pegboard

1983 and those who
air sampling.
Trail Making A and B,

(dominant hand), trail making A and B, and verbal

lived in the area

Immediate verbal recall,

recall (i.e., memory) were decreased in the

between 1983 and

POMS; Pulmonary

exposed subjects.

1993 during which

Function;

Litigation is suggested but not stated and study

time dumping of

The same examiners

paid by lawyers.

chlorinated solvents

who were blinded to the

Litigation status may introduce a bias, particularly

had supposedly ceased

subjects' exposure

if no validity tests were used.

and clean-up activities

status examined the



had been enacted;

Phoenix group, but the



Controls: 67 referents

Wickenburg referents'



from northeast

status was known to the



Phoenix, who had

examiners. Exact order



never resided near the

or timing of testing not



2 plants (mean

stated.



distance = 2,000 m,





range




-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
to
o
VO
= 1,400-3,600 mfrom
plants) and
161 regional referents
from Wickenburg, AZ
up-wind of Phoenix,
recruited via random
calls made to numbers
on voter registration
rolls, matched to
exposed subjects by
age and years of
education, records
showed no current or
past water
contamination in the
areas.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Kilburn
236 residents exposed
No discussion of
Simple reaction time,
Descriptive
Insignificant effects of longer duration of
(2002a)
environmentally from
exposure assessment
choice reaction time,
statistics,
residence. No effect of proximity and litigation.

a nearby microchip
methods and results.
Balance sway speed
Regression
Effects of longer duration of residence modest and

plant (exact number of
Solvents included TCE,
(with eyes open and
analysis; Similar
insignificant. No effect of proximity. No

litigants not stated);
DCE, PCE, and vinyl
eyes closed), color
study to the one
litigation effect. Zone A-100 clients were not

156 individuals
chloride; concluded
errors, blink reflex
reported above with
different from the 9 nonclients.

exposed for >10 y
exposure is primarily
latency, Supra orbital
the exception of
Zone B, nonclients were more abnormal in color

compared to 80
due to groundwater
tap (left and right),
looking at the
different than clients and right-sided blink was less

individuals <10 y of
plume rather than air
Culture Fair A,
effects of duration
abnormal in nonclients.

exposure; Controls: 58
releases.
Vocabulary, Pegboard,
of residence,
Zone C, 9 of the 13 measurements were not

nonclaimants in 3

Trail Making A and B,
proximity to the
significantly different.

areas within exposure

Immediate verbal recall,
microchip plant,
26 of the original 236 subjects re-tested in 1999:

zone (Zones A, B,

POMS.
and being involved
maintained impaired levels of functioning and

and C).


in litigation.
mood; No tests of effort and malingering used,
limiting interpretations.
Again, no tests of effort and malingering were
used, thus, limiting interpretation.
Litigation is suggested but not stated and study
paid by lawyers.
Litigation status may introduce a bias, particularly
if no validity tests were used.
>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Landrigan et
13 Pennsylvania
Community Evaluation:
Community evaluation,
Descriptive
Community Evaluation: No urinary TCA detected
al. (1987)
residents exposed
Nov 1979-
occupational
statistics
in community population except for 1 resident also

through drinking and
Questionnaires on TCE
evaluations; urine

working at plant and 1 resident with no exposure;

bathing water
and other chemical
evaluations for TCE

Occupational Evaluation: Range 117-357 mg/m3-

contaminated by
exposures, and
metabolites;

(21-64 ppm).

approximately 1,900
occurrence of signs and
Questionnaires to

Feb: airborne exposures exceeded NIOSH limit by

gallon TCE spill; Feb
symptoms of exposure
evaluate neurologic

up to 222 mg/m3 (40 ppm)(NIOSH TWA <135

1980: 9 workers
to TCE, morning urine
effects and symptoms;

mg/m3).

exposed to TCE while
samples, urine samples
ISO concentrations,

(24 ppm). Short term exposure exceeded NIOSH

degreasing metal in
analyzed
Map of TCE in

values of 535 mg/m3 (96 ppm) by up to 1,465

pipe manufacturing
coloreimetrically for
groundwater.

mg/m3 (264 ppm).

plant and 9 unexposed
total trichloro-


Personal breathing zone of other workers within

controls (mean ages
compounds.


recommended limits (0.5-125 mg/m3) (0.1-23

were 42.7 exposed and



ppm).

46.4-y old unexposed;
Occupational


7 exposed workers reported acute symptoms,

mean durations of
Evaluations (In


including fatigue, light-headedness, sleepiness,

employment = 4.4,
workers): breathing-


nausea, headache, consistent with TCE exposure;

exposed, and 9.4 y,
zone air samples( mean


No control workers reported such symptoms;

unexposed.;
205 mg/m3; 37 ppm);


Prevalence of 1 or more symptoms 78% in

May 1980: 10 exposed
medical evaluations, pre


exposed worker group, 0% in control worker

workers and same 9
and post shift spot urine


group; Symptoms decreased after

unexposed worker
samples in Feb and


recommendations were in place for 3 mos (may

controls from Feb
again in May, mid and


testing) for reduced exposures.

monitoring.
post shift venous blood
samples during the May
survey,



-------
>3
o
o
s
Co
I?
°s
VO
S"4
>}'
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
§-


Exposure assessment



a
Reference
Study population
and biomarkers
Tests used
Statistics
Results

Liu et al.
103 workers from
Exposed to TCE, mostly
Self-reported subjective
Prevalence of
Dose-response relationship established in
s*
£5
(1988)
factories in Northern
at less than 50 ppm;
symptom questionnaire.
affirmative
symptoms such as nausea, drunken feeling, light-
S*

China, exposed to TCE
concentration of

answers = total
headedness, floating sensation, heavy feeling of
<§>

(79 men, 24 women),
breathing zone air

number of
the head, forgetfulness, tremors and/or cramps in

during vapor
during entire shift

affirmative answers
extremities, body weight loss, changes in


degreasing production
measured by diffusive

divided by (number
perspiration pattern, joint pain, and dry mouth (all
'

or operation. The
samplers placed on the

of respondents x
>3 times more common in exposed workers);
t:
^3

unexposed control
chest of each worker;

number of
"bloody strawberry jam-like feces" was borderline
St
"3

group included 85 men
divided into three

questions); X2.
significant in the exposed group and"'frequent
8
to

and 26 women.
exposure groups;


flatus" was statistically significant. Exposure
Co
O


1-10 ppm, 11-50 ppm


ranged up to 100 ppm, however, most workers
S


and 51-100 ppm; Also,


were exposed below 10 ppm, and some at 11-50
1


hematology, serum


ppm. Contrary to expectations, production plant
a.


biochemistry, sugar,


men had significantly higher levels of exposure
o


protein, and occult


(24 had levels of 1-10 ppm, 15 had levels of
to
o


blood in urine were


11-50 ppm, 4 had levels of 51-100 ppm) than
o


collected.


degreasing plant men (31 had levels of 1-10 ppm,
2 had levels of 11-50 ppm, 0 had levels of 51-100
ppm); p < 0.05 by chi-square test. No significant
difference (p> 0.10) was found in women
workers. The differences in exposure intensity
between men and women was of borderline
significance (0.05 
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
McCunney
This is a case study
Case 1: TCE in air at the
Clinical evaluation of
There were no
Case 1 was a 25-yr old male, who presented with a
(1988)
conducted on 3 young
work place was
loss of balance, light
statistical analyses
loss of balance, light headedness, resting tremor,

white male workers
measured at 25 ppm, but
headedness, resting
of results presented.
blurred vision, and dysdiadochokinesia. The

exposed to TCE in
his TCA in urine was
tremor, blurred vision,

subject had been in a car accident and suffered

degreasing operations.
measured at 210 mg/L.
and

head injuries. He later returned with a change in

There were no controls
This is likely due to
dysdiadochokinesia,

demeanor and loss of coordination. He showed a

included. Case 1 was
dermal exposure while
change in demeanor and

normal CT scan, EEG, nerve conductivity, and

a 25-yr old male, Case
cleaning metal rods in
loss of coordination,

visual and somatosensory evoked response.

2 was a 28-yr old
TCE. Case 2: no TCE
cognitive changes were

Neurological exams revealed reduced sensitivity to

white male, Case 3
exposure data presented,
noted, as well as

pinprick over the face, deep tendon reflexes were

was a 45-yr old white
TCA at 9 mg/L after
depression; CT scan,

reduced, mild to moderate cognitive changes were

male.
6 mos; Case 3: no TCE
EEG, nerve

noted, as well as depression. Ophthalmic


exposure data presented.
conductivity, and visual

evaluation was normal. He was removed form the



and somatosensory

TCE exposure and appeared to recover.



evoked response.





Neurological exams

Case 2 was a 28-yr old white male who presented



included sensitivity to

with numbness and shooting pains in fingers. He



pinprick over the face;

exhibited anorexia, tiredness. He worked in a



Ophthalmic evaluation.

degreasing operation for a jeweler using open





containers filled with TCE in a small, unventilated





room. There were no exposure data provided, but





his TCA was 9 mg/L at 6 mos after exposure. He





had been hospitalized with hepatitis previously.





No neurological tests were administered.





Case 3 was a 45-yr old white male who presented





with numbness in hands and an inability to sleep.





He exhibited slurred speech. He was positive for





blood in stool, but had a history of duodenal





ulcers.
-------

§•
to
s
a
to
s
Co
a,
>r
>S
>S
to'
*
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Mhiri et al.
(2004)
23 phosphate industry
workers exposed to
TCE for 6 h/d for at
least 2 yrs while
cleaning walls to be
painted; Controls:
23 unexposed workers
from the department of
neurology.
Measurement of urinary
metabolites of TCE
were performed 3
times/worker. Blood
tests and hepatic
enzymes were also
collected.
Trigeminal
somatosensory evoked
potentials recorded
using Nihon-Kohden
EMG- evoked potential
system; baseline clinical
evaluations regarding
facial burn or
numbness, visual
disturbances,
restlessness,
concentration difficulty,
fatigue, mood changes,
assessment of cranial
nerves, quality of life;
biological tests
described under
biomarkers.
Paired or unpaired
Student's t-test as
appropriate.
/>-value set at
<0.05. Spearman
rank-correlation
procedure was used
for correlation
analysis.
Abnormal TSEP were observed in 6 workers with
clinical evidence of Trigeminal involvement and in
9 asymptomatic workers. A significant positive
correlation between duration of exposure and the
N2 latency (p < 0.01) and P2 latency (p < 0.02)
was observed. Only one subject had urinary TCE
metabolite levels over tolerated limits. TCE air
contents were over tolerated levels, ranging from
50-150 ppm.
Mitchell and
Parsons-
Smith
(1969)
This was a case study
of 1 male patient, age
33, occupational
exposed to TCE during
degreasing. There
were no controls.
No exposure data are
presented.
Trigeminal nerve, loss
of taste, X-rays of the
skull, EEG,
hemoglobin, and
Wassermann reaction.
No statistics
provided.
The patient had complete analgesia in the right
trigeminal nerve and complete loss of taste, patient
complained of loss of sensation on right side of
face, and uncomfortable right eye, as well as
vertigo and depression. X-rays of the skull, EEG,
hemoglobin, and Wassermann reaction were all
normal.
-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
to
o
VO
Nagaya et al.
(1990)
84 male workers ages
18-61 (mean 36.2)
constantly using TCE
in their jobs. Duration
of employment (i.e.,
exposure)
0.1-34.0 yrs, (mean
6.1 yrs; SD = 5.9).
Controls:
83 age-matched office
workers and students
with no exposure.
Workers exposed to
about 22-ppm TCE in
air. Serum dopamine-(3-
hydroxylase (DBH)
activity levels measured
from blood. Urinary
total trichloro-
compounds (U-TTC)
also measured.
Blood drawn during
working time and DBH
activities were
analyzed; Spot urine
collected at time of
blood sampling and
U-TTC determined by
alkaline-pyridine
method.
Student's t-test and
linear correlation
coefficient. Results
of U-TTC
presented by age
groups: <25;
26-40; >41.
A slight decrease in serum DBH activity with age
was noted in both groups. Significant inverse
correlation of DBH activity and age was found in
workers (r = -0.278, 0.01 
-------
IS*
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)




Exposure assessment




a
Reference
Study population
and biomarkers
Tests used
Statistics
Results


Reif et al.
143 residents of the
Hydraulic simulation
NCTB, tests of visual
Multivariate Model.
Statistical significance was approached as a result
o
s*
s
(2003)
Rocky Mountain
model used in
contrast sensitivity,

of high TCE exposure vs. referent group; poorer
o
s*

Arsenal community of
conjunction with a GIS
POMS.

performance on the digit symbol (p = 0.07),
s
Co


Denver whose water
estimated residential


contrast sensitivity C test (p = 0.06), and contrast
1?
>{

was contaminated with
exposures to TCE;


sensitivity D test (p = 0.07), and higher mean

—»

TCE and related
Approximately 80% of


scores for depression (p = 0.08). Alcohol was an
x


chemicals from nearby
the sample exposed to


effect modifier in high-exposed individuals—

t:
^3

hazardous waste sites
TCE exceeding MCL of


statistically significant on the Benton, digit
o
St
"3

between 1981 and
5 ppb and


symbol, digit span, and simple reaction time tests,
t
8
to

1986; Referent group
approximately 14%


as well as for confusion, depression, and tension.
Co
0
S
1

at lowest concentration
(<5 ppb).
exceeded 15 ppb. High
exposure group
>15 ppb, low exposure



VO
a.


referent group <5 ppb,



o
to
o


medium exposure group
5 < x < 15 ppb.



-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Rasmussen
368 metal workers
Questionnaire:
Questionnaire: 74 items
Chi-square; Odds
Neuropsychological symptoms significantly more
and Sabroe
working in degreasing
categorized in 4 groups;
about
ratios; t-test;
prevalent in the chlorinated solvents-exposed
(1986)
at various factories in
3 exposure groups plus
neuropsychological
logistic regression.
group; TCE caused the most "inconveniences and

Denmark (industries
control: (1) currently
symptoms (memory,

symptoms;" dose response between exposure to

not specified) with
working with
concentration,

chlorinated solvents and chronic

chlorinated solvents;
chlorinated solvents
irritability, alcohol

neuropsychological symptoms (memory

94 controls randomly
(n = 171; average.
intolerance, sleep

\p < 0.001], concentration \p < 0.02], irritability

selected semiskilled
duration: 7.3 yrs,
disturbance, fatigue).

\p < 0.004], alcohol intolerance [p < 0.004],

metal workers from
16.5 h/wk; 57% TCE


forgetfulness \p < 0.001], dizziness \p < 0.005],

same area; mean age:
and 37%


and headache \p < 0.01]); Significant associations

37.7 (range: 17-65+).
1,1,1-trichloroethane),


between previous exposure and consumption of

Total 443 men; 19
(2) currently working


alcohol with chronic neuropsychological

women.
with other solvents


symptom, s


(n = 131; petroleum,





gasoline, toluene,





xylene), (3) previously





(1-5 yrs) worked with





chlorinated or other





solvents (n = 66) (4)





never worked with





organic solvents





as
II



>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Rasmussen
96 Danish workers
Chronic exposure to
Medical interview,
Fisher's exact test,
After adjusting for confounders, the high exposure
et al.
involved in metal
TCE (n = 70); CFC
neurological exam,
Chi-square trend
group has significantly increased risk for
(1993d)
degreasing with
(« = 25); HC (n = 1);
neuropsychological
test, t-test,
psychoorganic syndrome following exposure (OR:

chlorinated solvents,
average duration: 7.1
exam; Tests: WAIS:
ANOVA, logistic
11.2); OR for medium exposed group = 5.6;

mostly TCE (n = 70);
yrs; range of full-time
Vocabulary, Digit
regression, odds
Significant increase in risk with age and with

(industries not
degreasing: 1 mo to
Symbol; Simple
ratios, Chi-square
decrease in WAIS Vocabulary scores; Prevalence

specified), age range:
36 yrs; occupational
Reaction Time,
goodness-of-fit test;
of psychoorganic syndrome: 10.5% in low

19-68; no external
history, blood and
acoustic-motor
Confounders
exposure group, 38.9 in medium exposure group,

controls.
urinary metabolites
function, discriminatory
examined: age,
63.4% in high exposure group; no significant


(TCA); biological
attention, Sentence
primary intellectual
interaction between age and solvent exposure.


monitoring for TCE and
Repetition, Paced
level,



TCE metabolites; CEI
Auditory Serial
arteriosclerosis,



calculated based on
Addition Test, Text
neurological/psychi



number of h/wk worked
Repetition, Rey's
atric disease,



with solvents x yr of
Auditory Verbal
alcohol abuse, and



exposure x 45 wk per
Learning, visual gestalt,
present solvent



yr; 3 groups: (1) low
Stone Pictures
exposure.



exposure: n = 19,
(developed for this




average full-time
study, nonvalidated),




exposure 0.5 yr; (2)
revised Santa Ana,




medium exposure:
Luria motor function,




n = 36, average full-time
Mira; Blind study.




exposure 2.1 yrs.; (3)





high exposure: n = 41,





average full-time





exposure 11 yrs; Mean





TCA in high exposure





group = 7.7 mg/L (max





= 26.1 mg/L); TWA





measurements of CFC





113 levels:





9^0— zL90 rmm m Q nnH



-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Rasmussen
96 Danish workers
Chronic exposure to
WAIS (original
Linear regression
Dose response with 9 of 15 tests; Controlling for
et al.
involved in metal
TCE (n = 70); CFC
version): Vocabulary,
analysis;
confounds, significant relationship of exposure
(1993c)
degreasing with
(« = 25); HC (n = 1);
Digit Symbol, Digit
Confounding
was found with Acoustic-motor function

chlorinated solvents
average duration: 7.1
Span; Simple Reaction
variables analyzed:
(p < 0.001), PASAT (p < 0.001), Rey AVLT

(industries not
yrs); range of full-time
Time, Acoustic-motor
age, primary
(p < 0.001), vocabulary (p < 0.001), and visual

specified), age range:
degreasing: 1 mo to 36
function (Luria),
intellectual
gestalts (p < 0.001); significant age effects.

19-68; No external
yrs; occupational
Discriminatory
function, word


controls.
history, blood and
attention (Luria),
blindness,



urinary metabolites
Sentence Repetition,
education,



(TCA); biological
PASAT, Text
arteriosclerosis,



monitoring for TCE and
Repetition, Rey's
neurological/psychi



TCE metabolites; CEI
Auditory Verbal
atric disease,



calculated based on
Learning, Visual
alcohol use, present



number of h/wk worked
Gestalts, Stone Pictures
solvent exposure.



with solvents x yr of
(developed for this




exposure x 45 wks per
study, nonvalidated),




yr; 3 groups: (1) low
revised Santa Ana,




exposure: n = 19,
Luria motor function,




average full-time expo
Mira; Blind study.




0.5 yr; (2) medium





exposure: n = 36,





average full-time





exposure 2.1 yrs; (3)





high exposure: n = 41,





average full-time





exposure 11 yrs; Mean





TCA in high exposure





group = 7.7 mg/L (max





= 26.1 mg/L); TWA





measurements of CFC





113 levels: 260-420





rvnm fTT Q nnH



-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Rasmussen
96 Danish workers
Chronic exposure to
Medical interview,
Multiple
Significant dose response between exposure and
et al.
involved in metal
TCE (n = 70); CFC
clinical neurological
regression; Fisher's
motor dyscoordination remained after controlling
(1993a)
degreasing with
(« = 25); HC (n = 1);
exam,
exact test; Mantel-
for confounders; Bivariate analysis showed

chlorinated solvents
average duration: 7.1
neuropsychological
Haenzel test for
increased vibration threshold with increased

(industries not
yrs); range of full-time
exam.
linear association.
exposure, but with multivariate analysis, age was a

specified), age range:
degreasing: 1 mo to 36


significant factor for the increase.

19-68; No external
yrs; occupational




controls.
history, blood and





urinary metabolites;





biological monitoring





for TCE and TCE





metabolites; CEI





calculated based on





number of h/wk worked





with solvents x yr of





exposure x 45 wk per





yr; 3 groups: (1) low





exposure: n = 19,





average full-time expo





0.5 yr; (2) medium





exposure: n = 36,





average full-time





exposure 2.1 yrs; (3)





high exposure: n = 41,





average full-time





exposure 11 yrs; Mean





TCA in high exposure





group = 7.7 mg/L (max





= 26.1 mg/L); TWA





measurements of CFC





113 levels:





9^0— zL90 rmm fTT Q nnH



-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Ruijten et al.
31 male printing
Relied on exposure data
General questionnaire,
Combined Z score
Slight reduction in Sural nerve conduction velocity
(1991)
workers exposed to
from past monitoring
cardiotachogram
= individual Z
was found and a prolongation of the Sural

TCE. Mean age 44;
activities conducted by
recorded on ink writer
scores of the FRSA
refractory period. Latency of the masseter reflex

Mean duration 16 yrs;
plant personnel using
to measure Autonomic
and MHR;
had increased. No prolongation of the blink reflex

Controls: 28; mean age
gas detection tubes.
nerve function,
ANCOVA to
was found; no impairment of autonomic or motor

45 yrs.
Estimated 17 ppmfor
including forced
calculate difference
nerve function were found. Long term exposure to


past 3 yrs, 35 ppm for
respiratory sinus
between
TCE at threshold limit values (approximately


preceding 8 yrs and
arrhythmia (FRSA),
exposed/nonexpose
35 ppm) may slightly affect the trigeminal and


70 ppm before that.
muscle heart reflex
d workers;
sural nerves.


Individual cumulative
(MHR), resting
Cumulative



exposure was calculated
arrhythmia; Trigeminal
exposure effect



as time spent in
nerve function
calculated by



different exposure
measured using
multiple linear



periods and the
masseter reflex and
regression analysis.



estimated exposure in
blink reflex;
Controlled for age,



those periods. Mean
electrophysiological
alcohol



cumulative exposure
testing of peripheral
consumption, and



= 704 ppm x yrs (SD
nerve functioning using
nationality by



583, range:
motor nerve conduction
including them as



160-2,150 ppm x yrs.
velocity of the peroneal
covariables.




nerve.
Quetelet-index





included for





autonomic nerve





parameters; Body





length and skin





temperature used





for all peripheral





nerve functions;





one-sided





significance level





of 5% used. Non-





nnrmcil

-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
to
o
VO
Smith
(1970)
130 (108 males,
22 females); Controls:
63 unexposed men
working at the same
factory matched by
age, marital status.
TCA metabolite levels
in urine were measured:
60.8% had levels up to
20 mg/L, and 82.1% had
levels up to 60 mg/L.
Cornell Medical Index
Questionnaire
(Psychiatric section),
Heron's Personality
Questionnaire, Fluency
Test, 13-Mistake Test,
Serial Sevens, Digit
Span, General
Knowledge Test, tests
of memory.
Descriptive
Statistics.
Of the 130 subjects exposed 27% had no
complaints of symptoms, 74.5% experienced
fatigue, 56.2% dizziness, 17.7% headache, 25.4%
gastro-intestinal problems, 7.7% autonomic
effects, and 24.9% had other symptoms. The
number of complaints reported by subjects were
statistically significant between those with 20
mg/L or less TCA (M = 1.8 complaints) and those
60 mg/L or more (M = 2.7). Each group, however,
had a similar proportion of subjects who reported
having only 'slight' symptoms. The total time of
continuous exposure to TCE (ranging from less
than 1 yr to more than 10 yrs) appeared to have
little influence on frequency of symptoms. No
results of the tests are reported; Author postulates
that symptom assessment raises the possibility of
"errors of subjective judgment."
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Triebig et al.
This study was
Measured TCE in air
Results were compared
Wilcoxon and
Mean values observed were 330-mg
(1977b)
conducted on
averaged 50 ppm
after exposure periods,
Willcox
trichloroethanol and 319-mg TCA/g creatinine,

8 subjects
(260 mg/m3). Length of
and compared to results
nonparametric tests.
respectively, at the end of a work shift. The

occupationally
occupational exposure
obtained after periods
Due to the small
psychological tests showed no statistically

exposed to TCE.
was not reported.
removed from
sample size a
significant difference in the results before or after

Subjects were 7 men

exposure. TCAand
significance level
the exposure-free time period.

and 1 woman with an

TCE metabolites in
of 1% was used.


age range from 23-38

urine and blood were



yrs. There was no

measured.



control group.

Psychological tests





included d2, MWT-A,





and short test.


>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
VO
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Triebig
This study was
Length of exposure
Nerve conduction
Data were analyzed
Results show no statistically significant difference
(1982)
conducted on
ranged from 1 to
velocities were
using parametric
in nerve conduction velocities between the

24 healthy workers (20
258 mos (mean 83 mos).
measured for sensory
and nonparametric
exposed and unexposed groups. This study has

males, 4 females)
TCE concentrations
and motor nerve fibers
tests, rank
measured exposure data, but exposures/responses

exposed to TCE
measured in air at work
using the following
correlation, linear
are not reported by dose levels.

occupationally at three
places ranged from
tests: MCVmax(U):
regression, with 5%


different plants. The
5-70 ppm. TCA, TCE,
Maximum NLG of the
error probability.


ages 17-56; length of
and trichloroethanol
motor fibers of the N.



exposure ranged from
were measured in blood,
ulnaris between the



1 to 258 mos (mean
and TCE and TCA were
wrist joint and the



83 mos). A control
measured in urine.
elbow; dSCV (U),



group of 144 controls

pSCV (U), and dSCV



used to establish

(M).



'normal' responses on





the nerve conduction





studies. The matched





control group





consisted of 24 healthy





nonexposed





individuals (20 males,





4 females), chosen to





match the subjects for





age and sex.




>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Triebig
The exposed group
Subjects were exposed
Nerve conduction
Data were analyzed
There was a dose-response relationship observed
(1983)
consists of 66 healthy
to a mixture of solvents,
velocities were
using parametric
between length of exposure to mixed solvents and

workers selected from
including TCE,
measured for sensory
and nonparametric
statistically significant reduction in nerve

a population of
specifically "ethanol,
and motor nerve fibers
tests, rank
conduction velocities observed for the medium and

112 workers. Workers
ethyl acetate, aliphatic
using the following
correlation, linear
long-term exposure groups for the NCV.

were excluded based
hydrocarbons
tests: MCVmax(U):
regression, with 5%


on polyneuropathy
(gasoline), MEK,
Maximum NLG of the
error probability.


(n = 46) and alcohol
toluene, and
motor fibers of the N.



consumption (n = 28).
trichloroethene."
ulnaris between the



The control group
Subjects were divided
wrist joint and the



consisted of 66 healthy
into 3 exposure groups
elbow; dSCV (U),



workers with no
based on length of
pSCV (U), and dSCV



exposures to solvents.
exposure, as follows: 20
(M).




employees with "short-





term exposure" (7-24





mos); 24 employees





with "medium-term





exposure" (25-60 mos);





22 employees with





"long-term exposure"





(over 60 mos). TCA,





TCE, and





trichloroethanol were





measured in blood, and





TCE and TCA were





measured in urine.



-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
Troster and
Ruff (1990)
3 occupationally
exposed workers to
TCE or TCA:
2 patients acutely
exposed to low levels
of TCE and 1 patient
exposed to TCA;
Controls: 2 groups of
n = 30 matched
controls; (all age and
education matched).
"Unknown amount of
TCE for 8 months."
SDNTB, "1 or more
of:" TAT, MMPI,
Rorschach, and
Interviewing
questionnaire, Medical
examinations (including
neurological, CT scan,
and/or Chemo-
pathological tests and
occupational history).
Not reported.
Case 1: Intelligence "deemed" to drop from
premorbid function at 1 y 10 mos after exposure.
Impaired functions improved for all but reading
comprehension, visuospatial learning and
categorization (abstraction). Case 2: Mild deficits
in motor speed, verbal learning, and memory;
"marked" deficits in visuospatial learning; good
attention; diagnosis of mild depression and
adjustment disorder, but symptoms subsided after
removal from exposure. Case 3: Manual dexterity
and logical thinking borderline impaired; no
emotional changes, cognitive function spared,
diagnosis of somatoform disorder.
vo
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
White et al.
Group 1:
Group 1: 2 wells tested
Occupational and
Data shown in
Group 1: Some individuals with subclinical
(1997)
28 individuals in
in 1979: 267 ppb TCE,
environmental
proportion in 3
peripheral neuropathy; 92.8% with reflex

Massachusetts exposed
21 ppb
questionnaire,
communities,
abnormalities; 75% total diagnosed with peripheral

to contaminated well
Tetrachloroethylene,
neurological exam,
clinical diagnostic
neuropathy; 88.9% with impairment in at least 1

water; source: tanning
12 ppb chloroform,
neuropsychological
categories, analysis
memory test; Impairments: attention and executive

factory and chemical
29 ppb dichloro-
exam: WAIS-R,
of central
function in 67.9%; motor function in 60.71%,

plant; age range: 9-55.
ethylene, 23 ppb
WISC-R, WMS,
tendencies, and
visuospatial in 60.71%, mild to moderate

Group 2:
T richlorotrifluoroethane
WMS-R, Wisconsin
descriptive
encephalopathy in 85.7%.

12 individuals in Ohio
; 2 yrs average TCE
Card Sorting, CO WAT,
statistics.
Group 2: 25% with abnormal nerve conduction;

exposed to
256 ppb for well G, and
Boston Naming, Boston

Impairments: attention and executive function in

contaminated well
111 ppb for well H.
Visuo spatial

83.33%, memory in 58.33%, language/verbal in

water; source:
Group 2:13 wells with
Quantitative Battery,

50%.

degreasing; age range:
1,1,1 -trichloroethane
Milner Facial

Group 3: 35.7% with peripheral neuropathy;

12-68 Group 3:
(up to 2,569 ppb) and
Recognition Test,

neuropsychological: all 6 tested had memory

20 individuals in
TCE (up to 760 ppb);
Sticks Visuospatial

impairment, attention and executive function

Minnesota exposed to
blood analysis of
Orientation Task, Word

impairment, 3 had manual motor slowing.

contaminated well
individuals 2 yrs after
triads, Benton Visual

Participants younger at time of exposure with

water; n = 14 for nerve
end of exposure and
Retention Test, Santa

wider range of deficits; Language deficits in

conduction studies and
soon after exposure
Ana, Albert's Famous

younger, but not in older participants.

n = 6 for
showed normal or mild
Faces, Peabody Picture



neuropsychological
elevations of TCE,
Vocabulary Test,



testing; source:
elevations of
WRAT, POMS, MMPI,



ammunition plant; age
1,1,1-trichloroethane,
Trail-making,



range: 8-62. No
ethylbenzene, and
Fingertapping, Delayed



controls.
xylenes. Group 3: mean
Recognition Span Test;




TCE for one well
Neurophysiological




261 ppb;
exam: eyeblink, evoked




1,1 -dichloroethylene
potentials, nerve




9.0 ppb;
conduction; Other:




1,2-dichloroethylene
EKG, EEG, medical




107 ppb.
tests.


-------

§•
to
s
a
to
s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
s
TO
Co
0
S
1
a.
to
o
VO
Winneke
(1982)
This is a review article
presenting multiple
studies that evaluated
neurological effects of
TCE, and other
solvents. Only the
TCE results are
summarized herein.
Experiment 1: 18
subjects [results taken
from Schlipkoter et al.
(1974) and summary is
based on informations
from Winneke (1982)]
Experiment 2: 12
subjects [results taken
from Winneke et al.
(1978) (1978; 1976)
and summary is based
on information from
Winneke (1982)]
Experiment 1: Subjects
were exposed to 50 ppm
TCE for 3.5 hours
Experiment 2:
Comparative study of
effects from (a) 50 ppm
TCE for 3.5 hours and
(b) 0.76 ml/kg ethanol.
For both experiments 1
and 2: critical flicker
fusion, sustained
attention task, auditory
evoked potentials
No statistical
details were
reported.
Significant decrease (p < 0.05) in auditory evoked
potentials in individuals (experiments 1 and 2)
exposed to 50 ppm TCE. No significant effects
were noted in the critical flicker fusion or the
sustained attention tasks.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
ATSDR
116 children from
Exposures were
Fisher Logemann test;
Screening results as
Exposed children had higher abnormalities for D-
(2003b)
registry of
modeled using tap water
OSME-R; CSP;
binary variables
COME-T (p < 0.002), CSP (p < 0.008),

14 hazardous waste
TCE concentrations and
D-COME-T; hearing
using logistic
velopharyngeal function (p < 0.04), high palatal

sites with TCE in
GIS for spatial
screening; DPOAE;
regression within
arch (p < 0.04), abnormal outer ear cochlear

groundwater; under 10
interpolation, and
SCAN.
SAS; independent
function; No difference observed in exposed and

yrs of age at time of
LaGrange for temporal

variables included
nonexposed populations for speech or hearing

registry; Control
interpolation to estimate

exposure measures,
function; No difference found in OSH function.

population (n = 177);
exposures from

age, gender, case


communities with no
gestation to 1990 across

history; chi-square


evidence of TCE in
the area of subject

test, Fisher's exact


groundwater
residences, modeled

test, t-tests, linear


(measured below
data were used to

models.


MCL); matched by age
estimate lifetime




and race; there were
exposures (ppb-yrs) to




other chlorinated
TCE in residential




solvents present in the
wells; 3 exposure level




exposed group wells.
groups; control = 0 ppb;
low exposure group = 0
<23 ppb-yrs; and high
exposure group =
>23 ppb-yrs;
confounding exposure
was a concern.



>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Epidemiological Studies: Controlled Exposure Studies; Neurological Effects of Trichloroethylene
Gamberale
15 healthy men aged
Exposed for TCE 70
RT addition, simple RT,
Friedman two-way
In the RT-Addition test the level of performance
etal. (1976)
20-31-yr old
mins via a breathing
choice RT and short
analysis by ranks to
varied significantly between the different exposure

employed by the
valve to 540 mg/m3 (97
term memory using an
evaluate difference
conditions (F[2.24] = 4.35; p < 0.051) and between

Department of
ppm), 1,080 mg/m3 (194
electronic panel.
between 3
successive measurement occasions (tF[2.24] =

Occupational
ppm), and during
Subjects also assessed
conditions,
19.25; p < 0.001); The level of performance

Medicine in
ordinary atmospheric
their own conditions on
nonsignificant
declined with increased exposure to TCE, whereas

Stockholm, Sweden;
air. Sequence was
a 7-pt scale.
when tested
repetition of the testing led to a pronounced

Controls: Within
counterbalanced

individually, but
improvement in performance as a result of the

Subjects
between the 3 groups,

significant when
training effect; No significant interaction effects

(15 self-controls).
days, and exposure

tested on the basis
between exposure to TCE and training.


levels. Concentration

of 6 variables.



was measured with a

Nearly half of the



gas chromatographic

subjects could



technique every third

distinguish



min for the 1st 50 mins,

exposure/nonexpos



then between tests

ure. ANOVAfor



thereafter.

4 performance tests





based on a 3 x 3





Latin square design





with repeated





measures.

>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)


Exposure assessment



Reference
Study population
and biomarkers
Tests used
Statistics
Results
Konietzko et
This is a controlled
Subjects were exposed
Evaluated for changes

The alpha segment increased over time of
al. (1975)
exposure study
to a constant TCE
in alpha waves

exposure (from 0800 to 0900 and 1000 h [military

conducted on
concentration of 95.3
(<14 Hz) in the EEG

time]) (P = 0.05). There were no significant

20 healthy male
ppm (520 mg/m3) for up
recordings; EEG

differences for the other time spans or for other

students and scientific
to 12 h, and Blood
recordings were

parameters. Subjects with highest and lowest TCE

assistants with a mean
concentrations of TCE
performed hourly for a

blood levels <2 (ig/mL and >5 (ig/mL were

age of 27.2 yrs.
were also analyzed at
hourly intervals.
period of 1 min with the
eyes closed. This was
used as a potential
measure of
psychomotor
disturbance.

compared to determine if they showed different
responses, but no case were the differences
statistically different.
Kylin et al.
12 subjects exposed to
1,000 ppm of TCE was
Optokinetic Nystagmus;
Ostwald's
"A number" of subjects showed reduction in
(1967)
1,000 ppm TCE for 2 h
blown into a chamber
Venus blood and
distribution factor
Fusion limit although more pronounced in the 2

in a 1.5 x 2 x 2 meters
via an infusion unit and
alveolar air specimens
for TCE (the
subjects who consumed alcohol. "Others,"

chamber; 2 subjects
vaporizing system.
were taken at various
quotient of the
however, showed little if any effect. No stats.

were given alcohol
Ostwald's distribution
times after exposure
amount of solvent


(0.7 gm of body
factor for TCE—the
and analyzed in a gas
in the blood in


weight); Controls: 7 of
quotient of the amount
chromatograph with a
mg/L by the


the 12 were tested
of solvent in the blood
flame ionization
amount of the


some days prior to
by the amount of
detector.
alveolar air in


exposure and 5 of the
alveolar air.

mg/L) = 9.7;


12 were tested some


Significant


days after exposure.


relationship
between TCE in air
and blood (0.88).

-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Salvini et al.
This is a controlled
TCE concentration was
Two sets of tests were
ANOVA
A decrease in function for all measured effects was
(1971)
exposure study
110 ppmfor4-h
performed for each

observed. Statistically significant results were

conducted on 6 male
intervals, twice per day.
subject corresponding

observed for perception tests learning (p < 0.001),

university students.
0-ppm control exposure
to exposure and control

mental fatigue (p < 0.01), subjects (p < 0.05); and

Each subject was
for all as self controls.
conditions. Perception

CRT learning (p < 0.01), mental fatigue (p < 0.01),

examined on 2

test with tachistoscopic

subjects (p < 0.05).

different days, once

presentation, Wechsler



under TCE exposure,

memory scale, complex



and once as self

reaction time test



controls, with no

(CRT), and manual



exposure.

dexterity test.


>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
VO
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Stewart et al.
13 subjects in
Ten chamber exposures
Physical examination
Descriptive
Ability to perceive TCE odor diminished as
(1970)
10 experiments
to TCE vapor (100 ppm
1 h prior to exposure.
statistics.
duration of expo increased; 40% had dry throat


and 200 ppm) for
Blood analysis for

after 30 min. exposure; 20% reported eye


periods of 1 h to a 5-day
complete blood cell

irritation; Urine specimens showed progressive


work week.
count (CBC),

increase in amounts of TCE metabolites over the


Experiments 1-7 were
sedimentation rate, total

5 consecutive exposures. Concentrations of TCA


for a duration of 7 h
serum lipid, total serum

and TCE decreased exponentially after last


with a mean TCE
protein, serum

exposure, but still present in abnormal amounts in


concentration of
electrophoresis, serum

urine specimens 12 d after exposure. Loss of


198-200 ppm.
glutamic oxaloacetic

smelling TCE: >1 h = 33%; >2 h = 80%; >6.5 h


Experiments 8 and
transaminase (SGOT)

= 100%; Symptoms of lightheadedness, headache,


9 exposed subjects to
and serum glutamic

eye, nose and throat irritation. Prominent fatigue


202 ppm TCE for a
pyruvic transaminase.

and sleepiness by all after 200 ppm. These


duration of 3.5 and 1 h,
24-h urine collection for

symptoms may be of clinical significance. All had


respectively.
urobilinogen, TCA and

normal neurological tests during exposure, but


Experiment 10 exposed
TCE. Also a

50% reported greater mental effort was required to


subjects to 100 ppm
preexposure

perform a normal modified Romberg test on more


TCE for 4 h.
expirogram, tidal

than one occasion.


Experiments 2-6 were
volume measurement,




carried out with the
and an alveolar breath




same subjects over
sample for TCE; Short




5 consecutive days; Gas
neurological exam




chromatography of
including modified




expired air; No self
Romberg test, heel-to-




controls.
toe test, finger-to-nose





test.


-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Triebig
This was a controlled
Subjects exposed for 6
Psychological tests
Regression
There was no correlation seen between exposed
(1976)
exposure study
h/d for 5 d to 100 ppm
were: the d2 test was an
analyses were
and unexposed subjects for any measured

conducted on 7 healthy
(550 mg/m3 TCE).
attention load test; the
conducted.
psychological test results. The biochemical data

male and female
Controls were exposed
short test is used to

did demonstrate that exposed subjects' exposures.

students (4 females, 3
in chamber to zero TCE.
record patient



males). The control
Biochemical tests
performance with



group was 7 healthy
included TCE, TCE, and
respect to memory and



students (4 females,
trichloroethanol in
attention; dailv



3 males).
blood. In this study the
Fluctuation




TCE concentrations in
Questionnaire measured




blood reported ranged
the difference between




from 4 to 14 (ig/mL. A
mental states at the start




range of 20 to 60 (ig/mL
of exposure and after




was obtained for TCA
the end of exposure is




in the blood.
recorded; The MWT-A





is a repeatable short





intelligence test; the





Freiburg Personality





Inventory is a test for





12 independent





personality traits;





CFT-3 is a nonverbal





intelligence test;





Erlanger Depression





Scale.


-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Triebig et al.
This was a controlled
Subjects exposed for 6
The testing consisted
Statistics were
Results indicated the anxiety values of the placebo
(1977a)
exposure study
h/d for 5 days to 100
of: the Syndrome Short
conducted using
random sample group dropped significantly more

conducted on 7 healthy
ppm (550 mg/m3 TCE).
Test; the "Attention
Whitney Mann.
during the course of testing (p < 0.05) than those

male and female
Controls were exposed
Load Test" or "d2

of the active random sample group. No

students (4 females, 3
in chamber to zero TCE.
Test;" Number recall

significantly different changes were obtained with

males) The control
Biochemical tests
test, letter recall test,

any of the other variables.

group was 7 healthy
included TCE, TCA and
The "Letter Reading



students (4 females,
trichloroethanol in
Test," "Word Reading



3 males).
blood. In this study the
Test," Erlanger




TCE concentrations in
Depression Scale.




blood reported ranged
Scale for Autonomic




from 4 to 14 (ig/mL. A
Dysfunction, Anxiety




range of 20 to 60 (ig/mL
Scale, Pain Short Scale,




was obtained for TCA
and Information on




in the blood.
Daily Fluctuations.


>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Vernon and
8 male volunteers age
TCE administered as
Flicker Fusion with
ANOVAs,
TCE did not produce any appreciable effects at
Ferguson
range 21-30; self
Trilene air-vapor
Krasno-Ivy Flicker
Dunnett's test.
lower concentrations. Compared to controls,
(1969)
controls: 0 dose.
mixtures through
Photometer,

participants exposed to 1,000 ppm of TCE had


spirometers
Howard-Dolman depth

adverse effects on the Howard-Dolman,


administered at random
perception apparatus,

steadiness, and part of the pegboard, but no effects


concentrations of 0, 100,
Muller-Lyer

on Flicker Fusion, from perception or code


300, or 1,000 ppmof
two-dimensional

substitution. No appreciable changes in CBC,


TCE for 2 h at a time,
illusion, groove-type

urinalysis, SGOT, or BUN.


during which testing
steadiness test, Purdue




took place.
Pegboard, Written




Concentrations were
"code substitution,"




measured with a halide
blood studies.




meter. Medical history,





exam including CBC,





urinalysis, BUN, and





SGOT.



>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
to
o
VO
-------
IS*
Table D-l. Epidemiological studies: neurological effects of trichloroethylene (continued)
Reference
Study population
Exposure assessment
and biomarkers
Tests used
Statistics
Results
Windemuller
and Ettema
(1978)
Pilot study: 24 healthy
male volunteers; age
range = 19-26 yr, 4
groups with 6
volunteers in each: (1)
control, (2) exposed to
TCE, (3) exposed to
alcohol, (4) exposed to
TCE and alcohol;
Final study: 15 other
volunteers, each
exposed to all
4 conditions.
Chamber study; Group 1
no exposure; Group 2
TCE exposure: 2.5 h
with 200 ppm; Group 3
alcohol exposure: 0.35
g/kg body weight;
Group 4 TCE and
alcohol: same as above
levels; Blood alcohol
levels taken with
breathalyzer; exhaled air
sampled for levels of
TCE and
trichloroethanol; TCE
exposure: average
measured TCE in
exhaled air = 29 |ig/L
(SD = 3); TCE and
alcohol expo: average
measured TCE in
exhaled air = 63 |ig/L
(SD = 12).
Binary Choice Task
(Visual); Pursuit Rotor;
Recording of heart rate,
sinus arrhythmia,
breathing rate;
Questionnaire (15 items
on subjective feelings).
K-sample trend
test; two-tailed
Wilcoxon test.
Pilot study: no systematic effect of exposure on
test perform. Alcohol group had higher heart rate
than TCE group, and TCE and alcohol group;
minimal effect of mental load on heart rate; sinus
arrhythmia suppressed as mental load increased
with higher suppression in exposed groups (all 3)
compared to controls (differences possibly due to
existing group differences); Final Study: pursuit-
rotor task "somewhat impaired by exposure
condition;" authors acknowledge possibility of
sequence effects; no significant difference between
conditions on questionnaire responses; performing
mental tasks resulted in higher heart rate in the
TCE + alcohol condition than in Alcohol alone
condition; Mental load suppressed sinus
arrhythmia, especially in TCE + alcohol condition;
Conclusion: TCE and alcohol together impair
mental capacity more than each one alone.
BUN = blood urea nitrogen, EEG = electroencephalograph, GI = gastrointestinal, NIOSH = National Instutute of Occupational Safety and
Health,
OR = odds ratio, PCE = perchloroethylene.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Albers et al.
(1999)
30 railroad workers with
toxic encephalopathy;
involved in litigation;
long-term exposure to
solvents (n = 20 yrs.;
range = 10-29 yrs.);
Historical controls
matched by gender, age,
and body mass.
Most common solvents
included trichloroethylene,
trichloroethane,
perchloroethylene;
respirator not typically
used.
Neurologic exams (cranial
nerves, motor function,
alternate motion range,
subjective sensory
function, Romberg test,
reflexes), occupational
history, medical history,
sensory and motor nerve
conduction studies (NCS).
Log
transformations of
amplitude data;
Mann-Whitney
U Test for NCS;
t-test; simple linear
regression and
stepwise regression
for dose response.
3 workers met clinical
polyneuropathy criteria; NCS values
not influenced by exposure duration
or job title; no significant difference
in NCS between presence or
absence of polyneuropathy
symptoms, disability status, severity
or type of encephalopathy, or prior
polyneuropathy diagnosis.
>3
o
o
s
Co
I?
£u 2
Oq ^
53
TO
*
«
8
TO
Co
0
s
1
a.
VO
to
o
-------
TO
Co
s
o
to
o
s
3"
rS"
°S
to
S

§•
TO
s
a
TO
S
Co
a,
>T
§ %
>S
^ 3
VO
TO
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Antti-Poika
(1982)
87 patients (painters,
paint and furniture
factory workers, carpet
and laundry workers)
diagnosed 3-9 yrs prior
with chronic solvent
exposure (mean age
38.6 yrs)
Control: 29 patients with
occupational asthma.
Mean duration of exposure
10.4 yrs; solvents:
trichloroethylene,
perchloroethylene, solvent
mixture; based on patients'
and/or employers' reports;
9 worksites visited for
environmental measures;
biological measures at
1 worksite; exposure
classified as low, moderate,
or high.
Interview, Neurologic
exam, EEG,
electroneuromyographs,
psychological examination
(intellectual, short-term
memory, sensory and
motor functions).
Correlation
coefficients for
prognosis and
factors influencing
diagnosis.
Reported symptoms: fatigue,
headaches, memory disturbances,
pain, numbness, paresthesias;
1st exam: 87 patients with objective
and subjective neurological signs,
61 with psychological disturbance,
58 abnormal EEG, 25 clinical
abnormalities, 57 PNS symptoms;
69 patients had neurophysiological
or psychological disturbances
identified by neurologist in only 4
patients; 2nd exam: 42 with clinical
neurological signs,; 21 patients
deteriorated, 23 improved, 43 same;
poor correlation between prognosis
of examinations; no significant
correlation between prognosis and
age, sex, exposure duration and
level, alcohol use, or other diseases.
-------

§•
to
s
a
to
s
Co
a,
>r
>S
>S
to'
*
«
s
TO
Co
0
S
1
a.
to
o
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Aratani et al.
(1993)
437 exposed workers
from various industries
(not specified);
394 males, 43 females
and 1,030 male clerical
workers as controls; age
range: 16-72.
Exposed to Thinner,
G/5100, TCE, xylene,
toluene, methylchloride,
gasoline.
Vibrometer (VPT);
Urinary Metabolites.
Spearman
correlation.
Positive correlations between age
and VPT 7; between job experience
and VPT; Urinary metabolites not
significantly correlated with VPT;
no dose-effect for subjective
symptoms and neurological signs.
Binaschi and
Cantu (1983)
35 patients with
occupational exposure to
organic solvents;
Industry not specified; no
controls.
Occupational history
provided by patients;
Descriptions of jobs and
conditions provided by
employer; Workplace
observations; Some
available measurements of
solvents in air; 9 patients
exposed to
trichloroethylene;
11 exposed to toluene and
xylene; 15 exposed to
mixtures of solvents; all
exposures described to be
under TLV-TWA, but short
exposure might have
exceeded ACGIF limit for
short time.
Examination of provoked
and spontaneous vestibular
symptoms; Pure tone
threshold measurement;
EEG; psychiatric
interviews and psychiatric
history; Prevalence of
37 psychiatric symptoms.
Not stated.
All patients had subjective
symptoms (fatigue, psychic
disturbances, dizziness, vegetative
symptoms, vertigo); Vestibular
system affected in most cases, with
lesions in nucleo-reticular substance
and brain stem; EEG change with
diffuse and focal slowing; 71% of
patients had mild neurasthenic
symptoms (fatigue, emotional
instability, memory and
concentration difficulties).
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Bowler et al.
67 former
Self-report and work history
California
t-test for matched
Exposed workers performed
(1991)
microelectronics workers
from microelectronics
Neuropsychological
pairs; Wilcoxon
significantly worse on tests of

exposed to multiple
workers. Exposures and
Screening Battery.
Signed Rank test.
attention, verbal ability, memory,

organic solvents;
risks were estimated.


visuospatial, visuomotor speed,

Controls (n = 157) were
Solvents include TCE,


cognitive flexibility, psychomotor

recruited from the same
TCA, benzene, toluene,


speed, and reaction time; no

region; 67 pairs were
methylene chloride,


significant differences in mental

matched on the basis of
n-hexane.


status, visual recall, learning, and

age, sex, ethnicity,



tactile function.

educational level, sex,





and number of children.




>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
VO
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Colvin et al.
Final sample: 67 workers
Chronic exposure was
Work and personal history
Division into
Exposed group performed worse
(1993)
(43 exposed;
assessed through
interview; brief
exposed and
than unexposed on 27 out of 33 test

24 unexposed) in a paint
self-reported detailed work
neurological evaluation,
unexposed;
results; only significant difference

manufacturing plant
history for each worker;
WHO Neurobehavioral
Student's t-test;
was on latency times of two

employed there for at
past and current industrial
Core Test Battery (all tests
Multiple linear
switching attention tests; no

least 5 yrs.; all black
hygiene measurements of
except POMS); Computer-
regression.
difference in subjects' symptom

males; exclusion criteria:
solvent levels in air; "total
administered tests:

reporting between groups when

encephalopathy, head
cumulative expo" in the
Reaction time,

questions analyzed separately or

injury with 24 + h
factory and "average
Fingertapping, Continuous

analyzed as a group; Average

unconsciousness,
lifetime exposures" were
Performance Test,

lifetime exposure was a significant

psychotropic medication,
calculated; visitations to
Switching attention,

predictor for Continuous

alcohol/drug dependence
establish areas with
Pattern Recognition Test,

performance latency time,

history, epilepsy, mental
"homogeneous exposure;"
Pattern Memory; UNISA

Switching attention latency time,

illness.
All exposures below the
Neuropsychological

Mean reaction time, Pattern


ACGIH limit. Solvents
Assessment Procedure:

Memory; fine visuomotor tracking


include MEK, benzene,
Four word memory test,

speed significantly associated with


TCE, MIBK, toluene, butyl
Paragraph memory,

cumulative exposure; effects of


acetate, xylene, cellosolve
Geometric Shape drawing;

exposure concluded to be "relatively


acetate, isophorone, and
symptom and health

mild" and subclinical.


white spirits.
questionnaires.


>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Daniell et al.
89 retired male workers
Chronic occupational
Psychiatric interview;
Odds ratio,
CEI was similar for painters and
(1999)
(62-74-yr old) with prior
exposure; Structured
questionnaires; physical
logarithmic
aerospace workers; Painters

long-term exposure to
clinical interview about past
exam; blood cell counts,
transformation of
reported greater alcohol use than

solvents including
and present exposure to
chemistry panel, blood
non-Gaussian data,
carpenters; painters also had lower

67 retired painters and
solvents; Cumulative
lead levels,
standardization of
scores on WAIS-R Vocabulary

22 aerospace
Exposure Index was
Neuropsychological: BDI,
test scores,
subtest; Controlling for age,

manufacturing workers;
constructed. Solvents not
verbal fluency test, WAIS-
ANCOVA,
education, alcohol use, and

Controls: 126 retired
specified.
R: Vocabulary,
Multiple Linear
vocabulary score, painters

carpenters with minimal

Similarities, Block Design,
regression; Kruskal
performed worse on motor,

solvent exposure.

Digit Span, Digit Symbol;
Wallis test for
memory, and reasoning ability tests;



Wisconsin Card Sorting;
differences in
painters reported more symptoms of



verbal aphasia screening
blood lead
depression and neurological



test, Trails A and B,
concentration.
symptoms; painters more likely to



Fingertapping; WMS-R:

have more abnormal test scores



logical memory and visual

(odds ratio: 3.1) as did aerospace



subtests; Rey Auditory

workers (odds ratio: 5.6); no dose



Verbal Learning; Benton

effect with increasing exposure and



Visual Retention test; d2

neuropsychological tests.



test; Stroop; Grooved





pegboard; simple reaction





time.


>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
ko	Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
is*
5}'

o
5S


Exposure assessment and




a
Reference
Study population
biomarkers
Tests used
Statistics
Results


Donoghue et
16 patients diagnosed
Average exposure duration
Visual acuity measured
Chi-square test.
6 participants (37.5%) with
o
S5
al. (1995)
with organic-solvent-
was 19 yrs (range = 5-36
with a 4-m optotype chart;

abnormal contrast sensitivity; 2 of
o
s
S*

induced toxic
yrs); Solvents include TCE,
Contrast sensitivity

the 6 (33%) had monocular
Co
?«•><'


encephalopathy with
MEK, toluene, thinners,
measured with Vistech

abnormalities; abnormalities
Si
r?
o

various occupations
unidentified hydrocarbons.
VCTS 6500 chart;

occurred at all tested spatial
Oq


compared to

monocular thresholds,

frequencies; significant difference
53
rS*
*

age-stratified normal

pupil diameter.

between groups at 3 cpd, 6 cpd,
^3
si

groups (n = 38): average



12 cpd frequencies.

"3
o

age: 43 y (range





>3
Co
O

= 31-58); Exclusion
criteria: diabetes mellitus,





1
a.

ocular disease impairing






vision, visual acuity with





o
to
o
o

existing refractive
correction of less than
4/6, abnormal direct
ophthalmoscopic exam.




-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)


Exposure assessment and



Reference
Study population
biomarkers
Tests used
Statistics
Results
Elofsson et
Epidemiologic study of
Long term, low level expo
Self-administered
Calculation of
Significant differences between
al. (1980)
car or industrial spray
to multiple solvents;
psychiatric questionnaires,
z values; Pearson
controls and exposed in symptoms

painters (male) exposed
Assessed by interviews, on-
Eysenck's Personality
correlation;
of neurasthenic syndrome, in

long-term to low levels
the-job measurements, and
Inventory, psychosocial
Multiple
reaction time, manual dexterity,

of organic solvents
a 1955 workshop model;
structured interview,
Regression
perceptual speed, and short-term

(,n = 80); 2 groups of
Blood analysis: mean
Comprehensive
Analysis.
memory; no significant differences

matched controls;
values were within normal
Psychopathological Rating

on verbal, spatial, and reasoning

80 nonexposed male
limits for both groups;
Scale; Visual Evoked

ability; Some differences on EEG,

industrial workers in each
Exposed group had
Responses; EEG;

VER, ophthalmologic, and CT.

control group.
significantly higher values
for alkaline phosphates,
hemoglobin, hematocrit,
and erythrocytes; early
exposure TLVs in Sweden
were significantly lower;
solvents include TCE, TCA,
methylene chloride, and
others.
Electroneurography;
Vibration Sense Threshold
estimations; Neurological
exam.


>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)


Exposure assessment and



Reference
Study population
biomarkers
Tests used
Statistics
Results
Gregersen
Workers exposed to
1st follow-up: data about
1st follow-up: structured
Wilcoxon-Mann-
More acute neurotoxic symptoms in
(1988)
organic solvents (paint,
working conditions,
interviews on
Whittney tests;
exposed group at both follow-ups,

lacquer, photogravure,
materials and exposure in
occupational, social,
Kruskal-Wallis
but fewer symptoms at

and polyester boat
prior 5 yrs used for
medical history; clinical
test; Chi-square;
2nd follow-up than at 1st follow-up;

industries); Controls:
exposure index; 2nd follow-
exam, neurological exam;
Spearman Rank
at both follow-ups exposed

warehousemen
up: 9 questions asking about
2nd follow-up: mailed
Partial Correlation
participants had more

electricians; 1st follow-up
exposure to solvents in the
questionnaire (49 follow-
Coefficient.
encephalopathy symptoms,

5.5 yrs after initial
prior 5 yrs; TCE, toluene,
up issues to 1st follow-up).

especially memory and

evaluation (59 exposed,
styrene, white spirits.


concentration; no encephalopathy

30 unexposed);



symptoms in control group;

2nd follow-up: 10.6 yrs



symptoms and signs of peripheral,

after initial evaluation



sensory, and motor neuropathy

(53 exposed,



significantly worse in participants

30 unexposed controls).



still exposed; Exposure index
showed dose-effect with memory
and concentration; Both follow-ups:
improvement in acute symptoms;
aggravation in CNS; more
symptoms of peripheral nervous
system and social consequences.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
ko	Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
is*

O
5S


Exposure assessment and




a
Reference
Study population
biomarkers
Tests used
Statistics
Results


Juntunen et
37 patients with
Patients were exposed to
Neurologic examination,
Descriptive
Clinical neurological findings of
o
S5
al. (1980)
suspected organic solvent
Carbon disulphide (n = 6),
pneumoencephalographic
Statistics.
slight psychoorganic alterations,
' J
o
s


poisoning (mean age
trichloroethylene (5),
exam, EEG, tests assessing

cerebellar dysfunction, and
Co

= 40.1 yrs.); selection
styrene (1), thinner (2),
intelligence, memory and

peripheral neuropathy; 63% had
Si
r?
>{
>5

based on
toluene (1), methanol (1),
learning, motor function,

indication of brain atrophy; 23 of
£u
Oq


pneumoencephalography;
and carbon tetrachloride (2),
and personality.

the 28 patients examined with
53
fV
*

no controls.
mixtures (19); Exposure


electroneuromyography showed
^3
si


was assessed by patients'


signs of peripheral neuropathy; 94%

"3
o


and employers' reports and


had personality changes, 80% had

>3
Co
O
53


measurements of air
concentrations when


psychomotor deficits, 69% had
impaired memory, and 57% had

1
a.


available.


intelligence findings; No dose-effect






found.
\Q ^
o
to
o
o
-------
ko	Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
is*

o
5S


Exposure assessment and




a
Reference
Study population
biomarkers
Tests used
Statistics
Results


Juntunen et
80 (41 women, 39 men)
Assessed by patients'
Neurologic examination;
Chi-square,
Significant correlations between
o
s
al. (1982)
Finnish patients
occupational history,
EEG and ENMG; tests of
Maxwell-Stuart,
prognosis of disturbances in gait
o
s
S*

diagnosed 3-9 yrs prior
employers' workplace
intellectual function,
Correlation and
(p < 0.05) and station and length of
Co


with chronic solvent
description, observations
memory, learning,
multiple linear
follow-up, duration and level of
Si
r?
o
>5

exposure (mean age
and data collected at
personality and
regression
exposure and multiplying the two;
Oq


= 38.6 yrs); 31 had slight
workplace, environmental
psychomotor performance.
analyses.
no gender effects; Common
53
fV
*

neurological signs; no
measurements, biological


subjective symptoms; headaches,
^3
si

controls.
tests; TCE, PCE, or mixed


fatigue, and memory problems;

"3
O


solvent exposures.


Impairment in fine motor skills,

Co
Co
O
53





gait, and cerebellar functions;
Subjective symptoms decreased

1
a.





during follow-up, but clinical signs






increased.
\Q ^
o
to
o
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)


Exposure assessment and



Reference
Study population
biomarkers
Tests used
Statistics
Results
Laslo-Baker
32 mothers with
Exposure information
Children: Wechsler
Power analysis,
Verbal IQ was lower (104) in
et al. (2004)
occupational exposure to
collected at 3 times:
Preschool and Primary
Multiple linear
children exposed in utero vs.

organic solvents during
(1) during pregnancy,
Scale of Intelligence,
regression.
unexposed children controls (110);

pregnancy and their
(2) when contacted for
WISC, Preschool

Children did not differ between

children (3-9 yrs of age);
study participation later in
Language Scale, Clinical

groups in birth weight, gestational

included if exposure
pregnancy, (3) at time of
Evaluations of Language

age, or developmental milestones;

started in 1st trimester
assessment; Information
Fundamentals, Beery -

Children in the exposed group had

and lasted for at least
collected included types of
Buktenica Developmental

significantly lower VIQ (108) and

8 wks of pregnancy
solvent, types of setting,
test of Visuo-Motor

Full IQ (108) than controls (VIQ

(32 mother-child pairs);
duration of exposure during
Integration, Grooved

= 116 and Full IQ= 114; No

Controls: 32 unexposed
pregnancy, use of
Pegboard Test, Child

significant difference in PIQ;

control mothers matched
protection, symptoms,
Behavior Checklist (Parent

Performance on expressive

on age, child age, child
ventilation; Solvents
Version), Connor's Rating

language, total language, and

sex, SES, and reported
include toluene (n= 12
Scale-Revised (Parent

receptive language was significantly

cigarette use and their
women), xylene (10),
Version), Behavioral Style

worse in children from exposed

children (32 mother-child
ethanol (7), acetone (6),
Questionnaire; Mothers:

group.

pairs).
methanol (5), TCE (3), etc.
(a total of 78 solvents were
reported).
WASI.


>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Lee et al.
40 Korean female shoe
4 workers wore passive
Questionnaire;
Multivariate
Significant differences between
(1998)
factory workers
personal air samplers for a
Neurobehavioral Core Test
ANOVA for tests
groups based on exposure index;

employed there for at
Ml 8-h shift; Detected
Battery (includes POMS,
with 2 outcomes;
Differences in performance between

least 5 yrs.; cases with
solvents: toluene, methyl
Simple Reaction Time,
ANOVA for tests
controls and participants on Santa

head injury, neurological
ethyl ketone, «-hexane,
Santa Ana Dexterity test,
with 1 outcome;
Ana were found only in the CEE

or psychological
c-hexane, cyclohexane,
Digit Span, Benton Visual
education was
(participants performed worse);

disorder, or hearing or
dichloroethylene,
Retention Test, Pursuit
adjusted in
CEE is a more sensitive measure of

visual impairment were
trichloroethylene, benzene,
aiming motor steadiness
analyses.
exposure to organic solvents.

excluded; Controls:
and xylene; In frame-
test); POMS was excluded



28 (housekeepers); no
making air concentration of
because of cultural



in-plant controls
solvents was 0.46-0.71; In
inapplicability.



available.
adhesive process solvent air





concentrations were





1.83-2.39; three exposure





indices were calculated:





current exposures, exposure





duration (yrs), and





Cumulative Exposure





Estimate (CEE) (yrs





x average exposures).



>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------

§•
to
s
a
to
s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
8
TO
Co
0
S
1
a.
to
o
VO
Lindstrom
(1973)
168 male workers with
suspected occupational
exposure to solvents
Group I with solvent
poisoning (n = 42);
Group II with solvent
exposure, undergoing
mandatory periodic
health check (n = 126);
Control-50 healthy
nonexposed male
volunteers working in a
viscose factory; Group
IV 50 male workers with
carbon disulfide
poisoning.
44 exposed to TCE, 8 to
tetrachloroethylene, 26 to
toluene, 25 to toluene and
xylene, 44 to thinners, 21 to
"miscellaneous;" Solvent-
exposed group had an
average of 6 y of expo; CS2
group had average of 9 yrs
of exposure.
WAIS: Similarities,
Picture Completion, Digit
Symbol; Bourdon-
Wiersma vigilance test,
Santa Ana, Rorschach
Inkblot test, Mira test.
Student's t-test.
The solvent-exposed group and CS2
group had significantly worse
"psychological performances" than
controls; Greatest differences in
sensorimotor speed and
psychomotor function;
solvent-exposed and CS2 groups had
deteriorated visual accuracy.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Lindstrom
56 male workers
Chronic "excessive"
WAIS subtests:
Factor analysis;
Significant decline in visuomotor
(1980)
diagnosed with
exposure: Mean duration of
Similarities, Digit Span,
Student's t-test;
performance and freedom from

occupational disease
exposure = 9.1 yrs (SD
Digit Symbol, Picture
Multivariate
distractibility (attention) in the

caused by solvents;
= 8.3); Exposed to;
Completion, Block Design;
Discriminant
solvent-exposed participants;

Controls:
halogenated and aromatic
WMS subtests: Visual
analysis.
significant relationship between

98 styrene-exposed
hydrocarbons, paint
Reproduction; Benton

duration of solvent exposure and

workers; 43 nonexposed
solvents, alcohols, and
Visual Retention test;

visuomotor performance; solvent

construction workers.
aliphatic hydrocarbons
Symmetry Drawing; Santa

exposure level was not significant;


(TCE n = 14); Individual
Ana Dexterity test; Mira

psychological test performance of


exposure levels estimated as
test.

styrene-exposed control was only


time-weighted averages,


slightly different from nonexposed


based on information


controls.


provided by subjects,





employer, or workplace





measurements, were





categorized as low (3





patients), intermediate (26





patients), and high (27





patients).



>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Lindstrom et
86 Patients with prior
Mean duration of exposure
Intellectual Function: from
Frequency
All patients grouped together
al. (1982)
diagnosis of solvent
10.4 yrs; solvents:
WAIS - Similarities,
distributions,
regardless of types of past solvent

intoxication (mean age
trichloroethylene,
Block Design, Picture
Student's t-test for
exposure; on follow-up, significant

38.6 yrs.); 40 male, 46
perchloroethylene, solvent
Completion; Short Term
paired data,
learning effects for Similarities

female; 52 exposed to
mixture; based on patients'
Memory: from WMS -
stepwise linear
when compared to results at initial

mixed solvents; 21
and/or employers' reports.
Digit Span, Logical
regression.
diagnosis; group mean for

exposed to TCE or PCE;

Memory, Visual

intellectual functioning increased;

13 exposed to both;

Reproduction; Benton

no significant change in memory

results at follow-up

Visual Retention test;

test results; group means for sensory

compared to those at

Sensory and Motor

and motor tasks were lower;

initial diagnosis.

Functions: Bourdon

prognosis was better for longer



Wiersma Vigilance Test,

follow-up and younger age and



Symmetry Drawing, Santa

poorer for users of medicines with



Ana Dexterity test, Mira

neurological effects.



test.


>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)


Exposure assessment and



Reference
Study population
biomarkers
Tests used
Statistics
Results
Marshall et
All singleton births in
Information on inactive

Odds ratios (OR),
13 CNS cases and 351 controls with
al. (1997)
1983-1986 in 188 New
waste sites was examined,

Fisher's exact test,
potential exposures; crude OR: 0.98;

York State counties (total
including air vapor, air

Chi-square,
When controlling for mother's

number not specified);
particulates, groundwater

unconditional
education, prenatal care, and

473 CNS-defect births
exposure via wells, and

logistic regression.
exposure to a TCE facility, OR was

and
groundwater exposure, via


0.84; CNS and solvents OR: 0.8;

3,305 musculoskeletal-
basements; exposure was


CNS and metals OR: 1.0,

defect births; Controls:
categorized as "high,"


musculoskeletal defects and

12,436 normal births;
"medium," "low," or


solvents OR: 0.9, musculoskeletal

Exclusion criteria:
unknown based on


defects and pesticides OR: 0.8;

Trisomy 13, 18, or 21,
probability of exposure;


higher risk for CNS defects when

birth weight of less than
proximity to waste sites was


living close to solvent-emitting

1,000 g, sole diagnosis of
also considered; Most


facilities.

hydrocephaly or
common solvents: TCE,




microencephalopathy,
toluene, xylenes,




hip subluxation.
tetrachloroethene,
1,1,1-trichloroethane; Most
common metals found lead,
mercury, cadmium,
chromium, arsenic, and
nickel.



>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------

§•
to
s
a
to
s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
o
o
s
Co
I?
Oq ^
53
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Co
a
a,
>r
<§>
>S
TO
*
8
TO
Co
0
S
1
a.
to
o
McCarthy
and Jones
(1983)
384 industrial workers
with solvent poisoning;
103 operated degreasing
baths, 62 maintained
degreasing baths, 37 used
TCE in portable form,
37 misc; no controls.
Individuals poisoned with
trichloroethylene,
perchloroethylene, and
methylchloroform were
examined retrospectively;
Medical record review; 288
exposed to TCE, 44 to
perchloroethylene, 52 to
1,1,1 -trichloroethane.
Symptoms reported in
occupational/medical
records from industrial
poisoning incidents; data
from 1961 to 1980 on
demographics, occupation,
work process, type of
industry, if incident caused
fatality.
17 fatality cases, with 10 in
confined spaces; Most common
symptoms include effects on CNS;
Gastrointestinal and Respiratory
symptoms; no strong evidence for
cardiac and hepatic toxicity; no
change in affected number of
workers in 1961 to 1980; greatest
effect due to narcotic properties.
vo
Mergler et al.
(1991)
54 matched pairs;
Matching on the basis of
age, sex, ethnicity,
educational level, sex,
and number of children
taken froml 80 former
microelectronics workers
exposed to multiple
organic solvents and
control population of
157 recruited from the
same region.
Average duration of
employment: 6.1 yrs (range:
1-15 yrs); information
about products used and
chemical make-up from
employer; chemicals:
chlorofluorocarbons,
chlorinated hydrocarbons,
glycol ethers, isopropanol,
acetone, toluene, xylene,
and ethyl alcohol.
Sociodemo graphic
questionnaire; Monocular
examination of visual
function: Far visual acuity
using a Snellen chart, near
visual acuity using a
National Optical Visual
Chart, color vision using
Lanthony D-15, near
contrast sensitivity using
Vistech grating charts.
Signed-rank
Wilcoxon test;
Mann-Whitney;
Chi-square test for
matched pairs;
Multiple
Regression;
Stepwise
regression.
Significant difference in near
contrast sensitivity: 75% of exposed
workers with poorer contrast
sensitivity at most frequencies than
the matched controls (no difference
in results based on smoking, alcohol
use, and near visual acuity loss);
Significant differences on near
visual acuity, color vision, and rates
of acquired dyschromatopsia for one
eye only; No difference between
groups in near or far visual acuity.
-------
ko	Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
is*
5}'

to
s


Exposure assessment and




a
to
a
Reference
Study population
biomarkers
Tests used
Statistics
Results

S3'
Morrow et al.
22 male patients with
Exposure assessed with
Exposure questionnaire,
Stepwise multiple
All profiles valid; 90% with at least
o

(1989)
exposure to multiple
questionnaire (duration,
Group form of the MMPI.
regression.
2 elevated scales above T score of
o
s
%¦
<¦&
Ss

organic solvents;
type of solvents, weeks


70 (clinically significant); Highest
Co

4 involved in litigation;
since last exposure, cases of


elevations on scales 1, 2, 3, and 8;
Si
r?
o
>S
>

Exclusion: neurologic or
excessive exposure);


only 1 case within normal limits;
£u
Oq
TO
;g

psychiatric disorder prior
Average exposure duration


when compared to a group of
53
fV
TO'
*

to assessment, alcohol
= 7.3 yrs (range: 2 mos-19


nonpsychiatric patients, exposed
^3
«

consumption more than
yrs); average weeks since


patients had more elevations,

O

2 drinks/day; Average yrs
last exposure was 19.8


although both groups have physical

>;
TO
Co
O
S

education 12 (range:
10-16 yrs); average age
(range: 1-84 wks); 28% had
at least one instance of


complaints; When compared with
WWII POW (1/2 diagnosed with

1
a.

38 yrs (range: 27-61);
excessive exposure.


PTSD) with similar SES and


compared to responses of



education, both groups have similar
VO
o
to
o
o

WWII prisoner of war
(POW) population with
posttraumatic stress
disorder (PTSD).



profiles; no age effects found;
significant positive correlation
between scale 8 and duration of
exposure; no significant difference
based on time since last exposure or
on experiencing excessive exposure.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Morrow et al.
9 men and 3 women
Exposure assessed with
Auditory event-related
Repeated measures
Exposed patients had significant
(1992)
occupationally exposed
occupational and
potentials under the
ANOVA.
delays in N250 and P300 compared

to multiple organic
environmental exposure
oddball paradigm:

to normal controls and in P300

solvents with CNS
questionnaire; mean
counting and choice

compared to psychiatric controls;

complaints; all met
duration of expo = 3 y
reaction time tasks.

Exposed patients had higher

criteria for mild toxic
(range = <1 d-30 y);


amplitudes forNlOO, P200, and

encephalopathy; exposed
average time between last


N250; no difference inP300

group average age was
exposure and assessment


amplitude between groups; for the

47 y; Controls:
was 2 y (range;


exposed group, P300 positively

19 (healthy male
2 mos-10 y); solvents


correlated with exposure duration;

volunteers);
toluene, TCE.


findings indicate that solvent

26 psychiatric controls



exposure affects neural networks.

(male patients with





chronic schizophrenia)





average age unexposed





controls: 34 yrs; average





age schizophrenic





patients.: 36 yrs.




>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Seppalainen
87 patients with solvent
Chronic exposure with
EEG using 10/20 system
Chi-square,
Significantly more ENMG
and
poisoning (40 male and
average duration of 10.7 yrs
with 25-30 mins of
Hypergeometric
abnormalities at follow-up than at
Antti-Poika
47 female) with
(range: 1-33); patients were
recording, 3 mins
distribution,
initial diagnosis; Most common
(1983)
occupational exposure to
exposed to TCE (n = 21),
hyperventilation and
McNemar test.
finding: slight polyneuropathy; 43%

solvents; Follow-up
perchloroethylene (n = 12),
intermittent photic

showed improved ENMG, 33% had

3-9 yrs after initial
mixtures of solvents
stimulation; ENMG.

deteriorated, and 18 pts. with similar

diagnosis; Mean age at
(« = 53), mixtures and TCE


ENMG findings (6 normal at both

diagnosis 38.6 (range:
or perchloroethylene


exams); at follow-up, slow-wave

20-59 yrs); no control
(,n = 13); Exposure of 54


abnormalities decreased and

population.
patients stopped after


paroxysmal abnormalities increased;


diagnosis, 33 continued to


41 with improved EEG, 28 with


be exposed; at follow-up,


similar EEG (19 had normal EEG at


only 5 working with


diagnosis), and 18 with deteriorated


potential of some exposure.


EEG; EEG pattern of change





compared to external head injuries.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Shlomo et
Male industrial workers;
Interview and record
Medical history,
Student's t-test,
Significant differences between
al., 2002
Mercury exposure group
review; Urine samples
Neurological tests
proportions test.
exposed and controls: 33.8% of CH

(n = 40); average age
collected at end of work
assessing cranial nerves

exposed workers with abnormal

49.7 (±6.4) yrs;
shift prior to testing and
and cerebellar function;

IPL I-III; 18% of controls; Authors

chlorinated hydrocarbons
tested for mercury and TCA
Otoscopy, review of

suggest ABRs are sensitive for

(CHs) exposure group
; chlorinated hydrocarbons:
archival data from pure-

detecting subclinical CNS effects of

(n = 37) average age 46.0
TCE (n = 7), PCE {n = 8),
tone audiometric tests;

CH and mercury.

(±4.73); Controls,
trichloroethane (n = 22);
Auditory brain stem



unexposed (n = 36)
Mean duration of chloral
responses (ABR).



average age 49.8 (±5.8),
hydrate (CH) exposure 15.8




matched by age;
(±7.2) yrs; Mean duration




(industries not specified).
of mercury exposure 15.5





(±6.4) yrs; Air sampling:





mercury: 0.008 mg/m3





(TLV = 0.025); TCE: 98





ppm (TLV = 350); PCE:





12.7 ppm (TLV = 25);





trichloroethane: 14.4 ppm





(TLV = 200); Blood levels:





mercury (B-hg) 0.5 gr%





(±0.3); TCA urine levels:





1-80% of Biologic





Exposure Index (BEI); CH





urine levels: 0.11-0.2 of





BEL



-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Till et al.
The children of mothers
Structured questionnaire
NEPSY: Visual Attention,
Mantel Haenszel
Lower composite neurobehavioral
(2001)
who had contacted a
about exposure; Method:
Statue, Tower, Body Part
test, t-test,
scores as exposure increased after

Canadian pregnancy risk
weight assigned to each
Naming, Verbal Fluency,
ANCOVA,
adjusting for demographics in

counseling program
exposure Parameter (length
Speeded Naming,
Hierarchical
Receptive language, expressive

during pregnancy and
of exposure, frequency of
Visuomotor Precision,
multiple linear
language, graphomotor ability;

reported occupational
exposure, symptoms); sum
Imitating Hand Positions,
regression.
Significantly more exposed children

exposure to solvents
of scores for each parameter
Block Construction,

rated with mild-severe problems;

(n = 33): children age
used as exposure index;
Design Copying, Arrows;

No significant difference between

range: 3-7; Mothers'
median split used to
Peabody Picture

groups in attention, visuo-spatial

occupations: lab
categorize in low (n= 19)
Vocabulary Test;

ability, and fine-motor skills; Mean

technicians, factory
and high (n= 14)
WRAVMA Pegboard test;

difference on broad- and

workers, graphic
exposures; solvents include
Child Behavior Checklist

narrow-band scales of Child

designers, artists, and dry
benzene, toluene, methane,
(Parent form); Continuous

Behavior Checklist scores not

cleaning; Controls:
ethane, TCE, methyl
Performance Test.

significant.

28 matched on age,
chloride, etc.




gender, parental SES, and





ethnicity; children of





mothers exposed to





nonteratogenic agents.




>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Till et al.
Children of mothers who
Structured questionnaire
Minimalist test to assess
Independent
Significantly higher number of
(2001)
had contacted a Canadian
about exposure; Method:
color vision; Cardiff Cards
samples t-tests,
errors on red-green and blue-yellow

pregnancy risk
weight assigned to each
to assess visual acuity.
Mantel Haenszel
discrimination in exposed children

counseling program
exposure parameter (length

Chi test;
compared to controls; exposed

during pregnancy and
of exposure, frequency of

Wilcoxon-Mann-
children had poorer visual acuity

reported occupational
exposure, symptoms); sum

Whitney test;
than controls; No significant

exposure to solvents
of scores for each parameter

Kruskal-Wallis Chi
dose-response relationship between

(n = 32): children age
used as exposure index;

square.
exposure index and color

range: 3-7; Mothers'
median split used to


discrimination and visual acuity.

occupations: lab
categorize in low (n= 19)




technicians, factory
and high (n= 14)




workers, graphic
exposures; solvents include




designers, artists, and dry
benzene, toluene, methane,




cleaning; Controls:
ethane, TCE, methyl




27 matched on age,
chloride, etc.




gender, parental SES, and





ethnicity; children of





mothers exposed to





nonteratogenic agents.




>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)


Exposure assessment and



Reference
Study population
biomarkers
Tests used
Statistics
Results
Till et al.
21 infants (9 male,
Structured questionnaire
1st visit: Sweep visual
Median split;
Significant decline of contrast
(2005)
12 female)of mothers
about exposure; Method:
evoked potentials (VEP) to
Multiple Linear
sensitivity in low and intermediate

who contacted a
weight assigned to each
assess contrast sensitivity
Regression;
spatial frequencies in exposed

Canadian pregnancy risk
exposure parameter (length
and grating acuity; 2nd visit
Chi-square, t-test,
infants when compared with

counseling program and
of exposure, frequency of
(2 wks after 1st): Transient
Mann-Whitney
controls; Significant effect of

reported occupational
exposure, symptoms); sum
VEPs to assess chromatic
U test, Multivariate
exposure level on grating acuity,

exposure to solvents
of scores for each parameter
and achromatic
ANCOVA,
26.3% of exposed (but 0% of

(occupations: factory,
used as exposure index;
mechanisms;
Pearson
controls) with abnormal VEP to

lab., dry cleaning;
median split used to
ophthalmological exam,
correlation,
red-green onset stimulus; No

Controls: 27 age-matched
categorize in low and high
physical and neurological
Logistic
differences between groups in

infants (17 male,
exposures; exposure groups:
exam; testers masked to
Regression.
latency and amplitude of chromatic

10 female) of mothers
(1) aliphatic and/or
exposure status of infant.

and achromatic response.

contacted the program
aromatic hydrocarbons




due to exposure during
(n = 9), (2) alcohols (n = 3),




pregnancy to
(3) multiple solvents




nonteratogenic
(n = 6), (4) PCE, (n = 3);




substances).
mean duration of exposure
during pregnancy 27.2 wks.
(SD 7.93, range = 12-40);
solvents include benzene,
toluene, methane, ethane,
TCE, methyl chloride, etc.



>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Valic et al.
138 occupationally
Solvents: TCE, PCE,
Lanthony D15.
Polytomous
Significant effect of age in exposed
(1997)
exposed and
toluene, xylene; Historical

logistic regression.
group; With alcohol of <250 g/wk

100 unexposed controls;
data on duration of


no significant correlation between

Exclusion criteria:
exposure protective


color confusion and solvent

congenital color vision
equipment use, subjective


exposure; Significant interaction

loss, severe ocular
evaluation of exposure,


between solvent exposure and

disease, significant vision
nonoccupational solvent


alcohol intake; Color Confusion

impairment, tainted
exposure, solvent-related


Index significantly higher in

glasses or contact lenses,
symptoms at work, alcohol


exposed group with alcohol use of

diabetes mellitus,
and smoking, drug intake;


>250 g/wk.

neurological disease,
Mean urinary levels of




prior severe head or eye
trichloroacetic acid: 1.55




injuries, alcohol abuse,
(±1.75) mg/L.




medication impairing





color vision.




>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Windham et
Children born in 1994 in
Birth addresses were
Archival data.
Pearson
Elevated adjusted odds ratios for
al. (2006)
San Francisco Bay Area
geocoded and linked to

correlation,
ASD (by 50%) in top quartile of

with Autism Spectrum
hazardous air pollutant

Logistic
chlorinated solvents, but not for

Disorders
database; Exposure levels

Regression.
aromatic solvents; AOR for TCE in

(ASDs)(« = 284) and
assigned for 19 chemicals;


4th quartile = 1.47; lessened when

controls (n = 657),
chemicals were grouped


adjusted for metals; correlation

matched on basis of
based on mechanistic and


between hydrocarbon and metals

gender and month of
structural properties;


exposures; when adjusted, increased

birth.
Summary index scores were


risk for metals (in 3rd quartile


calculated; risk of ASD


= 1.95; in 4th quartile = 1.7).


calculated in upper quartiles


Contributing compounds: mercury,


of groups or individual


cadmium, nickel, TCE, vinyl


chemical concentrations;


chloride; Results interpreted to


Adjustment for


suggest relationship between autism


demographic factors.


and estimated metal and solvent





concentrations in air around place of





birth residence.
>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Epidemiological Studies: Controlled Exposure Studies; Neurological Effects of Trichloroethylene/Mixed Solvents
>3
o
o
s
Co
I?
Oq ^
53
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
Levy et al.
(1981)
9 participants (8 males
and 1 female) recruited
through newspaper ad;
8 h fasting before testing;
no control.
Experiment 1: alcohol
consumption (3 doses)—
blood alcohol levels were
measured with breath
analyzer pre (multiple
baselines) and post test
(multiple).
Experiment 2: Chloral
hydrate administered orally
over 2 mins in either 500
mg or 1,500 mg dose;
multiple baseline smooth
pursuit eye movement
(SPEM) tests and multiple
posttests after exposure; No
control dose administered.
SPEM tests of following a
sinusoidally oscillated
target at 0.4 Hz; eye
movements were recorded
through electrodes at each
eye.
t-tests; ANOVA.
Experiment 1: prealcohol all
subjects had intact SPEM; no
significant effect for 1.5 mL/kg of
alcohol; significant decline in
SPEM at 2.0 and 3.0 mL/kg alcohol;
significant dose-effect.
Experiment 2: at 500 mg. chloral
hydrate, no significant change in
pursuit was noted; at 1,500 mg
chloral hydrate, qualitative
disruptions in pursuit in all
participants (4); at 500 mg
participants observed to be drowsy;
When number reading was added
SPEM impairment was 'attenuated'
in both alcohol and chloral hydrate
conditions.
-------
IS*
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents (continued)
Reference
Study population
Exposure assessment and
biomarkers
Tests used
Statistics
Results
Stopps and
McLaughlin
(1967)
Chamber study using
2 healthy male
volunteers exposed to
Freon-113; 1 volunteer
exposed to TCE; No
control.
Exposure booth was
constructed; TCE in air:
TCE concentrations: 100,
200, 300, 400 ppm (1965
TLV: 100 ppm for 8-h
exposure) in ascending and
descending order; total time
in chamber: 2.75 h; Freon-
113 concentrations: 1,500,
2,500, 3,500, 4,500 ppm
(1965 TLV: 1,000 ppm for
8-h exposure), duration 1.5
h;
TCE: (1) reduction of
weight of compound during
exposure was calculated, (2)
continuous air sampling in
the chamber; Freon-113 in
air: (1) and (2) same; (3) gas
chromatography on air
captured in bottles sealed in
the chamber; no control
dose given.
Crawford Small Parts
Dexterity Test, Necker
Cube Test, Card Sorting,
Card Sorting with an
Auxiliary Task, Dial
Display (TCE participant
only); Short Employment
Test-Clerical (Freon-113
participants only).
Descriptive
statistics for air
measurement plots
by % of TCE
change in groups.
No TCE effect at 100 ppm, but test
performance deteriorated with
increase of TCE concentration; No
effect of Freon-113 on psychomotor
function at 1,500 ppm, deterioration
at 2,500 ppm, as concentration
increased, performance deteriorated.
CNS = central nervous system, EEG = electroencephalograph, PCE = perchloroethylene, WHO = World Health Organization.
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
Table D-3. Literature review of studies of TCE and domains assessed with neurobehavioral/neurological methods
Authors
Year
Study
type
Participants no.
(N = exposed
C = nonexposed)
Dur
PM/RT
VM
Cogn
M&L
M&P
Sympf
Sen??
Resp
Dose effect
vv
urinary
metabolitesV
TCE levels
ATSDR
(2003b)
E
N = 116, C= 177
C
ne
ne
ne
ne
ne
ne
A
ne
ne
0 —> 23 ppb in
dg water
Barret et al.
(1984)
O
00
00
II
£
C
ne
ne
ne
ne
ne
H, D
T, N, V
ne
V
150 ppm
Barret et al.
(1987)
O
N = 104, C = 52
c
ne
ne
ne
ne
V
H, D, S, I
T, N
ne
V
ne
Barret, et al.
(1982)
o
N = 11, C = 2
c
ne
ne
ne
ne
ne
ne
T
ne
V
ne
Burg, et al.
(1995)
E
N = 4,281
c
ne
ne
ne
ne
ne
ne
A, N
V
V
ne
Burg and Gist
(1999)
E
N= 3915
c
ne
ne
ne
ne
ne
ne
A, N
V
vv
4 gps:
2-75,000 ppb
El Ghawabi et al.
(1973)
O
N = 30, C = 30
c
ne
ne
ne
ne
ne
H, S
(-)
ne
V
165 ppm
Feldman et al.
(1988)
E
N = 21, C = 27
c
ne
ne
ne
ne
ne
ne
T
ne
ne
ne
Feldman et al.
(1992)
O
N = 18, C = 30
A,C
ne
ne
ne
ne
ne
ne
T, N
ne
ne
ne
Gamberale, et al.
(1976)
C
N = 15
A
V
ne
V
(-)
ne
ne
ne
ne
ne
540-1,080 mg3
Gash et al.
(2008)
o
N= 30
c
V
ne
ne
ne
ne
M, N

ne
ne
ne
Grandjean et al.
(1955)
o
N= 80
c
ne
ne
ne
ne
ne
ne
N
ne
V, vv
6-1,120 ppm
Gun, et al.
(1978)
o
N = 8, C = 8
c
V
ne
V
ne
ne
ne
N
ne
ne
3-418ppm
Hirsch, et al.
(1996)
E
o
II
£
c
ne
ne
ne
ne
ne
H
ne
ne
ne
0-2,441 ppb
Kilburn and
Thornton
(1996)
E
N = 237, C = 264
c
V
ne
V
ne
ne
ne
ne
ne
ne
ne
Kilburn and
Warshaw
(1993)
E
N= 544, C= 181
c
V
V
V
V
V
M
T, N
ne
ne
6-500 ppb
Kilburn
(2002b)
E
N = 236, C = 228
c
ne
ne
V
ne
ne
M
B
ne
ne
6-500 ppb
Kilburn
(2002a)
E
N = 236, C = 58
c
(-)
ne
ne
ne
(-)
ne
ne
ne
ne
0.2-1,000 ppb
Konietzko, et al.
(1975)
C
2
II
to
o
A
ne
ne
ne
ne
ne
M
N
ne
V
953 ppm
>3
o
o
s
s ?

°S
TO
*
«
s
TO
Co
0
S
1
a.
to
o
VO
-------
>3
o
o
s
Co
I?
S"4
>}'
§•
TO
s
a
to
s
S"
a
a,
>r
<§>
>s
>s
to
TO'
*
^S
«
s
TO
Co
0
S
1
a.
Kylin, et al.
(1967)
C
N= 12
A
V
ne
ne
ne
ne
ne
N
ne
ne
1,000 ppm
Landrigan, et al.
(1987)
O
Residents and 12 W
A,C
ne
ne
V
ne
ne
H, D
ne
ne
vv
>183,000 ppb
Liu, et al.
(1988)
o
N = 103,C= 111
C
ne
ne
ne
V
ne
D, N
N
ne
vv
1-100 ppm
Mhiri et al.
(2004)
o
N = 23, C = 23
A
ne
ne
ne
ne
ne
ne
T
ne
V, vv
ne
Nagaya et al.
(1990)
o
N = 84, C = 83
C
ne
ne
ne
ne
ne
ne
N
ne
V
22 ppm
VO
to
o
-------
IS*
Table D-3. Literature review of studies of TCE and domains assessed with neurobehavioral/neurological methods
(continued)
Authors
Year
Study
type
Participants no. (N =
exposed
C = nonexposed)
Dur
PM/RT
VM
Cogn
M&L
M&P
Sympf
Senff
Resp
Dose effect
vv
urinary
metabolitesV
TCE levels
Rasmussen and
Sabroe
(1986)
O
N = 240, C = 350
C
ne
ne
ne

V
H,D, I, M
ne
ne
ne
ne
Rasmussen et al.
(1993d)
O
N= 96
C
ne
ne
V
ne
ne
ne
ne
ne

ne
Rasmussen et al.
(1993c)
o
N= 96
c
ne
V
V
ne
ne
ne
ne
ne

ne
Rasmussen et al.
(1993a)
o
N= 99
c
V
ne
ne
ne
ne
ne
N
ne
VV
ne
Reif et al.
(2003)
E
N = 143
c
V
V
ne
ne
V
M
M
ne
VV
5-15 ppb
Ruijten, et al.
(1991)
o
N = 31, C = 28
c
V
ne
ne
ne
ne
ne
ne
ne
ne
17-70 ppm
Smith
(1970)
o
N = 130, C = 63
c
ne
ne
ne
ne
ne
H, D
N
ne
V, -VV
ne
Stewart et al
(1970)
c
N = 13
A
ne
ne
V
ne
ne
H
ne
ne
V
100-202 ppm
Triebig, et al.
(1976)
c
N = 7, C = 7
A
ne
ne
V
V
V
(-)
ne
ne
V, vv
0-100 ppm
Triebig, et al.
(1977a)
c
N = 7, C = 7
A
ne
ne
V
V
V
M
(-)
ne
V, vv
0-100 ppm
Triebig, et al.
(1977b)
o
N = 8
A,C
ne
V
V
V
ne
ne
ne
ne
V
50 ppm
Triebig, et al.
(1982)
o
N = 24, C = 24
C
ne
ne
ne
ne
ne
ne
N
ne
V, -VV
5-70 ppm
Triebig, et al.
(1983)
o
N = 66, C = 66
c
ne
ne
ne
ne
ne
N,H
N
ne
V
10-600 mg/m3
Troster and Ruff
(1990)
o
N = 3, C = 60
A
V
V
V
V
V
ne
N
ne
ne
ne
Vernon and
Ferguson
(1969)
c
N = 8
A
V
V
ne
ne
ne
ne
N
ne
«
0-1000 ppm
Windemuller and
Ettema
(1978)
c
N= 39
A
V
ne
ne
ne
ne
ne
ne
ne
ne
200 ppm
Winneke
(1982)
o
Not reported
ne
(-)
(-)
ne
ne
ne
ne
ne
ne
ne
50 ppm
fH = Headaches; D = Dizziness; I = Insomnia; S = Sex Probls; M = Mood; N = Neurological.
-------
f f A = Audition; B = Balance; V = Vision; T = Trigeminal nerve; N = Other Neurological.
IS*
£< Study: C = Chamber; E = Environmental; O = Occupational.
C>
to
£
to Duration: A = Acute, C = Chronic
a, S
C> TO ,
S S ^ = positive findings; (-) = findings not significant; ne = not examined or reported; Dur = duration; PM/RT = psychomotor/reaction time; VM = visuo-motor; Cogn = cognitive;
>3
«S S3" M&L = memory and learning; M&P = mood and personality; Symp = symptoms; Sen = sensory; Resp = respiratory.
S. »
£
O >r
S vgs
>
S ?
^ 5
Oq 5.
to
^ ^3
^3 5
S. ^
.TO Co
V; to
Co
0
s
1
a.
VO
to
o
-------
D.2. CENTRAL NERVOUS TOXICITY IN ANIMAL STUDIES FOLLOWING
TRICHLOROETHYLENE (TCE) EXPOSURE
In vivo studies in animals and in vitro models have convincingly demonstrated that TCE
produces functional and physiological neurological changes. Overall, these effects collectively
indicate that TCE has central nervous system (CNS) depressant-like effects at lower exposures
and causes anesthetic-like effects at high exposures. Studies of TCE toxicity in animals have
generally not evaluated whether or not adverse effects seen acutely persist following exposure or
whether there are permanent effects of exposure. Exceptions to the focus on acute impairment
while under TCE intoxication include studies of hearing impairment and histopathological
investigations focused primarily on specific neurochemical pathways, hippocampal development,
and demyelination. These persistent TCE effects are discussed initially followed by the results
of studies that examined the acute effects of this agent. Summary tables for all the animal
studies are at the end of this section.
D.2.1. Alterations in Nerve Conduction
There is little evidence that TCE disrupts trigeminal nerve function in animal studies.
Two studies demonstrated TCE produces morphological changes in the trigeminal nerve at a
dose of 2,500 mg/kg-day for 10 weeks (1992; 1991). However, dichloroacetylene, a degradation
product formed during the volatilization of TCE was found to produce more severe
morphological changes in the trigeminal nerve and at a lower dose of 17 mg/kg-day (Barret et
al., 1992; Barret et al., 1991). Only one study (Albee et al., 2006) has evaluated the effects of
TCE on trigeminal nerve function, and a subchronic inhalation exposure did not result in any
significant functional changes. A summary of these studies is provided in Table D-4.
Barret et al. (1992; 1991) conducted two studies evaluating the effects of both TCE and
dichloroacetylene on trigeminal nerve fiber diameter and internodal length as well as several
markers for fiber myelination. Female Sprague Dawley rats (n = 7/group) were dosed with
2,500 mg/kg TCE or 17 mg/kg-day dichloroacetylene by gavage for 5 days/week for 10 weeks.
These doses were selected based upon the ratio of the LD50S (dose at which there is 50%
lethality) for these two agents. Two days after administration of the last dose, a morphometric
approach was used to study the diameter of teased fibers from the trigeminal nerve. The fibers
were classified as Class A or Class B and evaluated for internode length and fiber diameter.
TCE-dosed animals only exhibited changes in the smaller Class A fibers where internode length
increased marginally (<2%) and fiber diameter increased by 6%. Conversely, dichloroacetylene-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-109 DRAFT—DO NOT CITE OR QUOTE
-------
treated rats exhibited significant and more robust decreases in internode length and fiber
diameter in both fiber classes A and B. Internode length decreased 8% in Class A fibers and 4%
in Class B fibers. Fiber diameter decreased 10% in Class A fibers and 6% in Class B fibers.
Biochemical data are presented for fatty acid composition from total lipid extractions from the
trigeminal nerve. These two studies identify a clear effect of dichloroacetylene on trigeminal
nerve fibers, but the effect by TCE is quite limited.
Albee et al. (2006) evaluated the effects of a subchronic inhalation TCE exposure in
Fischer 344 rats (10/sex/group). Rats were exposed to 0-, 250-, 800-, and 2,500-ppm TCE for
6 hours/day, 5 days/week for 13 weeks. At the eleventh week of exposure, rats were surgically
implanted with epidural electrodes over the somatosensory and cerebellar regions, and TSEPs
were collected 2-3 days following the last exposure. TSEPs were generated using subcutaneous
needle electrodes to stimulate the vibrissal pad (area above the nose). The resulting TSEP was
measured with electrode previously implanted over the somatosensory region. The TCE
exposures were adequate to produce permanent auditory impairment even though TSEPs were
unaffected. While TCE appears to be negative in disrupting the trigeminal nerve, the TCE
breakdown product, dichloroacetylene, does impair trigeminal nerve function.
Albee et al. (1997) reported that dichloroacetylene disrupted trigeminal nerve
somatosensory evoked potentials in Fischer 344 male rats. The subjects were exposed to a
mixture of 300-ppm dichloroacetylene, 900-ppm acetylene, and 170-ppm TCE for a single
2.25-hour period. This dichloroacetylene was generated by decomposing TCE in the presence of
potassium hydroxide and stabilizing with acetylene. A second treatment group was exposed to a
175-ppm TCE/l,030-ppm acetylene mix with no potassium hydroxide present. Therefore, no
dichloroacetylene was present in the second treatment group, providing an opportunity to
determine the effects on the trigeminal nerve somatosensory evoked potential in the absence of
dichloroacetylene. Evoked potentials from the dichloroacetylene/TCE/acetylene-exposed rats
were about 17% smaller measured between peaks I and II and 0.13 msec slower in comparison to
the preexposure measurements. Neither latency nor amplitude of this potential changed
significantly between the preexposure and postexposure test in the air-exposed animals (control).
The dichloroacetylene-mediated evoked potential changes persisted at least until Day 4
postexposure. No changes in evoked potentials were observed in the 175-ppm TCE/l,030-ppm
acetylene mix group. It is noteworthy that dichloroacetylene treatment produced broader
evidence of toxicity as witnessed by a persistent drop in body weight among subjects over the
7-day postexposure measuring period. In light of the differences observed between the effects of
TCE and dichloroacetylene on the trigeminal nerve, it would be instructive to calculate the dose
of TCE that would be necessary to produce comparable tissue levels of dichloroacetylene
produced in the Albee et al. (1997) study.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-110 DRAFT—DO NOT CITE OR QUOTE
-------
Kulig (1987) also measured peripheral (caudal nerve) nerve conduction time in male
Wistar rats and failed to show an effect of TCE with exposures as high as 1,500 ppm for
16 hours/day, 5 days/week for 18 weeks.
D.2.2. Auditory Effects
D.2.2.1. Inhalation
The ability of TCE to disrupt auditory function and produce inner ear histopathology
abnormalities has been demonstrated in several studies using a variety of test methods. Two
different laboratories have identified NOAELs for auditory function of 1,600 ppm following
inhalation exposure for 12 hours/day for 13 weeks in Long Evans rats (n = 6-10) (Rebert et al.,
1991) and 1,500 ppm in Wistar-derived rats (n = 12) exposed by inhalation for 18 hours/day,
5 days/week for 3 weeks (Jaspers et al., 1993). The LOAELs identified in these and similar
studies are 2,500-4,000-ppm TCE for periods of exposure ranging from 4 hours/day for 5 days
to 12 hours/day for 13 weeks (e.g., Albee et al., 2006; Boyes et al., 2000; Crofton and Zhao,
1997; Crofton et al., 1994; Fechter et al., 1998; Muijser et al., 2000; Rebert et al., 1993; Rebert et
al., 1995). Rebert et al. (1993) estimated acute blood TCE levels associated with permanent
hearing impairment at 125 [j,g/mL by methods that probably underestimated blood TCE values
(rats were anaesthetized using 60% carbon dioxide). A summary of these studies is presented in
Table D-5.
Rebert et al. (1991) evaluated auditory function in male Long Evans rats (n = 10) and
F344 rats (n = 4-5) by measuring brainstem auditory-evoked responses (BAERs) following
stimulation with 4-, 8-, and 16-kHz sounds. The Long-Evans rats were exposed to 0-, 1,600-, or
3,200-ppm TCE, 12 hour/day for 12 weeks and the F344 rats were exposed to 0-, 2,000-, or
3,200-ppm TCE, 12 hours/day for 3 weeks. BAERs were measured every 3 weeks during the
exposure and then for an additional 6 weeks following the end of exposure. For the F344 rats,
both TCE exposure (2,000 and 3,200 ppm) significantly decreased BAER amplitudes at all
frequencies tested. In comparison, Long Evans rats exposed to 3,200-ppm TCE also had
significantly decreased BAER amplitude, but exposure to 1,600 ppm did not significantly affect
BAERs at any stimulus frequency. These data suggest a LOAEL at 2,000 ppm for the F344 rats
and a NOAEL at 1,600 ppm for the Long Evans rats. In subsequent studies, Rebert et al. (1993;
1995) again demonstrated TCE significantly decreases BAER amplitudes and significantly
increases the latency of the initial peak (identified as PI).
Jaspers et al. (1993) exposed Wi star-derived WAG-Rii/MBL rats (n = 12) to 0, 1,500 and
3,000-ppm TCE exposure for 18 hours/day, 5 days/week for 3 weeks. Auditory function for
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-111 DRAFT—DO NOT CITE OR QUOTE
-------
each frequency was assessed by reflex modification (recording the decibel threshold required to
generate a startle response from the rat). Three tones (5, 20, and 35 kHz) were used to test
auditory function. The startle measurements were made prior to exposure and at 1, 3, 5, and
6 weeks after exposure. A selective impairment of auditory threshold for animals exposed to
3,000-ppm TCE was observed at all postexposure times at 20 kHz only. No significant effects
were noted in rats exposed to 1,500-ppm TCE. This auditory impairment was persistent up
through 6 weeks after exposure, which was the last time point presented. There was no
impairment of hearing at either 5 or 25 kHz for animals exposed to 1,500- or 3,000-ppm TCE.
This study indicates TCE selectively produces a persistent mid-frequency hearing loss and
identifies a NOAEL of 1,500 ppm. Similarly, Crofton et al. (1994) exposed male Long Evans
rats (n = 7-8) to 3,500-ppm TCE, 8 hours/day for 5 days. Auditory thresholds were determined
by reflex modification audiometry 5-8 weeks after exposure. TCE produced a selective
impairment of auditory threshold for mid frequency tones, 8 and 16 kHz.
Muijser et al. (2000) evaluated the ability of TCE to potentiate the damaging effect of
noise on hearing. Wistar rats (n = 8 per group) were exposed by inhalation to 0 or 3,000-ppm
TCE alone for 18 hours/day, 5 days/week for 3 weeks (no noise) or in conjunction with 95-dB
broad band noise. The duration of noise exposure is not specified, but presumably was also
18 hours/day, 5 days/week for 3 weeks. Pure tone auditory thresholds were determined using
reflex modification audiometry 1 and 2 weeks following the exposures. Significant losses in
auditory sensitivity were observed for rats exposed to noise alone at 8, 16, and 20 kHz, for rats
exposed to TCE alone at 4, 8, 16, and 20 kHz and for combined exposure subjects at 4, 8, 16, 20,
and 24 kHz. The loss of hearing sensitivity at 4 kHz is particularly striking for the combined
exposure rats, suggesting a potentiation effect at this frequency. Impairment on this auditory test
suggests toxicity at the level of the cochlea or brainstem.
Fechter et al. (1998) exposed Long Evans rats inhalationally to 0 or 4,000-ppm TCE
6 hours/day for 5 days. Three weeks later auditory thresholds were assessed by reflex
modification audiometry (n = 12), and then 5-7 weeks later, cochlear function was assessed by
measuring compound action potentials (CAPs) and the cochlear microphonic response
(n = 3-10). Cochlear histopathology was assessed at 5-7 weeks (n = 4) using light microscopy.
Reflex modification thresholds were significantly elevated at 8 and 18 kHz, as were CAP
thresholds. The growth of the N1 evoked potential was reduced in the TCE group, and they
failed to show normal N1 amplitudes even at supra-threshold tone levels. There was no effect on
the sound level required to elicit a cochlear microphonic response of 1 |iV. Histological data
suggest that TCE produces a loss of spiral ganglion cells.
Albee et al. (2006) exposed male and female F344 rats to TCE at 250, 800, or 2,500 ppm
for 6 hours/day, 5 days/week, for 13 weeks. At 2,500-ppm TCE, mild frequency-specific
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-112 DRAFT—DO NOT CITE OR QUOTE
-------
hearing deficits were observed, including elevated tone-pip auditory brainstem response
thresholds. Focal loss of hair cells in the upper basal turn of the cochlea was observed in
2,500-ppm-exposed rats; this was apparently based upon midmodiolar sections, which lack
power in quantification of hair cell death. Except for the cochleas of 2,500-ppm-exposed rats, no
treatment-related lesions were noted during the neuro-histopathologic examination. The
NOAEL for this study was 800 ppm based on ototoxicity at 2,500 ppm.
The relationship between dose and duration of exposure with respect to producing
permanent auditory impairment was presented in Crofton and Zhao (1997) and again in Boyes et
al. (2000). The LOAELs identified in Long Evans rats (n = 10-12) were 6,000 ppm for a 1-day
exposure, 3,200 ppm per day for both the 1- and 4-week exposures, and 2,400 ppm per day for
the 13-week exposure. It was estimated from these data that the LOAEL for a 2-year long
exposure would be 2,100 ppm. Auditory thresholds were determined for a 16-kHz tone
3-5 weeks after exposure using reflex modification audiometry. Results replicated previous
findings of a hearing loss at 16 kHz for all exposure durations. One other conclusion reached by
this study is that TCE concentration and not concentration x duration of exposure is a better
predictor of auditory toxicity. That is, the notion that total exposure represented by the function,
concentration (C) x time (t), or Haber's law, is not supported. Therefore, higher exposure
concentrations for short durations are more likely to produce auditory impairment than are lower
concentrations for more protracted durations when total dosage is equated. Thus, consideration
needs to be given not only to total C x t, but also to peak TCE concentration.
Crofton and Zhao (1997) also presented a benchmark dose for which the calculated dose
of TCE would yield a 15-dB loss in auditory threshold. This benchmark response was selected
because a 15-dB threshold shift represents a significant loss in threshold sensitivity for humans.
The benchmark concentrations for a 15-dB threshold shift are 5,223 ppm for 1 day, 2,108 ppm
for 5 days, 1,418 ppm for 20 days, and 1,707 ppm for 65 days of exposure. While more sensitive
test methods might be used and other definitions of a benchmark effect chosen with a strong
rationale, these data provide useful guidance for exposure concentrations that do yield hearing
loss in rats.
These data demonstrate that the ototoxicity of TCE was less than that predicted by a strict
concentration x time relationship. These data also demonstrate that simple models of
extrapolation (i.e., C x t = k, Haber's Law) overestimate the potency of TCE when extrapolating
from short-duration to longer-duration exposures. Furthermore, these data suggest that, relative
to ambient or occupational exposures, the ototoxicity of TCE in the rat is a high-concentration
effect; however, the selection of a 15-dB threshold for detecting auditory impairment along with
tests at a single auditory frequency may not capture the most sensitive reliable measure of
hearing impairment.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-113 DRAFT—DO NOT CITE OR QUOTE
-------
With the exception of a single study performed in the Hartley guinea pig (n = 9-10)
(Yamamura et al., 1983), there are no data in other laboratory animals related to TCE-induced
ototoxicity. Yamamura et al. (1983) exposed Hartley guinea pigs to TCE at doses of 6,000,
12,000, and 17,000 ppm for 4 hours/day for 5 days and failed to show an acute impairment of
auditory function. However, despite the negative finding in this study, it should be considered
that auditory testing was performed in the middle of a laboratory and not in an audiometric sound
attenuating chamber. The influence of extraneous and uncontrolled noise on cochlear
electrophysiology is marked and assesses auditory detection thresholds in such an environment
unrealistic. Although the study has deficiencies, it is important to note that the guinea pig has
been reported to be far less sensitive than the rat to the effects of ototoxic aromatic hydrocarbons
such as toluene.
It may be helpful to recognize that the effects of TCE on auditory function in rats are
quite comparable to the effects of styrene (e.g., Campo et al., 2006; Crofton et al., 1994; Pryor et
al., 1987), toluene (e.g., Campo et al., 1999; Pryor et al., 1983), ethylbenzene (e.g, Cappaert et
al., 2000; Cappaert et al., 1999; Fechter et al., 2007), and /^-xylene (e.g., Gagnaire et al., 2001;
Pryor et al., 1987). All of these aromatic hydrocarbons produce reliable impairment at the
peripheral auditory apparatus (inner ear), and this impairment is associated with death of sensory
receptor cells, the outer hair cells. In comparing potency of these various agents to produce
hearing loss, it appears that TCE is approximactely equipotent to toluene and less potent than, in
order, ethylbenzene, p-xylene, and styrene. Occupational epidemiological studies do appear to
identify auditory impairments in workers who are exposed to styrene (Morata et al., 2002;
Morioka et al., 2000; Sliwinska-Kowalska et al., 1999) and those exposed to toluene (Abbate et
al., 1993; Morata et al., 1997), particularly when noise is also present.
D.2.2.2. Oral and Injection Studies
No experiments were identified in which auditory function was assessed following TCE
administration by either oral or injection routes.
D.2.3. Vestibular System Studies
The effect of TCE on vestibular function was evaluated by either (1) promoting
nystagmus (vestibular system dysfunction) and comparing the level of effort required to achieve
nystagmus in the presence and absence of TCE or (2) using an elevated beam apparatus and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-114 DRAFT—DO NOT CITE OR QUOTE
-------
measuring the balance. Overall, it was found that TCE disrupts vestibular function as presented
below. Summary of these studies is found in Table D-6.
Tham et al. (1984; 1979) demonstrated disruption in the stimulated vestibular system in
rabbits and Sprague Dawley rats during intravenous (i.v.) infusion with TCE. It is difficult to
determine the dosage of TCE necessary to yield acute impairment of vestibular function since
testing was performed under continuing infusion of a lipid emulsion containing TCE, and
therefore, blood TCE levels were increasing during the course of the study. Tham et al. (1979),
for example, infused TCE at doses of 1-5 mg/kg/min reaching arterial blood concentrations as
high as 100 ppm. They noted increasing numbers of rabbits experiencing positional nystagmus
as blood TCE levels increased. The most sensitive rabbit showed nystagmus at a blood TCE
concentration of about 25 ppm. Similarly, the Sprague Dawley rats also experienced increased
nystagmus with a threshold effect level of 120 ppm as measured in arterial blood (Tham et al.,
1984). Animals demonstrated a complete recovery in vestibular function when evaluated for
nystagmus within 5-10 minutes after the i.v. infusion was stopped.
Niklasson et al. (1993) showed acute impairment of vestibular function in male and
female pigmented rats during acute inhalation exposure to TCE (2,700-7,200 ppm) and to
trichloroethane (500-2,000 ppm). Both of these agents were able to promote nystagmus during
optokinetic stimulation in a dose related manner. While there were no tests performed to assess
persistence of these effects, Tham et al. (1984; 1979) did find complete recovery of vestibular
function in rabbits (n = 19) and female Sprague-Dawley rats (// = 11) within minutes of
terminating a direct arterial infusion with TCE solution.
The finding that trichloroethylene can yield transient abnormalities in vestibular function
is not unique. Similar impairments have been shown for toluene, styrene, along with
trichloroethane (Niklasson et al., 1993) and by Tham et al. (1984) for a broad range of aromatic
hydrocarbons. The concentration of TCE in blood at which effects were observed for TCE
(0.9 mM/L) was quite close to that observed for most of these other vestibulo-active solvents.
D.2.4. Visual Effects
Changes in visual function have also been demonstrated in animal studies following acute
(Boyes et al., 2003; Boyes et al., 2005) and subchronic exposure (Blain et al., 1994). Summary
of all TCE studies evaluating visual effects in animals can be found in Table D-6. In these
studies, the effect of TCE on visual-evoked responses to patterns (Boyes et al., 2003; Boyes et
al., 2005; Rebert et al., 1991) or a flash stimulus (Blain et al., 1994; Rebert et al., 1991) were
evaluated. Overall, the studies demonstrated that exposure to TCE results in significant changes
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-115 DRAFT—DO NOT CITE OR QUOTE
-------
in the visual evoked response, which is reversible once TCE exposure is stopped. Only one
study (Rebert et al., 1991) did not demonstrate changes in visual system function with a
subchronic TCE exposure, but visual testing was conducted 10 hours after each exposure.
Boyes et al. (2003; 2005) found significant reduction in the visual evoked potential
acutely while Long Evans male rats were being exposed to TCE concentrations of 500, 1,000,
2,000, 3,000, 4,000, and 5,000 ppm for intervals ranging from 4 to 0.5 hours, respectively. In
both instances, the degree of effect correlated more with brain TCE concentrations than with
duration of exposure.
Boyes et al. (2003) exposed adult, male Long-Evans rats to TCE in a head-only exposure
chamber while pattern onset/offset visual evoked potentials (VEPs) were recorded. Exposure
conditions were designed to provide C x t products of 0 ppm/hour (0 ppm for 4 hours) or
4,000 ppm/hour created through four exposure scenarios: 1,000 ppm for 4 hours; 2,000 ppm for
2 hours; 3,000 ppm for 1.3 hours; or 4,000 ppm for 1 hour (n = 9-10/concentration). Blood TCE
concentrations were assessed by GC with ECD, and brain TCE concentrations were estimated
using a physiologically based pharmacokinetic (PBPK) model. The amplitude of the VEP
frequency double component (F2) was decreased significantly (p < 0.05) by exposure. The mean
amplitude (VSEM in (j,V) of the F2 component in the control and treatment groups measured
4.4 V 0.5 (0 ppm/4 hours), 3.1V 0.5 (1,000 ppm/4 hours), 3.1V 0.4 (2,000 ppm/2 hours),
2.3 V 0.3 (3,000 ppm/1.3 hours), and 1.9 V 0.4 (4,000 ppm/1 hour). A PBPK model was used to
estimate the concentrations of TCE in the brain achieved during each exposure condition. The
F2 amplitude of the VEP decreased monotonically as a function of the estimated peak brain
concentration but was not related to the area under the curve of the brain TCE concentration.
These results indicate that an estimate of the brain TCE concentration at the time of VEP testing
predicted the effects of TCE across exposure concentrations and duration.
In a follow-up study, Boyes et al. (2005) exposed Long Evans male rats
(n = 8-10/concentration) to TCE exposures of 500 ppm for 4 hours, 1,000 ppm for 4 hours,
2,000 ppm for 2 hours, 3,000 ppm for 1.3 hours, 4,000 ppm for 1 hour and 5,000 ppm for
0.8 hour. VEP recordings were made at multiple time points, and their amplitudes were adjusted
in proportion to baseline VEP data for each subject. VEP amplitudes were depressed by TCE
exposure during the course of TCE exposure. The degree of VEP depression showed a high
correlation with the estimated brain TCE concentration for all levels of atmospheric TCE
exposure.
This transient effect of TCE on the peripheral visual system has also been reported by
Blain (1994) in which New Zealand albino rabbits were exposed by inhalation to 350- and
700-ppm TCE 4 hours/day, 4 days/week for 12 weeks. Electroretinograms (ERGs) and
oscillatory potentials (OPs) were recorded weekly under mesopic conditions. Recordings from
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-116 DRAFT—DO NOT CITE OR QUOTE
-------
the 350- and 700-ppm exposed groups showed a significant increase in the amplitude of the a-
and b-waves (ERG). The increase in the a-wave was dose related increasing 30% at the low dose
and 84% in the high dose. For the b-wave, the lower exposure dose yielded a larger change from
baseline (52%) than did the high dose (33%). The amplitude of the OPs was significantly
decreased at 350 ppm {51%) and increased at 700 ppm (117%). The decrease in the oscillatory
potentials (OPs) shown in the low-dose group appears to be approximately 25% from
9-12 weeks of exposure. These electroretinal changes were reversed to the baseline value within
6 weeks after the inhalation stopped.
Rebert et al. (1991) evaluated visual evoked potentials (flash evoked potentials and
pattern reversal evoked potentials) in male Long Evans rats that received 1,600- or 3,200-ppm
TCE for 3 weeks 12 hours/day. No significant changes in flash evoked potential measurements
were reported following this exposure paradigm. Limited shifts in pattern reversal visual evoked
potentials were reported during subchronic exposure, namely a reduction in the Nl-Pl response
amplitude that reached statistical significance following 8, 11, and 14 weeks of exposure. The
drop in response amplitude ranged from approximately 20% after 8 weeks to nearly 50% at
Week 14. However, this potential recovered completely during the recovery period.
D.2.5. Cognitive Function
There have been a number of reports (e.g., Kishi et al., 1993; Kjellstrand et al., 1980;
Kulig, 1987) showing alteration in performance in learning tasks such as a change in speed to
complete the task, but little evidence that learning and memory function are themselves impaired
by exposure. Table D-7 presents the study summaries for animal studies evaluating cognitive
effects following TCE exposure. Such data are important in efforts to evaluate the functional
significance of decreases in myelinated fibers in the hippocampus reported by Isaacson et al.
(1990) and disruption of long-term potentiation discovered through in vitro testing (Ohta et al.,
2001) since the hippocampus has been closely tied to memory formation.
Kjellstrand et al. (1980) exposed Mongolian gerbils (n = 12/sex) to 900-ppm TCE by
inhalation for 9 months. Inhalation was continuous except for 1-2 hours/week for cage cleaning.
Spatial memory was tested using the radial arm maze task. In this task, the gerbils had to visit
each arm of the maze and remember which arm was visited and unvisited in selecting an arm to
visit. The gerbils received training and testing in a radial arm maze starting after 2 months of
TCE exposure. There was no effect of TCE on learning or performance on the radial arm maze
task.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-117 DRAFT—DO NOT CITE OR QUOTE
-------
Kishi et al. (1993) acutely exposed Wistar rats to TCE at concentrations of 250, 500,
1,000, 2,000, and 4,000 ppm for 4 hours. Rats were tested on an active (light) signaled shock
avoidance operant response. Rats exposed to 250-ppm TCE showed a significant decrease both
in the total number of lever presses and in avoidance responses at 140 minutes of exposure
compared with controls. The rats did not recover their pre-exposure performance until
140 minutes after the exhaustion of TCE vapor. Exposures in the range 250- to 2,000-ppm TCE
for 4 hours produced concentration related decreases in the avoidance response rate. No
apparent acceleration of the reaction time was seen during exposure to 1,000- or 2,000-ppm
TCE. The latency to a light signal was somewhat prolonged during the exposure to 2,000- to
4,000-ppm TCE. It is estimated that there was depression of the central nervous system with
slight performance decrements and the corresponding blood concentration was 40 ng/mL during
exposure. Depression of the central nervous system with anesthetic performance decrements
was produced by a blood TCE concentration of about 100 [j,g/mL. In general, they observed
dose related reductions in total number of lever presses, but these changes may be more
indicative of impaired motor performance than of cognitive impairment. In any event, recovery
occurred rapidly once TCE exposure ceased.
Isaacson et al. (1990) studied the effects of oral TCE exposure in weanling rats at
exposure doses of 5.5 mg/day for 4 weeks, followed by an additional 2 weeks of exposure at
8.5 mg/day. No significant changes were observed in locomotor activity in comparison to the
control animals. This group actually reported improved performance on a Morris swim test of
spatial learning as reflected in a decrease in latency to find the platform from 14 seconds in
control subjects to 12 seconds in the lower dose TCE group to a latency of 9 seconds in the
higher TCE group. The high dose TCE group differed significantly from the control and low
TCE dose groups while these latter two groups did not differ significantly from each other. This
improvement relative to the control subjects occurred despite a loss in hippocampal myelination,
which approached 8% and was shown to be significant using Duncan's multiple range test.
Likewise, Umezu et al. (1997) exposed ICR strain male mice acutely to doses of TCE
ranging from 62.5-1,000 mg/kg depending upon the task. They reported a depressed rate of
operant responding in a conditioned avoidance task that reached significance with intraperitoneal
(i.p.) injections of 1,000 mg/kg. Increased responding during the signaled avoidance period at
lower doses (250 and 500 mg/kg) suggests an impairment in ability to inhibit responding or
failure to attend to the signal. However, all testing was performed under TCE intoxication.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-118 DRAFT—DO NOT CITE OR QUOTE
-------
D.2.6. Psychomotor Effects
Changes in psychomotor activity such as loss of righting reflex, functional observational
battery changes, and locomotor activity have been demonstrated in animals following exposure
to TCE. Summaries for some of these studies can be found below and are presented in detail in
Table D-8.
D.2.6.1. Loss of Righting Reflex
Kishi et al. (1993) evaluated the activity and performance of male Wistar rats in a series
of tasks following an acute 4-hour exposure to 250, 500, 1,000, 2,000, and 4,000 ppm. They
reported disruption in performance at the highest test levels with CNS depression and anesthetic
performance decrements. Blood TCE concentrations were about 100 ng/mL in Wistar rats (such
blood TCE concentrations were obtained at inhalation exposure levels of 2,000 ppm).
Umezu et al. (1997) studied disruption of the righting reflex following acute injection of
250, 500, 1,000, 2,000, 4000, and 5,000 mg/kg TCE in male ICR mice. At 2,000 mg/kg, loss of
righting reflex (LORR) was observed in only 2/10 animals injected. At 4,000 mg/kg,
9/10 animals experienced LORR, and 100% of the animals experienced LORR at 5,000 mg/kg.
Shih et al. (2001) reported impaired righting reflexes at exposure doses of 5,000 mg/kg in male
Mfl mic although lower exposure doses were not included. They showed, in addition, that
pretreatment prior to TCE with DMSO or disulfiram (which is a CYP2E1 inhibitor) in DMSO
could delay loss of the righting reflex in a dose related manner. By contrast, the alcohol
dehydrogenase inhibitor, 4-metylpyradine did not delay loss of the righting reflex that resulted
from 5,000 mg/kg TCE. These data suggest that the anesthetic properties of TCE involve its
oxidation via CYP2E1 to an active metabolite, a finding that is consistent with the anesthetic
properties of chloral hydrate.
D.2.6.2. Functional Observational Battery (FOB) and Locomotor Activity Studies
D.2.6.2.1. Functional observational battery (FOB) and locomotor activity studies with
trichloroethylene (TCE). A number of papers have measured locomotor activity and used
functional observational batteries (FOBs) in order to obtain a more fine grained analysis of the
motor behaviors that are impaired by TCE exposure. While exposure to TCE has been shown
repeatedly to yield impairments in neuromuscular function acutely, there is very little evidence
that the effects persist beyond termination of exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-119 DRAFT—DO NOT CITE OR QUOTE
-------
One of the most extensive evaluations of TCE on innate neurobehavior was conducted by
Moser et al. (2003; 1999) using FOB testing procedures. Moser et al. (1995) evaluated the
effects of acute and subacute (14-day) oral gavage administration of TCE in adult female Fischer
344 rats. Testing was performed both 4 hours post TCE administration and 24 hours after TCE
exposure, and a comparison of these two time points along with comparison between the first
day and the last day of exposure provides insight into the persistence of effects observed.
Various outcome measures were grouped into five domains: autonomic, activity, excitability,
neuromuscular, and sensorimotor. Examples of tests included in each of these groupings are as
follows: Autonomic—lacrimation, salivation, palpebral closure, pupil response, urination, and
defecation; Activity—rearing, motor activity counts home cage position. Excitability—ease of
removal, handling reactivity, arousal, clonic, and tonic movements; and Neuromuscular—gait
score, righting reflex, fore and hindlimb grip strength, and landing foot splay. Sensorimotor-tail-
pinch response, click response, touch response, and approach response. Scoring was performed
on a 4-point scale ranging from "1" (normal) to "4" (rare occurrence for control subjects). In the
acute exposure, the exposure doses utilized were 150, 500, 1,500, and 5,000 mg/kg TCE in corn
oil. These doses represent 3, 10, 30, and 56% of the limit dose. For the 14-day subacute
exposure, the doses used were 50, 150, 500, and 1,500 mg/kg. Such doses represent 1, 3, 10, and
30% of the limit dose for TCE.
The main finding for acute TCE administration is that a significant reduction in activity
level occurred after the highest dose of TCE (5,000 mg/kg) only. This effect showed substantial
recovery 24 hours after exposure though residual decrements in activity were noted.
Neuromuscular function as reflected in the gait score was also severely affected only at
5,000-mg/kg dose and only at the 4-hour test period. Sensorimotor function reflected in response
to a sudden click, was abnormal at both 1,500 and 5,000 mg/kg with a slight difference observed
at 1,500 mg/kg and a robust difference apparent at 5,000 mg/kg. Additional effects noted, but
not shown quantitatively were abnormal home-cage posture, increased landing foot splay,
impaired righting and decreased fore and hind limb grip strength. It is uncertain at which doses
such effects were observed.
With the exception of sensorimotor function, these same categories were also disrupted in
the subacute TCE administration portion of the study. The lack of effect of TCE on
sensorimotor function with repeated TCE dosing might reflect either habituation, tolerance, or an
unreliable measurement at one of the time points. Given the absence of effect at a range of
exposure doses, a true dose-response relationship cannot be developed from these data.
In the subacute study, there are no clearly reliable dose-related differences observed
between treated and control subjects. Rearing, a contributor to the activity domain, was elevated
in the 500-mg/kg dose group, but was normal in the 1,500-mg/kg group. The neuromuscular
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-120 DRAFT—DO NOT CITE OR QUOTE
-------
domain was noted as significantly affected at 15 days, but it is not clear which subtest was
abnormal. It appears that the limited group differences may be random among subjects unrelated
to exposure condition.
In a follow-up study, Moser et al. (2003) treated female Fischer 344 rats with TCE by
oral gavage for periods of 10 days at doses of 0, 40, 200, 800, and 1,200 mg/kg/d, and testing
was undertaken either 4 hours following the first or 10111 dose as well as 24 hours after these two
time points. The authors identified several significant effects produced by TCE administration
including a decrease in motor activity, tail pinch responsiveness, reactivity to handling, hind limb
grip strength, and body weight. Rats administered TCE also showed significantly more
piloerection, higher gait scores, lethality, body weight loss, and lacrimation compared to
controls. Only effects observed 4 hours after the 10th exposure dose were presented by the
authors, and no quantitative information of these measurements is provided.
Albee et al. (2006) exposed male and female Fischer 344 rats to 250-, 800-, and
2,500-ppm TCE for 6 hours/day, 5 days/week for 13 weeks. FOB was performed 4 days prior to
exposure and then monthly. Auditory impairments found by others (e.g., Boyes et al., 2000;
Crofton and Zhao, 1997; Crofton et al., 1994; Fechter et al., 1998; Muijser et al., 2000; Rebert et
al., 1995) were replicated at the highest exposure dose, but treatment related differences in grip
strength or landing foot splay were not demonstrated. The authors report slight increases in
handling reactivity among female rats and slightly more activity than in controls at an
intermediate time point, but apparently did not conduct systematic statistical analyses of these
observations. In any event, there were no statistically significant effects on activity or reactivity
by the end of exposure.
Kulig (1987) also failed to show significant effects of TCE inhalation exposure on
markers of motor behavior. Wistar rats exposed to 500, 1,000, and 1,500 ppm for 16 hours/day,
5 days/week for 18 weeks failed to show changes in spontaneous activity, grip strength, or
coordinated hind limb movement. Measurements were made every three weeks during the
exposure period and occurred between 45 minutes and 180 minutes following the previous TCE
inhalation exposure. This study establishes a NOAEL of 1,500-ppm TCE with an exposure
duration of 16 hours/day.
D.2.6.2.2. Acute and subacute oral exposure to dichloroacetic acid on functional
observational batteries (FOB). Moser et al. (1999) conducted a series of experiments on
DCA
ranging from acute to chronic exposures. The exposure doses used in the acute experiment were
100, 300, 1,000, and 2,000 mg/kg. In the repeated exposure studies (8 weeks-24 months), doses
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-121 DRAFT—DO NOT CITE OR QUOTE
-------
varied between 16 and 1,000 mg/kg/d. The authors showed pronounced neuromuscular changes
in Long Evans and F344 rats dosed orally with the TCE metabolite, DCA, over a period ranging
from 9 weeks to 24 months at different exposure doses. Using a multitude of exposure protocols
which most commonly entailed daily exposures to DCA either by gavage or drinking water the
authors identify effects that were "mostly limited" to the neuromuscular domain. These included
disorders of gait, grip strength, foot splay and righting reflex that are dose and duration
dependent. Data on gait abnormality and grip strength are presented in greatest detail. In adults
exposed to DCA by gavage, gait scores were "somewhat abnormal" at the 7-week test in both the
adult Long Evans rats receiving 300 and those receiving 1,000 mg/kg/d. There was no adverse
effect in the rats receiving 100 mg/kg/d. In the chronic study, which entailed intake of DCA via
drinking water yielding an estimated daily dose of 137 and 235 mg/kg/d "moderately to severely
abnormal" gait was observed within 2 months of exposure and dosing was either reduced or
discontinued because of the severity of toxicity. For the higher DCA dose, gait scores remained
"severely abnormal" at the 24-month test time even though the DCA had been discontinued at
the 6-month test time. Hindlimb grip strength was reduced to about V2 the control value in both
exposure doses and remained reduced throughout the 24 months of testing even though DCA
administration ceased at 6 months for the 235 mg/kg/d group. Forelimb grip strength showed a
smaller and apparently reversible effect among DCA treated rats.
D.2.6.3. Locomotor Activity
Wolff and Siegmund (1978) administered 182 mg/kg TCE (i.p.) in AB mice and
observed a decrease in spontaneous locomotor activity. In this study, AB mice were injected
with TCE 30 minutes prior to testing for spontaneous activity at one of 4 time points during a
24 hours/day (0600, 1200, 1800, and 2400 hours). Marked decreases (estimated 60-80% lower
than control mice) in locomotor activity were reported in 15-minute test periods. The reduction
in locomotion was particularly profound at all time intervals save for the onset of light (0600).
Nevertheless, even at this early morning time point, activity was markedly reduced from control
levels (60% lower than controls as approximated from a graph).
Moser et al. (1995; 2003) included locomotor activity as one of their measures of
neurobehavioral effects of TCE given by gavage over a 10-14 day period. In the 1995 paper,
female Fischer 344 rats were dosed either acutely with 150, 500, 1,500 or 5,000 mg/kg TCE or
for 14 days with 50, 150, 500 or 1,500 mg/kg. In terms of the locomotor effects, they report that
acute exposure produced impaired locomotor scores only at 5,000 mg/kg while in the subacute
study, locomotion was impaired at the 500 mg/kg dose, but not at the 1,500 mg/kg dose. In the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-122 DRAFT—DO NOT CITE OR QUOTE
-------
Moser (2003) study, it appears that 200 mg/kg TCE may actually have increased locomotor
activity while the higher test doses (800 and 1,200 mg/kg) decreased activity in a dose related
manner. What is common to both studies, however, is a depression in motor activity that occurs
acutely following TCE administration and which may speak to the anesthetic if not central
nervous system depressive effects of this solvent.
There are also a number of reports (Fredriksson et al., 1993; Kulig, 1987; Waseem et al.,
2001) that failed to demonstrate impairment of motor activity or ability following TCE exposure.
Waseem et al. (2001) failed to show effects of TCE given in the drinking water of Wistar rats
over the course of a 90 day trial. While nominal solvent levels were 350, 700, and 1,400 ppm in
the water, no estimate is provided of daily TCE intake or of the stability of the TCE solution over
time. However, assuming a daily water intake of 25 mL/day and body weight of 330 g, these
exposures would be estimated to be approximately 26, 52, and 105 mg/kg. These doses are far
lower than those studied by Moser and colleagues.
Fredriksson et al. (1993) studied the effects of TCE given by oral gavage to male NMRI
mice at doses of 50 and 290 mg/kg/d from postnatal Day 10-16 on locomotion assessed either
on the day following exposure or at age 60 days. They found no significant effect of TCE on
locomotor activity and no consistent effects on other motor behaviors (e.g., rearing).
Waseem et al. (2001) studied locomotor activity in Wistar rats exposed for up to 180 days
to 376-ppm TCE by inhalation for 4 hours/day, 5 days/week and acutely intoxicated with TCE.
Here the authors report seemingly inconsistent effects of TCE on locomotion. After 30 days of
exposure, the treated rats show an increase in locomotor activity relative to control subjects.
However, after 60 days of exposure they note a significant increase in distance traveled found
among experimental subjects, but a decrease in horizontal activity in this experimental group.
Moreover, the control subjects vary substantially in horizontal counts among the different time
periods. No differences between the treatment groups are found after 180 days of exposure. It is
difficult to understand the apparent discrepancy in results reported at 60 days of exposure.
D.2.7. Sleep and Mood Disorders
D.2.7.1. Effects on Mood: Laboratory Animal Findings
It is difficult to obtain comparable data of emotionality in laboratory studies. However,
Moser et al. (2003) and Albee et al. (2006) both report increases in handling reactivity among
rats exposed to TCE. In the Moser study, female Fischer 344 rats received TCE by oral gavage
for periods of 10 days at doses of 0, 40, 200, 800, and 1,200 mg/kg/d while Albee et al. (2006)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-123 DRAFT—DO NOT CITE OR QUOTE
-------
exposed Fischer 344 rats to TCE by inhalation at exposure doses of 250, 800, and 2,500 ppm for
6 hours/day, 5 days/week for 13 weeks.
D.2.7.2. Sleep Disturbances
Arito et al. (1994) exposed male Wistar rats to 50-, 100-, and 300-ppm TCE for
8 hours/day, 5 days/week for 6 weeks and measured electroencephalographic (EEG) responses.
EEG responses were used as a measure to determine the number of awake (wakefulness hours)
and sleep hours. Exposure to all the TCE levels significantly decreased amount of time spent in
wakefulness during the exposure period. Some carry over was observed in the 22-hour
postexposure period with significant decreases in wakefulness seen at 100-ppm TCE.
Significant changes in wakefulness-sleep elicited by the long-term exposure appeared at lower
exposure levels. These data seem to identify a low dose of TCE that has anesthetic properties
and established a LOAEL of 50 ppm for sleep changes.
D.2.8. Mechanistic Studies
D.2.8.1. Dopaminergic (DA) Neurons
In two separate animal studies, subchronic administration of TCE has resulted in a decrease
of dopaminergic (DA) cells in both rats and mice. Although the mechanism for DA neurons
resulting from TCE exposure is not elucidated, disruption of DA-containing neurons has been
extensively studied with respect to Parkinson's Disease and parkinsonism. In addition to
Parkinson's Disease, significant study of MPTP and of high-dose manganese toxicity provides
strong evidence for extrapyramidal motor dysfunction accompanying loss of dopamine neurons in
the substantia nigra. These databases may provide useful comparisons to the highly limited
database with regard to TCE and dopamine neuron effects. The studies are presented in Table D-9.
Gash et al. (2008) assessed the effects of subchronic TCE administration on
dopaminergic neurons in the central nervous system. Fischer 344 male rats were orally
administered by gavage 1,000 mg/kg TCE in olive oil, 5 days/week for 6 weeks. Degenerative
changes in DA containing neurons in the substantia nigra were reported as indexed by a 45%
decrease in the number of tyrosine hydroxylase positive cells. Additionally, there was a decrease
in the ratio of 3,4-dihydroxyphenylacetic acid, a metabolite of DA, to DA levels in the striatum.
This shift in ratio, on the order of 35%, was significant by Student's t-test, suggesting a decrease
in release and utilization of this neurotransmitter. While it is possible that long-term adaptation
might occur with regard to release rates for DA, the loss of DA cells in the substantia nigra is
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-124 DRAFT—DO NOT CITE OR QUOTE
-------
viewed as a permanent toxic effect. The exposure level used in this study was limited to one
high dose and more confidence in the outcome will depend upon replication and development of
a dose-response relationship. If the results are replicated, they might be important in
understanding mechanisms by which TCE produces neurotoxicity in the central nervous system.
The functional significance of such cellular loss has not yet been determined through behavioral
testing.
Guehl (1999) also reported persistent effects of TCE exposure on DA neurons. In this
study, OF1 male mice (n= 10) were injected i.p. daily for 5 days/week for 4 weeks with TCE
(400 mg/kg/d). Following a 7 day period when the subjects did not receive TCE, the mice were
euthanized and tyrosine hydroxylase immunoreactivity was used to measure neuronal death in
the substantia nigra pars compacta. Treated mice presented significant dopaminergic neuronal
death (50%) in comparison with control mice based upon total cell counts conducted by an
examiner blinded as to treatment group in six samples per subject. The statistical comparison
appears to be by Student's t-test (only means, standard deviations, and a probability ofp < 0.001
are reported). While this study appears to be consistent with that of Gash et al. (2008) there are
some limitations of this study. Specifically, no photomicrographs are provided to assess
adequacy of the histopathological material. Additionally, no dose-response data are available to
characterize dose-response relationships or identify either a benchmark dose or NOAEL.
Behavioral assessment aimed at determining functional significance was not determined.
The importance of these two studies suggesting death of dopaminergic neurons following
TCE exposure may be addressable by human health studies because they suggest the potential
for TCE to produce a parkinsonian syndrome.
D.2.8.2. Gamma-Amino Butyric Acid (GABA) and Glutamatergic Neurons
Disruption of GABAergic and glutamatergic neurons by toxicants can represent serious
impairment as gamma-amino butyric acid (GABA) serves as a key inhibitory neurotransmitter
while glutamate is equally important as an excitatory neurotoxicant. Moreover, elevations in
glutamatergic release have been identified as an important process by which more general
neurotoxicity can occur through a process identified as excitotoxicity. The data with regard to
TCE exposure and alteration in GABA and glutamate function is limited. The studies are
presented in Table D-10.
Briving et al. (1986) conducted a chronic inhalation exposure in Mongolian gerbils to
50-and 150-ppm TCE continuously for 12 months and reported the changes in amino acids levels
in the hippocampus and cerebellar vermis and on high affinity uptake of GABA and glutamate in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-125 DRAFT—DO NOT CITE OR QUOTE
-------
those same structures. A dose related elevation of glutamine in the hippocampus of
approximately 20% at 150 ppm was reported, but no other reliable changes in amino acids in
either of these two structures. With regard to high affinity uptake of glutamate and GAB A, there
were no differences in the hippocampal uptake between control and treated gerbils although in
the cerebellar vermis there was a dose related elevation in the high affinity uptake for both of
these neurotransmitter. Glutamate uptake was increased about 50% at 50 ppm and 100% at
150 ppm. The corresponding increases for GAB A were 69% and 74%. Since control tissue
uptake is identified as being 100% rather than as an absolute rate, the ability to assess quality of
the control data are limited. It is unclear if this finding in cerebellar vermis is also present in
other brain tissues and should be studied further. If these findings are reliable, the changes in
high affinity uptake in cerebellum for GABA and glutamate might represent alterations that
could have functional outcomes. For example, alteration in GABA release and reuptake from the
cerebellum might be consistent with acute alteration in vestibular function described below.
However, there are presently no compelling data to support such a relationship.
The change in hippocampal glutamine levels is not readily interpretable. What is not
clear from this paper is whether the alterations observed were acute effects observable only while
subjects were intoxicated with TCE or whether they would persist once TCE had been removed
from the neural tissue. This study used inhalation doses that were at least 1 order of magnitude
lower than those required to produce auditory impairment.
A study by Shih et al. (2001) provides indirect evidence in male Mfl mice that TCE
exposure by injection might alter GABAergic function. The mice were injected i.p. with 250,
500, 1,000 and 2,000 mg/kg TCE in corn oil and the effect of these treatments on susceptibility
to seizure induced by a variety of drugs was observed. Shih et al. report that doses of TCE as
low as 250 mg/kg could reduce signs of seizure induced by picrotoxin, bicuculline, and
pentylenetetrazol. These drugs are all GABAergic antagonists. TCE treatment had a more
limited effect on seizure threshold induced by non-GABAergic convulsant drugs such as
strychnine (glycine receptor antagonist), 4-aminopyridine (alcohol dehydrogenase inhibitor) and
N-methyl-d-aspartate (glutamatergic agonist) than was observed with the GABAergic
antagonists. While these data suggest the possibility that TCE could act at least acutely on
GABAergic neurons, there are no direct measurements of such an effect. Moreover, there is no
obvious relationship between these findings and those of Briving et al. (1986) with regard to
increased high affinity uptake of glutamate and GABA in cerebellum. Beyond that fact, this
study does not provide information regarding persistent effects of TCE on either seizure
susceptibility or GABAergic function as all measurements were made acutely shortly following a
single injection of TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-126 DRAFT—DO NOT CITE OR QUOTE
-------
D.2.8.3. Demyelination Following Trichloroethylene (TCE) Exposure
Because of its anesthetic properties and lipophilicity, it is hypothesized that TCE may
disrupt the lipid-rich sheaths that cover many central and peripheral nerves. This issue has also
been studied both in specific cranial nerves known to be targets of TCE neurotoxicity (namely
the trigeminal nerve) and in the central nervous system including the cerebral cortex,
hippocampus and cerebellum in particular. For peripheral and cranial nerves, there are limited
nerve conduction velocity studies that are relevant as a functional measure. For central
pathways, the most common outcomes studied include histological endpoints and lipid profiles.
A significant difficulty in assessing these studies concerns the permanence or persistence
of effect. There is a very large literature unrelated to TCE, which demonstrates the potential for
repair of the myelin sheath and at least partial if not full recovery of function. In the studies
where nerve myelin markers are assessed, it is not possible to determine if the effects are
transient or persistent.
There are two published manuscripts (Isaacson and Taylor, 1989; Isaacson et al., 1990)
that document selective hippocampal histopathology when Sprague Dawley rats are exposed to
TCE within a developmental model. Both of these studies employed oral TCE administration
via the drinking water. In Isaacson and Taylor (1989), a combined prenatal and neonatal
exposure was used while Isaacson's et al. (1990) report focused on a neonatal exposure. In
addition, Ohta et al. (2001) presented evidence of altered hippocampal function in an in vitro
preparation following acute in vivo TCE intoxication. The latter most manuscript details a shift
in long term potentiation elicited by tetanic shocks to hippocampal slices in vitro. In the two
developmental studies the exposure doses are expressed in terms of the concentration of TCE
placed in the drinking water and the total daily dose is then estimated based upon average water
intake by the subjects. However, since the subjects' body weight is not provided, it is not
possible to estimate dosage on a mg/kg body weight basis.
Isaacson and Taylor (1989) examined the development of the hippocampus in neonatal
rats that were exposed in utero and in the preweaning period to TCE via their dam. TCE was
added to the drinking water of the dam and daily maternal doses are estimated based upon water
intake of the dam as being 4 and 8.1 mg/day. Based upon body weight norms for 70-day old
female Sprague Dawley rats, which would predict body weights of about 250 g at that age, such
a dose might approach 16-32 mg/kg/d initially during pregnancy. Even if these assumptions
hold true, it is not possible to determine how much TCE was received by the pups although the
authors do provide an estimate of fetal exposure expressed as [j,g/mL of TCE, trichloroethanol,
and trichloroacetic acid. The authors reported a 40% decline in myelinated fibers in the CA1
region of the hippocampus of the weanling rats. There was no effect of TCE treatment on
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-127 DRAFT—DO NOT CITE OR QUOTE
-------
myelination in several other brain regions including the internal capsule, optic tract or fornix and
this effect appears to be restricted to the CA1 region of the hippocampus at the tested exposures.
In a second manuscript by that group (Isaacson et al., 1990), weanling rats were exposed
to TCE via their drinking water at doses of 5.5 mg/day for 4 weeks or 5.5 mg/day for 4 weeks, a
2 week period with no TCE and then a final 2 weeks of exposure to 8.5 mg/day TCE. Spatial
learning was studied using the Morris water maze and hippocampal myelination was examined
histologically starting 1 day postexposure. The authors report that the subjects receiving a total
of 6 weeks exposure to TCE showed better performance in the Morris swim test (p < 0.05) than
did controls while the 4 week exposed subjects performed at the same level as did controls.
Despite this apparent improvement in performance, histological examination of the hippocampus
demonstrated a dose dependent relationship with hippocampal myelin being significantly
reduced in the TCE exposed groups while normal myelin patterns were found in the internal
capsule, optic tract and fornix. The authors did not evaluate the signs of gross toxicity in treated
animals such as growth rate, which might have influenced hippocampal development.
Ohta et al. (2001) administered 300 or 1,000 mg/kg TCE, i.p., to male ddY mice.
Twenty-four hours after TCE administration, the mice were sacrificed and hippocampal sections
were prepared from the excised brains and long term potentiation was measured in the slices. A
dose related reduction in the population spike was observed following a tetanic stimulation
relative to the size of the population spike elicited in the TCE mice prior to tetany. The spike
amplitude was reduced 14% in the 300 mg/kg TCE group and 26% in the 1,000 mg/kg group.
Precisely how such a shift in excitability of hippocampal CA1 neurons relates to altered
hippocampal function is not certain, but it does demonstrate that injection with 300 mg/kg TCE
can have lingering consequences on the hippocampus at least 24 hours following i.p.
administration.
A critical area for future study is the potential that TCE might have to produce
demyelination in the central nervous system. While it is realistic to imagine that an anesthetic
and lipophilic agent such as TCE might interact with lipid membranes and produce alterations,
for example, in membrane fluidity at least at anesthetic levels, the data collected by Kyrklund
and colleagues suggest that low doses of TCE (50 and 150 ppm chronically for 12 months,
320 ppm for 90 days, 510 ppm 8 hours/day for 5 months) might alter fatty acid metabolism in
Sprague Dawley rats and Mongolian gerbils. Because they have not included high doses in their
studies and because the low doses produce only sporadic significant effects and these tend to be
of very small magnitude (5—10%) it is not certain that they are truly observing events with
biological significance or whether they are observing random effects. A key problem in
determining whether the effects under study are spurious or are due to ongoing exposure is that
the magnitude and direction of the effect does not grow larger as exposure continues. It could be
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-128 DRAFT—DO NOT CITE OR QUOTE
-------
hypothesized that the alterations in fatty acid metabolism could be an underlying mechanism for
demyelination. However, there is not enough evidence to determine if the changes in the lipid
profiles lead to demyelination or if the observed effects are purely due to chance. Similarly, the
size of statistically significant effects (5-12%) is generally modest. A broad dose-response
analysis or the addition of a positive control group that is treated with an agent well-known to
produce central demyelination would be important in order to characterize the potency of TCE as
an agent that disrupts central nervous system lipid profiles.
Kyrklund and colleagues (e.g., 1986) have generally evaluated the hippocampus, cerebral
cortex, cerebellum, and in some instances brainstem in adult gerbil. It is not apparent that one
brain region is more vulnerable to the effects of TCE than is another region. While this group
does not report significant changes in levels of cholesterol, neutral and acidic phospholipids or
total lipid phospholipids, they do suggest a shift in lipid profiles between treated and untreated
subjects. Similarly, inhalation exposure to trichloroethane at 1,200 ppm for 30 days (Kyrklund
and Haglid, 1991) leads to sporadic changes in fatty acid profiles in Sprague Dawley rats.
However, these changes are small and are not always in the same direction as the changes
observed following trichloroethylene exposure. In the case of trichloroethane, a NOAEL of
320 ppm for 30 days 24 hours/day was observed and no other doses were evaluated (Kyrklund et
al., 1988).
D.2.9. Summary Tables
Tables D-4 through D-8 summarize the animal studies by neurological domains
(Table D-4—trigeminal nerve; Table D-5—ototoxicity; Table D-6—vestibular and visual
systems; Table D-7—cognition; and Table D-8—psychomotor function and locomotor activity).
For each table, the reference, exposure route, species, dose level, effects and NOAEL/LOAEL
are provided. Tables D-9 through D-l 1 summarize mechanistic (Tables D-9 and D-l 1) and
neurochemical studies (Table D-10). Brief summaries of developmental neurotoxicity studies
are provided in Table D-12.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-129 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-4. Summary of mammalian in vivo trigeminal nerve studies
Reference
Exposure route
Species, strain,
sex, number
Dose level/
exposure
duration
NOAEL:
LOAEL
Effects
Barret et
al. (1991)
Direct Gastric
Administration
Rat,
Sprague-Dawley,
female, 21
0, 2.5 g/kg,
acute
administration
LOAEL:
2.5 g/kg
Morphometric analysis was
used for analyzing the
trigeminal nerve. Increase in
external and internal fiber
diameter as well as myelin
thickness was observed in the
trigeminal nerve after TCE
treatment.
Barret et
al. (1992)
Direct Gastric
Administration
Rat,
Sprague-Dawley,
female, 18
0, 2.5 g/kg;
1 dose/d,
5 d/wk, 10 wks
LOAEL:
2.5 g/kg
Trigeminal nerve analyzed
using morphometric analysis.
Increased internode length and
fiber diameter in class A fibers
of the trigeminal nerve
observed with TCE treatment.
Changes in fatty acid
composition also noted.
Albee et al.
(2006)
Inhalation
Rat,
Fischer 344,
male and
female,
10/sex/group
o,
250, 800,
2,500 ppm
N
OAEL:
2,500 ppm
No effect on trigeminal nerve
function was noted at any
exposure level.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 0 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-5. Summary of mammalian in vivo ototoxicity studies
Reference
Exposure
route
Species, strain,
sex, number
Dose level/
exposure
duration
NOAEL;
LOAEL
Effects
Rebert et
al. (1991)
Inhalation
Rat, Long Evans,
male, 10/group
Long Evans: 0,
1,600,
3,200 ppm;
12 h/d, 12 wk
Long Evans:
NOAEL:
1,600 ppm;
LOAEL:
3,200 ppm
BAERs were measured.
Significant decreases in B AER
amplitude and an increase in
latency of appearance of the
initial peak (PI).


Rat, F344, male,
4-5/group
F344: 0, 2000,
3200 ppm;
12 h/d, 3 wk
F344:
LOAEL:
2,000 ppm

Rebert et
al. (1993)

Rat, Long Evans,
male, 9/group
0, 2,500, 3,000,
3,500 ppm; 8 h/d,
5 d
NOAEL:
2,500 ppm
LOAEL:
3,000 ppm
BAERs were measured 1-2 wk
postexposure to assess auditory
function. Significant decreases
in BAERs were noted with TCE
exposure.
Rebert et
al. (1995)

Rat, Long Evans,
male, 9/group
0, 2,800 ppm;
8 h/d, 5 d
LOAEL:
2,800 ppm
BAER measured 2-14 d
postexposure at a 16-kHz tone.
Hearing loss ranged from
55-85 dB.
Crofton et
al. (1994)

Rat, Long Evans,
male, 7-8/group
0, 3,500 ppm
TCE; 8 h/d, 5 d
LOAEL:
3,500 ppm
BAER measured and auditory
thresholds determined 5-8 wk
postexposure. Selective
impairment of auditory function
for mid-frequency tones (8 and
16 kHz).
Crofton
and Zhao
(1997);
Boyes et
al. (2000)
Inhalation
Rat, Long Evans,
male, 9-12/group
0, 4,000, 6,000,
8,000 ppm; 6 h
NOAEL:
6,000 ppm
LOAEL:
8,000 ppm
Auditory thresholds as measured
by BAERs for the 16-kHz tone
increased with TCE exposure.


Rat, Long Evans,
male, 8-10/group
0, 1,600, 2,400,
3,200 ppm; 6 h/d,
5 d
NOAEL:
2,400 ppm




LOAEL:
3,200 ppm

This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-131 DRAFT—DO NOT CITE OR QUOTE
-------
Rat, Long Evans,
0, 800, 1,600,
NOAEL:
male, 8-10/group
2,400, 3,200
2,400 ppm

ppm; 6 h/d,


5 d/wk, 4 wk
LOAEL:


3,200 ppm
Rat, Long Evans,
0, 800, 1,600,
NOAEL:
male, 8-10/group
2,400, 3,200
1,600 ppm

ppm; 6 h/d,


5 d/wk, 13 wk
LOAEL:


2,400 ppm
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-132 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-5. Summary of mammalian in vivo ototoxicity studies (continued)



Dose level/



Exposure
Species, strain,
exposure
NOAEL;

Reference
route
sex, number
duration
LOAEL*
Effects
Fechter et
Inhalation
Rat, Long Evans,
0, 4,000 ppm;
LOAEL:
Cochlear function measured
al. (1998)

male, 12/group
6 h/d, 5 d
4,000 ppm
5-7 wk after exposure. Loss of
spiral ganglion cells noted.
Auditory function was
significantly decreased as
measured by compound action
potentials.
Jaspers et
Inhalation
Rat, Wistar derived
0, 1,500, 3,000
LOAEL:
Auditory function assessed
al. (1993)

WAG-Rii/MBL,
male, 12/group
ppm; 18 h/d,
5 d/wk, 3 wk
1,500 ppm
repeatedly 1-5 wk postexposure
for 5-, 20-, and 35-kHz tones; No
effect at 5 or 35 kHz; Decreased
auditory sensitivity at 20 kHz.
Muijseret
Inhalation
Rat, Wistar derived
0, 3,000 ppm
LOAEL:
Auditory sensitivity decreased
al. (2000)

WAG-Rii/MBL,
male, 8

3,000 ppm
with TCE exposure at 4-, 8-, 16-,
and 20-kHz tones.
A
In
Rat,
0, 250,
NO
Mild frequency specific hearing
lbee et al.
(2006)
halation
Fischer 344,
male and female,
10/sex/group
800, 2,500
ppm
AEL:
800 ppm
LO
AEL:
2,500 ppm
deficits; Focal loss of hair cells
and cochlear lesions.
Y
In
Guinea
0, 6,000, 12,000,
NO
No change in auditory sensitivity
amamura
et al.
(1983)
halation
Pig, albino
Hartley, male,
7-10/group
17,000 ppm;
4 h/d, 5 d
AEL:
17,000
ppm
at any exposure level as
measured by cochlear action
potentials and microphonics.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 3 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-6. Summary of mammalian sensory studies—vestibular and visual
systems
Reference
Exposure route
Species, strain,
sex, number
Dose level/
exposure duration
NOAEL;
LOAEL
Effects
Vestibular system studies
Tham et al.
(1979)
Intravenous
Rabbit, strain
unknown, sex
unspecified, 19
1-5 mg/kg/min

Positional nystagmus
developed once blood
levels reached 30 ppm.
Tham et al.
(1984)
Intravenous
Rat, Sprague-
Dawley, female,
11
80 (ig/kg/min

Excitatory effects on the
vestibule-oculomotor
reflex. Threshold effect at
blood (TCE) of 120 ppm or
0.9 mM/L.
Niklasson
et al.
(1993)
Inhalation
Rat, strain
unknown, male
and female, 28
0, 2,700, 4,200,
6,000, 7,200 ppm;
1 h
LOAEL:
2,700 ppm
Increased ability to produce
nystagmus.
Umezu et
al. (1997)
Intraperitoneal
Mouse, ICR,
male, 116
0, 250, 500,
1,000 mg/kg, single
dose and evaluated
30 min
postadministration
NOAEL:
250 mg/kg
LOAEL:
500 mg/kg
Decreased equilibrium and
coordination as measured by
the Bridge test (staying time
on an elevated balance
beam).
Visual system studies
Rebert et
al. (1991)
Inhalation
Rat, Long Evans,
male, 10/group
0, 1,600, 3,200 ppm;
12 h/d, 12 weeks
NOAEL:
3,200 ppm
No effect on visual function
as measured by visual
evoked potential changes.
Rat, F344, male,
4-5/group
0, 2,000, 3,200 ppm;
12 h/d, 3 wk
NOAEL:
3,200 ppm
Boyes et al.
(2003)
Inhalation
Rat, Long Evans,
male,
9-10/group
0 ppm, 4 h;
1,000 ppm, 4 h; 2,000
ppm, 2 h; 3,000 ppm,
1.3 h; 4,000 ppm, 1 h
LOAEL:
1,000 ppm,
4 h
Visual function
significantly affected as
measured by decreased
amplitude (F2) in
Fourier-transformed visual
evoked potentials.
Boyes et al.
(2005)
Inhalation
Rat, Long Evans,
male,
8-10/group
0 ppm, 4 h; 500 ppm,
4 h; 1,000 ppm, 4 h;
2,000 ppm, 2 h;
3,000 ppm, 1.3 h;
4,000 ppm, 1 h;
5,000 ppm, 0.8 h
LOAEL:
500 ppm,
4 h
Visual function
significantly affected as
measured by decreased
amplitude (F2) in
Fourier-transformed visual
evoked potentials.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-134 DRAFT—DO NOT CITE OR QUOTE
-------
Blain et al.
Inhalation
Rabbit, New
0, 350, 700 ppm;
LOAEL:
Significant effects noted in
(1994)

Zealand albino,
male, 6-8/group
4 h/d, 4 d/wk, 12 wk
350 ppm
visual function as measured
by ERG and OPs
immediately after exposure.
No differences in ERG or
OP measurements were
noted at 6 wk post-TCE
exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 5 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-7. Summary of mammalian cognition studies
Reference
Exposure route
Species, strain,
sex, number
Dose level/
exposure duration
NOAEL;
LOAEL
Effects
Kjellstrand
et al.
(1980)
Inhalation
Gerbil,
Mongolian,
males and
females,
12/sex/dose
0, 320 ppm; 9 mos,
continuous (24 h/d)
except 1-2 h/wk for cage
cleaning
NOAEL:
320 ppm
No significant effect
on spatial memory
(radial arm maze).
Kulig et al.
(1987)
Inhalation
Rat, Wistar,
male, 8/dose
0, 500, 1,000, 1,500 ppm;
16 h/d, 5 d/wk, 18 wk
NOAEL:
500 ppm
LOAEL:
1,000 ppm
Increased latency
time in the
two-choice visual
discrimination task
(cognitive disruption
and/or motor activity
related effect).
Isaacson et
al. (1990)
Oral, drinking
water
Rat,
Sprague
Dawley,
male,
(1)0
mg/kg/d, 8 wk
(2)	5.5 mg/d
NOA
EL: 5.5 mg/d,
4 wk spatial
learning
Decreased latency to
find platform in the
Morris water maze
(Group #3);
Hippocampal


12/dose
(47 mg/kg/d*), 4 wk
+ 0 mg/kg/d, 4 wk
(3) 5.5 mg/d,
4 wk (47 mg/kg/d*)
+ 0 mg/kg/d, 2 wk
+ 8.5 mg/d
(24 mg/kg/d),* 2 wk
LOAE
L: 5.5 mg/d
hippocampal
demyelination
demyelination
observed in all
TCE-treated groups.
Kishi et al.
(1993)
Inhalation
Rats,
Wistar, male,
number not
specified
0, 250,500,
1,000, 2,000, 4,000
ppm, 4 h
LOAE
L: 250 ppm
Decreased lever
presses and
avoidance responses
in a shock avoidance
task.
Umezu et
al. (1997)
Intraperitoneal
Mouse, ICR,
male,
6 exposed to all
treatments
0, 125, 250, 500,
1,000 mg/kg, single dose
and evaluated 30 min
postadministration
NOAEL:
500 mg/kg
LOAEL:
1,000 mg/kg
Decreased response
rate in an operant
response-cognitive
task.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 6 DRAFT—DO NOT CITE OR QUOTE
-------
Ohta et al.
(2001)
Intraperitoneal
Mous
e, ddY, male,
5/group
0, 300, 1,000
mg/kg, sacrificed 24
h after injection
LOAE
L: 300 mg/kg
Decreased response
(LTP response) to
tetanic stimulation in
the hippocampus.
Oshiro et
al. (2004)
Inhalation
Rat,
Long Evans,
male, 24
0, 1,600,
2,400 ppm; 6 h/d, 5
d/wk, 4 wk
NOA
EL:
2,400 ppm
No change in
reaction time in
signal detection task
and when challenged
with amphetamine,
no change in
response from
control.
*mg/kg/d conversion estimated from average male Sprague-Dawley rat body weight from ages 21-49 days (118 g)
for the 5.5 mg dosing period and ages 63-78 days (354 g) for the 8.5 mg dosing period.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 7 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-8. Summary of mammalian psychomotor function, locomotor
activity, and reaction time studies
Reference
Exposure
route
Species/strain/
sex/number
Dose level/
exposure
duration
NOAEL;
LOAEL
Effects
Savolainen et
al. (1977)
Inhalation
Rat, Sprague
Dawley, male,
10
0, 200 ppm; 6
h/d, 4 d
LOAEL: 200
ppm
Increased frequency of
preening, rearing, and
ambulation. Increased
preening time.
Wolff and
Siegmund
(1978)
Intraperitoneal
Mouse, AB,
male, 144
0, 182 mg/kg,
tested 30 min
after injection
LOAEL: 182
mg/kg
Decreased spontaneous
motor activity.
Kulig et al.
(1987)
Inhalation
Rat, Wistar,
male, 8/dose
0, 500, 1,000,
1,500 ppm; 16
h/d, 5 d/wk, 18
wk
NOAEL: 1,500
ppm
No change in spontaneous
activity, grip strength or
hindlimb movement.
Motohashi
and Miyazaki
(1990)
Intraperitoneal
Rat, Wistar,
male, 44
0, 1.2 g/kg,
tested 30 min
after injection
LOAEL: 1.2
g/kg
Increased incidence of rats
slipping in the inclined
plane test.
0, 1.2 g/kg/d, 3 d
LOAEL: 1.2
g/kg
Decreased spontaneous
motor activity.
Fredriksson
et al. (1993)
Oral
Mouse, NMRI,
male, 12 (3-4
litters)
0, 50, 290
mg/kg/d, at Days
10-16

Decreased rearing; No
evidence of dose response.
Moser et al.
(1995)
Oral
Rat, Fischer
344, female,
8/dose
0, 150, 500,
1,500, 5,000
mg/kg, 1 dose
NOAEL: 500
mg/kg
LOAEL: 1,500
mg/kg
Decreased motor activity;
Neuro-muscular and
sensorimotor impairment.
0, 50, 150, 500,
1,500 mg/kg/d,
14 d
NOAEL: 150
mg/kg/d
LOAEL: 500
mg/kg/d
Increased rearing activity.
Bushnell
(1997)
Inhalation
Rat, Long
Evans, male,
12
0, 400, 800,
1,200, 1,600,
2,000, 2,400
ppm, 1-h/test
day, 4
consecutive test
days, 2 wk
NOAEL: 800
ppm
LOAEL: 1,200
ppm
Decreased sensitivity and
increased response time in
the signal detection task.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 8 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-8. Summary of mammalian psychomotor function, locomotor
activity, and reaction time studies (continued)
Reference
Exposure
route
Species/strain/
sex/number
Dose level/
exposure
duration
NOAEL;
LOAEL*
Effects
Umezu et al.
(1997)
Intraperitoneal
Mouse, ICR,
male, 6
exposed to all
treatments
0, 2,000, 4,000,
5,000 mg/kg -
loss of righting
reflex measure
LOAEL: 2,000
mg/kg - loss of
righting reflex
Loss of righting reflex,
decreased operant
responses, increased
punished responding.



0, 62.5, 125,
250, 500, 1,000
mg/kg, single
dose and
evaluated 30 min
postadministrati
on
NOAEL: 500
mg/kg
LOAEL: 1,000
mg/kg - operant
behavior
NOAEL: 125
mg/kg
LOAEL: 250
mg/kg-
punished
responding

Bushnell and
Oshiro (2000)
Inhalation
Rat, Long
Evans, male,
32
0, 2,000, 2,400
ppm; 70 min/d, 9
d
LOAEL: 2,000
ppm
Decreased performance on
the signal detection task.
Increased response time
and decreased response
rate.
Nunes et al.
(2001)
Oral
Rat, Sprague
Dawley, male,
10/group
0, 2,000
mg/kg/d, 7 d
LOAEL: 2,000
mg/kg/d
Increased foot splay. No
change in any other FOB
parameter (e.g.,
piloerection, activity,
reactivity to handling).
Waseem et al.
(2001)
Oral
Rat, Wistar,
male, 8/group
0, 350, 700,
1,400 ppm in
drinking water
for 90 d
NOAEL: 1,400
ppm
No significant effect on
spontaneous locomotor
activity.

Inhalation
Rat, Wistar,
male, 6/group
0, 376 ppm for
up to 180 d
LOAEL: 376
ppm
Changes in locomotor
activity but not consistent
when measured over the
180-day period.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-13 9 DRAFT—DO NOT CITE OR QUOTE
-------
Moser et al.
(2003)
Oral
Rat, Fischer
344, female,
10/group
0, 40, 200, 800,
1,200 mg/kg/d,
10 d

Decreased motor activity;
Decreased sensitivity;
Increased abnormality in
gait; Adverse changes in
several FOB parameters.
Albee et al.
Inhalation
Rat,
0, 250,
NOA
No change in any FOB
(2006)

Fischer 344,
800, 2,500
EL: 2,500
measured parameter.


male and
ppm
ppm



female,





10/sex/grou
P



This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-140 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-9. Summary of mammalian in vivo dopamine neuronal studies
Reference
Exposure
route
Species/strain/
sex/number
Dose level/
exposure
duration
NOAEL;
LOAEL
Effects
Guehl et
al. (1999)
Intraperitoneal
administration
Mouse, OF1,
male, 10
0, 400 mg/kg
LOAEL:
400 mg/kg
Significant dopaminergic
neuronal death in substantia
nigra.
Gash et al.
(2008)
Oral
Rat, Fischer 344,
male, 17/group
0, 1,000 mg/kg
LOAEL:
1,000
mg/kg
Degeneration of dopamine-
containing neurons in
substantia nigra.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-141 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-10. Summary of neurochemical effects with TCE exposure

Exposure
Species/strain/
Dose level/
NOAEL;

Reference
route
sex/number
exposure duration
LOAEL
Effects
In vivo studies
Shih et al.
Intraperitoneal
Mouse, Mfl,
0, 250 500, 1,000,
—
Increased threshold for
(2001)

male, 6/group
2,000 mg/kg, 15

seizure appearance



min; followed by

with TCE pretreatment



tail infusion of PTZ

for all convulsants.



(5 mg/mL),

Effects strongest on the



picro toxin (0.8

GABAa antagonists,



mg/mL),

PTZ, picrotoxin, and



bicuculline (0.06

bicuculline suggesting



mg/mL), strychnine

GAB Aa receptor



(0.05 mg/mL), 4-

involvement. NMDA



AP (2 mg/mL), or

and glycine Rc



NMDA (8 mg/mL)

involvement also





suggested.
Briving et al.
Inhalation
Gerbils,
0, 50, 150 ppm,
NOAEL: 50
Increased glutamate
(1986)

Mongolian,
continuous, 24 h/d,
ppm; LOAEL:
levels in the


male and
12 mos
150 ppm for
hippocampus.


female, 6/group

glutamate levels





in hippocampus
Increased glutamate





and GABA uptake in




NOAEL: 150
the cerebellar vermis.




ppm for





glutamate and





GAB A uptake in





hippocampus





LOAEL: 50 ppm





for glutamate and





GAB A uptake in





cerebellar vermis

This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-142 DRAFT—DO NOT CITE OR QUOTE
-------
Subramoniam
etal. (1989)
Oral
Rat, Wistar,
female,
0, 1,000 mg/kg, 2
or 20 h
0, 1,000 mg/kg/d, 5
d/wk, 1 y

PI and PIP2 decreased
by 24 and 17% at 2 h.
PI and PIP2 increased
by 22 and 38% at 20 h.
PI, PIP, and PIP2
reduced by 52, 23, and
45% in 1-yr study.
Kjellstrand et
al. (1987)
Inhalation
Mouse, NMRI,
male
0, 150, 300 ppm, 24
h/d, 4 or 24 d
LOAEL: 150
ppm, 4 and 24 d
Sciatic nerve
regeneration was
inhibited in both mice
and rats.
Rat, Sprague-
Dawley, female
0, 300 ppm, 24 h/d,
4 or 24 d
NOAEL: 300
ppm, 4 d
LOAEL: 300
ppm, 24 d
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-143 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-10. Summary of neurochemical effects with TCE exposure (continued)
Reference
Exposure
route
Species/strain/
sex/number
Dose level/
exposure duration
NOAEL;
LOAEL*
Effects
Haglid et al.
(1981)
Inhalation
Gerbil,
Mogolian, male
and female,
6-7/group
0, 60, 320 ppm, 24
h/d, 7 d/wk, 3 mos
LOAEL: 60 ppm,
brain protein
changes
NOAEL: 60
ppm; LOAEL:
320 ppm, brain
DNA changes
(1)	Decreases in total
brain soluble protein
whereas increase in
S100 protein.
(2)	Elevated DNA in
cerebellar vermis and
sensory motor cortex.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-144 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-ll. Summary of in vitro ion channel effects with TCE exposure
Reference
Cellular
system
Neuronal channel/
receptor
Concentrations
Effects
In vitro studies
Shafer et al.
(2005)
PC12 cells
Voltage sensitive
calcium channels
(VSCC)
0, 500, 1,000,
1,500, 2,000 \M
Shift of VSCC activation to a more
hyperpolarizing potential.
Inhibition of VSCCs at a holding
potential of -70 mV.
Beckstead et
al. (2000)
Xenopus
oocytes
Human recombinant
Glycine receptor
al, GABAa
receptors, aipi,
aip2y2L
0, 390 \M
50% potentiation of the GABAa
receptors; 100% potentiation of the
glycine receptor.
Lopreato et al.
(2003)
Xenopus
oocytes
Human recombinant
serotonin 3 A
receptor
111
Potentiation of serotonin receptor
function.
Krasowski
and Harrison
(2000)
Human
embryonic
kidney 293
cells
Human recombinant
Glycine receptor al,
GABAa receptors
a2pi
Not provided
Potentiation of glycine receptor function
with an EC50 of 0.65 ± 0.05 mM.
Potentiation of GAB Aa receptor
function with an EC50 of 0.85 ± 0.2.
EC50 = median effective concentration.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-145 DRAFT—DO NOT CITE OR QUOTE
-------
Table D-12. Summary of mammalian in vivo developmental neurotoxicity
studies—oral exposures
Reference
Species/strain/
sex/number
Dose level/
exposure duration
Route/vehicle
NOAEL;
LOAELa
Effects
Fredriksson
etal. (1993)
Mouse, NMRI,
male pups, 12
pups from 3-4
different
litters/group
0, 50, or 290 mg/kg-d
PND 10-16
Gavage in a
20% fat
emulsion
prepared from
egg lecithin
and peanut oil
Dev.
LOAEL: 50
mg/kg/d
Rearing activity sig. 1 at
both dose levels on PND
60.
George et
al. (1986)
Rat, F334, male
and female, 20
pairs/treatment
group,
40 controls/sex
0,0.15,0.30, or
0.60%
microencapsulated
TCE.
Breeders exposed 1
wk premating, then
for 13 wk; pregnant
s throughout
pregnancy (i.e., 18-
wk total).
Dietary
LOAEL:
0.15%
Open field testing in pups:
a sig. dose-related trend
toward T time required for
male and female pups to
cross the first grid in the
test device.
Isaacson
and Taylor
(1989)
Rat, Sprague-
Dawley,
females, 6
dams/group
0, 312, or 625 mg/L.
(0, 4.0, or 8.1 mg/d)b
Dams (and pups)
exposed from 14 d
prior to mating until
end of lactation.
Drinking
water
Dev.
LOAEL: 312
mg/L
Sig. 1
myelinated fibers in
the stratum
lacunosum-
moleculare of pups.
Reduction in myelin
in the hippocampus.
Noland-
Gerbec et
al. (1986)
Rat, Sprague-
Dawley,
females, 9-11
dams/group
0, 312 mg/L
(Avg. total intake of
dams: 825 mg TCE
over 61 d.)b
Dams (and pups)
exposed from 14 d
prior to mating until
end of lactation.
Drinking
water
Dev. LOEL:
312 mg/L
Sig. i uptake of 3H-2-
DG in whole brains and
cerebella (no effect in
hippocampus) of exposed
pups at 7, 11, and 16 d,
but returned to control
levels by 21 d.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-146 DRAFT—DO NOT CITE OR QUOTE
-------
Ta
Rat,
0,312,
Drin
De
Exploratory behavior sig.
ylor et al.
(1985)
Sprague-
Dawley,
females, no.
dams/group
not reported
625, and
1,250 mg/L
Dams (and pups)
exposed from 14 d
prior to mating until
end of lactation.
king water
v.
LOAEL:
312 mg/L
t in 60- and 90-d old male
rats at all treatment levels.
Locomotor activity was
higher in rats from dams
exposed to 1,250-ppm
TCE.
aNOAEL, LOAEL, and LOEL (lowest-observed-effect level) are based upon reported study findings.
bDose conversions provided by study author(s).
PND = postnatal day.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	D-147 DRAFT—DO NOT CITE OR QUOTE
-------
APPENDIX E
Analysis of Liver and Coexposure Issues for
the TCE Toxicological Review
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-i DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS—Appendix E: Analysis of Liver and Coexposure Issues for the TCE
Toxicological Review
LIST OF TABLES	E-vi
LIST OF FIGURES	E-viii
FOREWORD	E-ix
AUTHORS, CONTRIBUTORS, AND REVIEWERS	E-x
ACKNOWLEDGMENTS	E-xi
APPENDIX E. ANALYSIS OF LIVER AND COEXPOSURE ISSUES FOR THE TCE TOXICOLOGICAL RI
E. 1. BASIC PHYSIOLOGY AND FUNCTION OF THE LIVER—A STORY OF
HETEROGENEITY	E-l
E. 1.1. Heterogeneity of Hepatocytes and Zonal Differences in Function and
Ploidy	E-l
E.1.2. Effects of Environment and Age: Variability of Response	E-8
E.2. CHARACTERIZATION OF HAZARD FROM TRICHLOROETHYLENE
(TCE) STUDIES	E-10
E.2.1. Acute Toxicity Studies	E-ll
E.2.1.1. Soni etal. (1998)	E-ll
E.2.1.2. Soni et al. (1999)	11-14
11.2.1.3.	Okino etal. (1991)	11-15
11.2.1.4.	Nunes etal. (2001)	11-16
E.2.1.5. Tao et al. (2000b)	11-17
11.2.1.6. Tucker etal. (1982)	11-18
E.2.1.7. Goldsworthy and Popp (1987)	E-20
E.2.1.8. Elcombe etal. (1985)	11-21
E.2.1.9. Dees and Travis (1993)	E-34
E.2.1.10. Nakajima et al. (2000)	E-39
11.2.1.1 l.Berman etal. (1995)	11-42
E.2.1.12. Melnick et al. (1987)	11-44
E.2.1.13. Laughter et al. (2004)	11-47
11.2.1.14. Ramdhan etal. (2008)	11-51
E.2.2. Subchronic and Chronic Studies of Trichloroethylene (TCE)	E-64
11.2.2.1. Merrick et al. (1989)	11-65
This document is a draft for review purposes only and does not constitute Agency policy.
E-ii DRAFT—DO NOT CITE OR QUOTE
-------
E.2.2.2. Goel et al. (1992)	E-69
E.2.2.3. Kj ell strand et al. (1981b)	11-71
E.2.2.4. Woolhiser et al. (2006)	I>75
E.2.2.5. Kj ell strand et al. (1983b)	I >76
E.2.2.6. Kj ell strand et al. (1983a)	I >81
E.2.2.7. Buben and O'Flaherty (1985)	11-85
11.2.2.8.	Channel et al. (1998)	11-88
11.2.2.9.	Dorfmueller et al. (1979)	11-92
11.2.2.10.	Kumar etal. (2001)	11-92
E.2.2.11. Kawamoto et al. (1988)	E-93
E.2.2.12. National Toxicology Program (NTP) (1990)	E-95
E.2.2.13.National Toxicology Program (NTP) (1988)	E-100
E.2.2.14. Fukuda et al. (1983)	11-102
11.2.2.15.1 Ienschler etal. (1980)	11-103
E.2.2.16.Maltoni etal. (1986)	11-105
E.2.2.17.Maltoni etal. (1988)	11-110
E.2.2.18. Van Duuren et al. (1979)	11-110
E.2.2.19. National Cancer Institute (NCI) (1976)	11-111
E.2.2.20. Herren-Freund et al. (1987)	E-116
11.2.2.21. Anna et al. (1994)	11-117
E.2.2.22. Bull et al. (2002)	11-119
E.2.3. Mode of Action: Relative Contribution of Trichloroethylene (TCE)
Metabolites	E-121
E.2.3.1. Acute studies of Dichloroacetic Acid (DCA)/Trichloroacetic
Acid (TCA)	11-122
E.2.3.2. Subchronic and Chronic Studies of Dichloroacetic Acid
(DCA) and Trichloroacetic Acid (TCA)	E-149
E.2.4. Summaries and Comparisons Between Trichloroethylene (TCE),
Dichloroacetic Acid (DCA), and Trichloroacetic Acid (TCA) Studies	E-214
E.2.4.1. Summary of Results For Short-term Effects of
Trichloroethylene (TCE)	E-215
E.2.4.2. Summary of Results For Short-Term Effects of
Dichloroacetic Acid (DCA) and Trichloroacetic Acid (TCA):
Comparisons With Trichloroethylene (TCE)	E-223
E.2.4.3. Summary Trichloroethylene (TCE) Subchronic and Chronic
Studies	E-249
E.2.4.4. Summary of Results For Subchronic and Chronic Effects of
Dichloroacetic Acid (DCA) and Trichloroacetic Acid (TCA):
Comparisons With Trichloroethylene (TCE)	E-261
E.2.5. Studies of Chloral Hydrate (CH)	E-286
E.2.6. Serum Bile Acid Assays	E-291
E.3. STATE OF SCIENCE OF LIVER CANCER MODES OF ACTION (MOAs) ...E-294
E.3.1. State of Science for Cancer and Specifically Human Liver Cancer	E-296
E.3.1.1. Epigenetics and Disease States (Transgenerational Effects,
Effects of Aging and Background Changes)	E-296
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-iii DRAFT—DO NOT CITE OR QUOTE
-------
E.3.1.2. Emerging Technologies, DNA and siRNA, miRNA
Microarrays—Promise and Limitations for Modes of Action
(MO As)	E-305
E.3.1.3. Etiology, Incidence and Risk Factors for Hepatocellular
Carcinoma (HCC)	E-316
E.3.1.4. Issues Associated with Target Cell Identification	E-320
E.3.1.5. Status ofMechanism of Action for Human Hepatocellular
Carcinoma (HCC)	E-325
E.3.1.6. Pathway and Genetic Disruption Associated with
Hepatocellular Carcinoma (HCC) and Relationship to Other
Forms of Neoplasia	E-327
E.3.1.7. Epigenetic Alterations in Hepatocellular Carcinoma (HCC)	E-329
E.3.1.8. Heterogeneity of Preneoplastic and Hepatocellular
Carcinoma (HCC) Phenotypes	E-332
E.3.2. Animal Models of Liver Cancer	E-339
E.3.2.1. Similarities with Human and Animal Transgenic Models	E-343
E.3.3. Hypothesized Key Events in HCC Using Animal Models	E-348
E.3.3.1. Changes in Ploidy	E-348
E.3.3.2. Hepatocellular Proliferation and Increased DNA Synthesis	E-354
E.3.3.3. Nonparenchymal Cell Involvement in Disease States
Including Cancer	E-3 5 7
E.3.3.4. Gender Influences on Susceptibility	E-365
E.3.3.5. EpigenomicModification	E-367
E.3.4. Specific Hypothesis for Mode of Action (MO A) of Trichloroethylene
(TCE) Hepatocarcinogenicity in Rodents	E-370
E.3.4.1. PPARa Agonism as the Mode of Action (MOA) for Liver
Tumor Induction—The State of the Hypothesis	E-370
E.3.4.2. Other Trichloroethylene (TCE) Metabolite Effects That May
Contribute to its Hepatocarcinogenicity	E-405
E.4. EFFECTS OF COEXPOSURES ON MODE OF ACTION (MOA)—
INTERNAL AND EXTERNAL EXPOSURES TO MIXTURES
INCLUDING ALCOHOL	E-417
E.4.1. Internal Coexposures to Trichloroethylene (TCE) Metabolites:
Modulation of Toxicity and Implications for TCE Mode of Action
(MOA)	I >420
E.4.2. Initiation Studies as Coexposures	E-420
E.4.2.1. Herren-Freund et al. (1987)	E-421
E.4.2.2. Parnell et al. (1986)	I>422
E.4.2.3. Pereira and Phelps (1996)	E-423
E.4.2.4. Tao et al (2000a)	I>428
E.4.2.5. Latendresse and Pereira (1997)	E-428
E.4.2.6. Pereira etal. (1997)	E-431
E.4.2.7. Tao et al. (1998)	I>433
E.4.2.8. Stauber et al. (1998)	11-434
E.4.3. Coexposures of Haloacetates and Other Solvents	E-436
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-iv DRAFT—DO NOT CITE OR QUOTE
-------
E.4.3.1. Carbon tetrachloride, Dichloroacetic Acid (DCA),
Trichloroacetic Acid (TCA): Implications for Mode of
Action (MOA) from Coexposures	E-436
E.4.3.2. Chloroform, Dichloroacetic Acid (DCA), and Trichloroacetic
Acid (TCA) Coexposures: Changes in Methylation Status	E-439
E.4.3.3. Coexposures to Brominated Haloacetates: Implications for
Common Modes of Action (MO As) and Background
Additivity to Toxicity	E-441
E.4.3.4. Coexposures to Ethanol: Common Targets and Modes of
Action (MOAs)	E-444
E.4.3.5. Coexposure Effects on Pharmacokinetics: Predictions Using
Physiologically Based Pharmacokinetic (PBPK) Models	E-447
E.5. POTENTIALLY SUSCEPTIBLE LIFE STAGES AND CONDITIONS
THAT MAY ALTER RISK OF LIVER TOXICITY AND CANCER	E-450
1 :.6. UNCERTAINTY AM) VARIABILITY	I>451
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-v DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES
Table E-l. Mice data for 13 weeks: mean body and liver weights	E-96
Table E-2. Prevalence and Multiplicity data from DeAngelo et al. (1999)	E-162
Table E-3. Difference in pathology by inclusion of unscheduled deaths from DeAngelo et al.
(1999)	E-l 63
Table E-4. Comparison of data from Carter et al. (2003) and DeAngelo et al. (1999)	E-169
Table E-5. Prevalence of foci and tumors in mice administered NaCl, DCA, or TCA from
Pereira (1996)	E-174
Table E-6. Multiplicity of foci and tumors in mice administered NaCl, DCA, or TCA from
Pereira (1996)	E-175
Table E-7. Phenotype of foci reported in mice exposed to NaCl, DCA, or TCA by Pereira (1996)
	I>176
Table E-8. Phenotype of tumors reported in mice exposed NaCl, DCA, or TCA by Pereira
(1996)	I >176
Table E-9. Multiplicity and incidence data (31 week treatment) from Pereira and Phelps (1996)
	E-179
Table E-10. Comparison of descriptions of control data between George et al. (2000) and
DeAngelo et al. (2008)	E-196
Table E-l 1. TCA-induced increases in liver tumor occurrence and other parameter over control
after 60 weeks (Study #1)	E-204
Table E-12. TCA-induced increases in liver tumor occurrence after 104 wks (Studies #2 and #3)
	I>208
Table E-13. Comparison of liver effects from TCE, TCA, and DCA (10-day exposures in mice)
	11-225
Table E-14. Liver weight induction as percent liver/body weight fold-of-control in male
B6C3F1 mice from DCA or TCA drinking water studies	E-230
Table E-l5. Liver weight induction as percent liver/body weight fold-of-control in male
B6C3F1 or Swiss mice from TCE gavage studies	E-232
Table E-16. B6C3F1 and Swiss (data sets combined)	E-234
Table E-17. Power calculations" for experimental design described in text, using Pereira et al. as
an example	E-273
Table E-l8. Comparison between results for Yang et al. (2007) and Cheung et al. (2004)a..E-399
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-vi DRAFT—DO NOT CITE OR QUOTE
-------
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-vii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
Figure E-l. Comparison of average fold-changes in relative liver weight to control and exposure
concentrations of 2 g/L or less in drinking water for TCA and DCA in male B6C3F1 mice for
14-30 days (Carter et al., 1995; DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-Weinstein et
al., 2001; Parrish et al., 1996; Sanchez and Bull, 1990)\. (Reproduced from Section 4.5.) ...E-236
Figure E-2. Comparisons of fold-changes in average relative liver weight and gavage dose of
(top panel) male B6C3F1 mice for 10-28 days of exposure (Dees and Travis, 1993; Elcombe et
al., 1985; Goldsworthy and Popp, 1987; Merrick et al., 1989) and (bottom panel) in male
B6C3F1 and Swiss mice. (Reproduced from Section 4.5.)	E-238
Figure E-3. Comparison of fold-changes in relative liver weight for data sets in male B6C3F1,
Swiss, and NRMI mice between TCE studies (Buben and O'Flaherty, 1985; Goel et al., 1992;
Kjellstrand et al., 1983a; Merrick et al., 1989) [duration 28-42 days] and studies of direct oral
TCA administration to B6C3 F1 mice (DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-
Weinstein et al., 2001; Parrish et al., 1996) [duration 14-28 days]. Abscissa for TCE studies
consists of the median estimates of the internal dose of TCA predicted from metabolism of TCE
using the PBPK model described in Section 3.5 of the TCE risk assessment. Lines show linear
regression with intercept fixed at 1. All data were reported fold-change in mean liver
weight/body weight ratios, except for Kjellstrand et al. (1983a), with were the fold-change in the
ratio of mean liver weight to mean body weight. In addition, in Kjellstrand et al. (1983a), some
systemic toxicity as evidence by decreased total body weight was reported in the highest dose
group. (Reproduced from Section 4.5.)	E-242
Figure E-4. Fold-changes in relative liver weight for data sets in male B6C3F1, Swiss, and
NRMI mice reported by TCE studies of duration 28-42 days (Buben and O'Flaherty, 1985; Goel
et al., 1992; Kjellstrand et al., 1983a; Merrick et al., 1989) using internal dose metrics predicted
by the PBPK model described in Section E.3.5: (A) dose metric is the median estimate of the
daily AUC of TCE in blood, (B) dose metric is the median estimate of the total daily rate of TCE
oxidation. Lines show linear regression. Use of liver oxidative metabolism as a dose metric
gives results qualitatively similar to (B), with R = 0.86. (Reproduced from Section 4.5.) ...E-244
Figure E-5. Comparison of Ito et al. and David et al. data for DEHP tumor induction from
(Guyton et al., 2009)	E-382
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-viii DRAFT—DO NOT CITE OR QUOTE
-------
FOREWORD
The purpose of this Appendix is to provide scientific support and rationale for the hazard
and dose-response sections of the Toxicological Review of Trichloroethylene (TCE) regarding
liver effects and those of coexposures. It is not intended to be a comprehensive treatise on the
chemical or toxicological nature of TCE. Please refer to the Toxicological Review of TCE for
characterization of EPA's overall confidence in the quantitative and qualitative aspects of hazard
and dose-response for TCE-induced liver effects. Matters considered in this appendix include
knowledge gaps, uncertainties, quality of data, and scientific controversies. This characterization
is presented in an effort to make apparent the scientific issues regarding the data and MOA
considerations for experimental animal data for liver effects in the TCE assessment.
For other general information about this assessment or other questions relating to IRIS,
the reader is referred to EPA's IRIS Hotline at (202) 566-1676 (phone), (202) 566-1749 (fax), or
hotline.iris@epa.gov (email address).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-ix DRAFT—DO NOT CITE OR QUOTE
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
CHEMICAL MANAGER
Weihsueh A. Chiu
National Center for Environmental Assessment—Washington Office
U.S. Environmental Protection Agency
Washington, DC
PRINCIPAL AUTHOR
Jane C. Caldwell
National Center for Environmental Assessment—Washington Office
U.S. Environmental Protection Agency
Washington, DC
CONTRIBUTORS
Weihsueh A. Chiu
National Center for Environmental Assessment—Washington Office
U.S. Environmental Protection Agency
Washington, DC
Marina V. Evans
National Health and Environmental Effects Research Laboratory
(on detail to National Center for Environmental Assessment—Washington Office)
U.S. Environmental Protection Agency
Research Triangle Park, NC
Kathryn Z. Guyton
National Center for Environmental Assessment—Washington Office
U.S. Environmental Protection Agency
Washington, DC
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-x DRAFT—DO NOT CITE OR QUOTE
-------
ACKNOWLEDGMENTS
The author and contributors would like to thank the NCEA management team for their
comments and support, and Terri Konoza and the TSS for their extensive technical editing
support.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-xi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
E.l. BASIC PHYSIOLOGY AND FUNCTION OF I II I LIVER—A STORY OF
HETEROGENEITY
The liver is a complex organ whose normal function and heterogeneity are key to
understanding and putting into context perturbations by trichloroethylene (TCE), cancer biology,
and variations in response observed and anticipated for susceptible life stages and background
conditions.
E. 1.1. Heterogeneity of Hepatocytes and Zonal Differences in Function and Ploidy
Malarkey et al. (2005) state that (1) the liver transcriptome (i.e., genes expressed as
measured by mRNA) is believed only second to the brain in its complexity and includes about
25-40% of the approximately 50,000 mammalian genes, (2) during disease states the
transcriptome can double or triple and its increased complexity is due not only to differential
gene expression (up- and down-regulation of genes) but also to the mRNA contributions from
the heterogeneous cell populations in the liver, and (3) when one considers that over a dozen cell
types comprise the liver in varying proportions, particularly in disease states, knowledge about
the cell types and cell-specific gene expression profiles help unravel the complex genomic and
protenomic data sets. Gradients of gene and protein activity varying from the periportal region
to the centrilobular region also exist for sinusoidal endothelial cells, Kuffper cells, hepatic
stellate cells, and the matrix in the space of Disse. Malarkey et al. (2005) also estimate that
hepatocytes constitute 60%, sinusoidal endothelial cells 20%, Kupffer cells 15%, and stellate
cells 5% of liver cells. Therefore, in experimental paradigms where liver homogenates are used
for the determination of "changes in liver," gene expression, or other parameters the individual
changes from cells residing in differing zones and by differing cell type is lost. Malarkey et al.
(2005) define the need to better characterize the histological cellular components of the tissues
from which mRNA and protein is extracted and referred to "phenotypic anchoring" and cite
acetaminophen as a "model hepatotoxicant under study to assess the strengths and weaknesses of
genomics and proteinomics technologies" as well as "a good example for understanding and
utilizing phenotypic anchoring to better understand genomics data." After acetaminophen
exposure "there is an unexplained and striking inter and intralobular variability in acute hepatic
necrosis with some regions having massive necrosis and adjacent areas within the same lobe or
other lobes showing no injury at all." Malarkey et al. (2005) go on to cite similar lobular
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-1 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
variability in response for "copper distribution, iron and phosphorous, chemical and spontaneous
carcinogenesis, cirrhosis and regeneration" and suggest that although uncertain "factors such as
portal streamlining of blood to the liver, redistribution of blood to core of the liver secondary to
nerve stimulation, and exposures during fetal development and possibly lobular gradients are
important." Hepatic interlobe differences exist for initiating agents in terms of DNA alkylation
and cell replication. In the rat, diethylnitrosamine (DEN) alkylation has been reported to occur
preferentially in the left and right median lobes, while cell replication was higher in the right
median and right anterior lobes (Richardson et al., 1986). Richardson et al. (1986) reported that
exposure to DEN induced a 100% incidence of hepatocellular carcinoma (HCC) in the left,
caudate, left median and right median lobes of the liver by 20 weeks versus only 30% in the right
anterior and right posterior hepatic lobes. There was a reported interlobe difference in adduct
formation, cell proliferation, liver lobe weight gain, number and size of y-glutamyltranspeptidase
(GGT)+ foci, and carbon 14 labeling from a single dose of DEN. Richardson et al. (1986)
suggest that many growth-selection studies utilizing the liver to evaluate the carcinogenic
potential of a chemical often focus on only one or two of the hepatic lobes, which is especially
true for partial hepatectomy, and that for DEN and possibly other chemicals this procedure
removes the lobes most likely to get tumors. Thus, the "distribution of toxic insult may not be
correctly assessed with random sampling of the liver tissue for microarray gene expression
analysis" (Malarkey et al., 2005) and certainly any such distributional differences are lost in
studies of whole-liver homogenates.
The liver is normally quiescent with few hepatocytes undergoing mitosis and, as
described below, normally occurring in the periportal areas of the liver. Mitosis is observed only
in approximately one in every 20,000 hepatocytes in adult liver (Columbano and Ledda-
Columbano, 2003). The studies of Schwartz-Arad et al. (1989), Zajicek et al. (1991), Zajicek
and Schwartz-Arad (1990), and Zajicek et al. (1989) have specifically examined the birth, death,
and relationship to zone of hepatocytes as the "hepatic streaming theory." They report that
hepatocytes and littoral cells continuously steam from the portal tract toward the terminal hepatic
vein and that the hepatocyte differentiates as it goes with biological age closely related to cell
differentiation. In other words, the acinus may be represented by a tube with two orifices: for
cell inflow situated at the portal tract rim and other for cell outflow, at the terminal hepatic vein
with hepatocytes streaming through the tube in an orderly fashion. In normal liver, cell
proliferation is suggested as the only driving force of this flow with each mitosis associated with
displacement of the cells by one cell location and the greater the cell production, the faster the
flow and visa versa (Zajicek et al., 1991). Thus, the microscopic section of the liver "displays an
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-2 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
instantaneous image of a tissue in flux" (Schwartz-Arad et al., 1989). Schwartz-Arad et al.
(1989) further suggest that
throughout its life the hepatocyte traverses three acinus zones; in each it is
engaged in different metabolic activity. When young it performs among other
functions gluconeogenesis, which is found in zone 1 hepatocytes (i.e. periportal),
and when old it turns into a zone 3 cell (i.e., pericentral), with a pronounced
glycolitic make up. The three zones thus represent differentiation stages of the
hepatocyte, and since they differ by their distance from the origin, e.g. zone 2
(i.e., midzonal) is more distant than zone 1, again, hepatocyte differentiation is
proportional to its distance.
Chen et al. (1995) report that
Hepatocytes are a heterogeneous population that are composed of cells expressing
different patterns of genes. For example, gamma-glutamyl transpeptidase and
genes related to gluconeogenesis are expressed preferential in periportal
hepatocytes, whereas enzymes related to glycolysis are more abundant in the
centrilobular area. Glutamine synthetase is expressed in a small number of
hepatocytes surrounding the central veins. Most cytochrome p450 enzymes are
expressed or induced preferentially in centrilobular hepatocytes relative to
periportal hepatocytes.
Along with changes in metabolic function, Vielhauer et al. (2001) reported that there is evidence
of zonal differences in carcinogen DNA effects and, also, chemical-specific differences for DNA
repair enzyme and that enhanced DNA repair is a general feature of many carcinogenic states
including the enzymes that repair alkylating agents but also oxidative repair. As part of this
process of differentiation and as livers age, the hepatocyte changes and increases its ploidy with
polyploid cells predominant in zone 2 of the acinus (Schwartz-Arad et al., 1989). The reported
decrease in DNA absorbance in zone 3 may be due to (1) a decline in chromatin affinity to the
dye, (2) cell death, and (3) DNA exit from intact cells and Zajicek and Schwartz-Arad (1990)
suggest that the fewer metabolic demands in Zone 3, under normal conditions, causes the cell to
"deamplify" its genes and for DNA excess to leak out cells adjacent to the terminal hepatic vein
or to be eliminated by apoptosis reflecting cell death. Thus, the three acinus zones represent
differentiation states of one and the same hepatocyte, which increase ploidy as functional
demands change. Zajicek and Schwartz-Arad (1990) also report that nuclear size is generally
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
proportional to DNA content and that as DNA accumulates, the nucleus enlarges. This has
import for histopathological descriptions of hepatocellular hypertrophy and attendant nuclear
changes after toxic insult as well.
The gene amplification associated with polyploidy is manifested by DNA accumulation
that involves the entire genome (Zajicek and Schwartz-Arad, 1990). Polyploidization is always
attended by the intensification of the transcription and translation and in rat liver the amino acid
label and activity of many enzymes increases proportionately to their ploidy. "Individual
chromosomes of a tetraploid genome of a hepatocyte reduplicate in the same sequence as in a
diploid one. In this case the properties of the chromosomes evidently remain unchanged and
polyploidy only means doubling the indexes of the diploid genome" (Brodsky and Uryvaeva,
1977). Polyploidy will be manifested in the liver by either increases in the number of
chromosomes per nucleus in an individual cell or by the appearance of two nuclei in a single cell.
Most cell polyploidization occurs in youth with mitotic polyploidization occurring
predominantly from 2 to 3 weeks postnatally and increases with age in mice (Brodsky and
Uryvaeva, 1977). Hepatocytes progress through a modified or polyploidizing cell cycle which
contains gaps and S-phases, but proceeds without cytokinesis. The result is the formation of the
first polyploidy cell, which is binucleated with diploid nuclei and has increased cell ploidy but
not cell number. The subsequent proliferation of bi-nucleated hepatocytes occurs with a fusion
of mitotic nuclei during metaphase that gives rise to mononucleated cells with higher levels of
ploidy. Thus, during normal liver ontogenesis, a polyploidizing cell cycle without cytokinesis
alternates with a mitotic cycle of binucleated cells and results in progressive and irreversible
increases in either cell or nuclear ploidy (Brodsky and Uryvaeva, 1977).
Polyploidization of the liver occurs during maturation in rodents and therefore,
experimental paradigms that treat or examine rodent liver during that period should take into
consideration the normally changing baseline of polyploidy in the liver. The development of
polyploidy has been correlated in rodents to correspond with maturation. (Brodsky and
Uryvaeva, 1977) report it is cells with diploid nuclei that proliferate in young mice, but that
among the newly formed cells, the percentage of those with tetraploid nuclei is high. By 1
month, most mice (CBA/C57BL mice) already have a polyploid parenchyma, but binucleate
cells with diploid nuclei predominate. In adult mice, the ploidy class with the highest percentage
of hepatocytes was the 4n X 2 class. The intensive proliferation of diploid hepatocytes occurs
only in baby mice during the first 2 weeks of life and then toward 1 month, the diploid cells
cease to maintain themselves and transform into polyploid cells. In aged animals, the
parenchyma retains only 0.02 percent of the diploid cells of the newborn animal. While the
weight of the liver increases almost 30 times within 2 years, the number of cells increase much
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
less than the weight or mean ploidy. Hence, the postnatal growth of the liver parenchyma is due
to cell polyploidization (Brodsky and Uryvaeva, 1977). In male Wistar rats fetal hepatocytes
(22 days gestation) were reported to be 85.3% diploid (2n) and 7.4% polyploid (4n + 8n) cells
with 7.3%) of cells in S-phase (SI and S2). By one month of age (25-day old suckling rats) there
were 92.9% diploid and 2.5% polyploid, at 2 months 47.5% diploid and 50.9% polyploid, at 6
months 29.1% diploid and 69.6% polyploid, and by 8 months 11.1% diploid and 87.3%
polyploidy (Sanz et al., 1996). However, mouse and rat differ in their polyploidization.
In the mouse, which has a higher degree of polyploidy than the rats, the scheme of
polyploidization differs in that each cell class, including mononucleate cells,
forms from the preceding one without being supplemented by self-maintenance.
Each cell class is regarded as the cell clone and it is implied that the cells of each
class have the same mitotic history and originate from diploid initiator cells with
similar properties. In this model 1 reproduction would give a 2n x 2 cell, the
second reproduction a 4n cell, and third reproduction a 4n X 2 cell all coming
from an originator diploid cell (Brodsky and Uryvaeva, 1977).
The cell polyploidy is most extensive in mouse liver, but also common for rat and
humans livers. The livers of young and aged mice differ considerably in the ploidy of the
parenchymal cells, but still perform fundamentally the same functions. In some mammals, such
as the mouse, rats, dog and human, the liver is formed of polyploid hepatocytes. In others, for
example, guinea pig and cats, the same functions are performed by diploid cells (Brodsky and
Uryvaeva, 1977). One obvious consequence of polyploidization is enlargement of the cells. The
volume of the nucleus and cytoplasm usually increases proportionately to the increased in the
number of chromosome sets with polyploidy reducing the surface/volume ratio. The labeling of
tritium doubles with the doubling of the number of chromosomes in the hepatocyte nucleus
(Brodsky and Uryvaeva, 1977). Kudryavtsev et al. (1993) have reported that the average levels
of cell and nuclear ploidy are relatively lower in humans than in rodent but the pattern of
hepatocyte polyploidization is similar and at maturity and especially during aging, the rate of
hepatocyte polyploidization increases with elderly individuals having binucleated and polyploid
hepatocytes constituting about one-half of liver parenchyma. Gramantieri et al. (1996) report
that in adult human liver a certain degree of polyploidization is physiological; the polyploidy
compartment (average 33% of the total hepatocytes) includes both mononucleated (28%) and
binucleated {12%) cells and the average percentage of binucleated cells in the total hepatocyte
population is 24% (Melchiorri et al., 1994).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Historically, aging in human liver has been characterized by fewer and larger
hepatocytes, increased nuclear polyploidy and a higher index of binucleate hepatocytes (Popper,
1986) but Schmucker (2005) notes that data concerning the effect of aging on hepatocyte volume
in rodent and humans are in conflict with some showing increases volume to be unchanged and
to increase by 25% by age 60 by others in humans. The irreversibility of hepatocyte polyploidy
has been used in efforts to identify the origin of tumor progenitor cells (diploid vs. polyploidy)
(see Section E.3.1.8, below). The associations with polyploidy and disease have been an active
area of study in cancer mode-of-action (MOA) studies (see Sections E.3.1.4 and E.3.3.1, below).
Not only are polyploid cells most abundant in zone 2 of the liver acinus and increase in
number with age, but polyploid cells have been reported to be more abundant following a
number of toxic insults and exposure to chemical carcinogens. Wanson et al. (1980)reported that
one of the earliest lesions obtained in the liver after A'-nitrosomorpholine treatment development
of hypertrophic parenchymal cells presenting a high degree of ploidy. Gupta (2000) reports
hepatic polyploidy is often encountered in the presence of liver disease and that for animals and
people, polyploidy is observed during advancement of liver injury due to cirrhosis or other
chronic liver disease (often described as large-cell dysplasia referring to nuclear and cytoplasmic
enlargement, nuclear pleomorphisms and multinucleation and probably representing increased
prevalence of polyploidy cells) and in old animals with toxic liver injury and impaired recovery.
Gorla et al. (2001) report that weaning and commencement of feeding, compensatory liver
hypertrophy following partial hepatectomy, toxin and drug-induced liver disease, and
administration of specific growth factors and hormones may induce hepatic polyploidy. They go
on to state that "although liver growth control has long been studied, whether the replication
potential of polyploidy hepatocytes is altered remains unresolved, in part, owing to difficulties in
distinguishing between cellular DNA synthesis and generation of daughter cells." Following
CCL4 intoxication, the liver ploidy rises and more cells become binucleate (Zajicek et al., 1989).
Minamishima et al. (2002) report that in 8-12 week old female mice before partial hepatectomy
there were 78.6% 2C, 19.1% 4C, and 2.3% 8C cells but 7 days after there were 42.0% 2C, 49.1%
4C, and 9.0% 8C. Zajicek et al. (1991) describe how hepatocyte streaming is affected after the
rapid hepatocyte DNA synthesis that occurs after the mitogenic stimulus of a partial
hepatectomy. These data are of relevance to findings of increased DNA synthesis and liver
weight gain following toxic insults and disease states. Zajicek et al. (1991) suggest that
following a mitogenic stimulus, not all DNA synthesizing cells do divide but accumulate newly
formed DNA and turn polyploid (i.e., during the first 3 days after partial hepatectomy in rats
50% of synthesized DNA was accumulated) and that since the acinus increased 15% and cell
density declined 10%, overall cell mass increased 5%. However, cell influx rose 1,300%. "In
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
order to accommodate all these cells, the 'acinus-tube' ought to swell 13-fold, while in reality it
increased only 5%" and that on day 3 "the liver remnant did not even double in its size." Zajicek
et al. (1991) conclude that apparently "cells were eliminated very rapidly, and may have even
been sloughed off, since the number of apoptotic bodies was very low" and therefore, "partial
hepatectomy triggers two processes: an acute process lasting about a week marked by massive
and rapid cell turnover during which most newly formed hepatocytes are eliminated, probably
sloughed off into the sinusoids; and a second more protracted process which served for liver
mass restoration mainly by forming new acini." Thus, a mitogenic stimulus may induce
increased ploidy and increased cell number as a result of increased DNA synthesis, and many of
the rapidly expanding number of cells resulting from such stimulation are purged and therefore,
do not participate in subsequent disease states of the liver.
Zajicek et al. (1989) note that the accumulation of DNA rather than proliferation of
hepatocytes "should be considered when evaluating the labeling index of hepatocytes labeled
with tritiated thymidine" as the labeling index, defined as the proportion of labeled cells, can
serve as a proliferation estimate only if it is assumed that a synthesizing cell will ultimately
divide. In tissues, such as the liver, "where cells also accumulate DNA, proliferation estimates
based on this index may fail" (Zajicek et al., 1989). The tendency to accumulate DNA is also
accompanied by a decreasing probability of a cell to proliferate, since young hepatocytes
generally divide after synthesizing DNA while older cells prefer instead to accumulate DNA.
However, polyploidy per se does not preclude cells from dividing (Zajicek et al., 1989). The
ploidy level achieved by the cell, no matter how high, does not, in itself, prevent it from going
through the next mitotic cycle and the reproduction of hepatocytes in the ploidy classes of 8n and
8n X 2 is common phenomenon (Brodsky and Uryvaeva, 1977). However, along with a reduced
capacity to proliferate, Sigal et al. (1999) report that the onset of polyploidy increases the
probability of cell death. The proliferative potentials of hepatocytes not only depend on their
ploidy, but also on the age of the animals with liver restoration occurring more slowly in aged
animals after partial hepatectomy (Brodsky and Uryvaeva, 1977). Species differences in the
ability of hepatocytes to proliferate and respond to a mitogenic stimulus have also been
documented (see Section E.3.4.2, below). The importance of the issues of cellular proliferation
versus DNA accumulation and the differences in ability to respond to a mitogenic stimulus
becomes apparent as identification of the cellular targets of toxicity (i.e., diploid vs. polyploidy)
and the role of proliferation in proposed MO As are brought forth. Polyploidization, as discussed
above, has been associated with a number of types of toxic injury, disease states, and
carcinogenesis by a variety of agents.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E. 1.2. Effects of Environment and Age: Variability of Response
The extent of polyploidization of the liver not only changes with age, but structural and
functional changes, as well as environmental factors (e.g., polypharmacy), affect the
vulnerability of the liver to toxic insult. In a recent review by Schmucker (2005), several of
these factors are discussed. Schmucker (2005) reports that approximately 13% of the population
of the United States is over the age of 65 years, that the number will increase substantially over
the next 50 years, and that increased age is associated with an overall decline in health and
vitality contributing to the consumption of nearly 40% of all drugs by the elderly. Schmucker
(2005) estimates that 65% of this population is medicated and many are on polypharmacy
regimes with a major consequence of a marked increase in the incidence of adverse drug
reactions (ADRs) (i.e., males and females exhibit 3- and 4-fold increases in ADRs, respectively,
when 20- and 60-year-old groups are compared). The percentage of deaths attributed to liver
diseases dramatically increases in humans beyond the age of 45 years with data from California
demonstrating a 4-fold increase in liver disease-related mortality in both men and women
between the ages of 45 and 85 years (Siegel and Kasmin, 1997). Furthermore, Schmucker
(2005) cites statistics from the United Stated Department of Health and Human Services to
illustrate a loss in potential lifespan prior to 75 years of age due to liver disease (i.e., liver disease
reduced lifespan to a greater extent than colorectal and prostatic cancers, to a similar extent as
chronic obstructive pulmonary disease, and nearly as much as HIV). Thus, the elderly are
predisposed to liver disease.
As stated above, the presence of high polyploidy cell in normal adults, nuclear
polyploidization with age, and increase in the mean nuclear volume have been reported in
people. Watanabe et al. (1978) reported the results from a cytophotometrical analysis of
35 cases of sudden death including 22 persons over 60 years of age that revealed that although
the nuclear size of most hepatocytes in a senile liver remains unchanged, there was an increase in
cells with larger nuclei. Variations in both cellular area and nucleocytoplasmic ratio were also
analyzed in the study, but the binuclearity of hepatocytes was not considered. No cases with a
clinical history of liver disease were included. Common changes in senile liver were reported to
include atrophy, fatty metamorphosis of hepatocytes, and occasional collapse of cellular cords in
the centrilobular area, slight cellular infiltration and proliferation of Kupffer cells in sinusoids,
and elongation of Glisson's triads with slight to moderate fibrosis in association with round cell
infiltration. Furthermore, cells with giant nuclei, with each containing two or more prominent
nucleoli, and binuclear cell were increased. There was a decrease in diploid populations with
age and an increase in tetraploid population and a tendency of polyploidy cells with higher
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
values than hexaploids with age. Cells with greater nuclear size and cellular sizes were observed
in livers with greater degrees of atrophy.
Schmucker notes that one of the most documented age-related changes in the liver is a
decline in organ volume but also cites a decrease in functional hepatocytes and that other studies
have suggested that the size or volume of the liver lobule increases as a function of increasing
age. Data are cited for rats suggesting sinusoidal perfusion rate in the rat liver remains stable
throughout the lifespan (Vollmar et al., 2002) but evidence in humans shows age-related shifts in
the hepatic microcirculation attributable to changes in the sinusoidal endothelium (McLean et al.,
2003) (i.e., a 60% thickening of the endothelial cell lining and an 80% decline in the number of
endothelial cell fenestrations, or pores, with increasing age in humans) that are similar in baboon
liver (Cogger et al., 2003). Such changes could impair sinusoidal blood flow and hepatic
perfusion, and the uptake of macromolecules such as lipoproteins from the blood. Schmucker
reports that there is a consensus that hepatic volume and blood flow decline with increasing age
in humans but that the effects of aging on hepatocyte structure are less clear. In rats, the volume
of individual hepatocytes was reported to increase by 60% during development and maturation,
but subsequently decline during senescence yielding hepatocytes of equivalent volumes in
senescent and very young animals (Schmucker, 2005).
The smooth surfaced endoplasmic reticulum (SER), which is the site of a variety of
enzymes involved in steroid, xenobiotic, lipid and carbohydrate metabolism, also demonstrated a
marked age-related decline rat hepatocytes (Schmucker et al., 1978; Schumucker et al., 1977).
Schmucker also notes that several studies have reported that the older rodents have less effective
protection against oxidative injury in comparison to the young animals, age-related decline in
DNA base excision repair, and increases in the level of oxidatively damaged DNA in the livers
of senescent animals in comparison to young animals. Age-related increases in the expression an
activity of stress-induced transcription factors (i.e., increased NF-kB binding activity but not
expression) were also noted, but that the importance of changes in gene expression to the role of
oxidative stress in the aging process remains unsolved. An age-related decline in the
proliferative response of rat hepatocytes to growth factors following partial hepatectomy was
noted, but despite a slower rate of hepatic regeneration, older livers eventually achieved their
original volume with the mechanism responsible for the age-related decline in the
posthepatectomy hepatocyte proliferative response unidentified.
As with other tissues, telomere length has been identified as a critical factor in cellular
aging with the sequential shortening of telomeres to be a normal process that occurs during cell
replication (see Sections E.3.1.1 and E.3.1.7, below). An association in telomere length and
strain susceptibility for carcinogenesis in mice has been raised. Herrera et al. (1999) examined
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
susceptibility to disease with telomere shortening in mice. However, this study only cites shorter
telomeres for C57BL6 mice in comparison to mixed C57BL6/129sv mice. The actual data are
not in this paper and no other strains are cited. Of the differing cell types examined, Takubo and
Kaminishi (2001) report that hepatocytes exhibited the next fastest rate of telomere shortening
despite being relatively long-lived cells raising the question of whether or not there are
correlations between age, hepatocyte telomere length and the incidence of liver disease
(Schmucker, 2005). Aikata et al. (2000) and Takubo et al. (2001) report that the mean telomere
length in healthy livers is approximately 10 kilobase pairs at 80 years of age and these
hepatocytes retain their proliferative capacity but that in diseased livers of elderly subjects was
approximately 5 kb pairs. Thus, short telomere length may compromise hepatic regeneration and
contribute to a poor prognosis in liver disease or as a donor liver (Schmucker, 2005).
Schmucker (2005) reports that interindividual variability in Phase I drug metabolism was
so large in human liver microsomes, particularly among older subjects, that the determination of
any statistically significant age or gender-related differences were precluded. In fact Schmucker
(2001) notes that "the most remarkable characteristic of liver function in the elderly is the
increase in interindividual variability, a feature that may obscure age-related differences."
Schumer notes that The National Institute on Aging estimates that only 15% of individuals aged
over 65 years exhibit no disease or disability with this percentage diminishing to 11 and 5% for
men and women respectively over 80 years. Thus, the large variability in response and the
presence of age-related increases in pharmacological exposures and disease processes are
important considerations in predicting potential risk from environmental exposures.
E.2. CHARACTERIZATION OF HAZARD FROM TRICHLOROETHYLENE (TCE)
STUDIES
The 2001 Draft assessment of the health risk assessment of TCE (U.S. EPA, 2001)
extensively cited the review article by Bull (2000) to describe the liver toxicity associated with
TCE exposure in rodent models. Most of the attention has been paid to the study of TCE
metabolites, rather than the parent compound, and the review of the TCE studies by Bull (2000)
was cursory. In addition, gavage exposure to TCE has been associated with a significant
occurrence of gavage-related accidental deaths and vehicle effects, and TCE exposure through
drinking water has been reported to decrease palatability and drinking water consumption, and to
have significant loss of TCE through volatilization, thus, further limiting the TCE database.
In its review of the draft assessment, U.S. Environmental Protection Agency (U.S.
EPA)'s Science Advisory regarding this topic suggested that in its revision, the studies of TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-10 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
should be more fully described and characterized, especially those studies considered to be key
for the hazard assessment of TCE. Although the database for studies of the parent compound is
somewhat limited, a careful review of the rodent studies involving TCE can bring to light the
consistency of observations across these studies, and help inform many of the questions
regarding potential MO As of TCE toxicity in the liver. Such information can inform current
MOA hypothesis (e.g., such as peroxisome proliferator activated receptor alpha [PPARa]
activation) as well. Accordingly, the primary acute, subchronic and chronic studies of TCE will
be described and examined in detail below with comments on consistency, major conclusions,
and the limitations and uncertainties in their design and conduct. Since all chronic studies were
conducted primarily with the goal of ascertaining carcinogenicity, their descriptions focus on that
endpoint, however, any noncancer endpoints described by the studies are described as well. For
details regarding evidence of hepatotoxicity in humans and associations with increased risk of
hepatocellular carcinoma, please refer to Sections 4.5.1 and 4.5.2. Some of the earlier studies
with TCE were contaminated with epichlorhydrin and are discussed in Sections 4.6, and 4.7 of
the TCE assessment document.
E.2.1. Acute Toxicity Studies
A number of acute studies have been undertaken to describe the early changes in the liver
after TCE administration with the majority using the oral gavage route of administration. Some
have been detailed examinations while others have reported primarily liver weight changes as a
marker of TCE-response. The matching and recording of age, but especially initial and final
body weight for control and treatment groups, is of particular importance for studies using liver
weight gain as a measure of TCE-response as difference in these parameters affect TCE-induced
liver weight gain. Most data are for exposures of at least 10 days.
E.2.2. Soni et al. (1998)
Soni et al. (1998) administered TCE in corn oil to male Sprague-Dawley (S-D) rats
(200-250 g, 8-10 weeks old) intraperitoneally at exposure levels of 250, 500, 1,250, and
2,500 mg/kg. Groups (4-6 animals per group) were sacrificed at 0, 6, 12, 24, 36, 48, 72, and
96 hours after administration of TCE or corn oil. Using this paradigm only 50% of rats survived
the 2,400 mg/kg intraperitoneal (i.p.) TCE administration with all deaths occurring between days
1 and 3 after TCE administration. Tritiated thymidine was also administered i.p. to rats 2 hours
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-11 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
prior to euthanasia. Light microscopic sections of the central lobe in 3-4 sections were
examined for each animal. The grading scheme reported by the authors was: 0, no necrosis; +1
minimal, defined as only occasional necrotic cells in any lobule; +2, mild, defined as less than
one-third of the lobular structure affected; +3, moderate, defined as between one-third and two-
thirds of the lobular structure affected; +4 severe, defined as greater than two-thirds of the
lobular structure affected. At the 2,500 mg/kg dose, histopathology data were obtained for the
surviving rats (50%). Lethality studies were done separately in groups of 10 rats. The survival
in the groups of rats administered TCE and sacrificed from 0 to 96 hours was given as 30%
mortality at 48 hours and 50% mortality by 72 hours.
The authors report that controls and 0-hour groups did not show sign of tissue injury or
abnormality. The authors only report a single number with one significant figure for each group
of animals with no means or standard deviations provided. In terms of the extent of necrosis
there was no difference between the 250 and 500 mg/kg/treated dose groups though 96 hours
with a single +1 given as the maximal amount of hepatocellular necrosis (minimal as defined by
occasional necrotic cells in any lobule). At the 1,250 mg/kg dose the maximal score was
achieved 24 hours after TCE administration and was reported as simply +2 (mild, defined as less
than one-third of lobular structure affected). The level of necrosis was reported to diminish to a
score of 0 by 72 hours after 250 mg/kg TCE with no decrease at 500 mg/kg. At 1,250 mg/kg, the
extent of necrosis was reported to diminish from +2 to +1 by 72 hours after administration. At
the 2,500 mg/kg dose (LD50 for this route) by 48 hours, the surviving rats were reported to have a
score of +4 (severe as defined by greater than two thirds of the lobular structure affected). The
authors report that
The necrosed cells were concentrated mostly in the midzonal areas and the cells
around central vein area were unaffected. Extensive necrosis was observed
between 24 and 48 hours for both 1250 and 2500 mg/kg groups. Injury was
maximal in the group receiving 2500 mg/kg between 36 and 48 hours as
evidenced by severe midzonal necrosis, vacuolization, and congestion.
Infiltration of polymorphonuclear cell was evident at this time as a mechanism for
cleaning dead cells and tissue debris from the lobules. At the highest dose, the
injury also started to spread toward the centrilobular areas. At the highest dose,
30 and 50% lethality was observed at 48 and 72 h, respectively. After 48 h, the
number of necrotic cells decreased and the number of mitotic cells increased. The
groups receiving 500 and 1250 mg/kg TCE showed relatively higher mitotic
activity as evidenced by cells in metaphase compared to other groups.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The authors do not give a quantitative estimate or indication as to the magnitude of the number
of cells going through mitosis. Although there was variability in the number of animals dying at
1,250 mg/kg TCE exposure though this route of exposure, no indication of variability in response
within these treatment groups was given by the authors in regard to extent of histopathological
changes. The authors do not comment on the manner of death using this paradigm or of the
effects of i.p. administration regarding potential peritonitis and inflammation.
TCE hepatotoxicity was "assessed by measuring plasma" sorbitol dehydrogenase (SDH)
and alanine aminotransferase (ALT) after TCE administration with vehicle treated control groups
reported to induce no increases in these enzymes. Plasma SDH levels were reported to increase
in a linear fashion after 250, 500, and 1,250 mg TCE/kg i.p. administration by 6 hours (i.e., ~3-,
10.5-, 22-, and 24.5-fold in comparison to controls from 250, 500, 1,250, and 2,500 mg/kg TCE,
respectively) with little difference between the 1,250 and 250 mg/kg dose. By 12 hours the 250,
500, and 1,250 levels had diminished to levels similar to that of the 250 mg/kg dose at 6 hours.
The 2,500 mg/kg levels were somewhat diminished from its 6 hour level. By 24 hours after TCE
administration by the i.p. route of administration, all doses were similar to that of the 250-mg/kg-
TCE 6-hour level. This pattern was reported to be similar for 5-, 36-, 48-, 72-, and 96-hour time
points as well. The results presented were the means and SE for four rats per group. The authors
did not indicate which rats were selected for these results from the 4-6 that were exposed in each
group. Thus, only SDH levels showed dose-dependence in results at the 6 hour time point and
such increases did not parallel the patterns reported for hepatocellular necrosis from
histopathological examination of liver tissues.
For ALT, the pattern of plasma concentrations after i.p. TCE administration differed both
from that of SDH but also from liver histopathology. Plasma ALT levels were reported to
increase in a nonlinear fashion and to a much smaller extent that SDH (i.e., -2.7-, 1.9-, 2.1-, and
4.0-fold of controls from 250, 500, 1,250, and 2,500 mg/kg TCE, respectively). The patterns for
12, 24, 36, 48, 72, and 96 hours were similar to that of the 6-hour exposure and did not show a
dose-response. The authors injected carbon tetrachloride (2.5.mL/kg) into a separate group of
rats and then incubated the resulting plasma with unbuffered trichloroacetic acid (TCA; 0, 200,
600, or 600 nmol) with decreases in enzyme activity in vitro at the two higher concentrations. It
is not clear whether in vitro unbuffered TCA concentrations of this magnitude, which could
precipitate proteins and render the enzymes inactive, are relevant to the patterns observed in the
in vivo data. The extent of extinguishing of SDH and ALT activity at the two highest TCA
levels in vitro were the same, suggestive of the generalized in vitro pH effect. However, the
enzyme activity levels after TCE exposure had different patterns, and thus, suggesting that in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
vitro TCA results are not representative of the in vivo TCE results. Neither ALT nor SDH levels
corresponded to time course or dose-response reported for the histopathology of the liver
presented in this study.
Tritiated thymidine results from isolated nuclei in the liver did not show a pattern
consistent with either the histopathology or enzyme results. These results were for whole-liver
homogenates and not separated by nuclear size or cell origin. Tritiated thymidine incorporation
was assumed by the authors to represent liver regeneration. There was no difference between
treated and control animals at 6 hours after i.p. TCE exposure and only a decrease (-50%
decrease) in thymidine incorporation after 12 hours of the 2,500 mg/kg TCE exposure level. By
24 hours, there as 5.6- and 2.8-fold tritiated thymidine incorporation at the 500 and 1,250 mg/kg
TCE levels with the 250 and 2,500 mg/kg levels similar to controls. For 36, 48, and 72 hours
after i.p. TCE exposure there continued to be no dose-response and no consistent pattern with
enzyme or histopathological lesion patterns. The authors presented "area under the curve" data
for tritiated thymidine incorporation for 0 to 95 hours, which did not include control values.
There was a slight elevation at 500 mg/kg TCE and slight decrease at 2,500 mg/kg from the
250 mg/kg TCE levels. Again, these data did not fit either histopathology or enzyme patterns
and also can include the contribution of nonparenchymal cell nuclei as well as changes in ploidy.
The use of an i.p. route of administration is difficult to compare to oral and inhalation
routes of exposure given that peritonitis and direct contact with TCE and corn oil with liver
surfaces may alter results. Whereas Soni et al. (1998) report the LD50 to be 2,500 mg/kg TCE
via i.p. administration, both Elcombe et al. (1985) and Melnick et al. (1987) do not report
lethality from TCE administered for 10 days at 1,500 mg/kg in corn oil, or up to 4,800 mg/kg/d
for 10-days in encapsulated feed. Also TCE administered via gavage or oral administration
through feed will enter the liver through the circulation with periportal areas of the liver the first
areas exposed with the entire liver exposed in a fashion dependent on blood concentration levels.
However, with i.p. administration, the absorption and distribution pattern of TCE will differ.
The lack of concordance with measures of liver toxicity from this study and the lack
concordance of patterns and dose-response relationships of toxicity reported from other more
environmentally and physiologically relevant routes of exposure make the relevance of these
results questionable.
E.2.2.1. Soni et al. (1999)
A similar paradigm and the same results were reported for Soni et al. (1999), in which
hepatocellular necrosis, tritiated thymidine incorporation, and in vitro inhibition of SDH and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-14 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
ALT data were presented along with dose-response studies with ally alcohol and a mixture of
TCE, Thioacetamine, allyl alcohol, and chloroform. The same issues with interpretation present
for Soni et al. (1998) also apply to this study as well.
E.2.2.2. Okino et al. (1991)
This study treated adult Wistar male rats (8 weeks of age) with TCE after being on a
liquid diet for 3 weeks and either untreated or pretreated with phenobarbital or ethanol. TCE
exposure was at 8,000 ppm for 2 hours, 2,000 or 8,000 ppm for 2 hours, and 500 or 2,000 ppm
for 8 hours. Each group contained 5 rats. Livers from rats, that were not pretreated with either
ethanol or Phenobarbital, were reported to show only a few necrotic hepatocytes around the
central vein at 6 and 22 hours after 2 hours of 8,000-ppm TCE exposure. At increased lengths
and/or concentrations of TCE exposure, the frequencies of necrotic hepatocytes in the
centrilobular area were reported to be increased but the number of necrotic hepatocytes was still
relatively low (out of-150 hepatocytes the percentages of necrotic pericentral hepatocytes were
0.2% ± 0.4%, 0.3% ± 0.4%, 2.7% ± 1.0%, 0.2% ± 0.4%, and 3.5% ± 0.4% for control,
2,000 ppm TCE for 2 hours, 8,000 ppm TCE for 2 hours, 500 ppm TCE for 8 hours, and 2,000
ppm TCE for 8 hours, respectively).
"Ballooned" hepatocytes were reported to be zero for controls and all TCE treatments
with the exception of 0.3% ± 0.6% ballooned midzonal hepatocytes after 8,000 ppm TCE for 2
hours exposure. Microsomal protein (mg/g/liver) was increased with TCE exposure
concentration and duration, but not reported to be statistically significant (i.e., mg/g/liver
microsomal protein was 21.2 ± 4.3, 22.0 ± 1.5, 25.9 ± 1.3, 23.3 ± 0.8, and 24.1 ±1.0 for control,
2,000 ppm TCE for 2 hours, 8,000 ppm TCE for 2 hours, 500 ppm TCE for 8 hours, and 2,000
ppm TCE for 8 hours, respectively).
The metabolic rate of TCE was reported to be increased after exposures over 2,000 ppm
TCE (i.e., metabolic rate of TCE in nmol/g/liver/min was 29.5 ± 5.7, 51.3 ± 6.0, 63.1 ± 16.0,
37.3 ± 3.3, and 69.5 ± 4.3 for control, 2,000 ppm TCE for 2 hours, 8,000 ppm TCE for 2 hours,
500 ppm TCE for 8 hours, and 2,000 ppm TCE for 8 hours, respectively). However, the
cytochrome P450 content of the liver was not reported to increase with TCE exposure
concentration or duration.
The liver/body weight ratios were reported to increase with all TCE exposures except
500 ppm for 8 hours (i.e., the liver/body weight ratio was 3.18%> ± 0.15%, 3.35% ± 0.10%,
3.39% ± 0.20%, 3.15% ± 0.10%, and 3.57% ± 0.14% for control, 2,000 ppm TCE for 2 hours,
8,000 ppm TCE for 2 hours, 500 ppm TCE for 8 hours, and 2,000 ppm TCE for 8 hours,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-15 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
respectively). These values represent 1.05-, 0.99-, 1.06-, and 1.12-fold of control in the 2,000
ppm TCE for 2 hours, 8,000 ppm TCE for 2 hours, 500 ppm TCE for 8 hours, and 2,000 ppm
TCE for 8 hours treatment groups, respectively. A statistically significant difference observed
after 8 hours of 2,000-ppm TCE exposure. Initial body weights and those 22 hours after
cessation of exposure were not reported, which may have affected liver weight gain. However,
these data suggest that TCE-related increases in metabolism and liver weight occurred as early as
22 hours after exposures of this magnitude from 2 to 8 hours of TCE with little concurrent
hepatic necrosis.
Ethanol and phenobarbital pretreatment were reported to enhance TCE toxicity. In
ethanol-treated rats a few necrotic hepatocytes were reported to be around the central vein along
with hepatocellular swelling without pyknotic nuclei at 6 hours after TCE exposure with no
pathological findings in the midzonal or periportal areas. At 22 hours centrilobular hepatocytes
were reported to have a few necrotic hepatocytes and cell infiltrations around the central vein,
but midzonal areas were reported to have ballooned hepatocytes with pyknotic nuclei frequently
accompanied by cell infiltrations. In phenobarbital treated rats 6 hours after TCE exposure,
centrilobular hepatocytes showed prenecrotic changes with no pathological changes reported to
be observed in the periportal areas. By 22 hours, zonal necrosis was reported in centrilobular
areas or in the transition zone between centrilobular and periportal areas. Treatment with
phenobarbital or ethanol induced hepatocellular necrosis primarily in centrilobular areas with
phenobarbital having a greater effect (89.1% ± 8.5% centrilobular necrosis) at the higher dose
and shorter exposure duration (8,000 ppm x 2 hours) with ethanol having a greater effect
(16.8%) ± 5.3%o centrilobular necrosis) at the lower concentration and longer duration of exposure
(2,000 ppm x 8 hours).
E.2.2.3. Nunes et al. (2001)
This study was focused on the effects of TCE and lead coexposure but treated male
75-day old S-D rats with 2,000 mg/kg TCE for 7 days via corn-oil gavage (n = 10). The rats
ranged in weight from 293 to 330 g (~12%>) at the beginning of treatment and were pretreated
with corn oil for 9 days prior to TCE exposure. TCE was reported to be 99.9% pure. Although
the methods section states that rats were exposed to TCE for 7 days, Table 1 of the study reports
that TCE exposure was for 9 days. The beginning body weights were not reported specifically
for control and treatment groups, but the body weights at the end of exposure were reported to be
342 ± 18 g for control rats and 323 ± 3 g for TCE exposed rats, and that difference (~6%>) to be
statistically significant. Because beginning body weights were not reported, it is difficult to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-16 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
distinguish whether differences in body weight after TCE treatment were treatment-related or
reflected differences in initial body weights. The liver weights were reported to be 12.7 ± 1.0 g
in control rats and 14.0 ± 0.8 g for TCE treated rats with the percent liver/body weight ratios of
3.7 and 4.3%, respectively. The increase in percent liver/body weight ratio represents 1.16-fold
of control and was reported to be statistically significant. However, difference in initial body
weight could have affected the magnitude of difference in liver weight between control and
treatment groups. The authors report no gross pathological changes in rats gavaged with corn oil
or with corn oil plus TCE but observed that one animal in each group had slightly discolored
brown kidneys. Histological examinations of "selected tissues" were reported to show an
increased incidence of chronic inflammation in the arterial wall of lungs from TCE-dosed
animals. There were no descriptions of liver histology given in this report for TCE-exposed
animals or corn-oil controls.
E.2.2.4. Tao et al. (2000b)
The focus of this study was to assess the affects of methionine on methylation and
expression of c-Jun and c-Myc in mouse liver after 5 days of exposure to TCE (1,000 mg/kg in
corn oil) and its metabolites. Female 8-week old B6C3F1 mice (n = 4-6) were administered
TCE ("molecular biology or HPLC grade") for 5 days with and without methionine (300 mg/kg
i.p.). Data regarding % liver/body weight was presented as a figure. Of note is the decrease in
liver/body weight ratio by methionine treatment alone (-4.6% liver/body weight for control and
-4.0% liver/body weight for control mice with methionine or -13% difference in liver/body
weight ratios between these groups). Neither initial body weights nor body weights after
exposure were reported by the authors so that the reported effects of treatment could have
reflected differences in initial body weights of the mice. TCE exposure was reported to increase
the percent liver/body weight ratio to —5.8% without methionine and to increase percent
liver/body weight ratio to -5.7% with methionine treatment. These values represent 1.26-fold of
control levels from TCE exposure without methionine and 1.43-fold of control from TCE
exposure with methionine. The number of animals examined was reported to be 4-6 per group.
The authors reported the differences between TCE treated animals and their respective controls
to be statistically significant but did not examine the differences between controls with and
without methionine. There were no descriptions of liver histology given in this report for TCE-
exposed animals or corn-oil controls.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-17 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.2.2.5. Tucker et al. (1982)
This study describes acute LD50, and 5- and 14-days studies of TCE in a 10% emulphor
solution administered by gavage. Screening level subchronic drinking water experiments with
TCE dissolved in 1% emulphor in mice were also conducted but with little detail reported. The
authors did describe the strains used (CD-I and ICR outbred albino) and that they were
"weanling mice," but the ages of the mice and their weights were not given. The TCE was
described as containing 0.004% diisopropylamine as the preservative and that the stabilizer had
not been found carcinogenic or overtly toxic. The authors report that "the highest concentration
a mouse would receive during these studies is 0.03 mg/kg/day." The main results are basically
an LD50 study and a short term study with limited reporting for 4 and 6-month studies of TCE
exposure. Importantly, the authors documented the loss of TCE from drinking water solutions
(less than 20% of the TCE was lost during the 3 or 4 days in the water bottles at 1.0, 2.5, and 5.0
mg/mL concentrations, but in the case of 0.1 mg/mL, up to 45% was lost over a 4-day period).
The authors also report that high doses of TCE in drinking water reduced palatability to such an
extent that water consumption by the mice was significantly decreased.
The LD50 with 95% confidence were reported to be 2,443 mg/kg (1,839 to 3,779) for
female mice and 2,402 mg/kg (2,065 to 2,771) for male mice. However, the number of mice
used in each dosing group was not given by the authors. The deaths occurred within 24 hours of
TCE administration with no animals recovering from the initial anesthetic effect of TCE dying
during the 14-day observation period. The authors reported that the only gross pathology
observed was hyperemia of the stomach of mice dying form lethal doses of TCE, and that mice
killed at 14 days showed not gross pathology.
In a separate experiment, male CD-I mice were exposed to TCE by daily gavage for 14
days at 240 and 24 mg/kg. These two doses did not cause treatment related deaths and body
weight and "most" organ weights were reported by the authors to not be significantly affected
but the data were not shown. The only effect noted was increased liver weight, which appeared
to be dose dependent but was reported to be significant only at the higher dose. The only
significant difference found in hematology was a 5% lower hematocrit in the higher dose group.
The number of animals tested in this experiment was not give by the authors.
Male CD-I mice (n= 11) were given TCE via gavage for 5 days (0.73 g/kg TCE twice on
Day 0, 1.46 g/kg twice on Day 1, 2.91 g/kg twice on Day 3, and 1.46 g/kg TCE on Days 4 and 5)
with only 4 of 11 mice treated with TCE surviving.
In a subchronic study, male and female CD-I mice received TCE in drinking water at
concentrations of 0, 0.1, 1.0, 2.5, and 5 mg/mL in 1% emulphor, and a naive group received
deionized water. There were 140 animals of each sex in the naive group and in each treatment
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-18 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
group, except for 260 mice in the vehicle groups. Thirty mice of each sex and treatment were
selected for recording body weights for 6 months. The method of "selection" was not given by
the authors. These mice were weighed twice weekly and fluid consumption was measured by
weighing the six corresponding water bottles. The authors reported that male mice at the two
highest doses of TCE consumed 41 and 66 mL/kg/day less fluid over the 6 months of the study
than mice consuming vehicle only and that this same decreased consumption was also seen in the
high dose (5 mg/mL) females. They report that weight gain was not affected except at the high
dose (5mg/mL) and even though the weight gain for both sexes was lower than the vehicle
control group, it was not statistically significant. However, these data were not shown. The
authors report that gross pathological examinations performed on mice killed at 4 and 6 months
were unremarkable and that a number of mice from all the dosing regimens had liver
abnormalities, such as pale, spotty, or granular livers. They report that 2 of 58 males at 4
months, and 11 of 59 mice at 6 months had granular livers and obvious fatty infiltration, and that
mice of both sexes were affected. Animals in the naive and vehicle groups were reported to
infrequently have pale or spotty livers, but exhibit no other observable abnormalities. No
quantitation or more detailed descriptions of the incidence of or severity of effects were given in
this report.
The average body weight of male mice receiving the highest dose of TCE was reported to
be 10% lower at 4 months and 11% lower at 6 months with body weights of female mice at the
highest dose also significantly lower. Enlarged livers (as percentage of body weight) were
observed after both durations of exposure in males at the three highest doses, and in females at
the highest dose. In the 4-month study, brain weights of treated females were significantly
increased when compared to vehicle control. However, the authors state
This increase is apparently because the values for the vehicle group were low,
because the naive group was also significantly increased when compared to
vehicle control. A significant increase in kidney weight occurred at the highest
dose in males at 6 months and in females, after both 4 and 6 months of TCE
exposure. Urinalysis indicated elevated protein and ketone levels in high-dose
females and the two highest dose males after 6 months of exposure (data not
shown).
The authors describe differences in hematology to include
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-19 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
a decreased erythrocyte count in the high dose males at 4 and 6 months (13% and
16%, respectively); decreased leukocyte counts, particularly in the females at 4
months and altered coagulation values consisting of increased fibrinogen in males
at both times and shortened prothrombin time in females at 6 months (data not
shown). No treatment-related effects were detected on the types of white cells in
peripheral blood.
It must be noted that effects reported from this study may have also been related to decreased
water consumption, this study did not include any light microscopic evaluation, and that most of
the results described are for data not shown. However, this study does illustrate the difficulties
involved in trying to conduct studies of TCE in drinking water, that the LD50s for TCE are
relatively high, and that liver weight increases were observed with TCE exposure as early as few
weeks and increased liver weight were sustained through the 6-month study period.
E.2.2.6. Goldsworthy and Popp (1987)
The focus of this study was peroxisomal proliferation activity after exposure to a number
of chlorinated solvents. In this study 1,000 mg/kg TCE (99+ % epoxide stabilizer free) was
administered to male F-344 rats (170-200 g or -10% difference) and B6C3F1 (20-25 g or -20%
difference) mice for 10 days in corn oil via gavage. The ages of the animals were not given. The
TCE-exposed animals were studied in two experiments (Experiments #1 and #3). In experiment
#2 corn oil and methyl cellulose vehicles were compared. Animals were killed 24 hours after the
last exposure. The authors did not show data on body weight but stated that the administration of
test agents (except WY-14,643 to rats which demonstrated no body weight gain) to rats and mice
for 10 days "had little or no effect on body weight gain." Thus, differences in initial body weight
between treatment and control groups, which could have affected the magnitude of TCE-induced
liver weight gain, were not reported. The liver/body weight ratios in corn oil gavaged rats were
reported to be 3.68% ± 0.06% and 4.52% ± 0.08% after TCE treatment which represented
1.22-fold of control (n = 5). Cyanide-(CN-)insenstive palmitoyl CoAl2 oxidation (PCO) was
reported to be 1.8-fold increased after TCE treatment in this same group. In B6C3F1 mice the
liver/body weight ratio in corn oil gavaged mice was reported to be 4.55% ± 0.13% and
6.83% ± 0.13% after TCE treatment which represented 1.50-fold of control (n = 7).
CN-insensitive PCO activity was reported to be 6.25-fold of control after TCE treatment in this
12CoA = coenzyme A.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-20 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
same group. The authors report no effect of vehicle on PCO activity but do not show the data
nor discuss any effects of vehicle on liver weight gain. Similarly the results for experiment #3
were not shown nor liver weight discussed with the exception of PCO activity reported to be
2.39-fold of control in rat liver and 6.25-fold of control for mouse liver after TCE exposure. The
number of animals examined in Experiment #3 was not given by the authors or the variation
between enzyme activities. However, there appeared to be a difference in PCO activity
Experiments #1 and #3 in rats. There were no descriptions of liver histology given in this report
for TCE-exposed animals or corn-oil controls.
E.2.2.7. Elcombe et al. (1985)
In this study, preservative free TCE was given via gavage to rats and mice for
10 consecutive days with a focus on changes in liver weight, structure, and hepatocellular
proliferation induced by TCE. Male Alderly Park rats (Wistar derived) (180-230 g), male
Osborne-Mendel rats (240-280 g), and male B6C3F1 or male Alderly Park Mice (Swiss)
weighing 30 to 35 g were administered 99.9% pure TCE dissolved in corn oil via gavage. The
ages of the animals were not given by the authors. The animals were exposed to 0, 500, 1,000,
or 1,500 mg/kg body wt TCE for 10 consecutive days. The number of mice and rats varied
widely between experiments and treatment groups and between various analyses. In some
experiments animals were injected with tritiated thymidine approximately 24 hours following the
final dose of TCE and killed one hour later. The number of hepatocytes undergoing mitosis was
identified in 25 random high-power fields (X40) for each animal with 5,000 hepatocyte per
animal examined. There was no indication by the authors that zonal differences in mitotic index
were analyzed. Sections of the liver were examined by light and electron microscopy by
conventional staining techniques. Tissues selected for electron microscopy included central vein
and portal tract so that zonal differences could be elucidated. Morphometric analysis of
peroxisomes was performed "according to general principles of Weibel et al (1964) on
electronphotomicrographs from pericentral hepatocytes." DNA content of samples and
peroxisomal enzyme activities were determined in homogenized liver (catalase and PCO
activity).
The authors reported that TCE treatment had no significant effect on body-weight gain
either strain of rat or mouse during the 10 days exposure period. However, marked increases (up
to 175% of control value) in the percent liver/body weight ratio were observed in TCE-treated
mice. Smaller increases (up to 130% of control) in relative liver weight were observed in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-21 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
TCE-treated rats. No significant effects of TCE on hepatic water content were seen so that the
liver weight did not represent increased water retention.
An interesting feature of this study was that it was conducted in treatment blocks at
separate times with separate control groups of mice for each experimental block. Therefore,
there were three control groups of B6C3F1 mice (n= 10 for each control group) and three
control groups for Alderly Park (n = 9 to 10 for each control group) mice that were studied
concurrently with each TCE treatment group. However, the percent liver/body weight ratios
were not the same between the respective control groups. There was no indication from the
authors as to how controls were selected or matched with their respective experimental groups.
The authors did not give liver weights for the animals so the actual changes in liver weights are
not given. The body weights of the control and treated animals were also not given by the
authors. Therefore, if there were differences in body weight between the control groups or
treatment groups, the liver/body weight ratios could also have been affected by such differences.
The percentage increase over control could also have been affected by what control group each
treatment group was compared to. There was a difference in the mean percent liver/body weight
ratio in the control groups, which ranged from 4.32 to 4.59% in the B6C3F1 mice (-6%
difference) and from 5.12 to 5.44% in the Alderly Park mice (-6% difference). The difference in
average percent liver/body weight ratio for untreated mice between the two strains was -16%.
Because the ages of the mice were not given, the apparent differences between strains may have
been due to both age or to strain.
After TCE exposure, the mean percent liver/body weight ratios were reported to be
5.53%) for 500 mg/kg, 6.50%> for 1,000 mg/kg, and 6.74%> for 1,500 mg/kg TCE-exposed
B6C3F1 mice. This resulted in 1.20-, 1.50-, and 1.47-fold values of control in percent liver
weight/body weight for B6C3F1 mice. For Alderly Park mice, the percent liver/body weight
ratios were reported to be 7.31, 8.50, and 9.54%> for 500, 1,000, and 1,500 mg/kg TCE treatment,
respectively. This resulted in 1.43-, 1.56-, and 1.75-fold of control values. Thus, there appeared
to be more of a consistent dose-related increase in liver/body weight ratios in the Alderly Park
mice than the B6C3F1 mice after TCE treatment. However, the variability in control values may
have distorted the dose-response relationship in the B6C3F1 mice. The Standard deviations for
liver/body weight ratio were as much as 0.52%> for the treated B6C3F1 mice and 0.91%> for the
Alderly Park treated mice. In regard to the correspondence of the magnitude of the TCE-induced
increases in percent liver/body weight with the magnitude of difference in TCE exposure
concentrations, in the B6C3F1 mice the increases were similar (~2-fold) between the 500 mg/kg
and 1,000 mg/k TCE exposure groups. For the Alderly Park mice, the increases in TCE
exposure concentrations were slightly less than the magnitude of increases in percent liver/body
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-22 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
ratios between all of the concentrations (i.e., ~1.3-fold of control vs. 2-fold for 500 and 1,000
mg/kg TCE dose and 1.3-fold of control vs. 1.5-fold for the 1,000 and 1,500 mg/kg TCE dose).
The DNA content of the liver varied greatly between control animal groups. For B6C3F1
mice it ranged from 2.71 to 2.91 mg/g liver. For Alderly Park mice it ranged from 1.57 to
2.76 mg/g liver. The authors do not discuss this large variability in baseline levels of DNA
content. The DNA content in B6C3F1 mice was mildly depressed by TCE treatment in a
nondose dependent manner. DNA concentration decrease from control ranged from 20-25%
between all three TCE exposure levels in B6C3F1 mice. For Alderly Park mice there was also
nondose related decrease in DNA content from controls that ranged from 18% to 34%. Thus, the
extent of decrease in DNA content of the liver from TCE treatment in B6C3F1 mice was similar
to the variability between control groups. The lack of dose-response for apparent treatment-
related effects in B6C3F1 mice and especially in the Alderly Park mice was confounded by the
large variability in the control animals. The changes in liver weight after TCE exposure for the
AP mice did not correlate with changes in DNA content further, raising doubt about the validity
of the DNA content measures. However, a small difference in DNA content due to TCE
treatment in all groups was reported for both strains and this is consistent with hepatocellular
hypertrophy.
The reported results for incorporation of tritiated thymidine in liver DNA showed large
variation in control groups and standard deviations that were especially evident in the Alderly
Park mice. For B6C3F1 mice, mean control levels were reported to range from 5,559 to
7,767 dpm/mg DNA with standard deviations ranging from 1,268 to 1,645 dpm/mg DNA. In
Alderly Park mice mean control levels were reported to range from 6,680 to 10,460 dpm/mg
DNA with standard deviations ranging from 308 to 5,235 dpm/mg DNA. For B6C3F1 mice,
TCE treatment was reported to induce an increase in tritiated thymidine incorporation with a
very large standard deviation, indicating large variation between animals. For 500 mg/kg TCE
treatment group the values were reported as 12,334 ± 4,038, for 1,000 mg/kg TCE treatment
group 21,909 ± 13,386, and for 1,500 mg/kg treatment TCE group 26,583 ± 10,797 dpm/mg
DNA. In Alderly Park mice TCE treatment was reported to give an increase in tritiated
thymidine incorporation also with a very large standard deviation. For 500 mg/kg TCE, the
values were reported as 19,315 ± 12,280, for 1,000 mg/kg TCE 21,197 ± 8,126 and for
1,500 mg/kg TCE 38,370 ± 13,961. As a percentage of concurrent control, the increase in
tritiated thymidine was reported to be 2.11-, 2.82-, and 4.78-fold of control in B6C3F1 mice, and
2.09-, 2.03-, and 5.74-fold of control in Alderly Park mice. Accordingly, the change in tritiated
thymidine incorporation did show a treatment related increase but not a dose-response.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Similar to the DNA content of the liver, the large variability in measurements between
control groups and variability between animals limit quantitative interpretation of these data.
The increase in tritiated thymidine, seen most consistently only at the highest exposure level in
both strains of mice, could have resulted from either a change in ploidy of the hepatocytes or cell
number. However, the large change in volume in the liver (75%) in the Alderly Park mice, could
not have resulted from only a 4-fold of control in cell proliferation even if all tritiated thymidine
incorporation had resulted from changes in hepatocellular proliferation. As mentioned in Section
E.l.l above, the baseline level of hepatocellular proliferation in mature control mice is very low
and represents a very small percentage of hepatocytes.
In the experiments with male rats, the same issues discussed above, associated with the
experimental design, applied to the rat experiments with the additional concern that the numbers
of animals examined varied greatly (i.e., 6 to 10) between the treatment groups. In Osborne-
Mendel rats, the control liver/body weight ratio was reported to vary from 4.26 to 4.36% with the
standard deviations varying between 0.22 to 021%. For the Alderly Park rats, the liver/body
weight ratios were reported to vary between 4.76 and 4.96% (in control groups) with standard
deviations varying between 0.24 to 0.47%. TCE treatment was reported to induce a dose-related
increase in liver/body weight ratio in Osborne-Mendel rats with mean values of 5.16, 5.35, and
5.53%) in 500, 1,000, and 1,500 mg/kg TCE treated groups, respectively. This resulted in 1.18-,
1.26-, and 1.30-fold values of control. In Alderly Park rats, TCE treatment was reported to result
in increased liver weights of 5.45, 5.83, and 5.65% for 500, 1,000, and 1,500 mg/kg TCE
respectively. This resulted in 1.14-, 1.17-, and 1.17-fold values of control. Again, the variability
in control values may have distorted the nature of the dose-response relationships in Alderly Park
rats. TCE treatment was reported to result in standard deviations that ranged from 0.31 to 0.48%
for Osborne-Mendel rats and 0.24 to 0.38% for Alderly Park rats. What is clear from these
experiments is that TCE exposure was associated with increased liver/body weight in rats.
The reported mean hepatic DNA concentrations and standard deviations varied greatly in
control rat liver as it did in mice. The variation in DNA concentration in the liver varied more
between control groups than the changes induced by TCE treatment. For Osborne-Mendel rats,
the mean control levels of mg DNA/g liver were reported to range from 1.99 to 2.63 mg
DNA/liver with standard deviations varying from 0.17 to 0.33 mg DNA/g. For Alderly Park
rats, the mean control levels of mg DNA/g liver were reported to be 2.12 to 3.16 mg DNA/g with
standard deviation ranging from 0.06 to 1.04 mg DNA/g. TCE treatment decreased the liver
DNA concentration in all treatment groups. For Osborne-Mendel rats, the decrease ranged from
8 to 13%) from concurrent control values and for Alderly Park rats the decrease ranged from 8 to
17%). There was no apparent dose response in the decreases in DNA content with all TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-24 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
treatment levels giving a similar decrease from controls and the same limitations discussed above
for the mouse data apply here. The magnitude of increases in liver/body ratios shown by TCE
treatment were not correlated with the changes in DNA content. However, as with the mouse
data, the small differences in DNA content due to TCE treatment in all groups and in both strains
was consistent with hepatocellular hypertrophy.
Incorporation of tritiated thymidine was reported to be even more variable between
control groups of rats than it was for mice and was reported to be especially variable between
control groups (i.e., 2.7-fold difference between control groups within strain) and differed
between the strains (average of 2.5-fold between strains). For Osborne-Mendel rats the mean
control levels were reported to range from 13,315 to 33,125 dpm/mg DNA, while for Alderly
Park rats tritiated thymidine incorporation ranged from 26,613 to 69,331 dpm/mg DNA for
controls. The standard deviations were also very large (i.e., for control groups of Osborne-
Mendel rats they were reported to range from 8,159 to 13,581 dpm/mg DNA, while for Alderly
Park rats they ranged from 9,992 to 45,789 dpm/mg DNA). TCE treatment was reported to
induce increases over controls of 110, 118, and 106% for 500, 1,000, and 1,500 mg/kg TCE-
exposed groups, respectively, in Osborne-Mendel rats with large standard deviations for these
treatment groups as well. In Alderly Park rats, the increases over controls were reported to be
206, 140, and 105% for 500, 1,000, and 1,500 mg/kg TCE, respectively. In general, these data
do indicate that TCE treatment appeared to give a mild increase in tritiated thymidine
incorporation but the lack of dose-response can be attributable to the highly variable
measurements of tritiated thymidine incorporation in control animal groups. The variation in the
number of animals examined between groups and small numbers of animals examined
additionally decrease the likelihood of being able to discern the magnitude of difference between
species- or strain-related effects for this parameter. Again, given the very low level of
hepatocyte turnover in control rats, this does not represent a large population of cells in the liver
that may be undergoing proliferation and cannot be separated from changes in ploidy.
The authors report that the reversibility of these phenomena was examined after the
administration of TCE to Alderly Park mice for 10 consecutive days. Effects upon liver weight,
DNA concentration, and tritiated thymidine incorporation 24 and 48 hours after the last dose of
TCE were reported to be still apparent. However, 6 days following the last dose of TCE, all of
these parameters were reported to return to control values with the authors not showing the data
to support this assertion. Thus, cessation of TCE exposure would have resulted in a 75%
reduction in liver weight by one week in mice exposed to the highest TCE concentration.
Analyses of hepatic peroxisomal enzyme activities were reported for catalase and
P-oxidation (PCO activity) following administration of TCE to B6C3F1 mice and Alderly Park
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-25 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
rats exposed to 1,000 mg/kg TCE for 10 days. The authors only used 5 control and 5 exposed
animals for these tests. An 8-fold of control value for PCO activity and a 1.5-fold of control
value for catalase activity were reported for B6C3F1 mice exposed to 1,000 mg/kg TCE. In the
Alderly Park rats, no significant changed occurred. It is unclear which mice or rats were selected
from the previous experiments for these analyses and what role selection bias may have played
in these results. The reduced number of animals chosen for this analysis also reduces the power
of the analysis to detect a change. In rats, there was a reported 13% increase in PCO; however,
the variation between the TCE treated rats was more than double that of the control animals in
this group and the other limitations described above limit the ability to detect a response. There
was no discussion given by the authors as to why only one dose was tested in half of the animals
exposed to TCE or why the strain with the lowest liver weight change due to TCE exposure was
chosen as the strain to test for peroxisomal proliferative activity.
The authors provided a description of the histopathology at the light microscropy level in
B6C3F1 mice, Alderly Park mice, Osborne-Mendel rats, and Alderly Park rats, but did not
provide a quantitative analysis or specific information regarding the variability of response
between animals within groups. There appeared to be 20 animals examined in the 1,000 mg/kg
TCE exposed group of B6C3F1 mice but no explanation as to why there were only 10 animals
examined in analyses for liver weight changes, DNA concentration, and tritiated thymidine
incorporation. There was no indication by the authors regarding how many rats were examined
by light microscopy.
Apart from a few inflammatory foci in occasional animals, hematoxylin and eoxin (H&E)
section from B6C3F1 control mice were reported to show no abnormalities. The authors suggest
that this is a normal finding in the livers of mice kept under "non-SPF conditions." A stain for
neutral lipid was reported to not be included routinely in these studies, but subsequent electron
microscopic examination of lipid was reported to show increases in the livers of corn-oil treated
control animals. The individual fat droplets were described as "generally extremely fine and are
not therefore detectable in conventionally process H&E stained sections, since both glycogen
and lipid are removed during this procedure." Thus, this study documents effects of using corn
oil gavage in background levels of lipid accumulation in the liver.
The finding of little evidence of gross hepatotoxicity in TCE-treated mice was reported,
even at a dose of 1,500 mg/kg. Specifically,
Of 19 animals examined receiving 1500 mg/kg body weight TCE, only 6 showed
any evidence of hepatocyte necrosis, and this pathology was restricted to single
small foci or isolated single cells, frequently occurring in a subcapsular location.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-26 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Examination of 20 animals receiving 1000 mg/kg body wt TCE demonstrated no
hepatocyte necrosis. Of 20 animals examined receiving 500 mg/kg body wt TCE,
1 showed necrosis of single isolated hepatocytes; however, this change was not a
treatment-related finding.
TCE-treated mice were reported to show
a change in staining characteristic of the hepatocytes immediately adjacent to the
central vein of the hepatocyte lobules, giving rise to a marked 'patchiness' of the
liver sections. Often this change consisted of increased eosinophilia of the central
cells. There was some evidence of cell hypertrophy in the centrilobular regions.
These changes were evident in most of the TCE treated animals, but there was a
dose-related trend, relatively few of the 500 mg/kg animals being affected, while
the majority of the 1,500 mg/kg animals showed central change. No other
significant abnormalities were seen in the liver of TCE treated mice compared to
controls apart from occasional mitotic figures and the appearance of isolated
nuclei with an unusual chromatin pattern. This pattern generally consisted of a
course granular appearance with a prominent rim of chromatin around the
periphery of the nucleus. These nuclei may have been in the very early stages of
mitosis. Similar changes were not seen in control mice.
The authors briefly commented on the findings in the Alderly Park mice stating that
H& E sections from Alderly Park mice gave similar results as for B6C3F1 mice.
No evidence of hepatotoxicity was seen at a dose of 500 mg/kg body wt TCE.
However, a few animals at the higher doses showed some necrosis and other
degenerative changes. This change was very mild in nature, being restricted to
isolated necrotic cells or small foci, frequently in subcapsular position.
Hypertrophy and increased eosinophilia were also noticed in the centrilobular
regions at higher doses.
Thus, from the brief description given by the authors, the centrilobular region is identified as the
location of hepatocellular hypertrophy due to TCE exposure in mice, and for it to be dose-related
with little evidence of accompanying hepatotoxicity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-27 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The description of histopathology for rats was even more abbreviated than for the mouse.
H& E sections from Osborne-Mendel rats showed that
livers from control rats contained large quantities of glycogen and isolated
inflammatory foci, but were otherwise normal. The majority of rats receiving
1,500 mg/kg body weight TCE showed slight changes in centrilobular
hepatocytes. The hepatocytes were more eosinophilic and contained little
glycogen. At lower doses, these effects were less marked and were restricted to
fewer animals. No evidence of treatment-related hepatotoxicity (as exemplified
by single cell or focal necrosis) was seen in any rat receiving TCE. H& E
sections from Alderly Park Rats showed no signs of treatment-related
hepatotoxicity after administration of TCE. However, some signs of dose-related
increase in centrilobular eosinophilia were noted.
Thus, both mice and rats exhibited pericentral hypertrophy and eosinophilia as noted from the
histopathological examination.
The study did report a quantitative analysis of the effects of TCE on the number of
mitotic figures in livers of mice. Few if any control mice exhibited mitotic figures. But, the
authors report
a considerable increase in both the numbers of figures per section was noted after
administration of TCE." The numbers of animals examined for mitotic figures
ranged from 75 (all control groups were pooled for mice) to 9 in mice, and ranged
from 15 animals in control rat groups to as low as 5 animals in the TCE treatment
groups. The range of mitotic figures found in 25 high-power fields was reported
and is equivalent to the number of mitotic figures per 5,000 hepatocytes examined
in random fields.
Thus, the predominance of mitotic figures in any zone of the liver cannot be ascertained.
For B6C3F1 mice the number of animals with mitotic figures was reported to be 0/75,
3/20, 7/20, and 5/20 for control, 500, 1,000, and 1,500 mg/kg TCE exposed mice, respectively.
The range of the number of mitotic figures seen in 5,000 hepatocytes was reported to be 0, 0-1,
0-5, 0-5 for those same groups with group means of 0, 0.15 ± 0.36, 0.6 ± 1.1, and 0.5 ± 1.2.
These results demonstrate a very small and highly variable response due to TCE treatment in
B6C3F1 mice in regard to mitosis. Thus, the highest percentage of cells undergoing mitosis
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-28 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
within the window of observation would be on average 0.012% with a standard deviation twice
that value. The data presented for mitotic figures also indicated no differences in results between
1,000 and 1,500 mg/kg treated B6C3F1 mice in regard to mitotic figure detection. However, the
tritiated thymidine incorporation data indicated that thymidine incorporation was ~2-fold greater
at 1,500 than 1,000 mg/kg TCE in B6C3F1 mice. For Alderly Park mice, the number of animals
with mitotic figures was reported to be 1/15, 0/9, 4/9, and 2/9 for control, 500, 1,000, and
1,500 mg/kg TCE exposed mice. The range of the number of mitotic figures seen in 5,000
hepatocytes was 0-1, 0, 0-2, 0-1 for those same groups with group means of 0.06 ± 0.25,
0.7 ± 0.9, and 0.2 ± 0.4. These results reveal the detection of at the most 2 mitotic figure in
5,000 hepatocytes for any mouse an any treatment group and no dose-related increased after
TCE treatment in Alderly Park mice. Thus, the highest percentage of cells with a mitotic figure
would be on average 0.014% with a standard deviation twice that value. The small number of
animals examined reduces the power of the experiment to draw any conclusions as to a dose-
response.
Similar to the B6C3F1 mice, there did not appear to be concordance between mitotic
figure detection and thymidine incorporation for Alderly park mice. Thymidine incorporation
showed a 2-fold increase over control for 500 and 1,000 mg/kg TCE and a 5.7-fold increase for
1,500 mg/kg TCE treated animals. However, in regard to mitotic figure detection, there were
fewer mitotic figures in 500 mg/kg TCE treated mice than controls, and fewer animals with
mitotic figures and fewer numbers of figures in the 1,500 mg/kg dose than the 1,000 mg/kg
exposed group. The inconsistencies between mitotic index data and thymidine incorporation
data in both strains of mice suggests that either thymidine incorporation is representative of only
DNA synthesis and not mitosis, an indication of changes in ploidy rather than proliferation, or
that this experimental design is incapable of discerning the magnitude of these changes
accurately. Data from both mouse strains show very little if any hepatocyte proliferation due to
TCE exposure with the mitotic figure index data having that advantage of being specific for
hepatocytes and to not to also include nonparenchymal cells or inflammatory cells in the liver.
The results for rats were similar to those for mice and even more limited by the varying
and low number of animals examined. For Osborne-Mendal rats the number of animals with
mitotic figures were reported to be 8/15, 2/9, 0/7, and 0/6 for control, 500, 1,000, and 1,500
mg/kg TCE exposed rats groups, respectively, with the range of the number of mitotic figures
seen in 5,000 hepatocytes to be 0-8, 0-3, 0, and 0. The group mean was 1.5 ± 2.0, 0.4 ± 1.0, 0,
and 0 for these groups. It would appear from these results that there are fewer mitotic figures
after TCE treatment with the highest percentage of cells undergoing mitosis to be on average
0.03%) in control rats. However, thymidine incorporation studies show a modest increase at all
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-29 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
treatment levels over controls in Osborne-Mendel rats rather than a decrease from controls. For
Alderly Park rats the number of animals with mitotic figures was reported to be 13/15, 5/9, 9/9,
and 4/9 for control, 500, 1,000, and 1,500 mg/kg TCE exposed rat groups with the range of the
number of mitotic figures seen in 5,000 hepatocytes to be 0-26, 0-5, 1-7, and 0-9. The group
mean was 7.2 ± 4.7, 1.6 ± 4.3, 3.8 ± 3.4, and 1.8 ± 2.9 for these groups.
It would appear that there are fewer mitotic figures after TCE treatment with the highest
percentage of cells to an average of 0.14% in control rats. However, thymidine incorporation
studies show 2-fold greater level at 500 mg/kg TCE than for control animals and a 40 and 5%
increase at 1,000 mg/kg and 1,500 mg/kg TCE exposure groups, respectively. Similar to the
results reported in mice, results in both rat strains show an inconsistency in mitotic index and
thymidine incorporation. The control rats appear to have a much greater mitotic index than any
of the mouse groups (treated or untreated) or the TCE-treatment groups. However, it is the mice
that were exhibiting the largest increased in liver weight after TCE exposure. By either
thymidine incorporation or mitosis, these data do provide a consistent result that at 10 days of
exposure very little sustained hepatocellular proliferation is occurring in either mouse or rat and
neither is correlated well with the concurrent changes in liver weight observed from TCE
exposure.
This study provided a qualitative discussion and quantitative analysis of structural
changes using electron microscopy. The qualitative discussion was limited and included
statements about increased observances without quantitative data shown other than the
morphometric analysis. The authors reported that
the ultrastructure of control mouse liver was essentially normal, although mild
dilatation of RER and SER was a frequent finding. Lipid droplets were also
usually present in the cell cytoplasm. The ultrastructural changes seen in mouse
liver following administration of up to 1,500 mg/kg body wt TCE for 10 days
were essentially similar in the B6C3F1 mouse and the Alderly Park mouse. The
most notable change in both strains of mouse was a dramatic increase in the
number of peroxisomes. This change was only apparent in the cells immediately
surrounding the central veins. Peroxisome proliferation was not noticeable in
periportal cells. The induced peroxisomes were generally small and very electron
dense and frequently lacked the characteristic nucleoid core found in peroxisomes
of control livers.
The authors conclude that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-30 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
morphometric analysis showed evidence of a dose-related response, peroxisomal
induction appearing to reach a maximum at 1,000 mg/kg in B6C3F1 mice.. .Lipid
was increased in the livers of treated mice at all doses and was present both as
free droplets in the cytoplasm and as liposomes (small lipid droplets in ER
cisternae). The centrilobular cell, which showed the greatest increase in numbers
of peroxisomes, showed no evidence of this lipid accumulation: fatty change was
more prominent in those cells away from the central vein (i.e., zone 2 of the liver
acinus). Accumulation of lipid, particularly in liposomes, was less marked in
Alderly Park mouse than in B6C3F1 mouse. Mild proliferation of smooth
endoplasmic reticulum was seen in both strains and both rough and smooth
endoplasmic reticulum was generally more dilated than in control mice.
Electron microscopic results for rat liver were reported
to show similar changes in Osborne-Mendel and Alderly Park rat treated with
TCE.. .Rats receiving either 1,000 or 1,500 mg/kg TCE for 10 days generally
showed mild proliferation of SER in centrilobular hepatocytes. The cisternae of
RER were frequently dilated, giving rise to a rather disorganized appearance in
contrast to the parallel stacks seen in control livers, although no detachment of
ribosomes was evident. The SER was also dilated. In contrast to mice,
peroxisomes were only very slightly and not significantly, increased in the liver of
TCE -treated rats. Morphometric analysis confirmed this observation, with the
volume density of peroxisomes in the cytoplasm of centrilobular hepatocytes
being only slightly increased in rats of both strains receiving 1,000 or 1,500
mg/kg body wt TCE.. .Lipid droplets were occasionally increased in some livers
obtained from rats receiving TCE, but the degree of fatty change generally
appeared similar to that found in control rats receiving corn oil. There were no
changes in membrane -bound liposomes, other organelles, or Golgi condensing
vesicles. Centrilobular glycogen was somewhat depleted in male rats receiving
1,500 mg/kg TCE. Periportal cells were ultrastructually normal in all rats.
For the morphometric analysis, the number of mice examined ranged from 7 in the
control group to 8 in the 1,500 mg/kg TCE exposed group. The authors did not indicate which
control animals were used for the morphometric analysis from the 75 animals examined for
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-31 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
mitotic index, the 20 examined by light microscopy, or the 30 mice used as concurrent controls
in the liver weight, DNA concentration, and tritiated thymidine incorporation studies. The
authors stated that morphometry was performed on three randomly selected photomicrographs
from each of three randomly selected pericentral hepatocytes for each animal (i.e., nine
photomicrographs per animal). A mean value representing the exposure group was reported with
the variability between photomitographs per animal or the variation between animals unclear.
The morphometric analysis did not examine all treatment groups (e.g., only the control and
500 mg/kg TCE group were examined in Alderly Park mice).
The percent cytoplasmic volume of the peroxisomal compartment (mean ± standard
deviation [SD]) was reported to be 0.6% ± 0.6% for controls, 4.8% ± 3.3% for 500 mg/kg TCE,
6.7% ± 1.9% for 1,000 mg/kg TCE, and 6.4% ± 2.5% for 1,500 mg/kg TCE in B6C3F1 mice. In
Alderly Park mice, only 12 control and 12 500 mg/kg TCE exposed mice were examined and,
similarly, their selection criteria was not given. The percent cytoplasmic volume of the
peroxisomal compartment was 1.2% ± 0.4% for control and 4.7 ± 2.8% for 500 mg/kg TCE
exposed mice.
For Osborne-Mendel rats control rats were reported to have a percent cytoplasmic
volume of the peroxisomal compartment for control rats (n = 9) of 1.8% ± 0.4%, 1,000 mg/kg
TCE (n = 5) 2.3%) ± 1.6%, and for 1,500 mg/kg exposed rats (n = 7) 2.3% ± 2.0%. For Alderly
Park rats only two groups were examined (control and 1,000 mg/kg TCE exposure). The percent
cytoplasmic volume of the peroxisomal compartment for control rats (n = 15) was reported to be
1.8% ± 0.8% and for 1,000 mg/kg TCE (n = 16) to be 2.4% ± 1.2%. The varying numbers of
animals examined, the varying and inconsistent number of treatment groups examined, the
limited number of photomitographs per animal, and the potential selection bias for animals
examined make quantitative conclusions regarding this analysis difficult. Although control
levels differed by a factor of 2 between the two strains of mice examined, as well as the number
of control animals examined (7 vs. 12), it appears that the 500-mg/kg TCE-exposed B6C3F1 and
Alderly Park mice had similar percentages of peroxisomal compartment in the pericentral cells
examined (-4.8%). There also appeared to be little difference between 1,000 mg/kg TCE treated
Osborne-Mendel and Alderly Park rats for this parameter (-2.4%). Although few animals were
examined, there was little difference reported between 500, 1,000, and 1,500 mg/kg TCE
exposure groups in regard to percentages of peroxisomal compartment in B6C3F1 mice
(4.8-6.7%)). For the few rats of the Osborne-Mendel strain examined, there also did not appear
to be a difference between 1,000 and 1,500 mg/kg TCE exposure for this parameter (2.3%).
Based on peroxisome compartment volume data, one would expect there to be little
difference between TCE exposure groups in mice or rats in regard to enzyme activity or other
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-32 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
"associated events." However, such comparisons are difficult due to limited power to detect
differences and the possibility of bias in selection of animals in differing assays. For the
B6C3F1 mice, only 5 animals per group were examined for enzyme analysis, 7 to 8 for
morphometric analysis, 75 animals in control, and 20 animals in 1,000 mg/kg TCE-exposed
groups for mitotic figure identification, and 10 animals per group for thymidine incorporation.
Since only a few animals were tested for enzyme activity the comparison between peroxisomal
compartment volume and that parameter is very limited. There was a reported 47% increase in
catalase activity between control (n = 5) and 1,000 mg/kg TCE exposed B6C3F1 mice (n = 5)
and 7.8-fold increase in PCO activity. The percent peroxisome compartment was reported to be
10.6-fold greater (0.6 vs. 6.4%). However, the B6C3F1 control percent volume of peroxisomal
compartment was reported to be half that of the AP mouse control. An accurate determination of
the quantitative differences in peroxisomal proliferation would be dependent on an accurate and
stable control value. For Alderly Park rats there was an 8% decrease in catalase activity between
control (n = 5) and 1,000 mg/kg TCE exposed rats (n = 5), and a 13% increase in PCO activity.
The percent peroxisome compartment was reported to be 33% greater in the TCE-exposed than
control group. Thus, for the very limited data that was available to compare peroxisomal
compartment volume with enzyme activity, there was consistency in result.
However, were such increases in peroxisomes associated with other events reported in
this study? Mouse peroxisome proliferation associated enzyme activities in B6C3F1 mice at
1,000 mg/kg TCE were reported to be 8-fold over control values in mice after 10 days of
treatment. However, this increase in activity was not accompanied by a similar increase in
thymidine incorporation (2.8-fold of control) or concordant with increases in mitotic figures
(7/20 mice having any mitotic figures at all with a range of 0-5 and a mean of 0.014% of cells
undergoing mitosis for 1,000 mg/kg TCE vs. 0 for control).
Although results reported in the rat showed discordance between thymidine incorporation
and detection of mitotic figures, there was also discordance with these indices and those for
peroxisomal proliferation. In comparison to controls, there was a reported 13% increase in PCO
activity in Alderly park rats exposed to 1,000 mg/kg TCE, a group mean of mitotic figures half
that in the TCE treated animals versus controls, and increase in thymidine incorporation of 40%.
Thus, these results are not consistent with TCE induction of peroxisome enzyme activity to be
correlated with hepatocellular proliferation by either mitotic index or thymidine incorporation.
Thymidine incorporation in liver DNA seen with TCE exposure also did not correlate with
mitotic index activity in hepatocytes and suggests that this parameter may be a reflection of
polyploidization rather than hepatocyte proliferation. More importantly, these data show that
hepatocyte proliferation, indicated by either measure, is confined to a very small population of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-33 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
cells in the liver after 10 days of TCE exposure. Hepatocellular hypertrophy in the centrilobular
region appears to be responsible for the liver weight gains seen in both rats and mice rather than
increases in cell number. These results at 10 days do not preclude the possibility that a greater
level of hepatocyte proliferation did not occur earlier and then had subsided by 10 days, as is
characteristic of many mitogens. Thymidine incorporation represents the status of the liver at
one time point rather than over a period of whole week and thus, would not capture the earlier
bouts of proliferation. However, there is no evidence of a sustained proliferative response, as
measured at the 10-day time period, in hepatocytes in response to TCE indicated from these data.
In regards to weight gain, although the volume of the peroxisomal compartment was
reported to be similar at 500 mg/kg TCE in B6C3F1 and Alderly Park mice (4.3%), the liver
weight./body weight gain in comparison to control was 20% higher in B6C3F1 mice versus 43%
higher in Alderly Park mice after 10 days of exposure. The liver/body weight ratio was 5.53% in
the B6C3F1 mice and 7.31% in the Alderly Park mice at 500 mg/kg TCE for 10 days.
Similarly, although the peroxisomal compartment was similar at 1,000 mg/kg TCE in
Osborne-Mendel (2.3%) and Alderly Park rats (2.4%), the liver weight/body weight gain was
26%) in Osborne-Mendel rats but 17% in Alderly Park rats at this level of TCE exposure. The
liver/body weight ratio was 5.35% in the Osborne-Mendel rats and 5.83% in the Alderly Park
mice at 1,000 mg/kg TCE for 10 days. Although there are several limitations regarding the
quantitative interpretation of the data, as discussed above, the data suggest that liver weight and
weight gain after TCE treatment was not just a function of peroxisome proliferation. This study
does clearly demonstrate TCE-induced changes at the lowest level tested in several parameters
without toxicity and without evidence of regenerative hyperplasia or sustained hepatocellular
proliferation. In regards to susceptibility to liver cancer induction in more susceptible (B6C3F1)
versus less susceptible (Alderly Park/Swiss) strains of mice (Maltoni et al., 1988), there was a
greater baseline level of liver weight/body weight ratio change, a greater baseline level of
thymidine incorporation as well as greater responses for those endpoints due to TCE exposure in
the "less susceptible" strain. However, both strains showed a hepatocarcinogenic response to
TCE induction and the limitations of being able to make quantitative conclusions regarding
species and strain susceptibility TCE toxicity from this study have been described in detail
above.
E.2.2.8. Dees and Travis (1993)
The focus of this study was to evaluate the nature of DNA synthesis induced by TCE
exposure in mice. The mitotic rate of liver cells was extrapolated using tritiated thymidine
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-34 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
uptake into DNA of male and female mice treated with HPLC grade (99 + pure) TCE. Male and
female hybrid B6C3F1 mice 8 weeks of age [male mice weighed 24-27 g (-12% difference) and
females weighing 18-21 g (-4% difference)] were dosed orally by gavage for 10 days with 100,
250, 500, and 1,000 mg/kg body weight TCE in corn oil (n = 4 per treatment group). 16 hours
after the last daily dose of TCE, mice received tritiated thymidine and were sacrificed 6 hours
later. Hepatic DNA was extracted form whole liver and standard histopathology was also
performed. Hepatic DNA content and cellular distributions were also determined for thymidine
uptake using autoradiography of tissue sections. Tritiated thymidine incorporation into DNA
was determined by microscopic observations of autoradiography slides and reported as positive
cells per 100 (200x power) fields.
Changes in the treatment groups were reported to
include an increase in eosinophilic cytoplasmic staining of hepatocytes located
near central veins, accompanied by loss of cytoplasmic vacuolization.
Intermediate zones appeared normal and no changes were noted in portal triad
areas. Male and female mice given 1,000 mg/kg body weight TCE exhibited
apoptosis located near central veins. No evidence of cellular proliferation was
seen in the portal areas. No evidence of increased lipofuscin was seen in liver
sections from male and female mice treated with TCE. Evaluation of cell death in
male and female mice receiving TCE was performed by enumerating apoptoses.
The apoptoses "did not appear to be in proportion to the applied TCE dose given to male or
female mice." The mean number of apopotosis per 100 (400x) fields in each group of 4 animals
(male mice) was 0, 0, 0, 1, and 8 for control, 100, 250, 500, and 1,000 mg/kg TCE treated
groups, respectively. Variations in number of apoptoses between mice were not given by the
authors. Feulgen stain was <1 for all doses except for 9 at 1,000 mg/kg.
Mitotic figure were reported to be
frequently seen in liver sections from both male and female mice treated with
TCE. Dividing cells were most often found in the intermediate zone and
resembled mature hepatocytes. Incorporation of the radiolabel into cells located
near the portal triad areas was rare. In general, mitotic figures were very rare, but
when found they were usually located in the intermediate zone. Little or no
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
incorporation of label was seen in areas near the bile duct epithelia or in areas
close to the portal triad.
No quantitative description of mitotic index was reported by the authors but this description is
consistent with there being replication of mature hepatocytes induced by TCE.
The distribution of tritiated thymidine was given for specific cell types in the livers of
5 animals per treatment group and radiolabel was reported to be predominantly associated with
peri sinusoidal cell in control mice. The authors state that the label was more often found in cells
resembling mature hepatocytes. The mean number of labeled cells in autoradiographs per 100
(200x power) fields was reported to be -125 and -150 labeled perisinusoidal cells in controls
male and female mice, respectively. The authors do not give any standard deviations for the
female perisinusoidal data except for the 1,000-mg/kg exposure group. For mature hepatocytes,
the mean baseline level of cell labeling for control male and female mice were reported to be -65
and -90 labeled cells, respectively. Although the baseline levels of hepatocyte labeling were
reported to differ between male and female mice, the mean peak level of labeling was similar at
-250 labeled cells for male and female mice treated with TCE. In male mouse liver, the number
of labeled cells increased -2-fold of control levels after 500 and 1,000 mg/kg TCE and in female
mouse liver increased -4-fold of control levels after 250, 500, and 1,000 mg/kg TCE over their
respective control levels.
Incorporation of tritiated thymidine into DNA extracted from whole liver in male and
female mice was reported to be significantly elevated after TCE treatment but, unlike the
autoradiographic data, there was no difference between genders and the mean peak level of
tritiated thymidine incorporation occurred at 250 mg/kg TCE treatment and remained constant
for the 500 and 1,000 mg/kg treated groups. Increased thymidine incorporation into DNA
extracted from liver of male and female mice were reported to show a very large standard
deviation with TCE treatment (e.g., at 100 mg/kg TCE exposure, male mice had a mean of
-130 dpm tritiated thymidine/microgram DNA with the upper bound of the standard deviation to
be 225 dpm). The increased thymidine incorporation peaked at a level that was a little less than
2-fold of control level. Thus, for both male and female mice both autoradiographs and total
hepatic DNA were reported to show that male and female mice had similar peaks of increased
thymidine incorporation after TCE exposure that reached a plateau at the 250 mg/kg TCE
exposure level and did not increase with increasing exposure concentration. These data also
indicate a very small population undergoing mitosis due to TCE exposure after 10 days of
exposure. If higher levels of hepatocyte replication had occurred earlier, such levels were not
sustained by 10 days of TCE exposure. More importantly, these data suggest that tritiated
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-36 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
thymidine levels were targeted to mature hepatocytes and in areas of the liver where greater
levels of polyploidization occur. The ages and weights of the mice were described by these
authors, unlike Elcombe et al, and a different strain was used. However, these results are
consistent with those of Elcombe in regard to the magnitude of thymidine incorporation induced
by TCE treatment and the lack of a dose response once a relative low level of exposure has been
exceeded.
The total liver DNA content of male and female mice treated with TCE were also
determined with the total micrograms DNA/g liver reported to be ~4 microgram/g for female
control mice and ~2 micrograms/g for male control mice. Although not statistically significant,
the total DNA concentration dropped from ~4 to ~3 at 100 mg/kg through 1,000 mg/kg exposure
to TCE in female mice. For male mice the total DNA rose slightly in the 250- and 500-mg/kg
groups to ~3 micrograms/gram and was similar to control levels at the 100 and 1,000 mg/kg TCE
treatment groups. The standard deviation in male mice was very large and the number of
animals small making quantitative judgments regarding this parameter difficult. The slight
decrease reported for female mice would be consistent with the results of Elcombe et al. (1985)
who describe a slight decrease in hepatic DNA in male mice. However, the reported slight
increase in hepatic DNA in male mice in this study is not consistent. Given the small number of
animals and the large deviations for female and male mice in the TCE treated groups, this study
may not have had the sensitivity to detect slight decreases reported by Elcombe et al. (1985).
In regard to clinical evaluation and weight analyses, both male and female mice given
TCE were reported "to appear clinically ill. These mice showed reduced activity and failed to
groom. Control mice showed no adverse effects. Female mice were markedly more affected by
TCE than their male counterparts. Several deaths of female mice occurred during the course of
the TCE treatment regimen." The authors do not give cause of deaths but state that two female
mice died in the group receiving 250 mg/kg TCE and one in the group receiving 1,000 mg/kg
during the gavage regimen of the female mice. This appears to be similar gavage error or
"accidental death" reported in National Toxicology Program (NTP) studies chronic studies of
TCE (see below).
The authors report
no significant difference in the absolute body weight of male and female mice
were noted in control groups. Body weight gain in female and males mice treated
with TCE was not significantly different from that of control mice. Liver weights
in male mice given 500 or 1,000 mg/kg and corrected for total body weight were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-37 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
significantly elevated. The corrected liver weights of female mice increase
proportionally with the applied dose of TCE.
For male mice, liver weights were reported to be 1.40 ±0.16, 1.38 ± 1.23, 1.48 ±0.09,
1.61 ± 0.07, and 1.63 ± 0.11 g for control, 100, 250, 500, and 1,000 mg/kg TCE in male mice
(n = 5), respectively. Body weights were smaller for the 100 mg/kg TCE treatment group
although not statistically significant. The liver weights after treatment had a much larger
reported standard deviation (1.23 g for 100 mg/kg group vs. <0.16 for all other groups). The
percent liver/body weight ratios were reported to be 5.40, 5.41, 5.42, 5.71, and 6.34% for the
same groups in male mice. This represents 1.06- and 1.17-fold of control at the 500 and
1,000 mg/kg dose. The authors report a statistically significant increase in percent liver/body
weight ratio only for the 500 mg/kg (i.e., 1.06-fold of control) and 1,000 mg/kg (i.e., 1.17-fold of
control) TCE exposure groups.
The results for female mice liver weights were reported in Table III of the paper, which
was mistakenly labeled as for male mice. The reported values for liver weight were 1.03 ± 0.07,
1.05 ± 0.10, 1.15 ± 0.98, 1.21 ±0.18, and 1.34 ± 0.08 g for control, 100, 250, 500, and 1,000
mg/kg TCE in female mice (n = 5, except for 250 mg/kg and 1,000 mg/kg groups), respectively.
The percent liver/body weight ratios were 5.26, 5.44, 5.68, 6.24, and 6.57% for the same groups.
These values represent 1.03-, 1.08-, 1.19-, and 1.25-fold of controls in percent liver/body weight.
The magnitude of increase in TCE-induced percent liver/body weight ratio in female mice is
reflective of the magnitude of the difference in dose up to 1,000 mg/kg where it is slightly lower.
The female mice were reported to have statistically significant increases in percent liver/body
ratios at the lowest dose tested (100 mg/kg TCE) after 10 days of TCE exposure that also
increased proportionately with dose. Male mice were not reported to have a significant increase
in percent liver/body weight until 500 mg/kg TCE but a statistically significant increase in liver
weight at 250 mg/kg TCE. Male mice had a much larger variation in initial body weight than did
female mice (range of means of 24.86 to 27.84 g between groups for males or ~11% difference
and range of means of 19.48 to 20.27 g for females or -4%) which may contribute to an apparent
lack of effect for a parameter that is dependent on body weight. Only 5 mice were used in each
group so the power to detect a change was relatively small.
The results from this experiment are consistent with those of Elcombe et al. (1985) in
showing a slight increase in thymidine incorporation (~2-fold of control) and mitotic figures that
are rare after TCE exposure. This study also records a lack of apoptosis with TCE treatment
except at the highest exposure level (i.e., 1,000 mg/kg). The increases in liver weight induced by
TCE were reported to be dose-related, especially in female mice where baseline body weights
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
were more consistent. However, the incorporation of tritiated thymidine reached a plateau at
250 mg/kg TCE in the DNA of both genders of mice. This study specifically identified where
thymidine incorporation and mitotic figures were occurring in TCE-treated livers and noted that
the mature hepatocyte that appeared to be primarily affected, as well as in the portion of the liver
where mature hepatocytes with higher ploidy are found. The authors note that the "lack of
thymidine incorporation in the periportal area, where the liver stem cells are reside," suggesting
that the mature hepatocyte is the target of TCE effects on DNA synthesis. This finding is
consistent with a change in ploidy accompanying hepatocellular hypertrophy and not just cell
proliferation after 10 days of TCE exposure. Like Elcombe et al. (1985), these data represent "a
snapshot in time" which does not show whether increased cell proliferation may have happened
at an earlier time point and then subsided by 10 days. However, like Elcombe et al. (1985) it
suggests that sustained proliferation is not a feature of TCE exposure and that the level of DNA
synthesis (which is very low in quiescent control liver) is increased in a small population of
hepatocytes due to TCE exposure that is not dose-dependent (only 2-fold increase over control in
animals exposed from 250 to 1,000 mg/kg TCE). In regards to toxicity, no evidence of increased
lipid peroxidation in TCE-treated animals was reported using histopathologic sections stained to
enhance observation of lipofuscin. No necrosis is noted by these authors and the deaths in
female mice are likely due to gavage error.
E.2.2.9. Nakajima et al. (2000)
This study focused on the effect of TCE treatment on PPARa-null mice in terms of
peroxisome proliferation but also included information on differences in liver weight between
null and wild-type mice, as well as gender-related effects. SV129 wild-type and PPARa-null
mice (10 weeks of age) were treated with corn oil or 750 mg/kg TCE in corn oil daily for
2 weeks via gavage (n = 6 per group). A small portion of the liver was removed for
histopathological examination but the lobe used was not specified by the authors. Liver
peroxisome proliferation was reported to be evaluated morphologically using
3,3'-diaminobenzidine (DAB) staining of sections and electron photomicroscopy to detect the
volume density of peroxisomes (percent of cytoplasm) in 15 micrographs of the pericentral area
per liver. A number of P-oxidation enzymes and P450s were analyzed by immunoblot of liver
homogenates.
The final body weights, liver weights, and percent liver/body weight ratios were reported
for all treatment groups. For male mice, vehicle treated PPARa-null mice had slightly lower
mean body weights (24.5 ± 1.8 g vs. 25.4 ± 1.9 g [SD]), slightly larger liver weights
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-39 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
(1.14 ± 0.13 g vs. 1.05 ± 0.15 g or -9%), and slightly higher percent liver/body weight ratios
(4.12% ± 0.32% vs. 4.10%) ± 031%) than wild-type mice. The mean values for final body
weights of the groups of mice in this study were reported and were similar which, as
demonstrated by the inhalation studies by Kjellstrand et al. (1983b) (see Section E.2.2.4), is
particularly important for determining the effects of TCE treatment on percent liver/body weight
ratios. For both groups of male mice, 2 weeks of TCE treatment significantly increased both
liver weight and percent liver/body weight ratios. For male wild-type mice the increase in
percent liver/body weight was 1.50-fold of vehicle control and for male PPARa-null mice the
increase was 1.26-fold of control after 2 weeks of TCE treatment.
For female mice, vehicle treated PPARa-null mice had slightly higher mean body
weights (22.7 ± 2.1 g vs. 22.4 ± 2.0 g), slightly larger liver weights (0.98 ± 0.15 g vs. 0.95 ±
0.14 g or ~3%>), and slightly higher percent liver/body weight ratios (4.32%> ± 0.35%> vs. 4.24%> ±
0.41%>) than wild-type mice. For both groups of female mice, 2 weeks of TCE treatment
significantly increased percent liver/body weight ratios. For liver weights there was a reporting
error for PPARa-null female treated with TCE so that liver weight changes due to TCE treatment
cannot be determined for this group. For female wild-type mice the increase in percent
liver/body weight was 1.24-fold of vehicle control and for female PPARa-null mice the increase
was 1.26-fold of control after 2 weeks of TCE treatment.
Thus, for both wild-type and PPARa-null mice, TCE exposure resulted in increased
percent liver/body weight over controls that was statistically significant after 2 weeks of oral
gavage exposure using corn oil as the vehicle. For male mice there was a greater TCE-induced
increase in percent liver/body weight in wild-type than PPARa-null mice (1.50- vs. 1.26-fold of
control) that was statistically significant, but for female mice the induction of increased liver
weight was statistically increased but the same in wild-type and PPARa-null mice (i.e., both
were ~1.25-fold of control). These date indicate that TCE-induced increases in mouse liver
weight were not dependent on a functional PPARa receptor in female mice and suggest that
some portion may be in male mice.
In regard to light and electron microscopic results, the numbers of peroxisomes in
hepatocytes of wild-type mice were reported to be increased, especially in the pericentral area of
the hepatic lobule, to a similar extent in both males and females (15 micrographs, n = 4 mice).
TCE exposure was reported to increase the volume density of peroxisomes 2-fold of control in
the pericentral area with no evident change in peroxisomes in the periportal areas, but data was
not shown for that area of the liver lobule. In contrast, no increase in peroxisomes was reported
to be observed in PPARa-null mice. Therefore, increases in liver weight observed in PPARa-
null mice after TCE treatment did not result from peroxisome proliferation. Similarly, the small
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-40 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
2-fold increase in peroxisome volume from 2 to 4% of cytoplasmic volume in the pericentral
area of the liver lobule in wild-type mice could not have been responsible for the 50% increase
liver weight observed in male wild-type mice.
Although no difference was reported between male and female wild-type mice in regard
to TCE-induced peroxisome proliferation in wild-type mice, the levels of hepatic enzymes
associated with peroxisomes (acyl-CoA [AOX], peroxisomal bifunctional protein [PH],
peroxisomal thiolase [PT], very long chain acyl-CoA synthetase, and D-type peroxisomal
bifunctional protein [DBF], cytosolic enzyme [cytosolic thioesterase II (CTEII)], mitochondrial
enzymes [mitochondrial trifunctional protein a subunits a and P(TPa and TPP)], and microsomal
enzymes [cytochrome P450 4A1 (CYP4A1)]) as measured by immunoblot analysis were
significantly elevated in male wild-type mice (n = 4) by a factor of-2-3, but except for a slight
elevation in PH and PT, were reported to not be elevated in female wild-type mice (n = 4). The
magnitude of increase in peroxisomal enzymes was similar to that of peroxisomal volume in
male mice. No TCE-induced increases in any of these enzymes were reported in male or female
PPARa-null mice by the authors. For CYP4A1, an enzyme reported to be induced by
peroxisomal proliferators, TCE exposure resulted in a much lower amount in female than male
wild-type mice (i.e., 2% of the level induced by TCE in males). However, the expression of
catalase was reported to be "nearly constant in all samples" (at most -30% change) which the
authors suggested resulted from induction by TCE that was independent of PPARa. The basis
for selection of 4 mice for this comparison out of the 6 studied per group was not given by the
authors. A comparison of control wild-type and PPARa-null mice showed that in males
background levels of the enzymes examined were generally similar except for DBF in which the
null mice had values -50% of the wild-type controls. A similar decrease was reported for female
PPARa-null mice. With regard to gender differences in wild-type mice, females had similar
values as males with the exceptions of TPa, TPP, and CYP2E1 which were in untreated female
wild-type mice at a 3.06-, 2.38-, and 1.63-fold for 1 TPa, TPP, and CYP2E1 levels over males,
respectively. Female PPARa-null mice had increases of 2.50-, 1.54-, and 2.07-fold over male
wild-type mice.
With regard to the induction of TCE metabolizing enzymes (CYP1A2, CYP2E1, and
ALDH), CYP1A2 was reported to be decreased by TCE treatment of both male and female wild-
type mice but liver CYP2E1 reported to be increased in male mice and constant in female mice
which resulted in similar expression level in both genders after TCE treatment. There was no
gender difference in ALDH activity reported after TCE exposure and activity was reported to be
independent of PPARa. The authors concluded that TCE metabolizing abilities of the liver of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-41 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
male and female mice were similar and therefore, poor induction of peroxisomal related enzymes
was not due to gender-related differences in TCE metabolism.
To investigate whether the a gender-related difference peroxisomal enzymes after TCE
exposure was due to a lower levels of PPARa and RXRa receptors, western blotting was
employed (n = 3). The level of PPARa protein was reported to be increased in both male wild-
type mice with less induction in females (control vs. TCE, 1.00 ± 0.20 vs. 2.17 ± 0.24 in males
and 0.95 ± 0.25 vs. 1.44 ± 0.09 in females) after TCE treatment. The hepatic level of RXRa was
also reported to be increased in the same manner as PPARa (control vs. TCE, 1.00 ± 0.33 vs.
1.92 ± 0.04 in males 0.81 ± 0.16 vs. 1.14 ± 0.10 in females). Northern blot analysis of hepatic
PPARa mRNA was reported to show greater TCE induction in male (2.6-fold of control) than in
female (1.5-fold of control) wild-type mice. Thus, males appeared to have higher induction of
the two receptor proteins as well as a greater response in peroxisomal enzymes and CYP4A1,
even though TCE-induced increases in peroxisomal volume was similar between male and
female mice. The increased response in males for induction of the two receptor proteins is
consistent with liver weight data that shows some portion of the induction of increased liver
weight response in male mice using this paradigm may be due to gender-specific differences in
PPARa response. However, as noted below (see Section E.2.2), corn oil vehicle has liver effects
alone, especially in the male liver, that have also been associated with PPARa responses.
E.2.2.10. Berman et al. (1995)
This study included TCE in a suite of compounds used to compare endpoints for
toxicological screening methods. Female Fischer 344 rats of 77 days of age (n = 8 per group)
were administered TCE in corn oil for 1 day (0, 150, 500, 1,500, or 5,000 mg/kg/d) or for
14 days (0, 50, 150, 500, or 1,500 mg/kg/d). Blood samples were taken 24 hours after the last
dose and livers were weighed and H&E sections were examined for evidence of parenchymal
cell degeneration, necrosis, or hypertrophy. No details were provided by the authors for the
extent or severity of the liver affects by histopathological examination. The serum chemistry
analysis included lactate dehydrogenase (LDH), alkaline phosphatase, ALT, aspartate
aminotrasferase (AST), total bilirubin, creatine, and blood urea nitrogen. The starting and
ending body weights of the animals or the absolute liver weights were not reported by the
authors.
The results of a multivariate analysis were reported to show a lowest effective dose of
1,500 mg/kg after 1 day of TCE exposure and 150 mg/kg after 14 days of TCE exposure that was
statistically significant. Liver weight and liver weight changes were not reported by the authors
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-42 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
but the percent liver to body weight ratios were. For the two control groups there was a
difference in percent liver/body weight of-8% (3.43% ± 0.74% for the 1-day control group and
3.16% ± 0.41%) for the 14-day control group, mean ± SEM). For the 1-day groups only the
5,000 mg/kg group was reported to show a statistically significant difference in percent
liver/body weight between control and TCE treatment (i.e., ~1.08-fold increase). Hepatocellular
necrosis was noted to occur in the 1,500 and 5,000 mg/kg groups in 6/7 and 6/8 female rats,
respectively but not to occur in lower doses. The extent of necrosis was not noted by the authors
for the two groups exhibiting a response after 1 day of exposure. Serum enzymes indicative of
liver necrosis were not presented and because only positive results were presented in the paper,
presumed to be negative. Therefore, the extent of necrosis was not of a magnitude to affect
serum enzyme markers of cellular leakage.
After 14 days of TCE exposure, there was a dose-related increase reported for percent
liver/body weight ratios that was statistically significant at all TCE dose levels although the
multivariate analysis indicated the lowest effective dose to be 150 mg/kg. The percent
liver/body weight ratio was 3.16% ± 0.41%, 3.38% ± 0.56%, 3.49% ± 0.69%, 3.82% ± 0.76%,
and 4.47%) ± 0.66% for control, 50, 150, 500, and 1,500 mg/kg TCE exposure levels,
respectively after 14 days of exposure. No hepatocellular necrosis was reported at any dose and
hepatocellular hypertrophy was reported only at the 1,500 mg/kg dose and in all rats. These rat
liver weights were 1.07-, 1.10-, 1.21-, and 1.41-fold of controls for the 50, 150, 500, and
1,500 mg/kg TCE dose groups, respectively. The 7% increase in liver weight at the 50 mg/kg
dose was approximately the same difference between the two control groups for Days 1 and
14 treatments. Without the data for starting and final body weights and an examination of
whether the control animals had similar body weight, it is impossible to discern whether the
reported effects at the low dose of TCE was also reflected differences between the control
groups. No serum enzyme levels changes were reported after 14 days of exposure to TCE for
any group.
The authors note that their study provided evidence of liver effects at lower levels than
other studies citing Elcombe et al. (1985) and Goldsworthy and Popp (1987). They suggest that
the differences in sensitivity to TCE between their results and those of these two studies may
reflect differences in strain or gender of the rats examined. However, they did not study male
rats of this strain concurrently so that differences in gender may have reflected differences
between experiments. The increase in liver weight without reporting increases in hepatocellular
hypertrophy as well as the lack of necrosis as low doses is consistent with the results of Melnick
et al. (1987) in male Fischer rats given TCE orally (see Section E.2.1.11, below).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-43 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.2.2. 11 Melnick et al. (1987)
The focus of this study was to assess microencapsulation as a way to expose rodents to
substances such as TCE that have issues related to volatilization in drinking water or apparent
gavage-related deaths. In this study, liver weight changes, extent of focalized necrosis, and
indicators of peroxisome proliferation were reported as metrics of TCE toxicity. TCE (99+ %)
was encapsulated in gelatin-sorbitol microcapsules and was 44.1% TCE w/w. The TCE
microcapsules were administered to male Fischer 344 rats (6-week old and weighing between 89
and 92 g or -3% difference) in the diet (0, 0.55, 1.10, 2.21, and 4.42% TCE in the diet) for
14 days. The number of animals in each group was 10. A parallel group of animals was
administered TCE in corn oil gavage for 14 consecutive days (corn oil control, 0.6, 1.2, and
2.8 g/kg/day TCE). The dosage levels of TCE in the gavage study were reported to be "adjusted
5 times during the 14-day" treatment period to be similar to the dosage levels of TCE in the feed
study. The time-weighted average dosage levels of TCE in the feed study were reported to be
0.6, 1.3, 2.2, and 4.8 g/kg/day.
There was less food consumption reported in the 2.2 and 4.8 g/kg/day dose feed groups,
which the authors attribute to either palatability or toxicity. There were no deaths in any of the
groups treated with microencapsulated TCE while, similar to many other gavage studies of TCE
reported in the literature, there were 4 deaths in the high-dose gavage group. Mean body weight
gains of the two highest dose groups of the feed study and of the highest dose group of the
gavage study were reported to be significantly lower than the mean body weight gains of the
respective control groups (i.e., -22 and -35% reduction at 2.2 and 4.8 g/kg/day in the feed study,
respectively, and -33% reduction at 2.8 g/kg/day TCE in the gavage study).
After 14 days of treatment, liver weights were reported to be 8.1 ± 0.8, 8.4 ± 0.8, 9.5 ±
0.5, 10.1 ± 1.2, 8.9 ± 1.3, and 7.4 ± 0.5 g for untreated control, placebo control, 0.6, 1.3, 2.2, and
4.8 g/kg TCE exposed feed groups, respectively. The corresponding percent liver/body weight
ratios were reported to be 5.2% ± 0.3%, 5.3% ± 0.2%, 6.0% ± 0.3%, 6.5% ± 0.5%, 7.0% ± 0.9%,
and 7.1%) ± 0.5% for untreated control, placebo control, 0.6, 1.3, 2.2, and 4.8 g/kg TCE exposed
groups, respectively. The increased percent liver/body weight ratio represents 1.13-, 1.23-, 1.32-
, and 1.34-fold of placebo controls, respectively.
For the gavage experiment, after 14 days of treatment liver weights were reported to be
7.1 ± 1.3, 9.3 ± 1.2, 9.1 ± 0.9, and 7.7 ± 0.4 g for corn oil control, 0.6, 1.2, and 2.8 g/kg TCE
exposed groups, respectively. The corresponding percent liver/body weight ratios were reported
to be 5.0% ± 0.4%, 6.0% ± 0.4%, 6.1% ± 0.3%, and 7.3% ± 0.5% for corn oil control, 0.6, 1.2,
and 2.8 g/kg TCE exposed groups, respectively. The percent liver/body weight ratios represent
1.20-, 1.22-, and 1.46-fold of corn oil controls, respectively. The 2.8 g/kg TCE gavage results
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-44 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
are reflective of the 6 surviving animals in the group rather than 10 animals in the rest of the
groups. There was no explanation given by the authors for the lower liver weights in the control
gavage group than the placebo control in the feed group (i.e., 20% difference) although the initial
and final body weights appeared to be similar. The decreased body weights in the feed and
gavage study are reflective if TCE systemic toxicity and appeared to affect the TCE-induced
liver weight increases in those groups.
The authors reported that the only treatment-related lesion observed microscopically in
rats from either dosed-feed or gavage groups was individual cell necrosis of the liver with the
frequency and severity of this lesion similar at each dosage levels of TCE administered
microencapsulated in the feed or in corn oil. Using a scale of minimal = 1-3 necrotic
hepatocytes/10 microscopic 200x fields, mild = 4-7 necrotic necrotic hepatocytes/10
microscopic 200x fields, and moderate = 8-12 necrotic hepatocytes/10 microscopic 200 x fields,
the frequency of lesion was 0-1/10 for controls, 2/10 for 0.6 and 1.3 g/kg and 9/10 for 2.2 and
4.8 g/kg feed groups. The mean severity was reported to be 0.0-0.1 for controls, 0.3-0.4 for 0.6
and 1.3 g/kg, and 2.0-2.5 for 2.2 and 4.8 g/kg feed groups. For the corn oil gavage study, the
corn oil control and 0.6 g/kg groups were reported to have a frequency of 0 lesions/10 animals,
the 1.2 g/kg group a frequency of 1/10 animals, while the 2.8 g/kg group to have a frequency of
5/6 animals. The mean severity score was reported to be 0 for the control and 0.6 g/kg groups,
0.1 for the 1.2 g/kg groups, and 1.8 for the remaining 6 animals in the 2.8 g/kg group. The
individual cell necrosis was reported to be randomly distributed throughout the liver lobule with
the change to not be accompanied by an inflammatory response. The authors also report that
there was no histologic evidence of cellular hypertrophy or edema in hepatic parenchymal cells.
Thus, although there appeared to be TCE-treatment related increases in focal necrosis after
14 days of exposure, the extent was even at the highest doses mild and involved few hepatocytes.
Microsomal NADPH cytochrome c-reductase was reported to be elevated in the 2.2 and
4.8 g/kg feed groups and in the 1.2 and 2.8 g/kg gavage groups. Cytochrome P450 levels were
reported to be elevated only in the two highest dose groups of the feed study. The authors
reported a dose-related increase in peroxisome PCO and catalase activities in liver homogenates
from rats treated with TCE microcapsules or by gavage and that treatment with corn oil alone,
but not placebo capsules, caused a slight increase in PCO activity.
After 14 days of treatment, PCO activities were reported to be 270 ± 12, 242 ± 17, 298 ±
64, 424 ± 55, 651 ± 148, and 999 ± 266 nmol H2O2 produced/min/g liver for untreated control,
placebo control, 0.6, 1.3, 2.2, and 4.8 g/kg TCE exposed feed groups, respectively. This
represents 1.23-, 1.75-, 2.69-, and 4.13-fold of placebo controls, respectively. After 14 days of
treatment, catalase activities were reported to be 8.49 ± 0.81, 7.98 ± 1.62, 8.49 ± 1.92, 8.59 ±
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-45 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
1.31, 13.03 ±2.01, and 15.76 ± 1.11 nmol H202 produced/min/g liver for untreated control,
placebo control, 0.6, 1.3, 2.2, and 4.8 g/kg TCE exposed groups, respectively. This represents
1.06-, 1.07-, 1.63-, and 1.97-fold of placebo controls, respectively. Thus, although reported to be
dose related, only the two highest exposure levels of TCE increased catalase activity and to a
smaller extent than PCO activity in microencapsulated TCE fed rats.
For the gavage experiment, after 14 days of treatment PCO activities were reported to be
318 ± 27, 369 ± 26, 413 ± 40, and 1,002 ± 271 nmol hydrogen peroxide (H2O2) produced/min/g
liver for corn oil control, 0.6, 1.2, and 2.8 g/kg TCE exposed groups, respectively. This
represents 1.16-, 1.29-, and 3.15-fold of corn oil controls. After 14 days of treatment, catalase
activities were reported to be 8.59 ± 0.91, 10.10 ± 1.82, 12.83 ± 3.43, and 13.54 ± 2.32 nmol
H2O2 produced/min/g liver for corn oil control, 0.6, 1.2, and 2.8 g/kg TCE exposed groups,
respectively. This represents 1.18-, 1.49-, and 1.58-fold of corn oil controls. As stated by the
authors the corn oil vehicle appeared to elevate catalase activities and PCO activities.
In regard to dose-response, liver and body weight were affected by decreased body
weight gain in the higher dosed animals in this experiment (i.e., 2.2 g/kg/day TCE exposure and
above) and by gavage related deaths in the highest-dosed group. The lower liver weight in the
gavage control group also may have affected the determination of the magnitude of TCE-related
liver weight gain at that dose. At the 2 doses, below which body weight gain was affected, there
appeared to be an approximately 20% increase in percent liver/body weight ratio in the gavage
study and a 13 and 23% weight increase in the feed study.
The extent of PCO activity appeared to increase more steeply with dose in the feed study
than did liver weight gain (i.e., a 1.23-fold of liver/body weight ratio at 1.3 g/kg/day
corresponded with a 1.75-fold PCO activity over control). At the two highest doses in the feed
study, the increase in PCO activity was 2.69- and 4.13-fold of control but the increase in liver
weight was not more than 34%. For the gavage study, there was also a steeper increase in PCO
activity than liver weight gain. For catalase activity, the increase was slightly less than that of
liver/body weight ratio percent for the two doses that did not decrease body weight gain in the
feed study. In the gavage study, they were about the same. In regard to what the cause of liver
weight gain was, the authors report that there was no histologic evidence of cellular hypertrophy
or edema in hepatic parenchymal cells and do not describe indicators of hepatocellular
proliferation or increased polyploidy. Accordingly, the cause of liver weight gain after TCE
exposure in this paradigm is not readily apparent.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-46 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E.2.2.12. Laughter et al. (2004)
Although the focus of the study was an exploration of potential MO As for TCE effects
through macroarray transcript profiling (see Section E.3.1.2 for discussions of limitations of this
approach and especially the need for phenotypic anchoring, Section E.3.4.1.3 for use of PPARa
knockout mice, and Section E.3.4.2.2 for discussion of genetic profiling data for TCE),
information was reported regarding changes in the liver weight of PPARa-null mouse and their
background strains. SV129 wild-type and PPARa-null male mice (9 ± 1.5 weeks of age) were
treated with 3 daily doses of TCE in 0.1% methyl cellulose for either 3 days or 3 weeks
(n = 4-5/group). Thus, this paradigm does not use corn oil, which has been noted to affect
toxicity (see Section E.2.2 below), but is not comparable to other paradigms that administer the
total dose in one daily gavage administration rather than to give the same cumulative dose but in
3 daily doses of lower concentration. The initial or final body weights of the mice were not
reported. Thus, the effects of systemic toxicity from TCE exposure on body weight and the
influence of differences in initial body weight on percent liver/body weight determinations
cannot be made.
For the 3-day study, mice were administered 1,500 mg/kg TCE or vehicle control. For
the 3-week study, mice were administered 0, 10, 50, 125, 500, 1,000, or 1,500 mg/kg TCE 5 days
a week except for 4 day/week on the last week of the experiment. In a separate study, mice were
given TCA or dichloroacetic acid (DCA) at 0.25, 0.5, 1, or 2 g/L (pH ~7) in the drinking water
for 7 days. For each animal a block of the left, anterior right, and median liver lobes was
reported to be fixed in formalin with 5 sections stained for H&E and examined by light
microscopy. The remaining liver samples were combined and used as homogenates for
transcript arrays. In the 3-week study, bromodeoxyuridine (BrdU) was administered via
miniosmotic pump on day one of Week 3 and sections of the liver assessed for BrdU
incorporation in at least 1,000 cells per animal in 10-15 fields.
Although initial body weights, final body weights, and the liver weights were not
reported, the percent liver/body ratios were. In the 3-day study, control wild-type and PPARa-
null mice were reported to have similar percent liver/body weight ratios of-4.5%. These
animals were -10 weeks of age upon sacrifice. However, at the end of the 3-week experiment
the percent liver/body weight ratios were increased in the PPARa-null male mice and were 5.1%.
There was also a slight difference in the percent liver/body weight ratios in the 1-week study
(4.3%) ± 0.4%) vs. 4.6%o ± 0.2%o for wild-type and PPARa-null mice, respectively). These results
are consistent with an increasing baseline of hepatic steatosis with age in the PPARa-null mice
and increase in liver weight.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-47 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In the 3-day study, the mean reported for the percent liver/body ratio was 1.4-fold of the
wild type animals tested with TCE in comparison to the control level. In the PPARa-null mice,
there was a 1.07-fold of control level reported by the authors to not be statistically significant.
However, given the low number of animals tested (the authors give only that 4-5 animals were
tested per group without identification as to which groups has 4 animals and which had 5), the
ability of this study to discern a statistically significant difference is limited.
In the 3-week study, wild-type mice exposed to various concentrations of TCE had
percent liver/body weights that were within -2% of control values except for the 1,000 mg/kg
and 1,500 mg/kg groups that were -1.18- and 1.30-fold of control levels, respectively. For the
PPARa-null mice exposed to TCE for 3 weeks, the variability in percent liver/body weight was
greater than that of the wild-type mice in most of the groups. The baseline level percent
liver/body weight was 1.16-fold in the PPARa-null mice in comparison to wild-type mice. At
the 1,500 mg/kg TCE exposure level percent liver/body weights were not recorded because of
the death of the null mice at this level. The authors reported that at the 1,500 mg/kg level all
PPARa-null mice were moribund and had to be removed from the study. However, at
1,000 mg/kg TCE exposure level there was a 1.10-fold of control percent liver/body weight
value that was reported to not be statistically significant. However, as noted above, the power of
the study was limited due to low numbers of animals and increased variability in the null mice
groups. The percent liver/body weight reported in this study was actually greater in the null
mice than the wild-type male mice at the 1,000 mg/kg TCE exposure level (5.6% ± 0.4% vs.
5.2% ± 0.5%), for null and wild-type mice, respectively).
Thus, at 1-week and at 3-weeks, TCE appeared to induce increases in liver weight in
PPARa-null mice, although not reaching statistical significance in this study, with concurrent
background of increased liver weight reported in the knockout mice. At 1,000 mg/kg TCE
exposure for 3 weeks, percent liver/body weight was reported to be 1.18-fold in wild-type and
1.10-fold in null mice of control values. As discussed above, Nakajima et al. (2000) reported
statistically significant increased liver weight in both wild-type and PPARa-null mice after 2
weeks of exposure with less TCE-induced liver weight increases in the knockout mice (see
Section E.2.1.10). They also used more mice, carefully matched to weights of their mice, and
used a single dose of TCE each day with corn oil gavage.
The authors noted that inspection of the livers and kidneys of the moribund null mice,
who were removed from the 3-week study, "did not reveal any overt signs of toxicity in this dose
group that would lead to morbidity" but did not show the data and did not indicate when the
animals were affected and removed. For the wild-type mice exposed to the same concentration
(1,500 mg/kg) but whose survival was not affected by TCE exposure, the authors reported that at
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-48 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
the 1,500 mg/kg dose these mice exhibited mild granuloma formation with calcification or mild
hepatocyte degeneration but gave not other details or quantitative information as to the extent of
the lesions or what parts of the liver lobule were affected. The authors noted that "wild-type
mice administered 1000 and 1500 mg/kg exhibited centrilobular hypertrophy" and that "the mice
in the other groups did not exhibit any gross pathological changes after TCE exposure." Thus,
the hepatocellular hypertrophy reported in this study for TCE appeared to be correlated with
increases in percent liver/body weight in wild-type mice. In regard to the PPARa-null mice, the
authors stated that "differences in the liver to body weights in the control PPARa-null mice
[between Study 1 and 2 the 3-day and 3-week studies] were noted and may be due to differences
in the degree of steatosis that commonly occurs in this strain." Further mention of the
background pathology due to knockout of the PPARa was not discussed. The increased percent
liver/body weight reported between control and 1,000 mg/kg TCE exposed mice (5.1 vs. 5.6%)
was not accompanied by any discussion of pathological changes that could have accounted for
the change.
Direct comparisons of the effects of TCE, DCA, and TCA cannot be made from this
study as they were not studied for similar durations of exposure. However, while TCE induced
increased in percent liver/body weight ratios after 3 days and 3 weeks of exposure in wild-type
mice at the highest dose levels, for TCA exposure percent liver/body weight after 1 week
exposure in drinking water was slightly elevated at all dose levels with no dose-response (-10%
increase), and for DCA exposure in drinking water a similar elevation in percent liver/body
weight was also reported for the 0.25, 0.5, and 1.0 g/L dose levels (~11%) and that was increased
at the 2.0 g/L level by -25% reaching statistical significance. The authors interpret these data to
show no TCA-related changes in wild-type mice but the limited power of the study makes
quantitative conclusions difficult.
For PPARa-null mice all there was a slight decrease in percent liver/body weight
between control and TCA treated mice at the doses tested (-2%). For DCA-treated mice, all
treatment levels of DCA were reported to induce a higher percent liver/body weight ratio of at
least ~5%> with a 13% increase at the 2.0 g/L level. Again the limited power of the study and the
lack of data for TCE at similar durations of exposure as those studied for TCA and DCA makes
quantitative conclusions difficult and comparisons between the chemicals difficult. However,
the pattern of increased percent liver/body weight appears to be more similar between TCE and
DCA than TCA in both wild-type and PPARa-null mice.
In terms of histological description of effects, the authors note that "livers from the 2 g/L
DCA-treated wild-type and PPARa-null mice had hepatocyte cytoplasmic rarefication probably
due to an increase in glycogen accumulation." However, no special procedures are staining were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-49 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
performed to validate the assumption in this experiment. No other pathological descriptions of
the DCA treatment groups were provided. In regard to TCA, the authors noted that "the livers
from wild-type but not PPARa-null mice exposed to 2.0g/L TCA exhibited centrilobular
hepatocyte hypertrophy." No quantitative estimate of this effect was given and although the
extent of increase of percent liver/body weight was similar for all dose levels of TCA, there is no
indication from the study that lower concentrations of TCA also increased hepatocellular
hypertrophy or why there was no concurrent increase in liver weight at the highest dose of TCA
in which hepatocellular hypertrophy was reported. Thus, reports of hepatocellular hypertrophy
for DCA and TCA in the 1-week study were not correlated with changes in percent liver/body
weight.
For control animals, BrdU incorporation in the last week of the 3-week study was
reported to be at a higher baseline level in PPARa-null mice than wild-type mice (~2.5-fold).
For wild-type mice the authors reported a statistically significant increase at 500 and
1,000 mg/kg TCE at levels of ~1 and -4.5% hepatocytes incorporating the label after 5 days of
BrdU incorporation. Whether this measure of DNA synthesis is representative of cellular
proliferation or of polyploidization was not examined by the authors. Even at 1,000 mg/kg TCE
the percent of cells that had incorporated BrdU was less than 5% of hepatocytes in wild-type
mice. The magnitude percent liver/body weight ratio change at this exposure level was 4-fold
greater than that of hepatocytes undergoing DNA synthesis (16% increase in percent liver/body
weight ratio vs. 4% increase in DNA synthesis). The -1% of hepatocytes undergoing DNA
synthesis at the 500 mg/kg TCE level, reported to be statistically significant by the authors, was
not correlated with a concurrent increase in percent liver/body weight ratio. Thus, TCE-induced
changes in liver weight were not correlated with increases in DNA synthesis in wild-type mice
after 3 weeks of TCE exposure.
For PPARa-null mice, there was a ~3-fold of control value for the percent of hepatocytes
undergoing DNA synthesis at the 1,000 mg/kg TCE exposure level. The higher baseline level in
the null mouse, large variability in response at this exposure level, and low power of this
experimental design limited the ability to detect statistical significance of this effect although the
level was greater than that reported for the 500 mg/kg TCE exposure in wild-type mice that was
statistically significant. Thus, TCE appeared to induce an increase in DNA synthesis in PPARa-
null mice, albeit at a lower level than wild-type mice. However, the -2% increase in percent of
rd
hepatocytes undergoing DNA synthesis during the 3 week of a 3-week exposure to 1,000
mg/kg TCE in PPARa-null mice was insufficient to account for the —10% observed increase in
liver weight. For wild-type and PPARa-null mice, the magnitude of TCE-induced increases in
liver weight were 4-5-fold higher than that of increases in DNA-synthesis under this paradigm
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-50 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
and in both types of mice, a relatively small portion of hepatocytes were undergoing DNA
synthesis during the last week of a 3-week exposure duration. Whether the increases in liver
weight could have resulted from and early burst of DNA synthesis as well as whether the DNA
synthesis results reported here represents either proliferation or polyploidization, cannot be
determined from this experiment. Because of the differences in exposure protocol (i.e., use of 3
daily doses in methylcellulose rather than one dose in corn oil) the time course of the transient
increase in DNA synthesis reported cannot be assumed to be the same for this experiment and
others.
Not only were PPARa-null mice different than wild-type mice in terms of background
levels of liver weights, and hepatic steatosis, but this study reported that background levels of
PCO activity to be highly variable and in some instances different between wild-type and null
mice. There was reported to be ~6-fold PCO activity in PPARa-null control mice in comparison
to wild-type control mice in the 1-week DCA/TCA experiment (-0.15 vs. 0.85 units of activity/g
protein). However, in the same figure a second set of data are reported for control mice for
comparison to WY-14,643 treatment in which PCO activity was slightly decreased in PPARa-
null control mice versus wild-type controls (-0.40 vs. 0.65 units of activity/g protein). In the
experimental design description of the paper, WY-14,643 treatment and a separate control were
not described as part of the 1-week DCA/TCA experiment. For the only experiment in which
PCO activity was compared between wild-type and PPARa-null mice exposed to TCE (i.e.,
3-day exposure study), there was a reported increased over the control value of-2.5-fold that
was reported to be statistically significant at 1,500 mg/kg TCE (1.5 vs. 0.60 units of activity/g
protein). For control mice in the 3-day TCE experiment, there was an increase in this activity in
PPARa-null mice in comparison to wild-type mice (-0.60 vs. 0.35 units of activity/g protein).
While not statistically significant, there appeared to be a slight increase in PCO activity after
1,500 mg/kg TCE exposure for 3 days in PPARa-null mice of-30%. However, as noted above
the background levels of this enzyme activity varied widely between the experiments with not
only values for control animals varying as much as 6-fold (i.e., for PPARa-null mice) but also
for WY-14,643 administration. There was a 6.6-fold difference in PCO results for WY-14,643
in PPARa-null mice at the same concentration of WY-14,643 in the 3-day and 1-week
experiment, and a 1.44-fold difference in results in wild-type mice in these two data sets.
E.2.2.13. Ramdhan et al. (2008)
Ramdhan et al. (2008) examined the role of CYP2E1 in TCE-induced hepatotoxicity,
using CYP2E1 +/+ (wild-type) and CYP2E1 -/- (null) Sv/129 male mice (6/group) which were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
exposed for 7 days to 0, 1,000, or 2,000-ppm TCE by inhalation for 8 hours/day. The exposure
concentrations are noted by the authors to be much higher than occupational exposures and to
have increased liver toxicity after 8 hours of exposure as measured by plasma AST levels. To
put this exposure concentration into perspective, the Kjellstrand et al. (1983a; 1983b) inhalation
studies for 30 days showed that these levels were well above the 150-ppm exposure levels in
male mice that induced systemic toxicity. Nunes also reported hepatic necrosis up to 4% in rats
at 2,000 ppm for just 8 hours not 7 days. AST and ALT were measured at sacrifice. Histological
changes were scored using a qualitative scale of 0 = no necrosis, 1 = minimal as defined as only
occasional necrotic cells in any lobule, 2 = mild as defined as less than one-third of the lobule
structure affected, 3 = moderate as defined as between one-third and two-thirds of the lobule
structure affected and 4 = severe defined as greater than two-thirds of the lobule structure
affected. Real-time polymerase chain reaction (PCR) was reported for mRNA encoding a
number of receptors and proteins. Total RNA and Western Blot analysis was obtained from
whole-liver homogenates. The changes in mRNA expression were reported as means for 6 mice
per group after normalization to a level of P-actin mRNA expression and were shown relative to
the control level in the CYP2E1 wild-type mice.
The deletion of the CYP2E1 gene in the null mouse had profound effects on liver weight.
The body were was significantly increased in control CYP2E1 -/- mice in comparison to wild-
type controls (24.48 ± 1.44 g for null mice vs. 23.66 ± 2.44 g, m ± SD). This represents a 3.5%
increase over wild-type mice. However, the liver weight was reported in the CYP2E1 -/- mice to
be 1.32-fold of that of CYP2E1 +/+ mice (1.45 ± 0.10 g vs. 1.10 ± 0.14 g). The percent
liver/body weight ratio was 5.47 versus 4.63% or 1.18-fold of wild-type control for the null
mice.
The authors report that 1,000-ppm and 2,000-ppm TCE treatment did induce a
statistically significant change body weight for null or wild-type mice. However, there was an
increase in body weight in the wild-type mice (i.e., 23.66 ± 2.44, 24.52 ± 1.17, and 24.99 ± 1.78
for control, 1,000 ppm, and 2,000-ppm groups, respectively) and an increase in the variability in
response in the null mice (i.e., 24.48 ± 1.44, 24.55 ± 2.26, and 24.99 ± 4.05, for control, 1,000
ppm, and 2,000 ppm exposure groups, respectively). The percent liver/body weight was 5.47%
± 0.23%), 5.51%) ± 0.27%o, and 5.58%> ± 0.10% for control, 1,000 ppm and 2,000 ppm the
CYP2E1 -/- mice, respectively. The percent liver/body weight was 4.63%> ± 0.13%>, 6.62%> ±
0.40%o, and 7.24% ± 0.84%> for control, 1,000 ppm, and 2,000 ppm wild-type mice, respectively.
Therefore, while there appeared to be little difference in the TCE and control exposures for
percent liver/body weights in the CYP2E1 -/- mice (2%>) there was a 1.56-fold of control level
after 2,000 ppm in the wild-type mice after 7 days of inhalation exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The authors reported that "in general, the urinary TCE level in CYP2E1 -/- mice was less
than half that in CYP2E1 +/+ mice: urinary TCA levels in the former were about one-fourth
those in the latter." Of note is the large variability in urinary TCE detected in the 2,000-ppm
TCE exposed wild-type mice, especially after Day 4, and that in general the amount of TCE in
the urine appeared to be greatest after the 1st day of exposure and steadily declined between 1
and 7 days (i.e., -45% decline at 2,000 ppm and a -70% decline at 1,000 ppm) in the wild-type
mice. The amount of TCE in the urine was proportional to the difference in dose at days 1 and 5
(i.e., a 2-fold difference in dose resulted in a 2-fold difference in TCE detected in the urine). As
the detection of TCE in the urine declined with time, the amount of TCA was reported to steadily
increase between days 1 and 7 (e.g., from -3 mg TCA after the 1st day to -5.5 mg after 7 days
after 2,000 ppm exposure in wild-type mice). However, unlike TCE, there was a much smaller
differences in response between the two TCE exposure levels (i.e., a 12—44% or 1.12- to 1.44-
fold difference in TCA levels in the urine at days 1-7 for exposure concentrations that differ by a
factor of 2). This could be indicative of saturation in metabolism and TCA clearance into urine
at these high concentrations levels. The authors note that their results suggest that the
metabolism of TCE in both null and wild-type mice may have reached saturation at 1,000 ppm
TCE.
For ALT and AST activities in CYP2E1 -/- or CYP2E1 +/+ mice, both liver enzymes
were significantly elevated only at the 2,000 ppm level in CYP2E1 +/+ mice. Although the
increases in excreted TCA in the urine differed by only -33% between the 1,000 and 2,000 ppm
levels, liver enzyme levels in plasma differed by a much greater extent after 7 days exposure
between the 1,000 and 2,000-ppm groups of CYP2E1 +/+ mice (i.e., 1.26- and 1.83-fold of
control [ALT] and 1.40- and 2.20-fold of control [AST] for 1,000 ppm and 2,000 ppm TCE
exposure levels, respectively). The authors reported a correlation between plasma ALT and both
TCE (r = 0.7331) and TCA (r = 0.8169) levels but do not report details of what data were
included in the correlation (i.e., were data from CYP2E1 +/+ mice combined with those of the
CYP2E1 -/- mice and were control values included with treated values?).
The authors show photomicrograph of a section of liver from control CYP2E1 +/+ and
CYP2E1 -/- mice and describe the histological structure of the liver to appear normal. This
raises the question as to the cause of the hepatomegaly for the CYP2E1 mice in which the liver
weight was increased by a third.
The qualitative scoring for each of the 6 animals per group showed that none of the
CYP2E1 -/- control or treated mice showed evidence of necrosis. For the CYP2E1 +/+ mice
there was no necrosis reported in the control mice and in 3/6 mice treated with 1,000 ppm TCE.
Of the 3 mice that were reported to have necrosis, the score was reported as 1-2 for 2 mice and 1
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-53 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
for the third. It is not clear what a score of 1-2 represented given the criteria for each score
given by the authors, which defined a score of 1 as minimal and one of 2 as mild. For the 2,000
ppm TCE-exposed mice, all mice were reported to have at least minimal necrosis (i.e., 4 mice
were reported to have scores of 1-2, one mouse a score of 3 and one mouse a score of 1).
What is clear from the histopathology data are that there appeared to be great
heterogeneity of response between the 6 animals in each TCE-exposure group in CYP2E1 +/+
mice and that there was a greater necrotic response in the 2,000-ppm-exposed mice than the
1,000 ppm mice. These results are consistent with the liver enzyme data but not consistent with
the small difference between the 1,000 ppm and 2,000 ppm exposure groups for TCA content in
urine and by analogy, metabolism of TCE to TCA. A strength of this study is that it reports the
histological data for each animal so that the heterogeneity of liver response can be observed (e.g.,
the extent of liver necrosis was reported to range from only occasional necrotic cells in any
lobule to between one-third and two-thirds of the lobular structure affected after 2,000 ppm TCE
exposure for 7 days). Immunohistochemical analysis was reported to show that CYP2E1 was
expressed mainly around the centrilobular area in CYP2E1 +/+ mice where necrotic changes
were observed after TCE treatment.
Given the large variability in response within the liver after TCE exposure in CYP2E1
mice, phenotypic anchoring becomes especially important for the interpretation of mRNA
expression studies (see Sections E.l.l and E.3.1.2 for macroarray transcript profiling limitations
and the need for phenotypic anchoring). However, the data for mRNA expression of PPARa,
peroxisomal bifunctional protein (hydratase+3-hydroxyacyl-CoA dehydrogenase),very long
chain acyl-CoA dehydrogenase (VLCAD), CYP4A10, NFkB (p65, P50, P52), and IkBoi was
reported at the means ± SD for 6 mice per group and represented total liver homogenates. A
strength of the study was that they did not pool their RNA and can show means and standard
deviations between treatment groups. The low numbers of animals tested however, limits the
ability to detect statistically significance of the response. By reporting the means, differences in
the responses within dose groups was limited and reflected differential response and involvement
for different portions of the liver lobule and for the responses of the heterogeneous group of liver
cells populating the liver.
The authors reported that they normalized values to the level of P-actin mRNA in same
preparation with a value of 1 assigned as the mean from each control group. The values for
mRNA and protein expression reported in the figures appeared to have all been normalized to the
control values for the CYP2E1 -/- mice. Although all of the CYP2E1 -/- control values were
reported as a value of 1, the control values for the CYP2E1+/+ mice differed with the greatest
difference being presented for the CYP4A10-mRNA (i.e., the control level of CYP4A10 mRNA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
was ~3-fold higher in the CYP2E1+/+ mice than the CYP2E1 -/- mice). Further characterization
of the CYP2E1 mouse model was not provided by the authors.
The mean expression of PPARa mRNA was reported slightly reduced after TCE
treatment in CYP2E1 -/- mice (i.e., 0.72- and 0.78-fold of control after 1,000 and 2,000 ppm
TCE exposure, respectively). The CYP2E1 -/- mice had a higher baseline of PPARa mRNA
expression than the CYP2E1+/+ mice (i.e., the control level of the CYP2E1 -/- mice was 1.5-fold
of the CYP2E1+/+ mice). After TCE exposure, the CYP2E1 +/+ had a similar increase in
PPARa mRNA (~2.3-fold) at both 1,000 ppm and 2,000 ppm TCE. Thus, without the presence
of CYP2E1 there did not appear to be increased PPARa mRNA expression. For PPARa protein
expression, there was a similar pattern with ~1.6-fold of control levels of protein in the
CYP2E1 -/- mice after both 1,000 ppm and 2,000 ppm TCE exposures.
In the CYP2E1 +/+ mice the control level of PPARa protein was reported to be ~1.5-fold
of the CYP2E1 -/- control level. Thus, while the mRNA expression was less, the protein level
was greater. After TCE treatment, there was a 2.9-fold of control level of protein at 1,000 ppm
TCE and a 3.1-fold of control level of protein at 2,000 ppm. Thus, the magnitude of mRNA
increase was similar to that of protein expression for PPARa in CYP2E1 +/+ mice. The
magnitude of both was 3-fold or less over control after TCE exposure. This pattern was similar
to that of TCA concentration formed in the liver where there was very little difference between
the 1,000 and 2,000 ppm exposure groups in CYP2E1 +/+ mice. However, this pattern was not
consistent with the liver enzyme and histopathology of the liver that showed a much greater
response after 2,000-ppm exposure than 1,000-ppm TCE. In addition, where the mean enzyme
markers of liver injury and individual animals displayed marked heterogeneity in response to
TCE exposure, there was a much smaller degree of variability in the mean mRNA expression
and protein levels of PPARa.
For peroxisomal bifunctional protein there was a greater increase after 1,000 ppm TCE-
treated exposure than after 2,000 ppm TCE-treatment for both the CYP2E1 -/- and CYP2E1 +/+
mice (i.e., there was a 2:1 ratio of mRNA expression in the 1,000- vs. 2,000-ppm-exposed
groups). The CYP2E1 +/+ mice had a much greater response than the CYP2E1 -/- mice (i.e., the
CYP2E1 -/- mice had a 2-fold of control and the CYP2E1 +/+ mice had a 7.8-fold of control
level after 1,000 ppm TCE treatment). For peroxisomal bifunctional protein expression, the
magnitude of protein induction after TCE exposure was much greater than the magnitude of
increase in mRNA expression. In the CYP2E1 -/- mice 1,000 ppm TCE exposure resulted in a
6.9-fold of control level of protein while the 2,000 ppm TCE group had a 2.3-fold level.
CYP2E1 +/+ mice had a -50% higher control level than CYP2E1 mice and after TCE exposure
the level of peroxisomal bifunctional protein expression was 44-fold of control at 1,000 ppm
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-55 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
TCE and 40-fold of control at 2,000 ppm. Thus, CYP2E1 -/- mice were reported to have less
mRNA expression and peroxisomal bifunctional protein formed than CYP2E1 +/+ mice after
TCE exposure. However, there appeared to be more mRNA expression after 1,000 ppm than
2,000 ppm TCE in both groups and protein expression in the CYP2E1 -/- mice. After 2,000 ppm
TCE, there was similar peroxisomal bifunctional protein expression between the 1,000 ppm and
2,000 ppm TCE treated CYP2E1 +/+ mice. Again this pattern was more similar to that of TCA
detection in the urine—not that of liver injury.
For VLCAD the expression of mRNA was similar between control and treated
CYP2E1 -/- mice. For CYP2E1 +/+ mice the control level of VLCAD mRNA expression was
half that of the CYP2E1 -/- mice. After 1,000 ppm TCE the mRNA level was 3.7-fold of control
and after 2,000 ppm TCE the mRNA level was 3.1-fold of control. For VLCAD protein
expression was 1.8-fold of control after 1,000 ppm and 1.6-fold of control after 2,000 ppm in
CYP2E1 -/- mice. The control level of VLCAD protein in CYP2E1 +/+ mice appeared to be
1.2-fold control CYP2E1 -/- mice. After 1,000-ppm TCE treatment the CYP2E1 -/- mice were
reported to have 3.8-fold of control VLCAD protein levels and after 2,000-ppm TCE treatment
to have 3.9-fold of control protein levels. Thus, although showing no increase in mRNA there
was an increase in VLCAD protein levels that was similar between the two TCE exposure
groups in CYP2E1 -/- mice. Both VLCAD mRNA and protein levels were greater in CYP2E1
+/+ mice than CYP2E1 -/- mice after TCE exposure. This was not the case for peroxisomal
bifunctional protein. The magnitudes of TCE-induced increases in mRNA and protein increases
were similar between the 1,000 and 2,000 ppm TCE exposure concentrations, a pattern more
similar to TCA detection in the urine but not that of liver injury.
Finally, for CYP4A10 mRNA expression, there was an increase in expression after TCE
treatment of 3-fold for 1,000 ppm and 5-fold after 2,000 ppm in CYP2E1 -/- mice. Thus,
although the enzyme assumed to be primarily responsible for TCE metabolism to TCA was
missing, there was still a response for the mRNA of this enzyme commonly associated with
PPARa activation. Of note is that urinary concentrations of TCA were not zero after TCE
exposure in CYP2E1 -/- mice. Both 1,000 and 2,000 ppm TCE exposure resulted in -0.44 mg
TCA after 1 day or about 15-22% of that observed in CYP2E1 +/+ mice. Thus, some
metabolism of TCE to TCA is taking place in the null mice, albeit at a reduced rate. For
CYP2E1 +/+ mice, 1,000 ppm TCE resulted in an 8.3-fold of control level of CYP4A10 mRNA
and 2,000 ppm TCE resulted in a 9.3-fold of control level.
The authors did not perform an analysis of CYP4A10 protein. The authors state that "in
particular, the mRNA levels of microsomal enzyme CYP4A10 significantly increased in
CYP2E1+/+ mice after TCE exposure in a dose-dependent manner." However, the 2-fold
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-56 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
difference in TCE exposure concentrations did not result in a similar difference in response as
shown above. Both resulted in ~9-fold of control response in CYP2E1 +/+ mice. As with
PPARa, peroxisomal bifunctional protein, and VLCAD, the response was more similar to that of
TCA detection in the urine and not measured of hepatic toxicity. These data are CYP2E1
metabolism of TCE is important in the manifestation of TCE liver toxicity, however, it also
suggests that effects other than TCA concentration and indicators of PPARa are responsible for
acute hepatotoxicity resulting from very high concentrations of TCE.
The NFkB family and IkBoi were also examined for mRNA and protein expression.
These cell signaling molecules are involved in inflammation and carcinogenesis and are
discussed in Section E.3.3.3.3 and E.3.4.1.4. Given that presence of hepatocellular necrosis in
some of the CYP2E1 +/+ mice to varying degrees, inflammatory cytokines and cell signaling
pathways would be expected to be activated. The authors reported that
overall, TCE exposure did not significantly increase the expression of p65 and
p50 mRNAs in either CYP2E1+/+ or CYP2E1 -/- mice... However, p52 mRNA
expression significantly increased in the 2,000 ppm group of CYP2E1+/+ mice,
and correlation analysis showed that a significant positive relationship existed
between the expression of NFkB p52 mRNA and plasma ALT activity.., while no
correlation was seen between NFkB p64 or p50 and ALT activity (data not
shown).
The authors also note that TCE treatments "did not increase the expression of TNFR1 and
TNFR2 mRNA in CYP2E1+/+ and CYP2E1 -/- mice (data not shown)."
A more detailed examination of the data reveals that there was a similar increases in p65,
p50, and p52 mRNA expression increases with TCE treatment in CYP2E1 +/+ mice at both TCE
exposure levels. However, only p52 levels for the 2,000 ppm-exposed mice were reported to be
statistically significant (see comment above about the statistical power of the experimental
design and variability between animals). For 1,000 ppm TCE exposure the levels of p65, p50,
and p52 mRNA expression were 1.5-, 1.8-, and 2.0-fold of control. For 2,000 ppm TCE the
levels of p65, p50, and p52 mRNA expression were 1.8-, 1.8-, and 2.1-fold of control. Thus,
there was generally a similar response in all of these indicators of NFkB mRNA expression in
CYP2E1 +/+ mice that was mild with little to no difference between the 1,000 ppm and
2,000 ppm TCE exposure levels. For IkBoi mRNA expression there was not difference between
control and treatment groups for either type of mice. For CYP2E1 -/- mice there appeared to be
a -50% decrease in P52 mRNA expression in mice treated with both exposure concentrations of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-57 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
TCE. The authors plotted the relationship between p52 mRNA and plasma ALT concentration
for both CYP2E1 -/- and CYP2E1 +/+ mice together and claimed the correlation coefficient
(/' = 0.5075) was significant. However, of note is that none of the CYP2E1 -/- mice were
reported to have either hepatic necrosis or significant increases in ALT detection.
For protein expression, the authors showed results for p50 and p42 proteins. The control
CYP2E1 -/- mice appeared to have a slightly lower level of p50 protein expression (-30%) with
a much larger increase in p52 protein expression (i.e., 2.1-fold) than CYP2E1 +/+ mice. There
appeared to be a 2-fold increase in p50 protein expression after both 1,000-ppm and 2,000 ppm
TCE exposures in the CYP2E1 +/+ mice and a similar increase in p52 protein levels (i.e., 1.9-
and 2.5-fold of control for 1,000- and 2,000-ppm TCE exposures, respectively). Thus, the
magnitude of mRNA and protein levels were similar for p50 and p52 in CYP2E1 +/+ mice and
there was no difference between the 1,000- and 2,000-ppm treatments. For the CYP2E1 -/- mice
there was a modest increase in p50 protein after TCE exposure (1.1- and 1.3-fold of control for
1,000 and 2,000 ppm respectively) and a slight decrease in p52 protein (0.76- and 0.79-fold of
control). There was little evidence that the patterns of either expression or protein production of
NFkB family and IicBa corresponded to the markers of hepatic toxicity or that they exhibited a
dose-response. The authors note that although he expression of p50 protein increased in
CYP2E1 +/+ mice, "the relationship between p50 protein and ALT levels was not significant
(data not shown)." For TNFR1 there appeared to be less protein expression in the CYP2E1 +/+
mice than the CYP2E1 -/- mice (i.e., the null mice levels were 1.8-fold of the wild-type mice
levels). Treatment with TCE resulted in mild decrease of protein levels in the CYP2E1 -/- mice
and a 1.4- and 1.7-fold of control level in the CYP2E1 +/+ mice for 1,000 ppm and 2,000 ppm
levels, respectively. For p65, although TCE treatment-related effects were reported, of note the
levels of protein were 2.4 higher in the CYP2E1 +/+ mice than the CYP2E1 -/- mice. Thus,
protein levels of the NFkB family appeared to have been altered in the knockout mice. Also, as
noted in Section E.3.4.1.4, the origin of theNF-KB is crucial as to its effect in the liver and the
results of this report are for whole liver homogenates that contain parenchymal as well as
nonparenchymal cell and have been drawn from liver that are heterogeneous in the magnitude of
hepatic necrosis. The authors suggest that "TCA may act as a defense against hepatotoxicity
cause by TCE-delivered reactive metabolite(s) via PPARa in CYP2E1+/+ mice." However, the
data from this do not support such an assertion.
E.2.2.14. E.2.1.15 Ramdhan et al. (2010)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-58 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Ramdhan et al. (2010) examined the role of mouse and human PPARa in TCE-induced
hepatic steatosis and toxicity using male wild type, PPARa-null and PPARa-null mice with
human PPARa inserted ( hPPARa) (Cheung et al.. 2004) on a Sv/129 male mice (6/group) which
were exposed for 7 days to 0, 1,000, or 2,000-ppm TCE by inhalation for 8 hours/day. This was a
similar paradigm as that used in Ramdhan et al. (2008) with results between wild type mice
directly comparable. The expression of human PPARa cDNA in the humanized mice was
limted to hepatocytes under the control of tetracycline regulatory system.
Plasma aminotransferase activities (AST and ALT) were measured in plasma as well as
triglycerides. Hepatic triglyceride levels were measured as well. Urinary metabolites were
measured simarily to Ramdhan et al. (2008). Hepatic steatosis was identified based on the
presence of vacuoles consistent with lipid accumution and classified as microvesicular steatosis
if the nucleus remained in the center of the hepatocyte. Hepatocyte proliferation was classified
based on the presence of large hepatocytes with prominent eosinophilic cytoplasm.
Histopathology findings were scored in 20 randomly selected 200x microscopic fields per
section with steatotic scores of 0-3: none, mild 5-44% of parenchymal involvement of steatosis),
moderate (33-66%) or severe (> 66%). Necrotic cells were scored as 0-4: no necrosis, minimal
(only occasional necrotic cells in any lobule), mild ( two-thirds of the
lobular structur affected). Hepatocyte proliferation was scored as 0 (absent) or 1 (present).
Real-time PCR analysis was performed on total RNA from whole liver. Western Blot
analysis was also performed on whole liver (derived from both hepatocytes and non-
parenchymal cells) for NFkB, p65,p50,p52 and PPARa.
Significant differences were observed among control mice for each genotype. The mean
body weight of hPPARa mice was 14% less and 8.5% less than wild type mouse and PPARa-
null mice, respectively. The mean liver weight of hPPARa mice was 11% less than PPARa-
null mice and the liver the liver/body weight ratio of PPARa-null mice was 11% higher than
wild type mice. TCE at both 1000 and 2000 ppm significantly increased liver weight in the three
mouse lines to a similar extent (i.e., 38 and 49% in wild type mice, 20 and 37% in PPAR-null
mice, and 28 and 32% in hPPARa mice). The increases were not statistically significant
between doses within each strain. Liver/body weight ratios were also significantly increased
with TCE exposure at 1000 and 2000 ppm relative to controls (i.e., 38 and 43% in wild type
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-59 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
mice, 24 and 36% in PPARa-null mice, and 27 and 39% in hPPARa mice, respectively). The
difference between 2000 and 1000 ppm TCE exposure was statistically significant in PPARa-
null mice.
The authors reported no differences in urinary volume by genotype or exposure but did
not show the data. TCA and trichloroethanol were detected in all exposed mice with no
significant differences between the 1000 and 2000 ppm TCE levels. TCA concentrations were
reported to be significantly lower and trichloroethanol levels significantly higher in PPARa-null
mice relative to wild type mice with no differences in genotype between the sum of total TCA
and trichloroethanol concentrations between genotypes.
AST and ALT liver injury biomarkers were reported to vary < 10% among control mice
of each strain and to be significantly increased in all exposed mice relative to controls (41-74%
and 36-79%) higher, respectively) with mean levels within each group higher, though not
statistically significantly different, with exposure to 2000 vs 1000 ppm TCE.
Higher levels of plasma triglycerides were reported in untreated hPPARa mice than wild
type mice (52%). Significantly higher liver triglyceride levels were reported in untreated
hPPARa mice than wild type mice or PPARa-null mice (77 and 30%, respectively) and between
untreated PPARa-null mice and wild type mice (36%). Exposure to 2000 ppm TCE was
reported to induce an even greater difference between the wild type and PPARa-null mice
(113%>) . Exposure to 1000 ppm TCE was reported to induce greater liver triglyceride level in
hPPARa mice (50%) compared to wild type mice as well as 2000 ppm TCE (87%). There were
no signicant difference in mean plasma or liver triglyceride levels between the 2000 and 1000
ppm TCE treatment groups wthin each genotype. Hepatic triglyceride levels were reported to be
significantly correlated with liver/body weight ratios of all mice used in the study ( r= 0.54).
Neither necrosis or inflammatory cells were reported in liver sections from unexposed
mice. The authors reported small cytoplasmic vacuoles in section from unexposed PPARa-null
mice and hPPARa mice that resulted in steatosis scores >0. Steatosis was reported to be absent
in unexposed wild type mice and significantly increased in exposed vs unexposed PPARa-null
and hPPARa mice. Steatosis scores were reported to be significantly higher in the 2000 vs 1000
ppm TCE exposures to PPARa-null mice. The authors reported steatosis scored to be
significantly correlated with liver triglyceride levels of all mice examined in the study (r=0.75).
Macrovesicular steatosis was reported to occur more frequently in hPPARa than PPARa-null
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
mice. Necrosis scores were reported to be significantly higher in TCE exposed mice relative to
controls in all three genotype mice and to be significantly higher with 2000 vs 1000 ppm TCE
exposure in wild type mice and hPPARa mice. Inflammation scores were reported to be
significantly higher with exposed group than control with 2000 ppm TCE exposure than controls
for each genotype group with a difference between the 2000 ppm and 1000 ppm exposure groups
in wild type mice. Hepatocyte proliferation was reported to be significantly increased with 2000
ppm TCE exposure in wild type mice but not in the other genotypes or exposure concentrations.
Of note, the criteria for "proliferation" did not employ quantitative methods of DNA synthesis
but phenotypic descriptions of enlarged hepatocytes that may be indicative of polyploidy.
Background expression levels of several genes were reported to differ significantly
between strains in control mice. Very long chain acyl-CoA dehydrogenase (VLCAD), medium
chain acyl-CoA dehydrogenase (MCAD), peroxisomal bifunctional protein (hydratase+3-
hydroxyacyl-CoA dehydrogenase) (PH), peroxisomal thiolase (PT), diacylflicerol
acyltransferase 1 (DGAT1) and p52 mRNA levels were reported to be higher in untreated
hPPARa mice than wild type mice and PPARa-null mice. PPARa, proliferation cell nuclear
antigen (PCNA), p50 and tumor necrosis factor alpha (TNFa) mRNA levels were reported to be
higher in untreated hPPARa mice than PPARa-null mice. VLCAD, PH and PT mRNA levels
were reported to be significantly lower in untreated PPARa-null mice than wild type mice and
p50, p52, PPARy and TNFa were higher in untreated PPARa-null mice than wild type mice.
Exposure to TCE was reported to not increase the expression of human PPARa mRNA in
hPPARa mice but 2000 TCE exposure did significantly increase mouse PPARa mRNA in wild
type mice. PCNA mRNA expression and mRNA expression of VLCAD, MCAD, PH, and PT
was increased in TCE exposed vs control wild type mice and hPPARa mice. More pronounced
induction of PH and PT mRNA was reported for exposed wild type mice. Significant differences
were not reported in gene expression between 1000 and 2000 ppm TCE exposures.
DGAT1 and DGAT2 mRNA was reported to be significantly increased in hPPARa mice
exposed to 2000 ppm TCE and PPARa-null mice exposed to 1000 and 2000 ppm TCE in
comparison to respective control mice. Exposure to 1000 and 2000 ppm TCE was reported to
significantly increase PPARy mRNA in PPARa-null and hPPARa mice. DGAT1 and DGAT2,
PPARy mRNA levels were not changed with TCE exposure in wild type mice.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-61 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
NFkB p65 mRNA was reported to be significantly increase after TCE exposure in
PPARa-null and hPPARa mice but not wild type mice. NFkB p50 mRNA expression was
reported to be significantly increased with exposure to TCE in PPARa-null mice only but NFkB
p52 and TNFa mRNA expression was increased significantly with exposure in all strains. The
authors reported that NFkB p52 mRNA levels were significantly correlated with plasma ALT
levels in all mice used in the study ( r= 0.54).
Protein expression levels were reported to differ between the genotypes of untreated
mice. PPARa levels were 10.4 times higher in untreated hPPARa mice than wild type mice.
VLCAD, PT, acyl-CoA(ACOX) A and ACOX B proteins were reported to be significantly
higher in untreated hPPARa mice than wild type and PPARa-null mice and NFkB p65 to be
lower in hPPARa mice than PPARa-null mice. VLCAD, MCAD, PH, PT, ACOX A and ACOX
B expression was reported to be slightly lower and p65 and p52 expression slightly higher in
untreated PPARa-null mice vs wild type mice.
TCE exposure was reported to increase VLCAD, PH, PT, ACOX A and ACOX B in wild
type and hPPARa mice but not to induced PPARa protein expression. MCAD protein was
significantly increased after TCE exposure in hPPARa mice only. PCNA protein was increased
in TCE exposed mice in comparison to controls in all strains. NFkB p52 and TNFa proteins
were also increased from TCE exposure in all strains but NFkB p50 and p65 proteins were
increased in TCE-exposed PPARa-null mice only. 4-Hydroxy-2- nonenal protein (a maker of
oxidative stress) was increased by 1000 ppm TCE exposure in PPARa-null mice and by 2000
ppm TCE exposure in wild type and hPPARa mice.
The authors reported that they measured hepatic protein expression of CYP2E1 and
ALDH2 enzymes and did not observe a significant difference among controls (data not shown)
and that TCE exposure did not alter hepatic CYP2E1 expression but did decrease ALDH2
expression to a comparable extent in all mouse lines (data not shown). Thus, changes in urinary
TCA levels in the differing strains were not related to changes in expression of these metabolic
enzymes.
While the authors of the paper suggested that the increased susceptibility of PPARa-null
mice and hPPATa mice to TCE toxicity they report is indicative of "protection" by having intact
and normal PPARa expression in mice, the disturbances they also reported in these genotypes
without treatment shows that an already compromised animal is more susceptible to additional
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-62 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
insult by high levels of TCE exposure. This study provides an extensive set of parameters
altered in the PPARa-null and hPPARa mice by such genetic manipulation alone. In particular
insertion of human PPAR in the null mice did not return the mice to a normal state. The authors
noted that hepatic triglyceride levels were the highest in untreated hPPARa among the 3 strains
suggesting that human PPARa insertion did not restore proper lipid regulation in the liver. The
humanized mice in particular exhibited a greater than 10-fold expression of PPAR in an
untreated state. Functional differences between the human and rodent versions of PPAR are
difficult to ascertain from this study given the large differences in PPAR protein expression
between wild type and humanized mice and the presence of human PPAR only in the
hepatocytes in this model. The authors noted that the replacement of human PPARa in the
humanized mouse may not have been sufficient to prevent steatosis and that the differences in
responses between wild type and humanized mice may reflect functional consequences related to
the use of an artificial construct of the reinserted gene without normal control elements in
addition to or instead of any differences between human or mouse PPARa. They stated that
because they used genetically modified mice with underlying dysregulation, and evaluated very
high TCE exposures, their findings may not directly reveal the differences in human PPARa
function between mice and humans. The increased toxicity from overexpression of human
PPARa in this model is also acknowledged as leading to greater background toxicity in
unexposed humanized mice.
Responses reported for gene expression are for liver homogenates so that NFkB and
TNFa mRNA expression changes could not be distinguished between Kuffer cell or hepatocytes
origin. The authors noted the similarity of TCE induced hepatomegally in PPARa null mice in
this study and that of Nakajima et al. (2000). They noted that TCE induction of PCNA protein
(cell proliferation marker) was increased in all three group but using their phenotypic marker of
increased cell size of evidence of increased hepatocyte proliferation in wild type mice.
The authors noted that they report differences in this study and their study of similar
design (Ramdhan et al.. 2008) for gene expression induced by TCE exposure in wild type mice.
Differences in TCE-induced effects between the two studies include less pronounced induction
of PPARa , more pronounced increases in PH protein and VLCAD mRNA expression, and ALT
and AST levels for this study than the previous one for wild type mice. They stated that urinary
TCA levels in wild type mice were incorrectly reported by Ramdhan et al. (2008) but have been
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
corrected in this study. They also noted discrepancies in mRNA and protein expression for some
genes in this study. Finally, the authors acknowledged that the small number of mice examined
in each group limits the power to identify statistically significant biological effects.
E.2.3. Subchronic and Chronic Studies of Trichloroethylene (TCE)
For the purposes of this discussion, studies of duration of 4 weeks or more are considered
subchronic. Like those of shorter duration, there is variation in the depth of study of liver
changes induced by TCE with many of the longer duration studies focused on the induction of
liver cancer. Many subchronic studies were conducted a high doses of TCE that caused toxicity
with limited reporting of effects. Similar to acute studies, some of the subchronic and chronic
studies have detailed examinations of the TCE-induced liver effects while others have reported
primarily liver weight changes as a marker of TCE-response. Similar issues also arise with the
impact of differences in initial and final body weights between control and treatment groups on
the interpretation of liver weight gain as a measure of TCE-response.
For many of the subchronic inhalation studies, issues associated with whole body
exposures make determination of dose levels difficult. For gavage experiments, death from
gavage dosing, especially at higher TCE exposures, is a recurring problem and, unlike inhalation
exposures, the effects of vehicle can also be at issue for background liver effects. Concerns
regarding effects of oil vehicles, especially corn oil, have been raised with Kim et al. (1990a)
noting that a large oil bolus will not only produce physiological effects, but alter the absorption,
target organ dose, and toxicity of volatile organic compounds (VOCs). Charbonneau et al.
(1991) reported that corn oil potentiates liver toxicity from acetone administration that is not
related to differences in acetone concentration. Several oral studies in particular document that
use of corn oil gavage induces a different pattern of toxicity, especially in male rodents (see
Merrick et al., 1989, Section E.2.2.1 below). Several studies listed below report the effects of
hepatocellular DNA synthesis and indices of lipid peroxidation (i.e., Channel et al., 1998) are
especially subject to background vehicle effects. Rusyn et al. (1999) report that a single dose of
dietary corn oil increases hepatocyte DNA synthesis 24 hours after treatment by ~3.5-fold,
activation of NF-kB to a similar extent ~2 hours after treatment almost exclusively in Kupffer
cells, a ~3-4-fold increase in hepatocytes after 8 hours, and increased in TNFa mRNA between
8 and 24 hours after a single dose in female rats. In regard to studies that have used the i.p. route
of administration, as noted by Kawamoto et al. (1988b) (see Section E.2.2.10 below), injection of
TCE may result in paralytic ileus and peritonitis and that subcutaneous treatment paradigm will
result in TCE not immediately being metabolized but retained in the fatty tissue. Wang and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-64 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Stacey (1990) state that "intraperitoneal injection is not particularly relevant to humans" and that
intestinal interactions require consideration in responses such as increase serum bile acid (see
Section E.2.3.5 below).
E.2.3.1. Merrick et al. (1989)
The focus of this study was the examination of potential differences in toxicity or orally
gavaged TCE administered in corn oil an aqueous vehicle in B6C3F1 mice. As reported by
Melnick et al. (1987) above, corn oil administration appeared to have an effect on peroxisomal
enzyme induction. TCE (99.5% purity) was administered in corn oil or an aqueous solution of
20% Emulphor to 14-17 week old mice (n = 12/group) at 0, 600, 1,200 and 2,400 mg/kg/d
(males) and 0, 450, 900, and 1,800 mg/kg/d (females) 5 times a week for 4 weeks. The authors
stated that due to "varying lethality in the study, 10 animals per dose group were randomly
selected (where possible) among survivors for histological analysis." Hepatocellular lesions
were characterized
as a collection of approximately 3-5 necrotic hepatocytes surrounded by
macrophages and polymorphonuclear cells and histopathological grading was
reported as based on the number of necrotic lesions observed in the tissue
sections: 0 = normal; 1 = isolated lesions scattered throughout the section; 2 = one
to five scattered clusters of necrotic lesions; 3 = more than five scattered clusters
of necrotic lesions; and 4 = clusters of necrotic lesions observed throughout the
entire section."
The authors described lipid scoring of each histological section as "0 = no Oil-
Red O staining present; 1 = less than 10% staining; 2 = 10-25%) staining; 3 = 25-30%)
staining; and 4 = greater than 50%> staining.
The authors reported dose-related increases in lethality in both males and females
exposed to TCE in Emulphor with all male animals dying at 2,400 mg/kg/d with 8/12
females dying at 1,800 mg/kg/d. In both males and females, 2/12 animals also died at the
next highest dose as well with no unscheduled deaths in control or lowest dose animals.
For corn oil gavaged mice, there were 1-2 animals in each TCE treatment groups of male
mice that died while there were no unscheduled deaths in female mice. The authors
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-65 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
stated that lethality occurred within the first week after chemical exposure. The authors
presented data for final body weight and liver/body weight values for 4 weeks of
exposure and listed the number of animals per group to be 10-12 for corn oil gavaged
animals. The reduced number of animals in the Emulphor gavaged animals are reflective
of lethality and limit the usefulness of this measure at the highest doses (i.e., 1,800
mg/kg/d for female mice). In mice treated with TCE in Emulphor gavage, the final body
weight of control male animals appeared to be lower than those that were treated with
TCE while for female mice the final body weights were similar between treated and
control groups. For male mice treated with Emulphor, body weights were 22.8 ± 0.8,
25.3 ± 0.5, and 24.3 ± 0.4 g for control, 600 mg/kg/d, and 1,200 mg/kg/d and for female
mice body weights were 20.7 ± 0.4, 21.4 ± 0.3, and 20.5 ± 0.3 g for control, 450 mg/kg/d,
and 900 mg/kg/d of TCE.
For percent liver/body weight ratios, male mice were reported to have 5.6% ± 0.2%,
6.6% ± 0.1%), and 7.2%> ± 0.2%> for control, 600, and 1,200 mg/kg/d and for female mice were
5.1%) ± 0.1%o, 5.8%o ± 0.1%o, and 6.5%> ± 0.2%> for control, 450 mg/kg/d, and 900 mg/kg/d of
TCE. These values represent 1.11- and 1.07-fold of control for final body weight in males
exposed to 600 and 1,200 mg/kg/d and 1.18- and 1.29-fold of control for percent liver/body
weight, respectively. For females, they represent 1.04- and 0.99-fold of control for final body
weights in female exposed to 450mg/kg/d and 900 mg/kg/d and 1.14- and 1.27-fold of control
for percent liver/body weight, respectively.
In mice treated with corn oil gavage the final body weight of control male mice was
similar to the TCE treatment groups and higher than the control value for male mice given
Emulphor vehicle (i.e., 22.8 ± 0.8 g for Emulphor control vs. 24.3 ± 0.6 g for corn oil gavage
controls or a difference of ~7%>). The final body weights of female mice were reported to be
similar between the vehicles and TCE treatment groups. The baseline percent liver/body weight
was also lower for the corn oil gavage control male mice (i.e., 5.6%> for Emulphor vs. 4.7%> for
corn oil gavage or a difference of—19% that was statistically significant). Although the final
body weights were similar in the female control groups, the percent liver/body weight was
greater in the Emulphor vehicle group (5.1%> ± 0.1%> in Emulphor vehicle group vs. 4.7%> ± 0.1%>
for corn oil gavage or a difference of ~9%> that was statistically significant). For male mice
treated with corn oil, final body weights were 24.3 ± 0.6, 24.3 ± 0.4, 25.2 ± 0.6, and 25.4 ± 0.5 g
for control, 600, 1,200, and 2,400 mg/kg/d, and for female mice body weights were 20.2 ± 0.3,
20.8 ± 0.5, 21.8 ± 0.3 g, and 22.6 ± 0.3 g for control, 450, 900, and 1,800 mg/kg/d of TCE.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-66 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
For percent liver/body weight ratios, male mice were reported to have 4.7% ± 0.1%,
6.4% ± O.P/o, 7.7% ± 0.1%), and 8.5%> ± 0.2%> for control, 600, 1,200, and 2,400 mg/kg/d and for
female mice were 4.7%> ± 0.1%>, 5.5%> ± 0.1%>, 6.0%> ± 0.2%>, and 7.2% ± 0.1%> for control, 450,
900, and 1,800 mg/kg/d of TCE. These values represent 1.0-, 1.04-, and 1.04-fold of control for
final body weight in males exposed to 600, 1,200, and 2,400 mg/kg/d TCE and 1.36-, 1.64-, and
1.81-fold of control for percent liver/body weight, respectively. For females, they represent
1.03-, 1.08-, and 1.12-fold of control for body weight in female exposed to 450, 900, and 1,800
mg/kg/d and 1.17-, 1.28-, and 1.53-fold of control for percent liver/body weight, respectively.
Because of premature mortality, the difference in TCE treatment between the highest
doses that are vehicle-related cannot be determined. The decreased final body weight and
increased percent liver/body weight ratios in the Emulphor control animals make comparisons of
the exact magnitude of change in these parameters due to TCE exposure difficult to determine as
well as differences between the vehicles. The authors did not present data for age-matched
controls which did not receive vehicle so that the effects of the vehicles cannot be determined
(i.e., which vehicle control values were most similar to untreated controls given that there was a
difference between the vehicle controls).
A comparison of the percent liver/body weight ratios at comparable doses between the
two vehicles shows little difference in TCE-induced liver weight increases in female mice.
However, the corn oil vehicle group was reported to have a greater increase in comparison to
controls for male mice treated with TCE at the two lower dosage groups. Given that the control
values were approximately 19%> higher for the Emulphor group, the apparent differences in TCE-
dose response may have reflected the differences in the control values rather than TCE exposure.
Because controls without vehicle were not examined, it cannot be determined whether the
difference in control values was due to vehicle administration or whether a smaller or younger
group of animals was studied on one of the control groups. The body weight of the animals was
also not reported by the authors at the beginning of the study so that the impact of initial
differences between groups versus treatment cannot be accurately determined.
Serum enzyme activities for ALT, AST and LDH (markers of liver toxicity) showed that
there was no difference between vehicle groups at comparable TCE exposure levels for male or
female mice. Enzyme levels appeared to be elevated in male mice at the higher doses (i.e., 1,200
and 2,400 mg/kg/d for ALT and 2,400 mg/kg/d for AST) with corn oil gavage inducing similar
increases in LDH levels at 600, 1,200, and 2,400 mg/kg/d TCE. For ALT and AST there
appeared to be a dose-related increase in male mice with the 2,400 mg/kg treatment group having
much greater levels than the 1,200 mg/kg group. In Emulphor treatment groups there was a
similar increase in ALT levels in males treated with 1,200 mg/kg TCE as with those treated with
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-67 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
corn oil and those increases were significantly elevated over control levels. For LDH levels
there were similar increase at 1,200 mg/kg TCE for male mice treated using either Emulphor or
corn oil.
The authors report that visible necrosis was observed in 30-40% of male mice
administered TCE in corn oil but not that there did not appear to be a dose-response (i.e., the
score for severity of necrosis was reported to be 0, 4, 3, and 4 for corn oil control, 600, 1,200,
and 2,400 mg/kg/d treatment groups from 10 male mice in each group). No information in
regard to variation between animals was given by the authors. For male mice given Emulphor
gavage the extent of necrosis was reported to be 0, 0, and 1 for 0, 600, and 1,200 mg/kg/d TCE
exposure, respectively. For female mice, the extent of necrosis was reported to be 0 for all
control and TCE treatment groups using either vehicle.
Thus, except for LDH levels in male mice exposed to TCE in corn oil there was not a
correlation with the extent of necrosis and the increases in ALT and AST enzyme levels.
Similarly, there was an increase in ALT levels in male mice treated with 1,200 mg/kg/d exposure
to TCE in Emulphor that did not correspond to increased necrosis.
For Oil-Red O staining there was a score of 2 in the Emulphor treated control male and
female mice while 600 mg/kg/d TCE exposure in Emulphor gavaged male mice and 900 mg/kg/d
TCE in corn oil gavaged female mice had a score of 0, along with the corn oil gavage controls in
male mice. For female control mice treated with corn oil gavage, the staining was reported to
have a score of 3. Thus, there did not appear to be a dose-response in Oil-Red oil staining
although the authors claimed there appeared to be a dose-related increase with TCE exposure.
The authors described lesions produced by TCE exposure as
focal and were surrounded by normal parenchymal tissue. Necrotic areas were
not localized in any particular regions of the lobule. Lesions consisted of central
necrotic cells encompassed by hepatocytes with dark eosinophilic staining
cytoplasm, which progressed to normal-appearing cells. Areas of necrosis were
accompanied by localized inflammation consisting of macrophages and
polymorphonuclear cells.
No specific descriptions of histopathology of mice given Emulphor were provided in terms of
effects of the vehicle or TCE treatment. The scores for necrosis were reported to be only a 1 for
the 1,200 mg/kg concentration of TCE in male mice gavaged with Emulphor but 3 for male mice
given the same concentration of TCE in corn oil. However, enzyme levels of ALT, AST, and
LDH were similarly elevated in both treatment groups.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-68 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
These results do indicate that administration of TCE for 4 weeks via gavage using
Emulphor resulted in mortality of all of the male mice and most of the female mice at a dose in
corn oil that resulted in few deaths. Not only was there a difference in mortality, but vehicle also
affected the extent of necrosis and enzyme release in the liver (i.e., Emulphor vehicle caused
mortality as the highest dose of TCE in male and female mice that was not apparent from corn
oil gavage, but Emulphor and TCE exposure induced little if any focal necrosis in males at
concentrations of TCE in corn oil gavage that caused significant focal necrosis). In regard to
liver weight and body weight changes, TCE exposure in both vehicles at nonlethal doses induced
increased percent liver/body weight changes male and female mice that increased with TCE
exposure level. The difference in baseline control levels between the two vehicle groups
(especially in males) make a determination of the quantitative difference vehicle had on liver
weight gain problematic although the extent of liver weight increase appeared to be similar
between male and female mice given TCE via Emulphor and female mice given TCE via corn
oil. In general, enzymatic markers of liver toxicity and results for focal hepatocellular necrosis
were not consistent and did not reflect dose-responses in liver weight increases. The extent of
necrosis did not correlate with liver weight increases and was not elevated by TCE treatment in
female mice treated with TCE in either vehicle, or in male mice treated with Emulphor. There
was a reported difference in the extent of necrosis in male mice given TCE via corn oil and
female mice given TCE via corn oil but the necrosis did not appear to have a dose-response in
male mice. Female mice given corn oil and male and female mice given TCE in Emulphor had
no to negligible necrosis although they had increased liver weight from TCE exposure.
E.2.3.2. Goel et al. (1992)
The focus of this study was the description of TCE exposure related changes in mice after
28 days of exposure with regard to TCE-induced pathological and liver weight change. Male
Swiss mice (20-22 g body weight or 9% difference) were exposed to 0, 500, 1,000 or 2,000
mg/kg/d TCE (BDH analytical grade) by gavage in groundnut oil (n = 6 per group) 5 days a
week for 28 days. The ages of the mice were not given by the authors. Livers were examined
for "free -SH contents," total proteins, catalase activity, acid phosphatase activity, and "protein
specific for peroxisomal origin of approx, 80 kd."
The authors report no statistically significant change in body weight with TCE treatment
but a significant increase in liver weight. Body weight (mean ± SE) was reported to be 32.67 ±
1.54, 31.67 ± 0.61, 33.00 ± 1.48, and 27.80 ± 1.65 g from exposure to oil control, 500, 1,000,
and 2,000 mg/kg/d TCE, respectively. There was a 15% decrease in body weight at the highest
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
exposure concentration of TCE that was not statistically significant, but the low number of
animals examined limits the power to detect a significant change. The percent relative
liver/body weight was reported to be 5.29% ± 0.48%, 7.00% ± 0.36%, 7.40% ± 0.39%, and
7.30%) ± 0.48%) from exposure to oil control, 500, 1,000, and 2,000 mg/kg/d TCE, respectively.
This represents 1.32-, 1.41-, and 1.38-fold of control in percent liver/body weight for 500, 1,000,
and 2,000 mg/kg/d TCE, respectively.
The "free -SH content" in [j,mol ~SH/g tissue was reported to be 5.47 ±0.17, 7.46 ±
0.21, 7.84 ± 0.34, and 7.10 ± 0.34 from exposure to oil control, 500, 1,000, and 2,000 mg/kg/d
TCE, respectively. This represents 1.37-, 1.44-, and 1.30-fold of control in -SH/g tissue weight
for 500, 1,000, and 2,000 mg/kg/d TCE, respectively. Total protein content in the liver in mg/g
tissue was reported to be 170 ±3, 183 ± 5, 192 ± 7, and 188 ± 3 from exposure to oil control,
500, 1,000, and 2,000 mg/kg/d TCE, respectively. This represents 1.08-, 1.13-, and 1.11-fold of
control in total protein content for 500, 1,000, and 2,000 mg/kg/d TCE, respectively. Thus, the
increases in liver weight, "free -SH content" and increase protein content were generally parallel
and all suggest that liver weight increases had reached a plateau at the 1,000 mg/kg/d exposure
concentration perhaps reflecting toxicity at the highest dose as demonstrated by decreased body
weight in this study.
The enzyme activities of S-ALA dehydrogenase ("a key enzyme in heme biosynthesis"),
catalase, and acid phosphatase were assayed in liver homogenates. Treatment with TCE
decreased S-ALA dehydrogenase activity to a similar extent at all exposure levels (32-35%>
reduction). For catalase the activity as units of catalase/mg protein was reported to be
25.01 ± 1.81, 32.46 ± 2.59, 41.11 ± 5.37, and 33.96 ± 3.00 from exposure to oil control, 500,
1,000, and 2,000 mg/kg/d TCE, respectively. This represents 1.30-, 1.64-, and 1.36-fold in
catalase activity for 500, 1,000, and 2,000 mg/kg/d TCE, respectively. The increasing variability
in response with TCE exposure concentration is readily apparent from these data as is the
decrease at the highest dose, perhaps reflective of toxicity. For acid phosphatase activity in the
liver there was a slight increase (5—11%>) with TCE exposure that did not appear to be dose-
related.
The authors report that histologically, "the liver exhibits swelling, vacuolization,
widespread degeneration/necrosis of hepatocytes as well as marked proliferation of endothelial
cells of hepatic sinusoids at 1000 and 2000 mg/kg TCE doses." Only one figure is given at the
light microscopic level in which it is impossible to distinguish endothelial cells from Kupffer
cells and no quantitative measures or proliferation were examined or reported to support the
conclusion that endothelial cells are proliferating in response to TCE treatment. Similarly, no
quantitation regarding the extent or location of hepatocellular necrosis is given. The presence or
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
absence of inflammatory cells was not noted by the authors as well. In terms of white blood cell
count, the authors noted that it was slightly increased at 500 mg/kg/d but decreased at 1,000 and
2,000 mg/kg/d TCE, perhaps indicating macrophage recruitment from blood to liver and kidney,
which was also noted to have pathology at these concentrations of TCE.
E.2.3.3. Kjellstrand et al. (1981b)
This study was conducted in mice, rats, and gerbils and focused on the effects of
150-ppm TCE exposure via inhalation on body and organ weight. No other endpoints other than
organ weights were examined in this study and the design of the study is such that quantitative
determinations of the magnitude of TCE response are very limited. NMRI mice (weighing -30 g
with age not given), S-D rats (weighing -200 g with age not given), and Mongolian gerbils
(weighing -60 g with age not given) were exposed to 150-ppm TCE continuously. Mice were
exposed for 2, 5, 9, 16, and 30 days with the number of exposed animals and controls in the 2, 5,
9, and 16 days groups being 10. For 30-day treatments, there were two groups of mice
containing 20 mice per group and one group containing 12 mice per group. In addition, there
was a group of mice (n = 15) exposed to TCE for 30 days and then examined 5 days after
cessation of exposure and another group (n = 20) exposed to TCE for 30 days and then examined
30 days after cessation of exposure. For rats, there were three groups exposed to TCE for 30
days, which contained 24, 12, and 10 animals per group. For gerbils, there were three groups
exposed to TCE for 30 days, which contained 24, 8, and 8 animals per group. The groups were
reported to consist of equal numbers of males and female but for the mice exposed to TCE for 30
days and then examined 5 days later, the number was 10 males and 5 females. Body weights
were reported to be recorded before and after the exposure period. However, the authors state
"for technical reasons the animals within a group were not individually identified, i.e., we did not
know which initial weight in the group corresponded to which final one." They authors stated
that this design presented problems in assessing the precision of the estimate. They go on to
state that rats and gerbils were partially identifiable as the animals were housed 3 to a cage and
cage averages could be estimated. Not only were mice in one group housed together but
even worse: at the start of the experiment, the mice in M2 [group exposed for 2
days] and M9 [group exposed for 9 days] were housed together, and similarly M5
[group exposed for 5 days] and Ml6 [group exposed for 16 days]. Thus, we had,
e.g., 10 initial weights for exposed female mice in M2 and M9 where we could
not identify those 5 that were M2 weights. Owing to this bad design (forced upon
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-71 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
us by the lack of exposure units), we could not study weight gains for mice and so
we had to make do with an analysis of final weights.
The problems with the design of this study are obvious from the description given by the
authors themselves. The authors stated that they assumed that the larger the animal the larger the
weight of its organs so that all organ weights were converted into relative weights as percentage
of body weight. The fallacy of this assumption is obvious, especially if there was toxicity that
decreased body weight and body fat but at the same time that caused increased liver weight as
has been observed in many studies at higher doses of TCE. In fact, Kjellstrand et al. (1983b)
reported that a 150-ppm TCE exposure for 30 days does significantly decreases body weight
while elevating liver weight in a group of 10 male NMRI mice. Thus, the body weight estimates
from this study are inappropriate for comparison to those in studies where body weights were
actually measured. The liver/body weight ratios that would be derived from such estimates of
body weights would be meaningless.
The group averages for body weight reported for female mice at the beginning of the 30-
day exposure varied significantly and ranged from 23.2 to 30.2 g (-24%). For males, the group
averages ranged from 27.3 to 31.4 g (-14%). For male mice there was no weight estimate for
the animals that were exposed for 30 days and then examined 30 days after cessation of
exposure.
The authors only report relative organ weight at the end of the experiment rather than the
liver weights for individual animals. Thus, these values represent extrapolations based on what
body weight may have been. For mice that were exposed to TCE for 30 days and examined after
30 days of exposure, male mice were reported to have "relative organ weight" for liver of 4.70%
± 0.10%) versus 4.27% ± 0.13% for controls. However, there were no initial body weights
reported for these male mice and the body weights are extrapolated values. Female mice
exposed for 30 days and examined 30 days after cessation of exposure were reported to have
"relative organ weights" for liver of 4.42% ± 0.11% versus 3.62% ± 0.09%. The group average
of initial body weights for this group was reported by the authors.
Although the initial body weight for female control mice as a group average was
reported to be similar between the female group exposed to 30 days of TCE and sacrificed 30
days later and those exposed for 30 days and sacrificed 5 days later (30.0 g vs. 30.8 g), the
liver/body weight ratio varied significantly in these controls (4.25 ± 0.19 vs. 3.62 ± 0.09) as did
the number of animals studied (5 female mice in the animals sacrificed after 5 days exposure
versus 10 female mice in the group sacrificed after 30 days exposure). In addition, although
there were differences between the 3 groups of mice exposed to TCE for 30 days and then
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-72 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
sacrificed immediately, the authors present the data for extrapolated liver/body weight as pooled
results between the 3 groups. In comparison to control values, the authors report 1.14-, 1.35-,
1.58-, 1.47-, and 1.75-fold of control for percent liver/body weight using body weight
extrapolated values in male mice at 2, 5, 9, 16, and 30 days of TCE exposure, respectively. For
females, they report 1.27-, 1.28-, 1.49-, 1.41-, and 1.74-fold of control at 2, 5, 9, 16, and 30 days
of TCE, respectively.
Although the authors combined female and male relative increases in liver weight in a
figure, assign error bars around these data point, and attempt to draw assign a time-response
curve to it, it is clear that these data, especially for female mice, do not display time-dependent
increase in liver/body weight from 5 to 16 days of exposure and that a comparison of results
between 5 animals and 26 is very limited in interpretation. Of note, is the wide variation in the
control values for relative liver/body weight.
For male mice there did not seem to be a consistent pattern with increasing duration of
the experiment with values of 4.61, 5.15, 5.05, 4.93, and 4.04% for 2, 5, 9, 16, and 30-day
exposure groups. This represented a difference of-27%. For female mice, the relative
liver/body weight was 4.14, 4.58, 4.61, 4.70, and 3.99% for 2, 5, 9, 16, and 30 day exposure
groups. Thus, it appears that the average relative liver/body weight percent was higher in the 5,
9, and 16 day treatment group for both genders than that to the 30 day group and was consistent
between these days. There is no apparent reason for there to be such large difference between 16
day and 30-day treatment groups due to increasing age of the animals. Of note is that for the
control groups pared with animals treated for 30 days and then examined 30 days later, the male
mice had increased relative liver/body weights (4.27 vs. 4.04%) but that the females had
decreases (3.62 vs. 3.99%). Such variation between controls does not appear to be age and size
related but to variations in measure or extrapolations, which can affect comparisons between
treated and untreated groups and add more uncertainty to the estimates. In addition, the number
of mice in the groups exposed to 2 though 16 days were only 5 animals for each gender in each
group while the number of animals reported in the 30-day exposure group numbered 26 for each
gender.
For animals exposed to 30 days and then examined after 5 or 30 days, male mice were
reported to have percent liver/body weight 1.26- and 1.10-fold of control after 5 and 30 days
cessation of exposure while female mice were reported to have values of 1.14- and 1.22-fold of
control after 5 and 30 days cessation of exposure, respectively. Again, the male mice exposed
for 30 days and then examined after 30 days of cessation of exposure did not have reported
initial body weights giving this value a great deal of uncertainty. Thus, while liver weights
appeared to increase during 30 days of exposure to TCE and decreased after cessation of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-73 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
exposure in both genders of mice, the magnitudes of the increases and decrease cannot be
determined from this experimental design. Of note is that liver weights appeared to still be
elevated after 30 days of cessation exposure.
In regard to initial weights, the authors reported that the initial weight of the rats were
different in the 3 experiments they conducted with them and state that "in those 2 where
differences were found in females, their initial weights were about 200 g and 220 g, respectively,
while the corresponding weights were only about 160 g in that experiment where no differences
were found." The differences in initial body weight of the rat groups were significant. In
females group averages were 198, 158, and 224 g, for groups 1, 2, and 3, respectively, and for
males group averages were 222, 166, and 248 g for groups 1,2, and 3 respectively. This
represents as much as a 50% difference in initial body weights between these TCE treatment
groups. Control values varied as well with group averages for controls ranging from 167 g for
group 2 to 246 g for group 3 at the start of exposure. For female rats control groups ranged from
158 to 219 g at the start of the experiment.
The number of animals in each group varied greatly as well making quantitative
comparison even more difficult with the numbers varying between 5 and 12 for each gender in
rats exposed for 30 days to TCE. The authors pooled the results for these very disparate groups
of rats in their reporting of relative organ weights. They reported 1.26- and 1.21-fold of control
in male and female rat percent relative liver/body weight after 30 days of TCE exposure.
However, as stated above, these estimates are limited in their ability to provide a quantitative
estimate of liver weight increase due to TCE.
There were evidently differences between the groups of gerbils in response to TCE with
one group reported to have larger weight gain than control and the other 2 groups reported to not
show a difference by the authors. Of the 3 groups of gerbils, group 1 contained 12 animals per
gender but groups 2 and 3 only 4 animals per gender. As with the rat experiments, the initial
average weights for the groups varied significantly (30% in females and males). The authors
pooled the results for these very disparate groups of gerbils in their reporting of relative organ
weights as well. They reported a nearly identical increase in relative liver/body weight increase
for gerbil (1.22-fold of control value in males and 1.25-fold in females) as for the rat after
30 days of TCE exposure. However, similar caveats should be applied in the confidence in this
experimental design to determine the magnitudes of response to TCE exposure.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-74 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.2.3.4. Woolhiser et al. (2006)
An unpublished report by Woolhiser et al. (2006) was received by the U.S. EPA to fill
the "priority data needed" for the immunotoxicity of TCE as identified by the Agency for Toxic
Substances and Disease Registry and designed to satisfy U.S. EPA OPPTS 870.7800
Immunotoxicity Test Guidelines. The study was conducted on behalf of the Halogenated
Solvents Industry Alliance and has been submitted to the U.S. EPA but not published. Although
conducted as an immunotoxicity study, it does contain information regarding liver weight
increases in female Sprague Dawley (S-D) female rats exposed to 0, 100, 300, and 1,000 ppm
TCE for 6 hours/day, 5 days/week for 4 weeks. The rats were 7 weeks of age at the start of the
study. The report gives data for body weight and food weight for 16 animals per exposure group
and the mean body weights ranged between 181.8 to 185.5 g on the first day of the experiment.
Animals were weighed pre-exposure, twice during the first week, and then "at least weekly
throughout the study." All rats were immunized with a single intravenous injection of sheep red
blood cells via the tail vein at Day 25. Liver weights were taken and samples of liver retained
"should histopathological examination have been deemed necessary." But, histopathological
analysis was not conducted on the liver.
The effect on body weight gain by TCE inhalation exposure was shown by 5 days and
continued for 10 days of exposure in the 300-ppm and 1,000-ppm-exposed groups. By Day 28,
the mean body weight for the control group was reported to be 245.7 g but 234.4 g, 232.4 g, and
232.4 g for the 100-ppm, 300-ppm, and 1,000-ppm exposure groups, respectively. Food
consumption was reported to be decreased in the dayl-5 measurement period for the 300-and
1,000-ppm exposure groups and in the 5-10 day measurement period for the 100-ppm group.
Although body weight and food consumption data are available for 16 animals per
exposure group, for organ and organ/body weight summary data, the report gives information for
only 8 rats per group. The report gives individual animal data in its appendix so that the data for
the 8 animals in each group examined for organ weight changes could be examined separately.
The final body weights were reported to be 217.2, 212.4, 203.9, and 206.9 g for the control, 100-,
300-, and 1,000-ppm exposure groups containing only 8 animals. For the 8-animal exposure
groups, the mean initial body weights were 186.6, 183.7, 181.6, and 181.9 g for the control, 100-,
300-, and 1,000-ppm exposure groups. Thus, there was a difference from the initial and final
body weight values given for the groups containing 16 rats and those containing 8 rats. The
ranges of initial body weights for the eight animals were 169.8-204.3, 162.0-191.2,
169.0-201.5, and 168.2-193.7 g for the control, 100-, 300 -, and 1,000-ppm groups. Thus, the
control group began with a larger mean value and large range of values (20% difference between
highest and lowest weight rat) than the other groups.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-75 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
In terms of the percent liver/body weight ratios, an increase due to TCE exposure is
reported in female rats, although body weights were larger in the control group and the two
higher exposure groups did not gain body weight to the same extent as controls. The mean
percent liver/body weight ratios were 3.23, 3.39, 3.44, and 3.65%, respectively for the control,
100-ppm, 300-ppm, and 1,000-ppm exposure groups. This represented 1.05-, 1.07-, and
1.13-fold of control percent liver/body weight changes in the 100-, 300-, and 1,000-ppm
exposure groups. However, the small number of animals and the variation in initial animal
weight limit the ability of this study to determine statistically significant increases and the
authors report that only the 1,000-ppm group had statistically significant increased liver weight
increases.
E.2.3.5. Kjellstrand et al. (1983b)
This study examined seven strains of mice (wild, C57BL, DBA, B6CBA, A/sn, NZB, and
NMRI) after continuous inhalation exposure to 150-ppm TCE for 30 days. "Wild" mice were
reported to be composed of "three different strains: 1. Hairless (HR) from the original strain, 2.
Swiss (outbred), and 3. Furtype Black Pelage (of unknown strain)." The authors did not state the
age of the animals prior to TCE exposure but stated that weight-matched controls were exposed
to air only chambers. The authors stated that "the exposure methods" have been described
earlier (Kjellstrand et al., 1980) but only reference (Kjellstrand et al., 1981b). In both of this and
the 1981 study, animals were continuously exposed with only a few hours of cessation of
exposure noted a week for change of food and bedding. Under this paradigm, there is the
possibility of additional oral exposure to TCE due to grooming and consumption of TCE on food
in the chamber.
The study was reported to be composed of two independent experiments with the
exception of strain NMRI which had been studied in Kjellstrand et al. (1983a; 1981b). The
number of animals examined in this study ranged from 3-6 in each treatment group. The authors
reported "significant difference between the animals intended for TCE exposure and the matched
controls intended for air-exposure were seen in four cases (Table 1.)," and stated that the
grouping effects developed during the 7-day adaptation period. Premature mortality was
attributed to an accident for one TCE-exposed DBA male and fighting to the deaths of two TCE-
exposed NZB females and one B6CBA male in each air exposed chamber. Given the small
number of animals examined in this study in each group, such losses significantly decrease the
power of the study to detect TCE-induced changes. The range of initial body weights between
the groups of male mice for all strains was between 18 g (as mean value for the A/sn strain) and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-76 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
32 g (as mean value for the B6CBA strain) or -44%. For females, the range of initial body
weights between groups for all strains was 15 g (as mean value for the A/sn strain) and 24 g (as
mean value for the DBA strain) or -38%.
Rather than reporting percent liver/body weight ratios or an extrapolated value, as was
done in Kjellstrand et al. (1981b), this study only reported actual liver weights for treated and
exposed groups at the end of 30 days of exposure. The authors reported final body weight
changes in comparison to matched control groups at the end of the exposure periods but not the
changes in body weight for individual animals. They reported the results from statistical
analyses of the difference in values between TCE and air-exposed groups.
A statistically significant decrease in body weight was reported between TCE exposed
and control mice in experiment 1 of the C57BL male mice (-20% reduction in body weight due
to TCE exposure). This group also had a slight but statistically significant difference in body
weight at the beginning of exposure with the control group having a -5% difference in starting
weight. There was also a statistically significant decrease in body weight of 20% reported after
TCE exposure in one group of male B6CBA mice that did not have a difference in body weight
at the beginning of the experiment between treatment and control groups. One group of female
and both groups of male A/sn mice had statistically significant decreases in body weight after
TCE exposure (10% for the females, and 22 and 26% decreases in the two male groups) in
comparison to untreated mice of the same strain. The magnitude of body weight decrease in this
strain after TCE treatment also reflects differences in initial body weight as there were also
differences in initial body weight between the two groups of both treated and untreated A/sn
males that were statistically significant, 17 and 10% respectively. One group of male NZB mice
had a significant increase in body weight after TCE exposure of 14% compared to untreated
animals. A female group from the same strain treated with TCE was reported to have a
nonsignificant but 7% increase in final body weight in comparison to its untreated group. The
one group of male NMRI mice (n= 10) in this study was reported to have a statistically
significant 12% decrease in body weight compared to controls.
For the groups of animals with reported TCE exposure-related changes in final body
weight compared to untreated animals, such body weight changes may also have affected the
liver weights changes reported. The authors did not explicitly state that they did not record liver
and body weights specifically for each animal, and thus, would be unable to determine liver/body
weight ratios for each. However, they did state that the animals were housed 4-6 in each cage
and placed in exposure chambers together. The authors only present data for body and liver
weights as the means for a cage group in the reporting of their results. While this approach lends
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-77 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
more certainty in their measurements than the approach taken by Kjellstrand et al. (1981b) as
described above, the relative liver/body weights cannot be determined for individual animals.
It appears that the authors tried to carefully match the body weights of the control and
exposed mice at the beginning of the experiment to minimize the effects of initial body weight
differences and distinguish the effects of treatment on body weight and liver weight. However,
there was no ability to determine liver/body weight ratios and adjust for difference in initial body
weight from changes due to TCE exposure. For the groups in which there was no change in
body weight after TCE treatment and in which there was no difference in initial body weight
between controls and TCE-exposed groups, the reporting of liver weight changes due to TCE
exposure is a clearer reflection of TCE-induced effects and the magnitude of such effects.
Nevertheless the small number of animals examined in each group is still a limitation on the
ability to determine the magnitude of such responses and there statistical significance.
In wild-type mice there were no reported significant differences in the initial and final
body weight of male or female mice before or after 30 days of TCE exposure. For these groups
there was 1.76- and 1.80-fold of control values for liver weight in groups 1 and 2 for female
mice, and for males 1.84- and 1.62-fold of control values for groups 1 and 2, respectively. For
DBA mice there were no reported significant differences in the initial and final body weight of
male or female mice before or after 30 days of TCE exposure. For DBA mice there was 1.87-
and 1.88-fold of control for liver weight in groups 1 and 2 for female mice, and for males 1.45-
and 2.00-fold of control for groups 1 and 2, respectively. These groups represent the most
accurate data for TCE-induced changes in liver weight not affected by initial differences in body
weight or systemic effects of TCE, which resulted in decreased body weight gain. These results
suggest that there is more variability in TCE-induced liver weight gain between groups of male
than female mice.
The C57BL, B6CBA, NZB, and NMRI groups all had at least one group of male mice
with changes in body weight due to TCE exposure. The A/sn group not only had both male
groups with decreased body weight after TCE exposure (along with differences between exposed
and control groups at the initiation of exposure) but also a decrease in body weight in one of the
female groups. Thus, the results for TCE-induced liver weight change in these male groups also
reflected changes in body weight. These results suggest a strain-related increased sensitivity to
TCE toxicity as reflected by decreased body weight.
For C57BL mice, there was 1.65- and 1.60-fold of control for liver weight after TCE
exposure was reported in groups 1 and 2 for female mice, and for males 1.28-fold (the group
with decreased body weight) and 1.82-fold of control values for groups 1 and 2, respectively.
For B6CB A mice there was 1.70- and 1.69-fold of controls values for liver weight after TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-78 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
exposure in groups 1 and 2 for female mice, and for males 1.21-fold (the group with decreased
body weight) and 1.47-fold of control values reported for groups 1 and 2, respectively. For the
NZB mice there was 2.09-fold (n = 3) and 2.08-fold of control values for liver weight after TCE
exposure in groups 1 and 2 for female mice and for males 2.34- and 3.57-fold (the group with
increased body weight) of control values reported for groups 1 and 2, respectively. For the
NMRI mice, whose results were reported for one group with 10 mice, there was 1.66-fold of
control value for liver weight after TCE exposure for female mice and for males 1.68-fold of
control value reported (a group with decreased body weight). Finally, for the A/sn strain that had
decreased body weight in all groups but one after TCE exposure and significantly smaller body
weights in the control groups before TCE exposure in both male groups, the results still show
TCE-related liver weight increases. For the As/n mice there was 1.56- and 1.72-fold (a group
with decreased body weight) of control value for liver weight in groups 1 and 2 for female mice
and for males 1.62-fold (a group with decreased body weight) and 1.58-fold (a group with
decreased body weight) of control values reported for groups 1 and 2, respectively.
The consistency between groups of female mice of the same strain for TCE-induced liver
weight gain, regardless of strain examined, is striking. The largest difference within female
strain groups occurred in the only strain in which there was a decrease in TCE-induced body
weight. For males, even in strains that did not show TCE-related changes in body weight, there
was greater variation between groups than in females. For strains in which one group had
TCE-related changes in body weight and another did not, the group with the body weight
decrease always had a lower liver weight as well. Groups that had increased body weight after
TCE exposure also had an increased liver weight in comparison to the groups without a body
weight change. These results demonstrate the importance of carefully matching control animals
to treated animals and the importance of the effect of systemic toxicity, as measured by body
weight decreases, on the determination of the magnitude of liver weight gain induced by TCE
exposure. These results also show the increased variation in TCE-induced liver weight gain
between groups of male mice and an increase incidence of body weight changes due to TCE
exposure in comparison to females, regardless of strain.
In terms of strain sensitivity, it is important not only to take into account differing effects
on body weight changes due to TCE exposure but also to compare animals of the same age or
beginning weight as these parameters may also affect liver weight gain or toxicity induced by
TCE exposure. The authors do not state the age of the animals at the beginning of exposure and
report, as stated above, a range of initial body weights between the groups as much as 44% for
males and 38% for females. These differences can be due to strain and age. The differences in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-79 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
final body weight between the groups of controls, when all animals would have been 30 days
older and more mature, was still as much as 48% for males and 44% for females.
The data for female mice, in which body weight was decreased by TCE exposure only in
on group in one strain, suggest that the magnitude of TCE-induced liver weight increase was
correlated with body weight of the animals at the beginning of the experiment. For the C57BL
and As/n strains, female mice starting weights were averaged 17.5 and 15.5 g, respectively,
while the average liver weights were 1.63- and 1.64-fold of control after TCE exposure,
respectively. For the B6CBA, wild-type, DBA, and NZB female groups the starting body
weights averaged 22.5, 21.0, 23.0, and 21.0 g, respectively, while the average liver weight
increases were 1.70-, 1.78-, 1.88-, and 2.09-fold of control after TCE exposure. Thus, groups of
female mice with higher body weights, regardless of strain, generally had higher increases in
TCE-induced liver weight increases.
The NMRI group of female mice, did not follow this general pattern and had the highest
initial body weight for the single group of 10 mice reported (i.e., 27 g) associated with a 1.66-
fold of control value for liver weight. It is probable that the data for these mice had been
collected from another study. In fact, the starting weights reported for these groups of 10 mice
are identical to the starting weights reported for 26 mice examined in Kjellstrand et al. (1981b).
However, while this study reports a 1.66-fold of control value for liver weight after 30 days of
TCE exposure, the extrapolated percent liver/body weight given in the 1981 study for 30 days of
TCE exposure was 1.74-fold of control in female NMRI mice. In the Kjellstrand et al. (1983a)
study, discussed below, 10 female mice were reported to have a 1.66-fold of control value for
liver weight after 30 days exposure to 150-ppm TCE with an initial starting weight of 26.7 g.
Thus, these data appear to be from that study. Thus, differences in study design, variation
between experiments, and strain differences may account for the differences results reported in
Kjellstrand et al. (1983b) for NMRI mice and the other strains in regard to the relationship to
initial body weight and TCE response of liver weight gain.
These data suggest that initial body weight is a factor in the magnitude of TCE-induced
liver weight induction rather than just strain. For male mice, there appeared to be a difference
between strains in TCE-induced body weight reduction, which in turn affects liver weight. The
DBA and wild-type mice appeared to be the most resistant to this effect (with no groups
affected), while the C57BL, B6CBA, and NZB strains appearing to have at least one group
affected, and the A/sn strain having both groups of males affected. Only one group of NMRI
mice were reported in this study and that group had TCE-induced decreases in body weight.
As stated above there appeared to be much greater differences between groups of males
within the same strain in regard to liver weight increases than for females and that the increases
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-80 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
appeared to be affected by concurrent body weight changes. In general the strains and groups
within strain, that had TCE-induced body weight decreases, had the smallest increases in liver
weight, while those with no TCE-induced changes in body weight in comparison to untreated
animals (i.e., wild-type and DBA) or had an actual increase in body weight (one group of NZB
mice) had the greatest TCE-induced increase in liver weight. Therefore, only examining liver
weight in males rather than percent liver/body weight ratios would not be an accurate predictor
of strain sensitivity at this dose due to differences in initial body weight and TCE-induced body
weight changes.
E.2.3.6. Kjellstrand et al. (1983a)
This study was conducted in male and female NMRI mice with a similar design as
Kjellstrand et al. (1983b). The ages of the mice were not given by the authors. Animals were
housed 10 animals per cage and exposed from 30 to 120 days at concentrations ranging from 37
to 3,600 ppm TCE. TCE was stabilized with 0.01% thymol and 0.03% diisopropylene. Animals
were exposed continuously with exposure chambers being opened twice a week for change of
bedding food and water resulting in a drop in TCE concentration of ~1 hour. A group of mice
was exposed intermittently with TCE at night for 16 hours. This paradigm results not only in
inhalation exposure but, also, oral exposure from TCE adsorption to food and grooming
behavior. The authors state that "the different methodological aspects linked to statistical
treatment of body and organ weights have been discussed earlier (Kjellstrand et al., 1981b). The
same air-exposed control was used in three cases." The design of the experiment, in terms of
measurement of individual organ and body weights and the inability to assign a percent
liver/body weight for each animal, and limitations are similar to that of Kjellstrand et al. (1983a).
The exposure design was for groups of male and female mice to be exposed to 37-, 75-,
150-, and 300-ppm TCE continuously for 30 days (n = 10 per gender and group except for the 37
ppm exposure groups) and then for liver weight and body weight to be determined. Additional
groups of animals were exposed for 150 ppm continuously for 120 days (n = 10). Intermittent
exposure of 4 hours/day for 7 days a week were conducted for 120 days at 900 ppm and
examined immediately or 30 days after cessation of exposure (n = 10). Intermittent exposures of
16 hours/day at 255-ppm group (n = 10), 8 hours/day at 450 ppm, 4 hours/day at 900 ppm,
2 hours/day at 1,800 ppm, and 1 hour/day at 3,600 ppm 7 days/week for 30 days were also
conducted (n = 10 per group).
As in Kjellstrand et al. (1983b), body weights for individual animals were not recorded in
a way that the initial and final body weights could be compared. The approach taken by the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-81 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
authors was to match the control group at the initiation of exposure and compare control and
treated average values. At the beginning of the experiment only one group began the experiment
with a statistically significant change in body weight between treated and control animals
(female mice exposed 16 hours a day for 30 days). In regard to final body weight, which would
indicate systemic TCE toxicity, 5 groups had significantly decreased body weight (i.e., males
exposed to 150 ppm continuously for 30 or 120 days, males and females exposed continuously to
300 ppm for 30 days) and 2 groups significantly increased body weight (i.e., males exposed to
1,800 ppm for 2 hours/day and 3,600 ppm for 1 hour/day for 30 days) after TCE exposure.
Thus, the accuracy of determining the effect of TCE on liver weight changes, reported
by the authors in this study for groups in which body weight were also affected by TCE
exposure, would be affected by similar issues as for data presented by Kjellstand et al. (1983b).
In addition, comparison in results between the 37-ppm exposure groups and those of the other
groups would be affected by difference in number of animals examined (10 vs. 20). As with
Kjellstrand et al (1983b), the ages of the animals in this study are not given by the author.
Difference in initial body weight (which can be affected by age and strain) reported by
Kjellstrand et al. (1983b) appeared to be correlated with the degree of TCE-induced change in
liver weight. Although each exposed group was matched to a control group with a similar
average weight, the average initial body weights in this study varied between groups (i.e., as
much as 14% in female control, 16% in TCE-exposed female mice, 12% in male control, and
16% in male exposed mice).
For female mice exposed from 37 ppm to 300 ppm TCE continuously for 30 days, only
the 300 pm group experienced a 16% decrease in body weight between control and exposed
animals. Thus, liver weight increased reported by this study after TCE exposure were not
affected by changes in body weight for exposures below 300 ppm in female mice. Initial body
weights in the TCE-exposed female mice were similar in each of these groups (i.e., range of
29.2-31.6 g, or 8%), with the exception of the females exposed to 150 ppm TCE for 30 days
(i.e., initial body weight of 27.3 g), reducing the effects of differences in initial body weight on
TCE-induced liver weight induction. Exposure to TCE continuously for 30 days resulted in a
dose-dependent change in liver weight in female mice with 1.06-, 1.27-, 1.66-, and 2.14-fold of
control values reported for liver weight at 37 ppm, 75 ppm, 150 ppm, and 300 ppm TCE,
respectively. In females, the increase at 300 ppm was accompanied by statistically significant
decreased body weight in the TCE exposed groups compared to control (-16%). Thus, the
response in liver weight gain at that exposure is in the presence of toxicity. However, the TCE-
induced increases in liver weight consistently increased with dose of TCE in a linear fashion.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
For male mice exposed to 37 to 300 ppm TCE continuously for 30 days, both the 150-
and 300-ppm-exposed groups experienced a 10 and 18% decrease in body weight after TCE
exposure, respectively. The 37- and 75-ppm groups did not have decreased body weight due to
TCE exposure but varied by 12% in initial body weight. Thus, there are more factors affecting
reported liver weight increases from TCE exposure in the male than female mice, most
importantly toxicity. Exposure to TCE continuously for 30 days resulted in liver weights of
1.15-, 1.50-, 1.69-, and 1.90-fold of control for 37, 75, 150, and 300 ppm, respectively. The
flattening of the dose-response curve for liver weight in the male mice is consistent with the
effects of toxicity at the two highest doses, and thus, the magnitude of response at these doses
should be viewed with caution. Consistent with Kjellstrand et al. (1983b) results, male mice in
this study appeared to have a higher incidence of TCE-induced body weight changes than female
mice.
The effects of extended exposure, lower durations of exposure but at higher
concentrations, and of cessation of exposure were examined for 150 ppm and higher doses of
TCE. Mice exposed to TCE at 150 ppm continuously for 120 days were reported to have
increased liver weight (i.e., 1.57-fold of control for females and 1.49-fold of control for males),
but in the case of male mice, also to have a significant decrease in body weight of 17% in
comparison to control groups. Increasing the exposure concentration to 900-ppm TCE and
reducing exposure time to 4 hours/day for 120 days also resulted in increased liver weight (i.e.,
1.35-fold of control for females and 1.49-fold of controls for males) but with a significant
decrease in body weight in females of 7% in comparison to control groups. For mice that were
exposed to 150-ppm TCE for 30 days and then examined 120 days after the cessation of
exposure, liver weights were 1.09-fold of control for female mice and the same as controls for
male mice.
With the exception of 1,800 ppm and 3,600 ppm TCE groups exposed at 2 and 1 hour,
respectively, exposure from 225 ppm, 450 and 900 ppm at 16, 8, and 4 hours, respectively for 30
days did not result in decreased body weight in males or female mice. These exposures did
result in increased liver weights in relation to control groups and for female mice the magnitude
of increase was similar (i.e., 1.50-, 1.54-, and 1.51-fold of control for liver weight after exposure
to 225-ppm TCE 16 hours/day, 450-ppm TCE 8 hours/day, and 900-ppm TCE 4 hours/day,
respectively). For these groups, initial body weights varied by 13% in females and 14% in
males. Thus, under circumstances without body weight changes due to TCE toxicity, liver
weight appeared to have a consistent relationship with the product of duration and concentration
of exposure in female mice.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
For male mice, the increases in TCE-induced liver weight were more variable (i.e., 1.94-,
1.74-, and 1.61-fold of control for liver weight after exposure to 225-ppm TCE 16 hours/day,
450-ppm TCE 8 hours/day, and 900-ppm TCE 4 hours/day, respectively) with the product of
exposure duration and concentration did not result in a consistent response in males (e.g., a lower
dose for a longer duration of exposure resulted in a greater response than a larger dose at a
shorter duration of exposure).
Kjellstrand et al. (1983a) reported light microscopic findings from this study and report
that
after 150 ppm exposure for 30 days, the normal trabecular arrangement of the
liver cells remained. However, the liver cells were generally larger and often
displayed a fine vacuolization of the cytoplasm. The nucleoli varied slightly to
moderately in size and shape and had a finer, granular chromatin with a varying
basophilic staining intensity. The Kupffer cells of the sinusoid were increased in
cellular and nuclear size. The intralobular connective tissue was infiltrated by
inflammatory cells. There was not sign of bile stasis. Exposure to TCE in higher
or lower concentrations during the 30 days produced a similar morphologic
picture. After intermittent exposure for 30 days to a time weighted average
concentration of 150 ppm or continuous exposure for 120 days, the trabecular
cellular arrangement was less well preserved. The cells had increased in size and
the variations in size and shape of the cells were much greater. The nuclei also
displayed a greater variation in basophilic staining intensity, and often had one or
two enlarged nucleoli. Mitosis was also more frequent in the groups exposed for
longer intervals. The vacuolization of the cytoplasm was also much more
pronounced. Inflammatory cell infiltration in the interlobular connective tissue
was more prominent. After exposure to 150 ppm for 30 days, followed by 120
days of rehabilitation, the morphological picture was similar to that of the air-
exposure controls except for changes in cellular and nuclear sizes.
Although not reporting comparisons between changes in male and female mice in the results
section of the paper, the authors stated in the discussion section that "However, liver mass
increase and the changes in liver cell morphology were similar in TCE-exposed male and female
mice."
The authors do not present any quantitative data on the lesions they describe, especially
in terms of dose-response. Most of the qualitative description is for the 150-ppm exposure level,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
in which there are consistent reports of TCE induced body weight decreases in male mice. The
authors suggest that lower concentrations of TCE give a similar pathology as those at the
150-ppm level, but did not present data to support that conclusion. Although stating that Kupffer
cells were increased in cellular and nuclear size, no differential staining was applied light
microscopy sections distinguish Kupffer from endothelial cells lining the hepatic sinusoid in this
study. Without differential staining such a determination is difficult at the light microscopic
level. Indeed, Goel et al. (1992) describe proliferation of sinusoidal endothelial cells after
1,000 mg/kg/d and 2,000 mg/kg/d TCE exposure for 28 days in male Swiss mice. However, the
described inflammatory cell infiltrates in the Kjellstrand et al. (1983a) study are consistent with
invasion of macrophages and well as polymorphonuclear cells into the liver, which could
activate resident Kupffer cells.
Although not specifically describing the changes as consistent with increased
polyploidization of hepatocytes, the changes in cell size and especially the continued change in
cell size and nuclear staining characteristics after 120 days of cessation of exposure are
consistent with changes in polyploidization induced by TCE. Of note is that in the histological
description provided by the authors, although vacuolization is reported and consistent with
hepatotoxicity or lipid accumulation, which is lost during routine histological slide preparation,
there is no mention of focal necrosis or apoptosis resulting from these exposures to TCE.
E.2.3.7. Buben and O'Flaherty (1985)
This study was conducted with older mice than those generally used in chronic exposure
assays (Male Swiss-Cox outbred mice between 3 and 5 months of age) with a weight range
reported between 34 to 45 g. The mice were administered distilled TCE in corn oil by gavage
5 times a week for 6 weeks at exposure concentrations of either 0, 100, 200, 400, 800, 1,600,
2,400, or 3,200 mg TCE/kg/day. While 12-15 mice were used in most exposure groups, the
100- and 3,200-mg/kg groups contained 4-6 mice and the two control groups consisted of 24
and 26 mice. Liver toxicity was determined by "liver weight increases, decreases in liver
glucose-6-phosphate (G6P) activity, increases in liver triglycerides, and increases in serum
glutamate-pyruvate transaminase (SGPT) activity." Livers were perfused with cold saline prior
to testing for weight and enzyme activity and hepatic DNA was measured.
The authors reported the mice to tolerate the 6-week exposed with TCE with few deaths
occurring except at the highest dose and that such deaths were related to central nervous system
depression. Mice in all dose groups were reported to continue to gain weight throughout the
6-week dosing period. However, TCE exposure caused "dose-related increases in liver weight to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-85 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
body weight ratio and since body weight of mice were generally unaffected by treatment, the
increases represent true liver weight increases." Exposure concentrations, as low as
100 mg/kg/d, were reported to be "sufficient to cause statistically significant increase in the liver
weight/body weight ratio," and the increases in liver size to be "attributable to hypertrophy of the
liver cells, as revealed by histological examination and by a decrease in the DNA concentration
in the livers."
Mice in the highest dose group were reported to display liver weight/body weight ratios
that were about -75% greater than those of controls and even at the lowest dose there was a
statistically significant increase (i.e., control liver/body weight percent was reported to be
5.22% ± 0.09% vs. 5.85%) ± 0.20%> in 100 mg/kg/d exposed mice). The percent liver/body ratios
were 5.22% ± 0.09%, 5.84% ± 0.20%, 5.99% ± 0.13%, 6.51% ± 0.12%, 7.12% ± 0.12%,
8.51% ± 0.20%, 8.82% ± 0.15%, and 9.12% ± 0.15% for control (n = 24), 100 (n = 5),
200 (n = 12), 400 (n = 12), 800 (n = 12), 1,600 (n = 12), 2,400 (n = 12), and 3,200 (n = 4)
mg/kg/d TCE. This represents 1.12-, 1.15-, 1.25-, 1.36-, 1.63-, 1.69-, and 1.75-fold of control
for these doses. All dose groups of TCE induced a statistically significant increase in liver/body
weight ratios. For the 200 through 1,600 mg/kg exposure levels, the magnitudes of the increases
in TCE exposure concentrations were similar to the magnitudes of TCE-induced increases in
percent liver/body weight ratios (i.e., a ~2-fold increase in TCE dose resulted in ~1.7-fold
increase change in percent liver/body weight).
TCE exposure was reported to induce a dose-related trend towards increased triglycerides
(i.e., control values of 3.08 ± 0.29 vs. 6.89 ± 1.40 at 2,400 mg/kg TCE) with variation of
response increased with TCE exposure. For liver triglycerides the reported values in mg/g liver
were 3.08 ± 0.29 (n = 24), 3.12 ± 0.49 (n = 5), 4.41 ± 0.76 (n = 12), 4.53 ± 1.05 (n = 12),
5.76 ± 0.85 (n = 12), 5.82 ± 0.93 (n = 12), 6.89 ± 1.40 (n = 12), and 7.02 ± 0.69 (n = 4) for
control, 100, 200, 400, 800, 1,600, 2,400, and 3,200 mg/kg/d dose groups, respectively.
For G6P the values in jag phosphate/mg protein/20 minutes were 125.5 ±3,2 (n = 12),
117.8 ± 6.0 (n = 5), 116.4 ± 2.8 (n = 9), 117.3 ± 4.6 (n = 9), 111.7 ±3.3 (n = 9), 89.9 ± 1.7
(n = 9), 83.8 ±2.1 (n = 8), and 83.0 ± 7.0 (n = 3) for the same dose groups. Only the
2,400 mg/kg/d dosing group was reported to be statistically significantly increased for
triglycerides after TCE exposure although there appeared to be a dose-response. For decreases
in G6P the 800 mg/kg/d and above doses were statistically significant.
The numbers of animals varied between groups in this study but in particular only a
subset of the animals were tested for G6P with the authors providing no rationale for the
selection of animals for this assay. The differences in the number of animals per group and small
number of animals per group affected the ability to determine a statistically significant change in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
these parameters but the changes in liver weights were robust enough and variation small enough
between groups that all TCE-induced changes were described as statistically significant. The
livers of TCE treated mice, although enlarged, were reported to appear normal.
A dose-related decrease in glucose-6-phophatase activity was reported with similar small
decreases (-10%) observed in the TCE exposed groups that did not reach statistical significance
until the dose reached 800 mg/kg TCE exposure. SGPT activity was not observed to be
increased in TCE-treated mice except at the two highest doses and even at the 2,400 mg/kg dose
half of the mice had normal values. The large variability in SGPT activity was indicative of
heterogeneity of this response between mice at the higher exposure levels for this indicator of
liver toxicity. However, the results of this study also demonstrate that hepatomegaly was a
robust response that was observed at the lowest dose tested, was dose-related, and was not
accompanied by toxicity.
Liver histopathology and DNA content were determined only in control, 400, and
1,600 mg/kg TCE exposure groups. DNA content was reported to be significantly decreased
from 2.83 ± 0.17 mg/g liver in controls to 2.57 ± 0.14 in 400 mg/kg TCE treated group, and to
2.15 ± 0.08 mg/kg liver in the 1,600 mg/kg exposed group. This result was consistent with a
decreased number of nuclei per gram of liver and hepatocellular hypertrophy.
Liver degeneration was reported as swollen hepatocytes and to be common with
treatment. "Cells had indistinct borders; their cytoplasm was clumped and a vesicular pattern
was apparent. The swelling was not simply due to edema, as wet weight/dry weight ratios did
not increase." Karyorhexis (the disintegration of the nucleus) was reported to be present in
nearly all specimens and suggestive of impending cell death. A qualitative scale of negative, 1,
2, 3, or 4 was given by the authors to rate their findings without further definition or criterion
given for the ratings. "No Karyorhexis, necrosis, or polyploidy was reported in controls, but a
score of 1 for Karyorhexis was given for 400 mg/kg TCE and 2 for 1600 mg/kg TCE." Central
lobular necrosis reported to be present only at the 1,600 mg/kg TCE exposure level and as a
score of 1. "Polyploidy was also characteristic in the central lobular region" with a score of 1 for
both 400 and 1,600 mg/kg TCE. The authors reported that "hepatic cells had two or more nuclei
or had enlarged nuclei containing increased amounts of chromatin, suggesting that a regenerative
process was ongoing" and that there were no fine lipid droplets in TCE exposed animals.
The finding of "no polyploidy" in control mouse liver is unexpected given that
binucleate and polyploid hepatocytes are a common finding in the mature mouse liver. It is
possible that the authors were referring to unusually high instances of "polyploidy" in
comparison to what would be expected for the mature mouse. The score given by the authors for
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
polyploidy did not indicate a difference between the two TCE exposure treatments and that it
was of the lowest level of severity or occurrence.
No score was given for centrolobular hypertrophy although the DNA content and liver
weight changes suggested a dose response. The "Karyorhexis" described in this study could
have been a sign of cell death associated with increased liver cell number or dying of maturing
hepatocytes associated with the increased ploidy, and suggests that TCE treatment was inducing
polyploidization. Consistent with enzyme analyses, centrilobular necrosis was only seen at the
highest dose and with the lowest qualitative score, indicating that even at the highest dose there
was little toxicity.
Thus, the results of this study of TCE exposure for 6 weeks, is consistent with acute
studies and show that the region of the liver affected by TCE is the centralobular region, that
hepatocellular hypertrophy is observed in that region, and that increased liver weight is induced
at the lowest exposure level tested and much lower than those inducing overt toxicity. These
authors suggest polyploidization is occurring as a result of TCE exposure although a quantitative
dose response cannot be determined from these data.
E.2.3.8. Channel et al. (1998)
This study was performed in male hybrid B6C3Fl/CrlBR mice (13 weeks-old,
25-30 grams) and focused on indicators of oxidative stress. TCE was administered by oral
gavage 5 days a week in corn oil for up to 55 days for some groups. Although the study design
indicated that water controls, corn oil controls, and exposure levels of 400, 800, and 1,200 mg/kg
day TCE in corn oil, results were not presented for water controls for some parameters measured.
Initial body weights and those recorded during the course of the study were not reported for
individual treatment groups. Liver samples were collected on study days 2, 3, 6, 10, 14, 21, 28,
35, 42, 49, and 56. Histopathology was studied from a single section taken from the median
lobe. Thiorbarbiturate acid-reactive substances (TBARS) were determined from whole liver
homogenates. Nuclei were isolated from whole liver homogenates and DNA assayed for
8-hydroxy-2' deoxyguanosine (8-OHdG). There was no indication that parenchymal cell and
nonparenchymal cells were distinguished in the assay. Free radical electron paramagnetic
resonance (EPR) for total radicals was analyzed in whole liver homogenates. For peroxisome
detection and analysis, livers from 3 mice from the 1,200 mg/kg TCE and control (oil and water)
groups were analyzed via electron microscopy. Only centrilobular regions, the area stated by the
authors to be the primary site of peroxisome proliferation, were examined. For each animal, 7
micrographs of randomly chosen hepatocytes immediately adjacent to the central vein were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-88 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
examined with peroxisomal area to cytoplasmic area, the number of peroxisomes per unit area of
cytoplasm, and average peroxisomal size quantified. Proliferation cell nuclear antigen (PCNA),
described as a marker of cell cycle except GO, was examined in histological sections for a
minimum of 18 fields per liver section. The authors did not indicate what areas of the liver
lobule were examined for PCNA. Apoptosis was detected on liver sections using a apoptosis kit
using a single liver section from the median lobe and based on the number of positively labeled
cells per 10 mm in combination with the morphological criteria for apoptosis of
Columbano et al. (1985). However, the authors did not indicate what areas of the liver lobule
were specifically examined.
The authors reported that body weight gain was not adversely affected by TCE dosing of
the time course of the study but did not show the data. No gross lesions were reported to be
observed in any group. For TBARS no water control data was reported by the authors. Data
were presented for 6 animals per group for the corn oil control group and the 1,200 mg/kg group
(error bars representing the SE). No data were presented without corn oil so that the effects of
corn oil on the first day of the study (Day 2 of dosing) could not be determined.
After 2 and 3 days of dosing the corn oil and 1,200-mg/kg TCE groups appeared to have
similar levels of TBAR detected in whole liver as nmol TBARS/mg protein. However, by Day 6
the corn oil treated control had a decrease in TBAR that continued until Day 15 where the level
was -50% of that reported on Days 2 and 3. The variation between animals as measured by SE
was reported to be large on Day 10. By Day 20 there was a slight increase in variation that
declined by Day 35 and stayed the same through Day 55. For the TCE exposed group the
TBARs remained relatively consistent and began to decline by about Day 20 to a level that
similar to the corn oil declines by Day 35. Therefore, corn oil alone had a significant effect on
TBAR detection inducing a decline by 6 days of administration that persisted thought 55 days.
TCE administration at the 1,200 mg/kg dose in corn oil appeared to have a delayed decline in
TBARS. The authors interpreted this pattern to show that lipid peroxidation was elevated in the
1,200 mg/kg TCE group at Day 6 over corn oil. However, corn oil alone induced a decrease in
TBARs. At no time was TBARS in TCE treatment groups reported to be greater than the initial
levels at days 2 and 3, a time in which TCE and corn oil treatment groups had similar levels.
Rather than inducing increasing TBARS over the time course of the study TCE, at the
1,200 mg/kg dose, appeared to delay the corn oil induced suppression of TBARs detection.
Because the authors did not present data for aqueous control animals, the time course of TBARS
detection in the absence of corn oil, cannot be established.
For the 800 and 400 mg/kg TCE data the authors presented a figure, without standard
error information, for up to 35 days that shows little difference between 400 mg/kg TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-89 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
treatment and corn oil suppression of TBAR induction. There was little difference between the
patterns of TBAR detection for 800 and 400 mg/kg TCE, indicating that both delayed TBAR
suppression by corn oil to a similar extent and did not induce greater TBAR than corn oil alone.
For 8-OHdG levels, the authors reported that elevations were modest with the greatest
increase noted in the 1,200 mg/kg day TCE treatment group of 196% of oil controls on Day 56.
Levels fluctuated throughout the study with most of the time points that were elevated showing
129% of control for the 1,200 mg/kg/d group. Statistically significant elevations were noted on
days 2, 10, 28, 49 and 56 with depression on Day 3. On all other days (i.e., Days 6, 14, 21, 35,
and 42) the 8-OHdG values were similar to those of corn oil controls. No statistically significant
effects were reported to be observed at lower doses.
The figure presented by the authors shows the percent of controls by TCE treatment at
1,200 mg/kg/d but not the control values themselves. The pattern by corn oil is not shown and
neither is the standard error of the data. As a percent of control values the variations were very
large for many of the data points and largest for the data given at Day 55 in which the authors
report the largest difference between control and TCE treatment. There was no apparent pattern
of elevation in 8-OHdG when the data were presented in this manner. Because the data for the
corn oil control was not given, as well as no data given for aqueous controls, the effects of corn
oil alone cannot be discerned.
Given that for TBARS corn oil had a significant effect and showed a pattern of decline
after 6 days, with TCE showing a delayed decline, it is especially important to discern the effects
of corn oil and to see the pattern of the data. At time points when TBARS levels were reported
to be the same between corn oil and TCE (Days 42, 49 and 56) the pattern of 8-OHdG was quite
different with a lower level at Day 42 a slightly increased level at Day 49 and the highest
difference reported at Day 56 between corn oil control and TCE treated animals. The authors
reported that the pattern of "lipid peroxidation" to be similar between the 1,200 and 800 mg/kg
doses of TCE but for there to be no significant difference between 800 mg/kg TCE and corn oil
controls. Thus, the pattern of TBARS as a measure of lipid peroxidation and 8-OHdG level in
nuclear DNA did not match.
In regard to total free radical levels as measured by EPR, results were reported for the
1,200 mg/kg TCE as a signal that was subtracted from control values with the authors stating that
only this dose level induced an elevation significantly different from controls. Again, aqueous
control values were not presented to discern the effects of corn oil or the pattern that may have
arisen with time of corn oil administration.
The pattern of total free radical level appeared to differ from that of lipid peroxidation
and for that of 8-OHdG DNA levels with no changes at days 2, 3, a peak level at Day 6, a rapid
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-90 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
drop at Day 10, mild elevation at Day 20, and a significant decrease at Day 49. The percentage
differences between control and treated values reported at Day 6 and 20 by the authors was not
proportional to the fold-difference in signal indicating that there was not a consistent level for
control values over the time course of the experiment. While differences in lipid peroxidation
detection between 1,200 mg/kg TCE and corn oil control were greatest at Day 14, total free
radicals showed their biggest change between corn oil controls and TCE exposure on Day 6, time
points in which 8-OHdG levels were similar between TCE treatment and corn oil controls.
Again, there was no reported difference between corn oil control and the 800 mg/kg TCE
exposed group in total free radical formation but for lipid peroxidation the 800 mg/kg TCE
exposed group had a similar pattern as that of 1,200 mg/kg TCE.
Only the 1,200 mg/kg group was evaluated for peroxisomal proliferation at days 6, 10,
and 14. Thus, correlations with peroxisome proliferation and other parameters in the report at
differing times and TCE exposure concentrations could not be made. The authors reported that
there was a treatment and time effect for percent peroxisomal area, a "treatment only" effect for
number of peroxisome and no effect for peroxisomal size. They also reported that hepatocytes
examined from corn oil control rats were no different that those from water control rats for all
peroxisomal parameter, thus, discounting a vehicle effect.
However, there was an effect on peroxisomal size between corn oil control and water
with corn oil decreasing the peroxisomal size in comparison to water on all days tested. The
highest TCE-induced percent peroxisomal area and number occurred on Day 10 of the 3 time
points measured for this dose and the fold increase was -4.5- and -3.1 -fold increase,
respectively. The day-10 peak in peroxisomal area and number did not correlate with the
reported pattern of free radical or 8-OHdG generation.
For cell proliferation and apoptosis, data were given for days 2, 6, 10, 14, and 21 in a
figure. PCNA cells, a measure of cells that have undergone DNA synthesis, was elevated only
on Day 10 and only in the 1,200 mg/kg/d TCE exposed group with a mean of -60 positive nuclei
per 1,000 nuclei for 6 mice (-6%). Given that there was little difference in PCNA positive cells
at the other TCE doses or time points studied, the small number of affected cells in the liver
could not account for the increase in liver size reported in other experimental paradigms at these
doses.
The PCNA positive cells as well as "mitotic figures" were reported to be present in
centrilobular, midzonal, and periportal regions with no observed predilection for a particular
lobular distribution. No data were shown regarding any quantitative estimates of mitotic figures
and whether they correlated with PCNA results. Thus, whether the DNA synthesis phases of the
cell cycle indicated by PCNA staining were indentifying polyploidization or increased cell
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-91 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
number cannot be determined. The authors reported that there was no cytotoxicity manifested as
hepatocellular necrosis in any dose group and that there was no significant difference in
apoptosis between treatment and control groups with data not shown. The extent of apoptosis in
any of the treatment groups, or which groups and timepoints were studied for this effect cannot
be determined. No liver weight or body weight data were provided in this study.
These results confirm that as a vehicle corn oil is not neutral in its affects in the liver.
The TBARS results indicate a reduction in detection of TBARS in the liver with increasing time
of exposure to corn oil alone. Although control animals "treated with water" gavage were
studied, only the results for peroxisome proliferation were presented by the study so that the
effects of corn oil gavage were not easy to discern. In addition, the data were presented in such a
way for 8-OHdG and total free radical changes that the pattern of corn oil administration was
obscured. It is not apparent from this study that TCE exposure induces oxidative damage.
E.2.3.9. Dorfmueller et al. (1979)
The focus of this study was the evaluation of "teratogenicity and behavioral toxicity with
inhalation exposure of maternal rats" to TCE. Female Long-Evans hooded rats (n= 12) of
-210 g weight were treated with 1,800 ± 200-ppm TCE for 6 hours/day, 5 days/week, for
22 ± 6 days (until pregnancy confirmation) continuing through Day 20 of gestation. Control
animals were exposed 22 ± 3 days before pregnancy confirmation. The TCE used in this study
contained 0.2% epichlorhydrin. Body weights were monitored as well as maternal liver weight
at the end of exposure. Other than organ weight, no other observations regarding the liver were
reported in this study. The initial weights of the dams were 212 ± 39 g (mean ± SD) and
204 ± 35 g for treated and control groups, respectively. The final weights were 362 ± 32 g and
337 ± 48 g for treated and control groups, respectively. There was no indication of maternal
toxicity by body weight determinations as a result of TCE exposure in this experiment and there
was also no significant difference in absolute or relative percent liver/body weight between
control and treated female rats in this study.
E.2.3.10. Kumar et al. (2001a)
In this study, adult male Wistar rats (130 ± 10 g body weight) were exposed to
376 ± 1.76 ppm TCE ("AnalaR grade") for 8, 12, and 24 weeks for 4 hours/day 5 days/week.
The ages of the rats were not given by the authors. Each group contained 6 rats. The animals
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-92 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
were exposed in whole body chambers and thus, additional oral exposure was probable. Along
with histopathology of light microscopic sections, enzymatic activities of alkaline phosphatase
and acid phosphatase, glutamic oxoacetate transaminase, glutamic pyruvate transaminase,
reduced glutathione and "total sulphydryl" were assayed in whole liver homogenates as well as
total protein. The authors stated that "the size and weight of the liver were significantly
increased after 8, 12, and 24 weeks of TCE exposure." However, the authors did not report the
final body weight of the rats after treatment nor did they give quantitative data of liver weight
changes. In regard to histopathology, the authors stated
After 8 weeks of exposure enlarged hepatocytes, with uniform presence of fat
vacuoles were found in all of the hepatocytes affecting the periportal, midzonal,
and centrilobular areas, and fat vacuoles pushing the pyknosed nuclei to one side
of hepatocytes. Moreover congestion was not significant. After exposure of 12
and 24 weeks, the fatty changes became more progressive with marked necrosis,
uniformly distributed in the entire organ.
No other description of pathology was provided in this report. In regard to the description of
fatty change, the authors only did conventional H&E staining of sections with no precautions to
preserve or stain lipids in their sections. The authors provided a table with histological scoring
of simply + or - for minimal, mild or moderate effects and do not define the criteria for that
scoring. There was also no quantitative information given as to the extent, nature, or location of
hepatocellular necrosis. The authors reported "no change was observed in GOT and GPT levels
of liver in all the three groups. The GSH level was significantly decreased while TSH level was
significantly increased during 8, 12, and 24 weeks of TCE exposure. The acid and alkaline
phosphatases were significantly increased during 8, 12, and 24 weeks of TCE exposure." The
authors presented a series of figures that are poor in quality to demonstrate histopathological
TCE-induced changes. No mortality was observed from TCE exposure in any group despite the
presence of liver necrosis.
E.2.3 .11. Kawamoto et al. (1988b)
The focus of this study was the long-term effects of TCE treatment on induction of
metabolic enzymes in male adult Wistar rats. The authors reported that 8 rats weighing 200 g
were treated with 2.0 g/kg TCE in olive oil administered subcutaneously twice a week for
15 weeks with 7 rats serving as olive oil controls. In a separate experiment, 5 rats were injected
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
with 1.0 g/kg TCE in olive oil i.p. once a day for 5 continuous days. For comparative purposes
groups of 5 rats each were administered 3-methylcholanthrene (20 mg/kg in olive oil i.p.),
Phenobarbital (80 mg/kg in saline i.p.) for 4 days as well as ethanol was administered in drinking
water containing 10% ethanol for 14 days. Microsomes were prepared one week after the last
exposure from rats administered TCE for 15 weeks and 24 hours after the last exposure for the
other treatments.
Body weights were reported to be slightly less for the TCE treated group than for controls
with the initial weights, shown in a figure, to be similar for the first weeks of exposure. At
15 weeks there appeared to be -7.5% difference in mean body weights between control and TCE
treated rats which the authors reported to not be significantly different. Organ weights at the
termination of the experiment were reported to only be different for the liver with a 1.21-fold of
control value reported as a percentage of body weight with TCE treatment. The authors reported
their increase in liver weights in male rats from subcutaneous exposure to TCE in olive oil
(2.0 g/kg) to be consistent with the range of liver weight gain in rats reported by Kjellstrand et al.
(1981b) for 150-ppm TCE inhalation exposure (see comments on that study above). The 5-day
i.p. treatment with TCE was also reported to only produce increased liver weight but the data
were not shown and the magnitude of the percentage increase was not given by the authors. No
liver pathology results were studied or reported as well.
Along with an increase in liver weight, 15-week treatment with TCE was reported to
cause a significant increase of microsomal protein/g liver of -20% (10.64 ± 0.88 vs.
12.58 ± 0.71 mg/g liver for olive oil controls and TCE treatment, respectively). Microsomal
cytochrome P450 content was reported to show a mild increase that was not statistically
significant of 1.08-fold (1.342 ± 0.205 vs. 1.456 ±0.159 nmol/mg protein for olive oil controls
and TCE treatment, respectively) of control. However, cytochrome P450 content showed
1.28-fold of control value (14.28 ± 2.41 vs. 18.34 ± 2.31 nmol/g liver for olive oil controls and
TCE treatment, respectively) in terms of g/liver. Chronic treatment of TCE was also reported to
cause a significant increase in cytochrome b-5 level (-1.35-fold of control) and NADPH-
cytochrome c reductase activity (-1.50-fold of control) in g/liver.
The 5-day TCE treatment via the i.p. route of administration was reported to cause a
significant increase in microsomal protein (-20%), induce cytochrome P450 (—50% increase
g/liver and 22% increase in microsomal protein), but to also increase cytochrome b-5 and
NADPH-cytochrome c reductase activity by 50 and 70% in g/liver, respectively. Although
weaker, 5-day i.p. treatment with TCE induced an enzyme pattern more similar to that of
Phenobarbital and ethanol rather methylcholanthrene (i.e., increased cytochrome P450 but not
microsomal protein and NADPH-cytochrome c reductase). Direct quantitative comparisons of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-94 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
vehicle effects and potential impact on response to TCE treatments for 15 weeks subcutaneous
exposure and 5-day i.p. exposure could not be made as baseline levels of all enzyme and protein
levels changed as a function of age.
Of note is that, in the discussion section of the paper, the authors disclosed that injection
of TCE 2.0 or 3.0 g/kg i.p. for 5 days resulted in paralytic ileus from TCE exposure as
unpublished observations. They noted that the rationale for injecting TCE subcutaneously was
not only that it did not require an inhalation chamber but also guarded against peritonitis that
sometimes occurs following repeated i.p. injection. In terms of comparison with inhalation or
oral results, the authors noted that the subcutaneous treatment paradigm will result in TCE not
immediately being metabolized but retained in the fatty tissue and that after cessation of
exposure TCE metabolites continued to be excreted into the urine for more than 2 weeks.
E.2.3.12. National Toxicology Program (NTP) (1990)
E.2.3 .12.1. 13-week studies.
The NTP conducted a 13-week study of 7-week old F344/N rats (10 rats per group) that
received doses of 125 to 2,000 mg/kg (males [0, 125, 250, 500, 1,000, or 2,000 mg/kg]) and 62.5
to 1,000 mg/kg (females [0, 62.5, 125, 250, 500, or 1,000 mg/kg] TCE via corn oil gavage 5 days
per week (see Table E-l). For 7-week old B6C3Flmice (n = 10 per group), the dose levels were
reported to be 375 to 6,000 mg/kg TCE (0, 375, 750, 1,500, 3,000, or 6,000 mg/kg). Animals
were exposed via corn oil gavage to TCE that was epichlorhydrin free.
All rats were reported to survive the 13-week study, but males receiving 2,000 mg/kg
exhibited a 24% difference in final body weight. However, there was great variation in initial
weights between the dose groups with mean initial weights at the beginning of the study reported
to be 87, 88, 92, 95, 101, and 83 grams for the control, 125, 250, 500, 1,000, and 2,000 mg/kg
dose groups in male rats, respectively. This represents a 22% difference between the highest and
lowest initial weights between groups. Thus, changes in final body weight after TCE treatment
also reflect differences in starting weights between the groups that, in the case of the 500 and
1,000 mg/kg groups, would results in an lower than expected change in weight due to TCE
exposure.
For female rats, the mean initial starting weights were reported to be 81, 72, 74, 75, 73,
and 76 g, respectively for the control, 62.5, 125, 250, 500, and 1,000 mg/kg dose groups. This
represents a -13% difference between initial weights. In the case of female rats the larger mean
initial weight in the control group would tend to exaggerate the effects of TCE exposure on final
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-95 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
body weight. The authors did not report the variation in initial or final body weights within the
dose groups. At the lowest doses for male and female rats body mean weights were reported to
be decreased by 6 and 7% in male and female rats, respectively. Organ weight changes were not
reported for rats.
For male mice, mean initial body weights ranged from 19 to 22 g (-16% difference) and
for female mice ranged between 18 and 15 g (20% difference), and thus, similar to rats, the final
body weights in the groups dose with TCE reflect not only the effects of the compound but also
differences in initial weights. For male mice, the mean final body weights were reported to be 3
to 17%) less than controls for the 375 to 3,000 mg/kg dose. For female mice the percent
difference in final body weight was reported to be the same except for the 6,000 mg/kg dose
group but this lack of difference between controls and treated female mice reflected no change in
mice that started at differing weights.
Male mice started to exhibit mortality at 1,500 mg/kg with 8/10 surviving the 1,500
mg/kg dose, 3/10 surviving the 3,000 mg/kg dose, and none surviving the 6,000 mg/kg dose of
TCE until the end of the study. For females, 1 animal out of 10 died in the 750, 1,500, and 3,000
mg/kg dose groups and one surviving the 6,000 mg/kg group.
In general, the magnitude of increase in TCE exposure concentration was similar to the
magnitude of increase in percent liver/body weight for the 750 and 1,500 mg/kg TCE exposure
groups in male B6C3F1 mice and for the 750 to 3,000 mg/kg TCE exposure groups in female
mice (i.e., a 2-fold increase in TCE exposure resulted in ~2-fold increase in percent liver/body
weight).
Table E-l. Mice data for 13 weeks: mean body and liver weights
Dose (mg/kg
TCE)
Survival
Body weight
(mean in g)
Liver weight
(mean final in g)
% liver weight/BW
(fold change vs.
control)
Initial
Final
Male
0
10/10
21
36
2.1
5.8
375
10/10
20
35
1.74
5.0 (0.86)
750
10/10
21
32
2.14
6.8 (1.17)
1,500
8/10
19
29
2.27
7.6 (1.31)
3,000
3/10
20
30
2.78
8.5 (1.46)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-96 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
6,000
0/10
22
-
-
-
Female
0
10/10
18
26
1.4
5.5
375
10/10
17
26
1.31
5.0 (0.91)
750
9/10
17
26
1.55
5.8 (1.05)
1,500
9/10
17
26
1.8
6.5 (1.18)
3,000
9/10
15
26
2.06
7.8 (1.42)
6,000
1/10
15
27
2.67
9.5 (1.73)
The descriptions of pathology in rats and mice given by this study were not very detailed.
For rats only control and high dose rats were examined histologically. For mice only controls
and the two highest dose groups were examined histologically. Only mean liver weights were
reported with no statistical analyses provided to ascertain quantitative differences between study
groups.
Pathological results were reported to reveal that 6/10 males and 6/10 female rats had
pulmonary vasculitis at the highest concentration of TCE. This change was also reported to have
occurred in 1/10 control male and female rats. Most of those animals were also reported to have
had mild interstitial pneumonitis. The authors report that viral titers were positive during this
study for Sendai virus.
In mice, liver weights (both absolute and as a percent of body weight) were reported to
increase with TCE-exposure level. Liver weights were reported to have increased by more than
10% relative to controls for males receiving 750 mg/kg or more and for females receiving
1,500 mg/kg or more. The most prominent hepatic lesions detected in the mice were reported to
be centrilobular necrosis, observed in 6/10 males and 1/10 females administered 6,000 mg/kg.
Although centrilobular necrosis was not seen in either males or females
administered 3000 mg/kg, 2/10 males had multifocal areas of calcifications
scattered throughout their livers. These areas of calcification were considered to
be evidence of earlier hepatocellular necrosis. Multifocal calcification was also
seen in the liver of a single female mouse that survived the 6000 mg/kg dosage
regime. One female mouse administered 3000 mg/kg also had a hepatocellular
adenoma, an extremely rare lesion in female mice of this age (20 weeks).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-97 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
There appeared to be consistent decrease in liver weight at the lowest dose in both female and
male mice after 13 weeks of TCE exposure. Liver weight was increased at exposure
concentrations in which there was not increased mortality due to TCE exposure at 13 weeks of
TCE exposure.
E.2.3 .12.2. 2-year studies. In the 2-year phase of the NTP study, TCE was administered by
corn oil gavage to groups of 50 male and 50 female F344/N rats, and B6C3F1 mice. Dosage
levels were 500 and 1,000 mg/kg for rats and 1,000 mg/kg for mice. TCE was administered
5 times a week for 103 weeks and surviving animals were killed between weeks 103 and
107. The same number of animals receiving corn oil gavage served as controls. The
animals were 8 weeks old at the beginning of exposure. The focus of this study was to
determine if there was a carcinogenic response due to TCE exposure so there was little
reporting of non-neoplastic
pathology or toxicity. There was no report of liver weight at termination of the study, only body
weight.
The authors reported that there was no increase in necrosis in the liver from TCE
exposure in comparison to control mice. In control male mice, the incidence of hepatocellular
carcinoma (tumors with markedly abnormal cytology and architecture) was reported to be 8/48
in controls, and 31/50 in TCE-exposed male mice. For females control mice hepatocellular
carcinomas were reported in 2/48 of controls and 13/49 of TCE-exposed female mice.
Specifically, the authors described liver pathology in mice as follows:
Microscopically the hepatocellular adenomas were circumscribed areas of
distinctive hepatic parenchymal cells with a perimeter of normal appearing
parenchyma in which there were areas that appeared to be undergoing
compression from expansion of the tumor. Mitotic figures were sparse or absent
but the tumors lacked typical lobular organization. The hepatocellular
carcinomas had markedly abnormal cytology and architecture. Abnormalities in
cytology included increased cell size, decreased cell size, cytoplasmic
eosinophilia, cytoplasmic basophilia, cytoplasmic vacuolization, cytoplasmic
hyaline bodies, and variations in nuclear appearance. In many instance, several
or all of the abnormalities were present in different areas of the tumor. There
were also variations in architecture with some of the hepatocellular carcinomas
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-98 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
having areas of trabecular organization. Mitosis was variable in amount and
location.
The authors reported that the non-neoplastic lesion in male mice differing from controls was
focal necrosis in 4 versus 1 animal in the dosed group (8 vs. 2%). There was no fatty
metamorphosis in treated male mice versus 2 animals in control. In female mice there was focal
inflammation in 29 versus 19% of animals (dosed vs. control) and no other changes. Therefore,
the reported pathological results of this study did not show that the liver was showing signs of
toxicity after two years of TCE exposure except for neoplasia.
For hepatocellular adenomas the incidence was reported to be "7/48 control vs. 14/50
dosed in males and 4/48 in control vs. 16/49 dosed female mice." The administration of TCE to
mice was reported to cause increased incidences of hepatocellular carcinomas in males (control,
8/48; dosed, 31/50: p = 0.001) and in females (control 2/48; dosed 13/49; p < 0.005).
Hepatocellular carcinomas were reported to metastasize to the lungs in five dosed male mice and
one control male mouse, while none were observed in females. The incidences of hepatocellular
adenomas were reported to be increased in male mice (control 7/48; dosed 14/50) and in female
mice (control 4/48; dosed 16/49; p < 0.05).
The survival of both low and high dose male rats and dosed male mice was reported to
be less than that of vehicle controls with body weight decreases dose dependent. Female mice
body weights were comparable to controls. The authors report adjusted rates of 20.6% for
control versus 53.1% for dosed males for adenoma, 22.1% control, and 92.9% for carcinoma in
males, and liver carcinoma or adenoma adjusted rates of 100%. For female mice the adjusted
rates were reported to be 12.5% adenoma for control versus 55.6% for dosed, and 6.2% control
carcinoma versus 43.9% dosed, with liver carcinoma or adenoma adjusted rates of 18.7% for
control versus 69.1% for dosed. All of the liver results for male and female mice were reported
to be statistically significant. The administration of TCE was reported to cause earlier
expression of tumors as the first animals with carcinomas were 57 weeks for TCE-exposed
animals and 75 weeks for control male mice.
In male rats there was no reported treatment related non-neoplastic liver lesions. In
female rats a decrease in basophilic cytological change was reported to be of note in TCE treated
rats (~50%> in controls but -5% in TCE treatment groups). However, the authors reported that
"the results in male F344/N rats were considered equivocal for detecting a carcinogenic response
because both groups receiving TCE showed significantly reduced survival compared to vehicle
controls (35/70, 70%; 20/50, 40%; 16/50, 32%) and because 20% of the animals in the high-dose
group were killed accidently by gavage error." Specifically 1 male control, 3 low-dose males,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-99 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
10 high-dose males, 2 female controls, 5 low-dose females and 5 high-dose female rats were
killed by gavage error.
E.2.3.13. National Toxicology Program (NTP) (1988)
The studies described in the NTP (1988) TCE report were conducted "to compare the
sensitivities of four strains of rats to diisopropylamine-stabilized TCE." However, the authors
concluded
that because of chemically induced toxicity, reduced survival, and incomplete
documentation of experimental data, the studies are considered inadequate for
either comparing or assessing TCE-induced carcinogenesis in these strains of rats.
TCE (more than 99% pure, stabilized with 8ppm diisopropylamine) was
administered via corn oil gavage at exposure concentrations of 0, 500 or 1000
mg/kg per day, 5 days per week, for 103 weeks to 50 male and female rats of each
strain. The survival of "high-dose male Marshal rats was reduced by a large
number of accidental deaths (25 animals were accidentally killed).
However, the report stated survival was decreased at both exposure levels of TCE
because of mortality that occurred during the administration of the chemical. The number of
animals accidently killed were reported to be: 11 male ACI rats at 500 mg/kg, 18 male ACI rats
at 1,000 mg/kg, 2 vehicle control female ACI rats, 14 female ACI rats at 500 mg/kg, 12 male
ACI rats at 1,000 mg/kg, 6 vehicle control male August rats, 12 male August rats at 500 mg/kg,
11 male August rats at 1,000 mg/kg, 1 vehicle control female August rats, 6 female August rats
at 500 mg/kg, 13 male August rats at 1,000 mg/kg, 2 vehicle control male Marshal rats, 12 male
Marshal rats at 500 mg/kg, 25 male Marshal rats at 1,000 mg/kg, 3 vehicle control female
Marshal rats, 14 female Marshal rats at 500 mg/kg, 18 female Marshal rats at 1,000 mg/kg,
1 vehicle control male Osborne-Mendel rat, 6 male Osborne-Mendel rats at 500 mg/kg, 7 male
Osborne-Mendel rats at 1,000 mg/kg, 8 vehicle control female Osborne-Mendel rats, 6 female
Osborne-Mendel rats at 500 mg/kg, and 6 female Osborne-Mendel rats at 1,000 mg/kg. The ages
of the rats "when placed on the study" were reported to differ and were for ACI rats (6.5 weeks),
August rats (8 weeks), Marshal rats (7 weeks), and Osborne-Mendel rats (8 weeks). The ages of
sacrifice also varied and were 17-18 weeks for the ACI and August rats, and 110-111 weeks for
the Marshal rats.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-100 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Results from a 13-week study were briefly mentioned in the report. For the 13-week
duration of exposure, groups of 10 male ACI and August rats were administered 0,125, 250, 500,
1,000, or 2,000 mg/kg TCE in corn oil gavage. Groups of 10 female ACI and August rats were
administered 0, 62.5, 125, 250, 500, or 1,000 mg/kg TCE. Groups of 10 male Marshal rats
received 0, 268, 308, 495, 932, or 1,834 mg/kg and groups of female Marshal rats were given 0,
134, 153, 248, 466, or 918 mg/kg TCE. With the exception of 3 male August rats receiving
2,000 mg/kg TCE, all animals survived to the end of the 13-week experimental period. "The
administration of the chemical for 13 weeks was not associated with histopathological changes."
In the 2-year study the report noted that there
was no evidence of liver toxicity described as non-neoplastic changes in male
ACI rats due to TCE exposure with 4% or less incidence of any lesion in control
or treated animals. For female ACI rats, the incidence of fatty metamorphosis
was 6% in control vehicle, 9% in low dose TCE, and 13% in high dose TCE
groups. There was also a 2%, 11%, and 8% incidence of clear cell change,
respectively. A 6% incidence of hepatocytomegaly was reported in vehicle
control and 15% incidence in the high dose group.
All other descriptors had reported incidences of less than 4%.
For August rats there was also little evidence of liver toxicity. In male August rats there
was a reported incidence of 8, 4, and 10% focal necrosis in vehicle control, low dose, and high
dose, respectively. Fatty metamorphosis was reported to be 8% in control, and 2 and 4% in low
and high dose. All other descriptors were reported to be less than 4%. In female August rats, all
descriptors of pathology were reported to have a 4% or less incidence except for hepatomegaly,
which was 10% for vehicle control, 6% for the low dose and 2% for high dose TCE.
For male Marshal rats there was a reported 63% incidence of inflammation, NOS in
vehicle control, 12% in low dose and values not recorded at the high dose. There was a reported
6 and 14% incidence of fatty metamorphosis in control and low dose male rats. Clear cell
change was 8% in vehicle with all other values 4% or less. For female Marshal rats, all values
were 4% or less except for fatty metamorphosis in 6% of vehicle controls.
For male Osborne-Mendel rats, there was a reported 4, 10, and 4% incidence of focal
necrosis in vehicle control, low and high dose respectively. For "cytoplasmic change/NOS,"
there were reported incidences of 26, 32, and 27% in vehicle, low dose, and high dose animals,
respectively. All other descriptors were reported to be 4% or less. In female Osborne-Mendel
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-101 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
rats there was a reported incidence of 10% of focal necrosis at the low dose with all other
descriptors reported at 4% or less.
Obviously the negative results in this bioassay are confounded by the killing of a large
portion of the animals accidently by experimental error. Still, these large exposure
concentrations of TCE did not seem to be causing overt liver toxicity in the rat. Organ weights
were not reported in this study, which would have been hard to interpret if they had been
reported because of the mortality.
E.2.3.14. Fukuda et al. (1983)
In this 104-week bioassay designed primarily to determine a carcinogenic response,
female noninbred Crj:CD-l (ICR) mice and female Crj:CD (S-D) rats 7 weeks of age were
exposed to "reagent grade" TCE at 0, 50, 150, and 450 ppm for 7 hours a day, 5 days a week.
During the 2-year duration of the experiment inhalation concentrations were reported to be
within 2% of target values. The numbers of animals per group were reported to be 49-50 mice
and 49-51 rats at the beginning of the experiment. The impurities in the TCE were reported to
be 0.128% carbon tetrachloride benzene, 0.019% epichlorohydrin and 0.019%
1,1,2-trichloroethane. After 107 weeks from commencement of the exposure, surviving animals
were reported to be killed and completely necropsied. "Tumors and abnormal organs as well as
other major organs were excised and prepared for examination in H&E sections." No other
details of the methodologies used for pathological examination of tissues were given including
what areas of the liver and number of sections examined by light microscopy.
Body weights were not given but the authors reported that "body weight changes of the
mice and rats were normal with a normal range of standard deviation." It was also reported that
there were no significant differences in average body weight of animals at specified times during
the experiments and no significant difference in mortality between the groups of mice. The
report included a figure showing, that for the first 60 weeks of the experiment, there was a
difference in cumulative mortality at the 450 ppm dose in ICR mice and the other groups. The
authors reported that significantly increased mortalities in the control group of rats compared to
the other dosed groups were observed at 85 weeks and after 100 weeks reflecting many deaths
during the 81-85 week and 96-100 week periods for control rats. No significant comparable
clinical observations were reported to be noted in each group but that major symptoms such as
bloody nasal discharge (in rats), local alopecia (in mice and rats), hunching appearance (in mice)
and respiratory disorders (in mice and rats) were observed in some animals mostly after 1 year.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-102 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
The authors reported that "the numbers of different types of tumors were counted and
only malignant tumors were counted when both malignant and benign tumors were observed
within one organ." They also reported that "all animals were included in the effective numbers
except for a few that were killed accidently, severely autolyzed or cannibalized, and died before
the first appearance of tumors among the groups."
In mice the first tumors were observed at 286 days as thymic lymphoma and most of the
malignant tumors appearing later were described as lymphomas or lymphatic leukemias. The
incidences of mice with tumors were 37, 36, 54, and 52% in the control, 50-, 150- and 450-ppm
groups, respectively, by the end of the experiment. "Tumors of the ovary, uterus, subcutaneous
tissue, stomach, and liver were observed in the dose groups at low incidences (2-7%) but not in
the controls." For the liver, the control, 50- and 150-ppm groups were all reported to have no
liver tumors with one animal (2%) having an adenoma at the 450 ppm dose.
For rats the first tumor was reported to be observed at 410 days and for the incidences of
animals with tumors to be 64, 78, 66, and 63% for control, 50-ppm, 150-ppm, and 450-ppm
TCE, respectively, by the end of the experiment. Most tumors were distributed in the pituitary
gland and mammary gland with other tumors reported at a low incidence of 2-4% with none in
the controls. For the liver there were no liver tumors in the control or 150-ppm groups but 1
animal (2%) had a cystic cholangioma in 50-ppm group and one animal (2%) had a
hepatocellular carcinoma in the 450-ppm group of rats. No details concerning the pathology of
the liver or organ weight changes were given by the authors, including any incidences of
hepatomegaly or preneoplastic foci. Of note is that in these strains, there were no background
liver tumors in either strain, indicative of the relative insensitivity of these strains to
hepatocarcinogenicity. However, the carcinogenic potential of TCE was reflected by a number
of other tumor sites in this paradigm.
E.2.3.15. Henschler et al. (1980)
This report focused on the potential carcinogenic response of TCE in mice (NMRI
random bred), rats (WIST random bred) and hamsters (Syrian random bred) exposed to 0, 100,
and 500-ppm TCE for 6 hours/day 5 days/week for 18 months. The TCE used in the experiment
was reported to be pure with the exception of trace amounts of chlorinated hydrocarbons,
epoxides and triethanolamines (<0.000025%) w/w) and stabilized with 0.0015%) triethanolamine.
The number of animals in each group was 30 and the ages and initial and final body weights of
the animals were not provided in the report. For the period of exposure (8 am-2 pm), animals
were deprived of food and water. The exposure period was for 18 months with mice and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-103 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
hamsters sacrificed after 30 months and rats after 36 months. "Deceased animals" were reported
to be autopsied, spleen, liver, kidneys, lungs and heart weighed, and these organs, as well as
stomach, central nervous system, and tumorous tissues, examined in H&E sections.
Body weight gain was reported to be normal in all species with no noticeable differences
between control and exposed groups but data were not shown. However, a "clearly dose-
dependent decrease in the survival rate for both male and female mice" was reported to be
statistically significant in both sexes and concentrations of TCE with no other significant
differences reported in other species. The increase in mortality was more pronounced in male
mice, especially after 50 weeks of exposure. Hence the opportunity for tumor development was
diminished due to decreased survival in TCE treated groups.
No organ weights were provided for the study due to the design, in which a considerable
period of time occurred between the cessation of exposure and the sacrifice of the animals. Liver
weights changes due to TCE may have been diminished with time.
For the 30 autopsied male mice in the control group, 1 hepatocellular adenoma and 1
hepatocellular carcinoma was reported. Whether they occurred in the same animal cannot be
determined from the data presentation. In the 29 animals in the 100-ppm TCE exposure group, 2
hepatocellular adenomas and 1 mesenchymal liver tumor were reported but no hepatocellular
carcinomas also without a determination as whether they occurred in the same animal or not. In
the 30 animals autopsied in the 500-ppm-exposure group, no liver tumors were reported. In
female mice, of the 29 animals autopsied in the control group, 30 animals autopsied in the 100
group, and the 28 animals autopsied in the 500-ppm group, there were also no liver tumors
reported.
In both the 100- and 500-ppm-exposure groups, of male mice especially, low numbers of
animals studied, abbreviated TCE exposure duration, and lower numbers of animals surviving to
the end of the experiment, limit the power of this study to determine a treatment-related
difference in liver carcinogenicity. As discussed in Section E.2.3.2 below, the use of an
abbreviated exposure regime or study duration and low numbers of animals examined limits the
power of a study to detect a treatment-related response. The lack of any observed background
liver tumors in the female mice and a very low background level of 2 tumors in the male mice
are indicative of a low sensitivity to detect liver tumors in this paradigm, which may have
occurred either through its design, or a low sensitivity of mouse strain used for this endpoint.
However, the carcinogenic potential of TCE in mice was reflected by a number of other tumor
sites in this paradigm.
For rats and hamsters the authors reported "no dose-related accumulation of any kind of
tumor in either sex of these species." For male rats there was only 1 hepatocellular
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-104 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
adenoma reported at 100 ppm in the 30 animals autopsied and no carcinomas. For female rats
there were no liver tumors reported in control animals but, more significantly, at 100 ppm there
was 1 adenoma and 1 cholangiocarcinoma reported at 100 ppm and at 500 ppm
2 cholangioadenomas. Although not statistically significant, the occurrence of this relatively rare
biliary tumor was observed in both TCE dose groups in female rats. The difference in survival,
as reported in mice, did not affect the power to detect a response in rats, but the low numbers of
animals studied, abbreviated exposure duration and apparent low sensitivity to detect a
hepatocarcinogenic response suggest a study of low power. Nevertheless, the occurrence of
cholangioadenomas and 1 cholangiocarcinoma in female rats after TCE treatments is of concern,
especially given the relationship in origin and proximity of the bile and liver cells and the low
incidence of this tumor. For hamsters the low background rate of tumors of any kind suggests
that in this paradigm, the sensitivity for detection of this tumor is relatively low.
E.2.3.16. Maltoni et al. (1986)
The report by Maltoni et al. (1986) included a series of "systematic and integrated
experiments (BT 301, 302, 303, 304, 304bis, 305, 306 bis) started in sequence, testing TCE by
inhalation and by ingestion." The first experiment (BT 301) was begun in 1976 and the last in
1983 with this report representing the completed summary of the findings and results of project.
The focus of the study was detection of a neoplastic response with only a generalized description
of tumor pathology phenotype given and no reporting of liver weight changes induced by TCE
exposure.
In experiment BT 301, TCE was administered in male and female S-D rats (13 weeks at
start of experiment) via olive oil gavage at control, 50 mg/kg or 250 mg/kg exposure levels for
52 weeks (4-5 days weekly). The animals (30 male, 30 female for each dose group) were
examined during their lifetime. In experiment BT 302, male and female S-D rats (13 weeks old
at start of the experiment) were exposed to TCE via inhalation at 0, 100, and 600 ppm, 7 hours a
day, 5 days a week, for 8 weeks. The animals (90 animals in each control group, 60 animals in
each 100-ppm group, and 72 animals in each 600-ppm group) were examined during their
lifetime. In experiment BT 304, male and female Sprague Dawley (S-D) rats (12 weeks old at
start of the experiment) were exposed TCE via inhalation at 0, 100, 300, and 600 ppm 7 hours a
day, 5 days a week, for 104 weeks. The animals (95 male, 100 female rats control groups, 90
animals in each 100-ppm group, 90 animals in each 300-ppm group, and 90 animals in each 600-
ppm group) were examined during their lifetime. In experiment BT304bis, male and female S-D
rats (12 weeks old at start of the experiment) were exposed to TCE via inhalation at 0, 100, 300,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-105 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
and 600 ppm for 7 hours a day, 5 days a week, for 104 weeks. The animals (40 male, 40 female
rats control groups, 40 animals in each 100-ppm group, 40 animals in each 300-ppm group, and
40 animals in each 600-ppm group) were examined during their lifetime.
In experiment BT 303, Swiss mice (11 weeks old at the start of the experiment) were
exposed to TCE via inhalation in for 8 weeks using the same exposure concentrations as for
experiment BT 302. The animals (100 animals in each control group, 60 animals in the
100-ppm-exposed group, and 72 animals in each 600-ppm group) were examined during their
lifetime. In experiment BT 305, Swiss mice (11 weeks old at the start of the experiment) were
exposed to TCE via inhalation in for 78 weeks, 7 hours a day, 5 days a week. The animals
(90 animals in each control group, 90 animals in the 100-ppm-exposed group, 90 animals in the
300-ppm group, and 90 animals in each 600-ppm group) were examined during their lifetime. In
experiment BT 306, B6C3F1 mice (from NCI source) (12 weeks old at the start of the
experiment) were exposed to TCE via inhalation in for 78 weeks, 7 hours a day, 5 days a week.
The animals (90 animals in each control group, 90 animals in the 100-ppm-exposed group,
90 animals in the 300-ppm group, and 90 animals in each 600-ppm group) were examined during
their lifetime. In experiment BT 306bis B6C3F1 mice (from Charles River Laboratory as
source) (12 weeks old at the start of the experiment) were exposed to TCE via inhalation for
78 weeks, 7 hours a day, 5 days a week. The animals (90 animals in each control group,
90 animals in the 100-ppm-exposed group, 90 animals in the 300-ppm group, and 90 animals in
each 600-ppm group) were examined during their lifetime.
In all experiments, TCE was supplied, tested, and reported by the authors of the study to
be was highly purified and epoxide free with butyl-hydroxy-toluene at 20 ppm used as a
stabilizer. Extra virgin olive oil was used as the carrier for ingestion experiments and was
reported to be free of pesticides. The authors described the treatment of the animals and running
of the facility in detail and reported that:
Animal rooms were cleaned every day and room temperature varied from 19
degrees to 22 degrees and was checked 3 times daily. Bedding was changed
every two days and cages changes and washed once weekly. The animals were
handled very gently and, therefore, were neither aggressive nor nervous.
Concentrations of TCE were checked by continuous gas-chromatographic
monitoring. Treatment was performed by the same team. In particular, the same
person carried out the gavage of the same animals. This is important, since
animals become accustomed to the same operators. The inhalation chambers
were maintained at 23 ± 2 degrees C and 50 ± 10% relative humidity. Ingestion
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-106 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
from Monday to Friday was usually performed early in the morning. The status
and behavior of the animals were examined at least three times daily and
recorded. Every two weeks the animals were submitted to an examination for the
detection of the gross changes, which were registered in the experimental records.
The animals which were found moribund at the periodical daily inspection were
isolated in order to avoid cannibalism. The animals were weight every two weeks
during treatment and then every eight weeks. Animals were kept under
observation until spontaneous death. A complete necropsy was performed.
Histological specimens were fixed in 70% ethyl alcohol. A higher number of
samples was taken when particular pathological lesions were seen. All slides
were screened by a junior pathologist and then reviewed by a senior pathologist.
The senior pathologist was the same throughout the entire project. Analysis of
variance was used for statistical evaluation of body weights. Results are
expressed as means and standard deviations. Survival time is evaluated using the
Kruskal-Wallis test. For different survival rates between groups, the incidence of
lesions is evaluated by using the Log rank test. Non-neoplastic, preneoplastic,
and neoplastic lesions were evaluated using the Chi-square of Fisher' exact test.
The effect of different doses was evaluated using the Cochran-Armitage test for
linear trends in proportions and frequencies.
The authors stated that: "Although the BT project on TCE was started in 1976 and most of the
experiments were performed from the beginning of 1979, the methodological protocol adopted
substantially met the requirements of the Good Laboratory Practices Act." Finally, it was
reported that "the experiments ran smoothly with no accidents in relation to the conduct of the
experiment and the health of the animals, apart from an excess in mortality in the male B6C3F1
mice of the experiment BT 306, due to aggressiveness and fighting among the animals." This is
in contrast to the description of the gavage studies conducted by NTP (1988, 1990) in which
gavage error resulted in significant loss of experimental animals.
Questions have been raised about the findings, experimental conditions, and experimental
paradigm of the European Ramazzini Foundation (ERF) from which the Maltoni et al. (1986)
experiments were conducted (EFSA, 2006). However, these concerns were addressed by
Caldwell et al. (2008b), who concluded that the ERF bioassay program produced credible results
that were generally consistent with those of NTP
In regards to effects of TCE exposure on survival,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-107 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
a nonsignificant excess in mortality correlated to TCE treatment was observed
only in female rats (treated by ingestion with the compound) and in male B6C3F1
mice. In B6C3F1 mice of the experiment BT 306 bis, the excess in mortality in
treated animals was higher (p < 0.05 after 40 weeks) but was not dose correlated.
No excess in mortality was observed in the other experiments.
The authors reported that "no definite effect of TCE on body weight was observed in any of the
experiments, apart from experiment BT 306 bis, in which a slight nondose correlated decrease
was found in exposed animals."
In mice, "hepatoma" was the term used by the authors of these studies to describe all
malignant tumors of hepatic cells, of different subhistotypes, and of various degrees of
malignancy. The authors reported that the hepatomas induced by exposure to TCE
may be unique or multiple, and have different sizes (usually detected grossly at
necropsy). Under microscopic examination these tumors proved to be of the
usual type observed in Swiss and B6C3F1 mice, as well as in other mouse strains,
either untreated or treated with hepatocarcinogens. They frequently have
medullary (solid), trabecular, and pleomorphic (usually anaplastic) patterns. The
hepatomas may produce distant metastases, more frequently in the lungs.
In regard to the induction of "hepatomas" by TCE exposure, the authors report that in
Swiss mice exposed to TCE by inhalation for 8 weeks (BT303), the percentage of animals with
hepatomas was 1.0% in male mice and 1.0% in female mice in the control group (n = 100 for
each gender). For animals exposed to 100 ppm TCE, the percentage in female mice was 1.7%
and male mice 5.0% (n = 60 for each gender). For animals exposed to 600 ppm TCE, the
percentage in female mice was 0% and in male mice 5.5% (n = 72 for each gender).
The relatively larger number of animals used in this bioassay, in comparison to NTP
standard assays, allows for a greater power to detect a response. It is also apparent from these
results that Swiss mice in this experimental paradigm are a "less sensitive" strain in regard to
spontaneous liver cancer induction over the lifetime of the animals. These results suggest that 8
weeks of TCE exposure via inhalation at 100 ppm or 600 ppm may have been associated with a
small increase in liver tumors in male mice in comparison to concurrent controls.
In Swiss mice exposed to TCE via inhalation for 78 weeks (BT 305), the percentage of
animals with hepatomas was reported to be 4.4% in male mice and 0% in female mice in the
control group (n = 90 for each gender). For animals exposed to 100 ppm TCE, the percentage in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-108 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
female mice was reported to be 0% and male mice 2.2% (n = 90 for each gender). For animals
exposed to 300 ppm TCE, the percentage in female mice was reported to be 0% and in male
mice 8.9% (n = 90 for each gender). For animals exposed to 600 ppm TCE, the percentage in
female mice was reported to be 1.1% and in male mice 14.4%. As with experiment BT303, there
is a consistency in the relatively low background level of hepatomas reported for Swiss mice in
this paradigm. After 78 weeks of exposure there appears to be a dose-related increase in
hepatomas in male but not female Swiss mice via inhalation exposure.
In B6C3F1 mice exposed to TCE by inhalation for 78 weeks (BT306) the percentage of
animals with hepatomas was reported to be 1.1% in male mice and 3.3% in female mice in the
control group (n = 90 for each gender). For animals exposed to 100 ppm TCE, the percentage in
female mice was reported to be 4.4% and in male mice 1.1% (n = 90 for each gender). For
animals exposed to 300 ppm TCE, the percentage in female mice was reported to be 3.3% and in
male mice 4.4% (// = 90 for each gender). For animals exposed to 600 ppm TCE, the percentage
in female mice was reported to be 10.0% and in male mice 6.7%. This was the experimental
group with excess mortality in the male group due to fighting. The excess mortality could have
affected the results. The authors reported that there was a difference in the percentage of males
bearing benign and malignant tumors that was due to early mortality among males in experiment
BT306. It is unexpected for the liver cancer incidence to be less in male mice than female mice
and not consistent with the results reported for the Swiss mice.
In B6C3F1 male mice exposed to TCE via inhalation (BT 306 bis) the percentage of
animals with hepatomas was reported to be 18.9% in male mice in the control group (n = 90).
For animals exposed to 100 ppm TCE, the percentage in male mice was reported to be 21.1%
(n = 90). For animals exposed to 300 ppm TCE, the percentage in male mice was reported to be
30.0%) (n = 90). For animals exposed to 600 ppm TCE, the percentage in male mice was
reported to be 23.3%. This experiment did not examine female mice. The authors reported a
decrease in survival in mice from this experiment that could have affected results. It is apparent
from the BT 306 and BT 306 bis experiments that the background level of liver cancer was
significantly different in male mice, although they were supposed to be of the same strain. The
finding of differences in response in animals of the same strain but from differing sources has
also been reported in other studies for other endpoints (see Section E.3.1.2, below).
The authors reported 4 liver angiosarcomas: 1 in an untreated male rat (BT 304); 1 in a
male and 1 in a female rat exposed to 600 ppm TCE for 8 weeks (experiment BT302); and 1 in a
female rat exposed to 600 ppm TCE for 104 weeks (BT 304). The authors concluded that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-109 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
the tumors observed in the treated animals cannot be considered to be correlated
to TCE treatment, but are spontaneously arising. These findings are underlined
because of the extreme rarity of this tumor in control Sprague Dawley rats,
untreated or treated with vehicle materials. The morphology of these tumors is of
the liver angiosarcoma type produced by vinyl chloride in this strain of rats.
In rats treated for 104 weeks, TCE was reported to not affect the percentages of animals
bearing benign and malignant tumor and of animals bearing malignant tumors. Moreover, it did
not affect the number of total malignant tumors per 100 animals. This study did not report a
treatment related increase in liver cancer in rats. The report only explicitly described positive
findings so it is assumed that there were no increases in "hepatomas" in rat liver associated with
TCE treatment. The authors concluded that "under the tested experimental conditions, the
evidence of TCE (without epoxide stabilizer) carcinogenicity, gives the result of TCE treatment-
related hepatomas in male Swiss and B6C3F1 mice. A borderline increased frequency of
hepatomas was also seen after 8 weeks of exposure in male Swiss mice." Thus, the increase in
liver tumors in both strains of mice exposed to TCE via inhalation reported in this study is
consistent with the gavage results from the NTP (1990) study in B6C3F1 mice, where male mice
had a higher background level and greater response from TCE exposure than females.
E.2.3.17. Maltoni et al. (1988)
This report was an abbreviated description of an earlier study (Maltoni et al., 1986)
focusing on the identification of a carcinogenic response in rats and mice by chronic TCE
exposure.
E.2.3.18. Van Duuren et al. (1979)
This study exposed male and female noninbred HA ICR Swiss mice at 6-8 weeks of age
to distilled TCE with no further descriptions of purity. Gavage feeding of TCE was once weekly
in 0.1 mL trioctanoin. Neither initial nor final body weights were reported by the authors. The
authors reported that, at the termination of the experiments or at death, animals were completely
autopsied with specimens of all abnormal-appearing tissues and organs excised for
histopathologic diagnosis. Tissues from the stomachs, livers, and kidneys were reported to be
taken routinely for the intragastric feeding experiments. Tissues were reported to be stained for
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-110 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
H&E for pathologic examination, but no further description of the lobe(s) of the liver examined
or the sections examined was provided by the authors.
Results were only reported for the no of mice with forestomach tumors exposed to 0.5
mg/mouse of TCE treatment given once a week in 0.1 mL trioctanoin. Mouse body weights
were not given so the dose in mg/kg for the mice cannot be ascertained. The protocol used in
this experiment kept the mg/mouse constant with a 1 week dosing schedule so that as the mice
increased weight with age, the dose as a function of body weight was decreased. The days on
test were reported to be 622 for 30 male and female mice.
2 male and 1 female mice were reported as having forestomach tumors. For 30 mice
treated with trioctanoin alone the number of forestomach tumors was reported to be zero. For
mice with no TCE treatment, 5 of 100 male mice were reported to have forestomach tumors and
of 8 of 60 female mice were reported to have forestomach tumors for 636 and 649 days on test.
No results for liver were presented by the authors by the intragastric route of administration
including background rates of the incidences of liver tumors or treatment results. The authors
noted that except for repeated skin applications of certain chemicals, no significant difference
between the incidence of distant tumors in treated animals compared with no-treatment and
vehicle control groups was noted. Given the uncertainties in regard to dose, the once-a week
dosing regime, the low number of animals tested with resulting low power, and the lack of
reporting of experimental results, the ability to use the results from this experiment in regard to
TCE carcinogenicity is very limited.
E.2.3.19. National Cancer Institute (NCI) (1976)
This bioassay was "initiated in 1972 according to the methods used and widely accepted
at that time" with the design of carcinogenesis bioassays having "evolved since then in some
respects and several improvements" having been developed. The most notable changes reported
in the foreward of the report are changes "pertaining to preliminary toxicity studies, numbers of
controls used, and extent of pathological examination." Industrial grade TCE was tested (99%
TCE, 0.19% 1,2,-epoxybutane, 0.04%v ethyl acetate, 0.09% epichlorhydrin, 0,02% A'-methyl
pyrrole, and 0.03% diisobutylene) with rats and mice exposed via gavage in corn oil
5 times/week for 78 weeks using 50 animals per group at 2 doses with both sexes of Osborne-
Mendel rats and B6C3F1 mice. However, for control groups only 20 of each sex and species
were used. Rats were killed after 110 weeks and mice after 90 weeks. Rats and mice were
initially 48 and 35 days of age, respectively, at the start of the experiment with control and
treated animals born within 6 days of each other. Initial weight ranges were reported for treated
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-111 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
and control animals to be 168-229 g for male rats, 130-170 g for female rats, 11-22 g for male
mice, and 11-18 g for female mice. Animals were reported to be "randomly assigned to
treatment groups so that initially the average weight in each group was approximately the same."
Mice treated with TCE were reported to be
maintained in a room housing other mice being treated with one of the following
17 compounds: 1,1,2-2-tetrachloroethane, chloroform, 3-chloropropene,
chloropicrin, 1,2-dibromochloropropane, 1,2, dibromoethane, ethylene dichloride,
1,1-diochloroethane, 3-sulfolene, idoform, methyl chloroform, 1,1,2-
trichloroethane, tetrachloroethylene, hexachloroethane, carbon disulfide,
trichlorofluoromethane, and carbon tetrachloride. Nine groups of vehicle controls
and 9 groups of untreated controls were also housed in this same room.
The authors noted that
TCE-treated rats and their controls were maintained in a room housing other rats
being treated with one of the following compounds: dibromochloropropane,
ethylene dichloride, 1,1-dichloroethane, and carbon disulfide. Four groups of
vehicle-treated controls were in the same room." Thus, there was the potential of
co-exposure to a number of other chemicals, especially for the mice, resulting
from exhalation in treated animals housed in the same room, including the control
groups, as noted by the authors. The authors also noted that "samples of ambient
air were not tested for presence of volatile materials" but state that "although the
room arrangement is not desirable as is stated in the Guidelines for Carcinogen
Bioassay in Small Rodents, there is not evidence the results would have been
different with a single compound in a room.
The initial doses of TCE for rats were reported to be 1,300 and 650 mg/kg. However,
these levels were changed based on survival and body weight data "so that the time-weighted
average doses were 549 and 1097 mg/kg for both male and female rats." For mice, the initial
doses were reported to be 1,000 and 2,000 mg/kg for males and 700 and 1,400 mg/kg for
females. The "doses were increased so that the time weighted average doses were 1169 mg/kg
and 2339 mg/kg for male mice and 869 and 1739 mg/kg for female mice."
The authors reported that signs of toxicity, including reduction in weight, were evident in
treated rats, which, along with increased mortality, "necessitated a reduction in doses during the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-112 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
test." In contrast "very little evidence of toxicity was seen in mice, so doses were increased
slightly during the study." Doses were "changed for the rats after 7 and 16 weeks of treatment,
and for the mice after 12 weeks." At 7 weeks of age, male and female rats were dosed with
650mg/kg TCE, at 14 weeks they were dosed with 750 mg/kg TCE, and at 23 weeks of age 500
mg/kg TCE. For the high exposure level, the exposure concentrations were 1,300, 1,500, and
1,000 mg/kg TCE, respectively, for the same changes in dosing concentration. For rats the
percentage of TCE in corn oil remained constant at 60%. For female mice, the TCE exposure at
the beginning of dosing was 700 mg/kg TCE (10% in corn oil) at 5 weeks of age for the "lower
dose" level. The dose was increased to 900 mg/kg day (18% in corn oil) at 17 weeks of age and
maintained until 83 weeks of age. For male mice, the TCE exposure at the beginning of dosing
was 1,000 mg/kg TCE (15% in corn oil) at 5 weeks of age for the "lower dose" level. At 11
weeks, the level of TCE remained the same but the percentage of TCE in corn oil was reduced to
10%). The dose was increased to 1,200 mg/kg day at 17 weeks of age (24% in corn oil) and
maintained until 83 weeks of age. For the "higher dose," the TCE exposure at the beginning of
dosing was 1,400 mg/kg TCE (10% in corn oil) at 5 weeks of age in female mice. At 11 weeks
of age the exposure level of TCE was kept the same but the percentage of TCE in corn oil
increased to 20%. By 17 weeks of age the exposure concentration of TCE in corn oil was
increased to 1,800 mg/kg (18% in corn oil) in female mice. For the "higher dose" in male mice,
the TCE exposure at the beginning of dosing was 2,000 mg/kg (15% in corn oil) which was
maintained at 11 weeks in regard to TCE administered but the percent of TCE corn oil was
increased to 20%. For male mice the exposure concentration was increased to 2,400 mg/kg
(24% in corn oil). For all of the mice treatment continued on a 5 days/week schedule of oral
gavage dosing throughout the timecourse of treatment (78 weeks of treatment). Thus, not only
did the total dose administered to the animals change, but the volumes of vehicle in which TCE
was administered changed throughout the experiment.
The authors stated that at 37 weeks of age, "To help assure survival until planned
termination the dosing schedule was changed for rats to a cycle of 1 week of no treatment
followed by 4 weeks of treatment." for male and female rats. Thus, the duration of exposure in
rats was also changed. All lobes of the liver were reported to be taken including the free margin
of each lobe with any nodule or mass represented in a block 10x5x3 mm cut from the liver
and fixed in a marked capsule.
Body weights (mean ± SD) were reported to be 193 ± 15.0 g (n = 20), 193 ± 15.8 g
(n = 50), and 195 ± 16.7 g (n = 50) for control, low, and high dose male rats at initiation of the
experiment. By 1 year of exposure (50 weeks), 20/20 control male rats were still alive to be
weighed, 42/50 of the low dose rats were alive and 34/50 of high dose rats were still alive. The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-113 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
body weights of those remaining were decreased by 6.2 and 17% in the low and high dose
animals in comparison with the controls. For female rats, the mean body weights were reported
to be 146 ± 11.4 g (n = 20), 144 ± 11.0 g (n = 50), and 144 ± 9.5 g (n = 50) for control, low, and
high dose female rats at initiation of the experiment. By 1 year of exposure (50 weeks),
17/20 control female rats were still alive, 28/50 low dose and 39/50 of the high dose rats were
alive. The body weights of those remaining were decreased by 25 and 30% in the low and high
dose animals in comparison with the controls.
For male mice the initial body weights were 17 ± 0.5 g (n = 20), 17 ± 2.0 g (n = 50), and
17± 1.1 g (n = 50) for control, low and high doses. By 1 year of exposure (50 weeks), 18/20
control male mice were still alive, 47/50 or the low dose, and 34/50 of the high-dose groups were
still alive. The body weights of those remaining were unchanged in comparison to controls. For
female mice the initial body weights were 14 ± 0.0 g (n = 20), 14 ± 0.6 g (n = 50), and 14 ± 0.7 g
(n = 50) for control, low and high doses. By 1 year of exposure (50 weeks), 18/20 control male
mice were still alive, 45/50 or the low dose, and 41/50 of the high-dose groups were still alive.
The body weights of those remaining were unchanged in comparison to controls.
A high proportion of rats were reported to die during the experiment with 17/20 control,
42/50 low dose, and 47/50 high dose animals dying prior to scheduled termination. For female
rats, 12/20 control, 35/48 low dose, and 37/50 high dose animals were reported to die before
scheduled termination with two low dose females reported to be missing and not counted in the
denominator for that group. The authors reported that earlier death was associated with higher
TCE dose. A decrease in the percentage of tumor-bearing animals was reported to be lower in
treated animals and attributed by the authors to be likely related to the decrease in their survival.
A high percentage of respiratory disease was reported to be observed among the rats
without any apparent difference in the type, severity, or morbidity as to sex or group. The
authors reported that "no significant toxic hepatic changes were observed" but no other details
regarding results in the liver of rats.
Carbon tetrachloride was administered to rats as a positive control. A low incidence of
both hepatocellular carcinoma and neoplastic nodule was reported to be found in both colony
controls (1/99 hepatocellular carcinoma and 0/99 neoplastic nodule in male rats and 0/98
hepatocellular carcinoma and 2/98 neoplastic nodules in female rats) and carbon-tetrachloride-
treated rats. Hepatic adenomas were included in the description of neoplastic nodules in this
study with the diagnosis of hepatocellular carcinoma to be "based on the presence of less
organized architecture and more variability in the cells comprising the neoplasms."
The authors reported that "increased mortality in treated male mice appears to be related
to the presence of liver tumors." For mice both male and female mice the incidences of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-114 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
hepatocellular carcinoma were reported to be high from TCE treatment with 1/20 in age matched
controls, 26/50 in low dose and 31/48 in high dose males. Colony controls for male mice were
reported to be 5/77 for vehicle and 5/70 for untreated mice. For females mice hepatocellular
carcinomas were reported to be observed in 0/20 age matched controls, 4/50 low dose, and
11/47 high-dose female mice. Colony controls for female mice were reported to be 1/80 for
vehicle and 2/75 for untreated mice. In male mice, hepatocellular carcinomas were reported to
be observed early in the study with the first seen at 27 weeks. Hepatocellular carcinomas were
not observed so early in low dose male or female mice.
The diagnosis of hepatocellular carcinoma was reported to be based on histologic
appearance and the presence of metastasis especially to the lung with not other lesions
significantly elevated in treated mice. The tumors were reported to be
varied from those composed of well differentiated hepatocytes in a relatively
uniform trabecular arrangement to rather anaplastic lesions in which mitotic
figures occurred in cells which varied greatly in size and tinctorial characteristics.
Many of the tumors were characterized by the formation of relatively discrete
areas of highly anaplastic cells within the tumor proper which were, in turn,
surrounded by relatively well differentiated neoplastic cells. In general, various
arrangements of the hepatocellular carcinoma occurred, as described in the
literature, including those with an orderly cord-like arrangement of neoplastic
cells, those with a pseudoglandular pattern resembling adenocarcinoma, and those
composed of sheets of highly anaplastic cells with minimal cord or gland-like
arrangement. Multiple metaplastic lesions were observed in the lung, including
several neoplasms which were differentiated and relative benign in appearance."
The authors noted that almost all mice treated with carbon tetrachloride exhibited
liver tumors and that the "neoplasms occurring in treated [sic carbon tetrachloride
treated] mice were similar in appearance to those noted in the trichloroethylene-
treated mice.
Thus, phenotypically this study reported that the liver tumors induced in mice by TCE were
heterogeneous and typical of those arising after carbon tetrachloride administration. The
descriptions of liver tumors in this study and the tendency of metastasis to the lung are similar to
the descriptions provided by Maltoni et al. (1986) for TCE-induced liver tumors in mice via
inhalation.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-115 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
In terms of noncancer pathology of the liver, 1 control male rat was reported to display
fatty metamorphosis of the liver at 102 weeks. However, for the low dose, 3 male rats were
reported to display fatty metamorphosis (90, 110, and 110 weeks), 2 rats to display cystic
inflammation (76, 110 weeks), and one rat to display general inflammation (110 weeks). At the
high dose, 6 rats were reported to display fatty metamorphosis (12, 35, 49, 52, 52, and
58 weeks), 1 rat was reported to display cytomegaly (42 weeks), 2 rats were reported to display
centrilobular degeneration (53 and 58 weeks), 1 rats to display diffuse inflammation (62 weeks),
1 rat to display congestion (Week 12), and 5 rats to display angiectasis or abnormally enlarged
blood vessels which can be manifested by hyperproliferation of endothelial cells and dilatation of
sinusoidal spaces (35, 42, 52, 54, and 65 weeks). One control female rat was reported o display
fatty metamorphosis of the liver at 110 weeks, and one control female rats to display
"inflammation" of the liver at 110 weeks. Of the TCE dosed female rats, only 1 high dose
female rat displayed fatty metaphorphosis at Week 96.
Thus, for male rats, there was liver pathology present in some rats due to TCE exposure
examined from 12 weeks to a year at their time of their premature death. For mice the liver
pathology was dominated by the presence of hepatocellular carcinoma with additional
hyperplasia noted in 2 mice of the high dose male and female groups and 1 or less mouse
exhibiting hyperplasia in the control or low-dose groups.
The authors noted that "while the absence of a similar effect in rats appears most likely
attributable to a difference in sensitivity between the Osborne-Mendel rat and B6C3F1 mouse,
the early mortality of rats due to toxicity must also be considered." They concluded that "the test
in rats is inconclusive: large numbers of rats died prior to planned termination; in addition, the
response of this rat strain to the hepatocarcinogenicity of the positive control compound, carbon
tetrachloride, appeared relatively low." Finally, the authors noted that "while the results
obtained in the present bioassay could possibly have been influenced by an impurity in the TCE
used, the extremely low amounts of impurities found make this improbable."
E.2.3.20. Herren-Freund et al. (1987)
This study gave results primarily in initiated male B6C3F1 mice that were also exposed
to TCE metabolites in drinking water for 61 weeks. However, in Table 1 of the report, results
were given for mice that received no initiator but were given 40 mg/L TCE or 2 g/L NaCl as
control. The mice were reported to be 28 days of age when placed on drinking water containing
TCE. The authors reported that concentrations of TCE fell by about '/2 at the 40 mg/L dose of
TCE during the twice a week change in drinking water solution. For control animals (n = 22)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-116 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
body weight at termination was reported to be 32.93 ± 0.54 g, liver weight 1.80 ± 0.05 g, and
percent liver/body weight 5.47% ± 0.16%. For TCE treated animals (n = 32), body weight at
termination was reported to be 35.23 ± 0.66 g, liver weight 1.97 ± 0.10 g, and percent liver/body
weight 5.57%) ± 0.24%>. Thus, hepatomegaly was not reported for this paradigm at this time of
exposure. The study reported that for 22 control animals the prevalence of adenomas was 2/22
animals (or 9%>), with the mean number of adenomas per animal to be 0.09 ± 0.06 (SEM). The
prevalence of carcinomas in the control group was reported to be 0/22. For 32 animals exposed
to 40 mg/L TCE, the prevalence of adenomas was 3/32 animals (or 9%>), with the mean number
of adenomas per animal to be 0.19 ± 0.12 (SEM). The prevalence of animals with hepatocellular
carcinomas was 3/32 animals (or 9%>) with the mean number of hepatocellular carcinomas to be
0.10 ±0.05 (SEM).
Thus, similar to the acute study of Tucker et al. (1982), significant loss of TCE is a
limitation for trying to evaluate TCE hazard in drinking water. However, despite difficulties in
establishing accurately the dose received, an increase in adenomas per animal and an increase in
the number of animals with hepatocellular carcinomas were reported to be associated with TCE
exposure after 61 weeks of exposure. Also of note is that the increase in tumors was reported
without significant increases in hepatomegaly at the end of exposure. The authors did not report
these increases in tumors as being significant but did not do a statistical test between TCE
exposed animals without initiation and control animals without initiation. The low numbers of
animal tested limits the statistical power to make such a determination. However, for
carcinomas, there was none reported in controls but 9%> of TCE-treated mice had hepatocellular
carcinomas.
E.2.3.21. Anna et al. (1994)
This report focused on presenting incidence of cancer induction after exposure to TCE or
its metabolites and included a description of results for male B6C3F1 mice (8 weeks old at the
beginning of treatment) receiving 800 mg/kg/d TCE via gavage in corn oil, 5 days/week for
76 weeks. There was very limited reporting of results other than tumor incidence. There was no
reporting of liver weights at termination of the experiment. Although the methods section of the
report gives 800 mg/kg/d as the exposure level, Table 1 in the results section reports that TCE
was administered at 1,700 mg/kg/d. This could be a typographical error in the table as a
transposition with the dose of "perc" administered to other animals in the same study. The
methods section of the report states that the authors based their dose in mice that used in the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-117 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
1990 (NTP) study. The NTP study only used al,000 mg/kg/d in mice suggesting that the table is
mislabeled and suggests that the actual dose is 800 mg/kg/d in the Anna et al. (1994) study.
All treated mice were reported to be alive after 76 weeks of treatment. For control
animals, 10 animals exposed to corn oil, and 10 untreated controls were killed in a 9-day period.
The remaining controls were killed at 96, 103, 134 weeks of treatment. Therefore, the control
group (all) contains a heterogeneous group of animals that were sacrificed from 76-134 weeks
and were not comparable to the animals sacrificed at 76 weeks.
At 76 weeks 3 of 10 the untreated and two of the 10 corn oil treated controls were
reported to have one small hepatocellular adenoma. None of the controls examined at 76 weeks
were reported to have any observed hepatocellular carcinomas. The authors reported no
cytotoxicity for TCE, corn oil, and untreated control group. At 76 weeks, 75 mice treated with
800 mg/kg/d TCE were reported to have a prevalence of 50/75 animals having adenomas with
the mean number of adenomas per animal to be 1.27 ± 0.14 (SEM). The prevalence of
carcinomas in these same animals was reported to be 30/70 with the mean number of
hepatocellular carcinomas per animal to be 0.57 ± 0.10 (SEM).
Although not comparable in terms of time till tumor observation, Corn oil control animals
examined at much later time points did not have as great a tumor response as did those exposed
to TCE. At 76-134 weeks 32 mice treated with corn oil were reported to have a prevalence of
4/32 animals having adenomas with the mean number of adenomas per animal to be 0.13 ± 0.06
(SEM). The prevalence of carcinomas in these same animals was reported to be 4/32 with the
mean number of hepatocellular carcinomas per animal to be 0.12 ± 0.06 (SEM). Despite only
examining one exposure level of TCE and the limited reporting of findings other than incidence
data, this study also reported that TCE exposure in male B6C3F1 mice to be associated with
increased induction of adenomas and hepatocellular carcinoma, without concurrent cytotoxicity.
In terms of liver tumor phenotype, Anna et al. reported the percent of H-ras codon 61
mutations in tumors from concurrent control animals (water and corn oil treatment groups
combined) examined in their study, historical controls in B6C3 Flmice, and in tumors from TCE
or DCA (0.5% in drinking water) treated animals. From their concurrent controls they reported
that H-ras codon 61 mutations in 17% (n = 6) of adenomas and 100% (n = 5) of carcinomas. For
historical controls (published and unpublished) they reported mutations in 73% (n = 33) of
adenomas and mutations in 70% (n = 30) of carcinomas. For tumors from TCE treated animals
they reported mutations in 35% (// = 40) of adenomas and 69% (n = 36) of carcinomas, while for
DCA treated animals they reported mutations in 54% (n = 24) of adenomas and in 68% (n = 40)
of carcinomas. The authors reported that "in this study, the H-ras codon 61 mutation frequency
was not statistically different in liver tumors from dichloroacetic acid and trichloroethylene-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-118 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
treated mice and combined controls (62%, 51% and 69%, respectively)." In regard to mutation
spectra in H-ras oncogenes detected B6C3F1 mouse liver "tumors," the authors reported
combined results for concurrent and historical controls of 58% AAA, 27% CGA, and 14% CTA
substitutions for CAA at Codon 61 out of 58 mutations. For TCE "tumors" the substitution
pattern was reported to be 29% AAA, 24% CGA, and 40% CTA substitutions for CAA at Codon
61 out of 39 mutations and for DCA 28% AAA, 35% CGA, and 38% CTA substitutions for
CAA at Codon 61 out of 40 mutations.
E.2.3.22. Bull et al. (2002)
This study primarily presented results from exposures to TCE, DCA, TCA and
combinations of DCA and TCA after 52 weeks of exposure with some animals examined at
87 weeks. It only examined and described results for liver. In a third experiment, 1,000 mg/kg
TCE was administered once daily 7 days a week for 79 weeks in 5% alkamuls in distilled water
to 40 B6C3F1 male mice (6 weeks old at the beginning of the experiment). At the time of
euthanasia, the livers were removed, tumors identified, and the tissues section of for examination
by a pathologist and immunostaining. Liver weights were not reported. For the TCE gavage
experiment there were 6 gavage-associated deaths during the course of this experiment among a
total of 10 animals that died with TCE treatment. No animals were lost in the control group.
The limitations of this experiment were discussed in Caldwell et al. (2008b).
Specifically, for the DCA and TCA exposed animals, the experiment was limited by low
statistical power, a relatively short duration of exposure, and uncertainty in reports of lesion
prevalence and multiplicity due to inappropriate lesions grouping (i.e., grouping of hyperplastic
nodules, adenomas, and carcinomas together as "tumors"), and incomplete histopatholology
determinations (i.e., random selection of gross lesions for histopathology examination).
For the TCE results, Bull et al. (2002) reported a high prevalence (23/36 B6C3F1 male
mice) of adenomas and hepatocellular carcinoma (7/36) and gave results of an examination of
approximately half of the lesions induced by TCE exposure. Tumor incidence data were
provided for only 15 control mice and reported as 2/15 (13%) having adenomas and 1/15 (7%)
carcinomas. Thus, this study presents results that are consistent with other studies of chronic
exposure that show TCE induction of hepatocellular carcinoma in male B6C3F1 mice.
For determinations of immunoreactivity to c-Jun as a marker of differences in "tumor"
phenotype, Bull et al. (2002) did include all lesions in most of their treatment groups, decreasing
the uncertainty of his findings. The exceptions were the absence of control lesions and inclusion
of only 16/27 and 38/72 lesions for 0.5 g/L DCA + 0.05 g/L TCA and 1 g/kg/day TCE exposure
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-119 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
groups, respectively. Immunoreactivity results were reported for the group of hyperplastic
nodules, adenomas, and carcinomas. Thus, changes in c-Jun expression between the differing
types of lesions were not determined.
Bull et al. (2002) reported lesion reactivity to c-Jun antibody to be dependent on the
proportion of the DCA and TCA administered after 52 weeks of exposure. Given alone, DCA
produced lesions in mouse liver for which approximately half displayed a diffuse
immunoreactivity to a c-Jun antibody, half did not, and none exhibited a mixture of the two.
After TCA exposure alone, no lesions were reported to be stained with this antibody. When
given in various combinations, DCA and TCA coexposure induced a few lesions that were only
c-Jun+, many that were only c-Jun-, and a number with a mixed phenotype whose frequency
increased with the dose of DCA. For TCE exposure of 79 weeks, TCE-induced lesions also had
a mixture of phenotypes (42% c-Jun+, 34% c-Jun-, and 24% mixed) and were most consistent
with those resulting from DCA and TCA coexposure but not either metabolite alone.
Mutation frequency spectra for the H-ras codon 61 in mouse liver "tumors" induced by
TCE (n = 37 tumors examined) were reported to be significantly different than that for TCA
(n = 41 tumors examined), with DCA-treated mice tumors giving an intermediate result
(n = 64 tumors examined). In this experiment, TCA-induced "tumors" were reported to have
more mutations in codon 61 (44%) than those from TCE (21%) and DCA (33%). This frequency
of mutation in the H-ras codon 61 for TCA is the opposite pattern as that observed for a number
of peroxisome proliferators in which the mutation spectra in tumors has been reported to be
much lower than spontaneously arising tumors (see Section E.3.4.1.5).
Bull et al. (2002) noted that the mutation frequency for all TCE-, TCA- or DCA- induced
tumors was lower in this experiment than for spontaneous tumors reported in other studies (they
had too few spontaneous tumors to analyze in this study), but that this study utilized lower doses
and was of shorter duration than that of Ferreira-Gonzalez et al. (1995). These are additional
concerns along with the effects of lesion grouping in which a lower stage of progression is group
with more advanced stages. In a limited subset of tumor that were both sequenced and
characterized histologically, only 8 of 34 (24%) TCE-induced adenomas but 9/15 (60%) of TCE-
induced carcinomas had mutated H-ras at codon 61, which the authors suggest is evidence that
this mutation is a late event.
The issues involving identification of MO A through tumor phenotype analysis are
discussed in detail below for the more general case of liver cancer as well as for specific
hypothesized MO As (see Sections E.3.1.4, E.3.1.8, E.3.2.1, and E.3.4.1.5). In an earlier paper,
Bull (2000) suggested that "the report by Anna et al (1994) indicated that TCE-induced tumors
possessed a different mutation spectra in codon 61 of the H-ras oncogene than those observed in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-120 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
spontaneous tumors of control mice." Bull (2000) stated that "results of this type have been
interpreted as suggesting that a chemical is acting by a mutagenic mechanism" but went on to
suggest that it is not possible to a priori rule out a role for selection in this process and that
differences in mutation frequency and spectra in this gene provide some insight into the relative
contribution of different metabolites to TCE-induced liver tumors. Bull (2000) noted that data
from Anna et al. (1994), Ferreira-Gonzalez et al. (1995), and Maronpot et al. (1995a) indicated
that mutation frequency in DCA-induced tumors did not differ significantly from that observed
in spontaneous tumors, that the mutation spectra found in DCA-induced tumors has a striking
similarity to that observed in TCE-induced tumors, and that DCA-induced tumors were
significantly different than that of TCA-induced liver tumors.
What is clear from these observations is the phenotype of TCE-induced tumors appears
to be more like DCA-induced tumors (which are consistent with spontaneous tumors), or those
resulting from a coexposure to both DCA and TCA, than from those induced by TCA. More
importantly, these data suggest that using measures other than dysplasticity and tincture indicate
that mouse liver tumors induced by TCE are heterogeneous in phenotype. The descriptions of
tumors in mice reported by the NTP and Maltoni et al studies are also consistent with phenotypic
heterogeneity as well as consistency with spontaneous tumor morphology.
E.2.4. Mode of Action: Relative Contribution of Trichloroethylene (TCE) Metabolites
Several metabolites of TCE have also been shown to induce liver cancer in rodents with
DCA and TCA having been the focus of study as potential active agent(s) of TCE liver toxicity
and/or carcinogenesis and both able to induce peroxisome proliferation (Caldwell and Keshava,
2006). A variety of DCA effects from exposure have been noted that are consistent with
conditions that increase risk of liver cancer [e.g., effects on the cytosolic enzyme glutathione
(GST)-S-transferase-zeta, diabetes, and glycogen storage disease], with the pathological changes
induced by DCA on whole liver consistent with changes observed in preneoplastic foci from a
variety of agents (Caldwell and Keshava, 2006). Chloral hydrate (CH) is one of the first
metabolites from oxidative metabolism of TCE with a large fraction of TCE metabolism
appearing to go through CH and then subsequent metabolism to TCA and trichloroethanol (Chiu
et al., 2006b). Similarities in toxicity may indicate that common downstream metabolites may
be toxicologically important, and differences may indicate the importance of other metabolic
pathways.
Although both induce liver tumors, DCA and TCA have distinctly different actions
(Caldwell and Keshava, 2006) and apparently differ in induced tumor phenotype (see discussion
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-121 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
above in Section E.2.2.8) and many studies have been conducted to try to elucidate the nature of
those differences (Caldwell et al., 2008b). Limitations of all of the available chronic studies of
TCA and most of the studies of DCA include less than lifetime exposures, varying and small
numbers of animals examined, and few exposure concentrations that were relatively high.
E.2.4.1. Acute studies of Dichloroacetic Acid (DCA)/Trichloroacetic Acid (TCA)
The studies in this section focus on studies of DCA and TCA that examine, to the extent
possible, similar endpoints using similar experimental designs as those of TCE examined above
and that give insight into proposed MO As for all three. Of note for any experiment involving
TCA, is whether exposure solutions were neutralized. Unbuffered TCA is commonly used as a
reagent to precipitate proteins so that any result from studies using unbuffered TCA could
potentially be confounded by the effects on pH.
E.2.4.1.1. Sanchez and Bull (1990). In this report TCA and DCA were administered to
male B6C3F1 mice (9 weeks of age) and male and female Swiss-Webster mice (9 weeks of
age) for up to 14 days. At 2, 4, or 14 days, mice were injected with tritiated thymidine.
Experiments were replicated at least once but results were pooled so that variation between
experiments could not be determined. B6C3F1 male mice were given DCA or TCA at 0,
0.3 g/L, 1.0 g/L, or 2.0 g/L in drinking water (n = 4 for each group for 2 and 5 days, but n =
15 for control and n = 12 for treatment groups at Day 14). Swiss-Webster mice (n = 4) at
were exposed to DCA only on Day 14 at 0,1.0 or 2.0 g/L. Mice were injected with tritiated
thymidine 2 hours prior to sacrifice. The pH of the drinking water was adjusted to 6.8-7.2
with sodium hydroxide. Concentrations of TCA and DCA were reported to be stable for a
minimum of 3 weeks.
E.2.4.1.2. Hepatocyte diameters were reported to be determined by randomly selecting 5
different high power fields (400 x) in five different sections per animals (total of 25
fields/animal with "cells in and around areas of necrosis, close to the edges of the section, or
displaying mitotic figures were not included in the cell diameter measurements." PAS
staining was reported to be done for glycogen and lipofuscin determined by
autofluorescence. Tritiated thymidine was reported to be given to the animals 2 hours
prior to sacrifice. In 2 of 3 replications of the 14-day experiment, a portion of the liver was
reported to be set aside for DNA extraction with the remaining group examined
autoradiographically for tritiated thymidine incorporation into individual hepatocytes.
Autoradiographs were also reported to be examined in the highest dose of either DCA or
TCA for the 2- and 5-day treatment groups. Autoradiographs were reported to be
analyzed in randomly selected fields (5 sections per animal in 10 different fields) for a total
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-122 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
of 50 fields/animal and reported as percentage of cells in the fields that were labeled. There
was no indication by the authors that they characterized differing zones of the liver for
preferential labeling. DNA thymidine incorporation results were not examined in the same
animals as those for individual hepatoctye incorporation and also not examined at 2- or 5-
day time periods. The only analyses reported for the Swiss-Webster mice were of hepatic
weight change and histopathology. Variations in results were reported as standard error of
the mean.
E.2.4.1.3. Liver weights were reported but not body weights so the relationship of
liver/body weight ratio could not be determined for the B6C3F1 mice. For liver weight, the
numbers of animals examined varied greatly between and within treatment groups. The
number of control animals examined were reported to be n = 4 on Day 2, n = 8 on day 5
and n = 15 on Day 14. There was also a large variation between control groups in regard to
liver weight. Control liver weights for Day 2 were reported to be 1.3 ± 0.1, Day 5 to be 1.5
± 0.05 and for Day 14 to be 1.3 ± 0.04 g. Liver weights in Day 5 control animals were much
greater than those for Day 2 and Day 14 animals and thus, the means varied by as much as
15%.
E.2.4.1.4. For DCA, there was no reported change in liver weights compared to controls
values at any exposure level of DCA after 2 days of exposure. After 5 days of exposure
there was no difference in liver weight between controls and 0.3 g/L exposed animals.
However, the animals exposed at 1.0 or 2.0 g/L DCA had identical increases in liver weight
of 1.7 ± 0.13 and 1.7 ± 0.8 g, respectively. Due to the low power of the experiment, only the
2.0 g/L DCA result was identified by the authors as significantly different from the control
value. For TCA there was a slight decrease reported between control values and the 0.3
g/L treatment group (1.2 ± 0.1 g vs. 1.3 ± 0.1 g), but the 1.0 and 2.0 g/L treatment groups
had similar slight increases over control (for 1.0 g/L liver weight was 1.5 ± 0.1 and for 2.0
g/L liver weight was 1.4 ± 0.1 g). The same pattern was apparent for the 5-day treatment
groups for TCA as for the 2-day treatment groups.
For 14 days exposure periods the number of animals studied was increased tol2 for the
TCA and DCA treatment groups. After 14 days of DCA treatment, there was a reported dose-
related increase in liver weight that was statistically significant at the two highest doses (i.e., at
0.3 g/L DCA liver weight was 1.4 ± 0.04, at 1.0 g/L DCA liver weight was 1.7 ± 0.07 g, and at
2.0 g/L DCA liver weight was 2.1 ± 0.08 g). This was 1.08-, 1.31-, and 1.62-fold of controls,
respectively. After 14 days of TCA exposure there was a dose-related increase in liver weight
that the authors reported to be statistically significant at all exposure levels (i.e., at 0.3 g/L liver
weight was 1.5 ± 0.06, at 1.0 g/L liver weight was 1.6 ± 0.07 g, and at 2.0 g/L liver weight was
1.8 ± 0.10 g). This represents 1.15-, 1.23-, and 1.38-fold of control.
The authors note that at 14 days that DCA-associated increases in hepatic liver weight
were greater than that of TCA. What is apparent from these data are that while the magnitude of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-123 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
difference between the exposures was ~6.7-fold between the lowest and highest dose, the
differences between TCA exposure groups for change in liver weight was -2.5. For DCA the
slope of the dose-response curve for liver weight increases appeared to be closer to the
magnitude of difference in exposure concentrations between the groups (i.e., a difference of 7.7-
fold between the highest and lowest dose for liver weight induction). Given that the control
animal weights varied as much as 15%, the small number of animals examined, and that body
weights were also not reported, there are limitations for making quantitative comparisons
between TCA and DCA treatments. However, after 14 days of treatment it is apparent that there
was a dose-related increase in liver weight after either DCA or TCA exposure at these exposure
levels. For male and female Swiss-Webster mice 1 g/L and 2 g/L DCA treatment (n = 4) was
reported to also induce an increase in percent liver/body weight that was similar to the magnitude
of exposure difference (see below).
Grossly, livers of B6C3F1 mice treated with DCA for 1 or 2 g/L were reported to have
"pale streaks running on the surface" and occasionally, discrete, white, round areas were also
observed on the surface of these livers. Such areas were not observed in TCA-treated or control
B6C3F1 mice. Pale streaks on the surface of the liver were not observed in Swiss-Webster mice.
Again there was no significant effect on total body or renal weights (data not shown)."
Swiss-Webster mice were reported to have "dose-related increases in hepatic weight and
hepatic/body weight ratios were observed. DCA-associated increases in relative hepatic weights
in both sexes were comparable to those in B6C3F1 mice. The authors report liver weights for
the Swiss-Webster male mice (n = 4 for each group) to be 2.1 ± 0.1 g for controls, 2.1 ± 0.1 g for
1.0 g/L DCA and 2.4 ± 0.2 g for 2.0 g/L DCA 14-day treatment groups. The percent liver/body
weights for these same groups were reported to be 6.4% ± 0.4%, 6.9% ± 0.2%, and 8.1% ± 0.3%,
respectively. For female Swiss-Webster mice (n = 4 for each group) the liver weights were
reported to be 1.1 ± 0.1 g for controls, 1.5 ± 0.1 g for 1.0 g/L DCA and 1.7 ± 0.2 g for 2.0 g/L
DCA 14-day treatment groups. The percent liver/body weights for these same groups of Swiss
mice were reported to be 4.8% ± 0.2%, 6.0% ± 0.2%, and 6.8% ± 0.4%, respectively.
Thus, while there was no significant difference in "liver weight" between the control and
the 1.0 g/L DCA treatment group for male or female Swiss-Webster mice, there was a
statistically significant difference in liver/body weight ratio reported by the authors. These data,
illustrate the importance of reporting both measures and the limitations of using small numbers
of animals (n = 4 for the Swiss Webster vs. n = 12-14 for B6C3F1 14-days experiments).
Relative liver weights were reported by the authors for male B6C3F1 mice only for the
14-day groups, as a function of calculated mean water consumption, as pooled data from the
three experiments, and as a figure that was not comparable to the data reported for Swiss-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-124 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Webster mice. The liver weight data indicate that male mice of the same age appeared to differ
in liver weight between the two strains without treatment (i.e., male B6C3F1 mice had control
liver weights at 14 days of 1.3 ± 0.04 g for 15 mice, while Swiss-Webster mice had control
values of 2.1 ± 0.1 for 4 mice). While the authors report that results were "comparable" between
the B6C3F1 mice in regard to DCA-induced changes in liver weight, the increase in percent
liver/body weight ratios were 1.27-fold of control for Swiss-Webster male mice (n = 4) and 1.42-
fold of control for female while the increase in liver weight for B6C3F1 male mice (n= 12-14)
was 1.62-fold of controls after 14 days of exposure to 2 g/L DCA.
The concentration of DNA in the liver was reported as mg hepatic DNA/g of liver. This
measurement can be associated with hepatocellular hypertrophy when decreased, or increased
cellularity (of any cell type), increased DNA synthesis, and/or increased hepatocellular ploidy in
the liver when increased. The number of animals examined for this parameter varied. For
control animals there were 4 animals reported to be examined at 2 days, 8 animals examined at
5 days, and at 14 days 8 animals were examined.
The mean DNA content in control livers were not reported to vary greatly, however, and
the variation between animals was relatively low in the 5- and 14-day control groups (i.e., 1.67
± 0.27 mg DNA/g, 1.70 ± 0.05 mg DNA/g, and 1.69 mg DNA/g, for 2-, 5-, or 14-day control
animals, respectively). For treatment groups the number of animals reported to be examined
appeared to be the same as the control animals.
For DCA treatment there did not appear to be a dose-response in hepatic DNA content
with the 1 g/L exposure level having the same reported value as control but the 0.3 g/L and 2.0
g/L values reported to be lower (mean values of 1.49 and 1.32 mg DNA/g, respectively). After 5
days of exposure, all treatment groups were reported to have a lower DNA content that the
control value (i.e., 1.44 ± 0.06 mg DNA/g, 1.47 ± mg DNA/g, and 1.30 ± 0.14 mg DNA/g, for
0.3, 1.0, and 2.0 g/L exposure levels of DCA, respectively). After 14 days of exposure, there
was a reported increase in hepatic DNA at the 0.3 g/L exposure level but significant decreases at
the 1.0 g/L and 2.0 g/L exposure levels (i.e., 1.94 ± 0.20 mg DNA/g, 1.44 ± 0.14 mg DNA/g, and
1.19 ± 0.16 mg DNA/g for the 0.3, 1.0, and 2.0 g/L exposure levels of DCA, respectively).
Changes in DNA concentration in the liver were not correlated with the pattern of liver
weight increases after DCA treatment. For example, while there was a clear dose-related
increase in liver weight after 14 days of DCA treatment, the 0.3 g/L DCA exposed group was
reported to have a higher rather than lower level of hepatic DNA than controls. After 2 or 5 days
of DCA treatment, liver weights were reported to be the same between the 1.0 and 2.0 g/L
treatment groups but hepatic DNA was reported to be decreased.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-125 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
For TCA, there appeared to be a dose-related decrease in reported hepatic DNA after
2 days of treatment (i.e., 1.63 ± 0.07 mg DNA/g, 1.53 ± 0.08 mg DNA/g, and 1. 43 ± 0.04 mg
DNA/g for the 0.3 g/L, 1.0 g/L, and 2.0 g/L exposure levels of TCA, respectively). After 5 days
of TCA exposure there was a reported decrease in hepatic DNA for all treatment groups that was
similar at the 1.0 g/L and 2.0 g/L exposure groups (i.e., 1.45 ± 0.17 mg DNA/g, 1.29 ± 0.18 mg
DNA/g, and 1. 26 ± 0.22 mg DNA/g for the 0.3 g/L, 1.0 g/L, and 2.0 g/L exposure levels of
TCA, respectively). After 14 days of TCA treatment, there was a reported decrease in all
treatment groups in hepatic DNA content that did not appear to be dose-related (i.e.,
1.31 ± 0.17 mg DNA/g, 1.21 ± 0.17 mg DNA/g, and 1. 33 ± 0.18 mg DNA/g for the 0.3 g/L,
1.0 g/L, and 2.0 g/L exposure levels of TCA, respectively).
Thus, similar to the results reported for DC A, the patterns of liver weight gain did not
match those of hepatic DNA decrease for TCA treated animals. For example, although there
appeared to be a dose-related increase in liver weight gain after 14 days of TCA exposure, there
was a treatment but not dose-related decrease in hepatic DNA content.
In regard to the ability to detect changes, the low number of animals examined after
2 days of exposure (n = 4) limited the ability to detect a significant change in liver weight and
hepatic DNA concentration. For hepatic DNA determinations, the larger number of animals
examined at 5 and 14 day time points and the similarity of values with relatively smaller standard
error of the mean reported in the control animals made quantitative differences in this parameter
easier to determine. However, animals varied in their response to treatment and this variability
exceeded that of the control groups. For DCA results reported at 14 days and those for TCA
reported at 5 and 14 days, the standard errors for treated animals showed a much greater
variability than those of the control animals (range of 0.04-0.05 mg DNA/g for control groups,
but ranges of 0.17 to 0.22 mg DNA/g for TCA at 5 days and 0.14 to 0.20 mg DNA/g for DCA or
TCA at 14 days). The authors stated that
the increases in hepatic weights were generally accompanied by decreases in the
concentration of DNA. However, the only clear changes were in animals treated
with DCA for 5 or 14 days where the ANOVAs were clearly significant (P<0.020
and 0.005, respectively). While changes of similar magnitude were observed in
other groups, the much greater variation observed in the treated groups resulted in
not significant differences by ANOVA ( p = 0.41, 0.66. 0.26, 0.15 for DCA - 2
days, and TCA for 2,5, and 14 days, respectively).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-126 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
The size of hepatocytes is heterogeneous and correlated with its ploidy, zone, and age of
the animal (see Section E. 1.1 above). The authors did not indicate if there was predominance in
zone or ploidy for hepatocytes included in their analysis of average hepatocyte diameter in the
random selection of 25 fields per animal (n = 3 to 7 animals). There appeared to be a dose-
related increase in cell diameter associated with DCA exposure and a treatment but not dose-
related increase with TCA treatment after 14 days of treatment. For control B6C3F1 male mice
(n = 7) the hepatocyte diameter was reported to be 20.6 ± 0.4 microns. For mice exposed to
DCA hepatocyte diameter was reported to be 22.2 ± 0.2, 25.2 ± 0.6, and 26.0 ±1.0 microns for
0.3 g/L, 1.0 g/L, and 2.0 g/L treated mice (n = 4 for each group), respectively. For mice exposed
to TCA hepatocyte diameter was reported to be 22.2 ± 0.2, 22.4 ± 0.6, and 23.2 ± 0.4 microns for
0.3 g/L, 1.0 g/L, and 2.0 g/L treated mice (n = 4 for the 0.3 g/L and 1.0 g/L groups and n = 3 for
the 2.0 g/L group), respectively.
The small number of animals examined limited the power of the experiment to determine
statistically significant differences with the authors reporting that only the 1.0 g/L DCA and 2.0
g/L DCA and TCA treated groups statistically significant from control values. The dose-related
increases in reported cell diameter were consistent with the dose-related increases in liver weight
reported for DCA after 14 days of exposure. However, the pattern for hepatic DNA content did
not. For TCA, the dose-related increases in cell diameter were also consistent with the dose-
related increases in liver weight after 14 days of exposure. Similar to DCA results, the changes
in hepatic DNA content did not correlate with changes in cell size. In regard to the magnitude of
increases over control values, the 68 versus 38% increase in liver weight for DCA versus TCA at
2.0 g/L, was less than the 26 and 13% increases in cell diameter for the same groups,
respectively. Therefore, for both DCA and TCA exposure there appeared to be dose-related
hepatomegaly and increased cell size after 14-days of exposure.
The authors reported PAS staining for glycogen content as an attempt to examine the
nature of increased cell size by DCA and TCA. However, they did not present any quantitative
data and only provided a brief discussion. The authors reported that
hepatic sections of DCA-treated B6C3F1 mice (1 and 2 g/L) contained very large
amounts of perilobular PAS-positive material within hepatocytes. PAS stained
hepatic sections from animals receiving the highest concentration of TCA
displayed a much less intense staining that was confined to periportal areas.
Amylase digesting confirmed the majority of the PAS-positive material to by
glycogen. Thus, increased hepatocellular size in groups receiving DCA appears
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-127 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
to be related to increased glycogen deposition. Similar increases in glycogen
deposition were observed in Swiss-Webster mice.
There is no way to discern whether DCA-induced glycogen deposition was dose-related and
therefore, correlated with increased liver weight and cell diameter. While the authors suggest
that Swiss-Webster mice displayed "similar increased in glycogen deposition" the authors did
not report a similar increase in liver weight gain after DCA exposure at 14 days (1.27-fold of
control percent liver/body weight ratio in Swiss male mice and 1.42-fold in female Swiss-
Webster mice vs. 1.62-fold of control in B6C3F1 mice after 14 days of exposure to 2 g/L DCA).
Thus, the contribution of glycogen deposition to DCA-induced hepatomegaly and the nature of
increased cell size induced by acute TCA exposure cannot be determined by this study.
However, this study does show that DCA and TCA differ in respect to their effects on glycogen
deposition after short-term exposure.
The authors report that
localized areas of coagulative necrosis were observed histologically in both
B6C3F1 and Swiss-Webster mice treated with DCA at concentrations of 1 and 2
g/L for 14 days. The necrotic areas corresponded to the pale streaked areas seen
grossly. These areas varied in size, shape and location within sections and
occupied up to several mm . An acute inflammatory response characterized by
thin rims of neutrophils was associated with the necrosis, along with multiple
mitotic figures. No such areas of necrosis were observed in animals treated at
lower concentrations of DCA, or in animals receiving the chemical for 2 or 5
days. Mice treated with 2 g/L TCA for 14 days have some necrotic areas, but at
such low frequency that it was not possible to determine if it was treatment-
related (2 lesions in a total of 20 sections examined). No necrosis was observed
in animals treated at the lower concentrations of TCA or at earlier time points.
Again there were no quantitative estimates given of the size of necrotic areas, variation between
animals, variation between strain, or dose-response of necrosis reported for DCA exposure by
the authors. The lack of necrosis after 2 and 5 days of exposure at all treatment levels and at the
lower exposure level at 14 days of exposure is not correlated with the increases in liver weight
reported for these treatment groups.
Autoradiographs of randomly chosen high powered fields (400x) (50 fields/animal) were
reported as the percentage of cells in the fields that were labeled. There was significant variation
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-128 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
in the number of animals examined and in the reported mean percent of labeled cells between
control groups. The number of control animals was not given for the 2-day group but for the
5-day and 14 day groups were reported to be n = 4 and n =11, respectively. The mean percent
of labeling in control animals was reported at 0.11 ± 0.03, 0.12 ± 0.04, and 0.46 ± 0.07% of
hepatocytes for 2-day, 5-day, and 14-day control groups, respectively. Only the 2.0 g/L
exposures of DCA and TCA were examined at all 3 times of exposure while all groups were
examined at 14 days. However, the number of animals examined in all treatment groups
appeared to be only 4 animals in each group.
There was not an increase over controls reported in the 2.0 g/L DCA or TCA 2- and 5-
day exposure groups in hepatocyte labeling with tritiated thymidine. After 14 days of exposure,
there was a statistically significant but very small dose-related increase over the control value
after DCA exposure (i.e., 0.46% ± 0.07%, 0.64% ± 0.15%, 0.75% ± 0.22%, and 0.94% ± 0.05%
labeling of hepatocytes in control, 0.3, 1.0, and 2.0 g/L DCA treatment groups, respectively).
For TCA, there was no change in hepatocyte labeling except for a 50% decrease from control
values at after 14 days of exposure to 2.0 g/L TCA (i.e., 0.46% ± 0.07%, 0.50% ± 0.14%, 0.52%
± 0.26%), and 0.26% ± 0.14% labeling of hepatocytes in control, 0.3, 1.0, and 2.0 g/L TCA
treatment groups, respectively). The authors report that
labeled cells were localized around necrotic areas in these [sic DCA treated]
groups. Since counts were made randomly, the local increased in DCA-treated
animals at concentrations of 1 and 2 g/L are in fact much higher than indicated by
the data. Labeling indices in these areas of proliferation were as high as 30%.
Labeled hepatocytes in TCA-treated and the control animals were distributed
uniformly throughout the sections. There was an apparent decrease in the
percentage of labeled cells in the group of animals treated with the highest dose of
TCA. This is because no labeled cells were found in any of the fields examined
for one animal.
The data for control mice in this experiment are consistent with others showing that the
liver is quiescent in regard to hepatocellular proliferation with few cells undergoing mitosis (see
Section E. 1.1). For up to 14 days of exposure with either DCA or TCA, there was little increase
in hepatocellular proliferation except in instances and in close proximity to areas of proliferation.
The increases in liver weight reported for this study were not correlated with and cannot be a
result of hepatocellular proliferation as only a very small population of hepatocytes is
undergoing DNA synthesis. For TCA, there was no increase in DNA synthesis in hepatocytes,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-129 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
even at the highest dose, as shown by autoradiographic data of tritiated thymidine incorporation
in random fields.
Whole liver sections were examined for tritiated thymidine incorporation from DNA
extracts. The number of animals examined varied (i.e., n = 4 for the 2-day exposure groups and
n = 8 for 5- and 14-day exposure groups) but the number of control animals examined was the
same as the treated groups for this analysis. The levels of tritiated thymidine incorporation in
"3
hepatic DNA (dpm/mg DNA expressed as mean x 10 ± SE of n animals) were reported to be
similar across control groups (i.e., 56 ± 11, 56 ± 6, and 56 ± 7 dpm/mg DNA, for 2-, 5-, and
14-day treatment groups, respectively).
After two days of DC A exposure, there appeared to be a slight treatment-related but not
dose-related increase in reported tritiated thymidine incorporation into hepatic DNA (i.e., 72 ±
23, 80 ± 6, and 68 ± 7 dpm/mg DNA for 0.3, 1.0, or 2.0 g/L DCA, respectively). After 5 days of
DCA exposure, there appeared to be a dose-related increase in reported tritiated thymidine
incorporation into hepatic DNA (i.e., 68 ± 18, 110 ± 20, and 130 ± 7 dpm/mg DNA for 0.3, 1.0,
or 2.0 g/L DCA, respectively). However, after 14 days of DCA exposure, levels of tritiated
thymidine incorporation were less than those reported at 5 days and the level for the 0.3 g/L
exposure group was less than the control value (i.e., 33 ± 11, 77 ± 9, and 81 ± 12 dpm/mg DNA
for 0.3, 1.0, or 2.0 g/L DCA, respectively).
After two days of TCA exposure there did not appear to be a treatment-related increase in
tritiated thymidine incorporation into hepatic DNA (i.e., 82 ± 16, 52 ± 7, and 54 ± 7 dpm/mg
DNA for 0.3,1.0, or 2.0 g/L TCA, respectively). Similar to the reported results for DCA, after 5
days of TCA exposure there appeared to be a dose-related increase in reported tritiated thymidine
incorporation into hepatic DNA (i.e., 79 ± 23, 86 ± 17, and 158 ± 33 dpm/mg DNA for 0.3, 1.0,
or 2.0 g/L TCA, respectively). After 14 days of TCA exposure there were treatment related
increases but not a dose-related increase in reported tritiated thymidine incorporation into hepatic
DNA (i.e., 71 ± 10, 73 ± 14, and 103 ± 14 dpm/mg DNA for 0.3, 1.0, or 2.0 g/L TCA,
respectively).
It would appear that for both TCA and DCA the increase in tritiated thymidine
incorporation into hepatic DNA was dose related and peaked after 5 days of exposure. The
authors report that the decrease in incorporation into hepatic DNA observed after 14 days of
DCA treatment at 0.3 g/L to be statistically significant as well as the increases after 5 and
14 days of TCA exposure at the 2.0 g/L level. The small numbers of animals examined, the
varying number of animals examined, and the degree of variation in treatment-related effects
limits the statistical power of this experiment to detect quantitative changes.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 0 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Given the limitations of this experiment, determination of an accurate measure of the
quantitative differences in tritiated thymidine incorporation into whole liver DNA or that
observed in hepatocytes are hard to determine. In general, the results for tritiated thymidine
incorporation into hepatic DNA were consistent with those for tritiated thymidine incorporation
into hepatocytes in that they show that there were at most a small population of hepatocytes
undergoing DNA synthesis after up to 14 days of exposure at relative high levels of exposure to
DCA and TCA (i.e., the largest percentage of hepatocytes undergoing DNA synthesis for any
treatment group was less than 1% of hepatocytes). The highest increases over control levels for
hepatic DNA incorporation for the whole liver were reported at the highest exposure level of
TCA treatment after 5 days of treatment (3-fold of control) and after 14 days of TCA treatment
(2-fold of control).
Although the authors report small areas of focal necrosis with concurrent localized
increases in hepatocyte proliferation in DCA treated animals exposed tol.O g/L and 2.0 g/L
DCA, the levels of whole liver tritiated thymidine incorporation were only slightly elevated over
controls at these concentrations, and were decreased at the 0.3 g/L exposure concentration for
which no focal necrosis was reported. The whole liver DNA incorporation of tritiated thymidine
was not consistent with the pattern of tritiated thymidine incorporation observed in individual
hepatocytes. The authors state that "at present, the mechanisms for increased tritiated thymidine
uptake in the absence of increased rates of cell replication with increasing doses of TCA cannot
be determined." The authors do not discuss the possibility that the difference in hepatocyte
labeling and whole liver DNA tritiated thymidine incorporation could have been due to the
labeling representing increased polyploidization rather than cell proliferation, as well as
increased numbers of proliferating nonparenchymal and inflammatory cells. The increased cell
size due from TCA exposure without concurrent increased glycogen deposition could have been
indicative of increased polyploidization. Finally, although both TCA- and DCA-induced
increases in liver weight were generally consistent with cell size increases, they were not
correlated with patterns of change in hepatic DNA content, incorporation of tritiated thymidine
in DNA extracts from whole liver, or incorporation of tritiated thymidine in hepatocytes. In
regard to cell size, although increased glycogen deposition with DCA exposure was noted by the
authors of this study, lack of quantitative analyses of that accumulation precludes comparison
with DCA-induced liver weight gain.
E.2.4.1.5. Nelson et al. (1988; 1989). Nelson and Bull (1988) administered TCE (0, 3.9,11.4,
22.9, and 30.4 mmol/kg) in Tween 80® via gavage to male Sprague Dawley rats and male
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-131 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
B6C3F1 mice, sacrificed them fours hours after treatment (n = 4-7), and measured the rate
of DNA unwinding under alkaline conditions. They assumed that this assay represented
increases in single-strand breaks.
E.2.4.1.6. For rats there was little change from controls up to 11.4 mmol/kg (1.5 g/kg TCE)
but a significantly increased rate of unwinding at 22.9 and 30.4 mmol/kg TCE (~2-fold
greater at 30.4 mmol). For mice there was a significantly increased level of DNA
unwinding at 11.4 and 22.9 mmol. Concentrations above 22.9 mmol/kg were reported to be
lethal to the mice.
E.2.4.1.7. In this same study, TCE metabolites were administered in unbuffered solution
using the same assay. DCA was reported to be most potent in this assay with TCA being
the lowest, while CH closely approximated the dose-response curve of TCE in the rat. In
the mouse the most potent metabolite in the assay was reported to be TCA followed by
DCA with CH considerably less potent.
The focus of the Nelson et al. (1989) study was to examine whether reported single strand
breaks in hepatic DNA induced by DCA and TCA (Nelson and Bull, 1988) were secondary to
peroxisome proliferation also reported to be induced by both. Male B6C3F1 mice (25-30 g but
no age reported) were given DCA (10 mg/kg or 500 mg/kg) or TCA (500 mg/kg) via gavage in
1% aqueous Tween 80® with no pH adjustment. The animals were reported to be sacrificed 1, 2,
4, or 8 hours after administration and livers examined for single strand breaks as a whole liver
homogenate. In a separate experiment (experiment #2) treatment was parallel to the first
(500 mg/kg treatment of DCA or TCA) but levels of PCO activity were measured as an
indication of peroxisome proliferation and expressed as [j,mol/min/g liver. In a separate
experiment (experiment #3) mice were administered 500 mg/kg DCA or TCA for 10 days with
Clofibrate administered at a dose of 250 mg/kg as a positive control. 24 hours after the last dose,
animals were killed and liver examined by light microscopy and PCO activity. Finally, in an
experiment parallel in design to experiment #3, single strand breaks were measured in total
hepatic DNA after 500 mg/kg exposure to TCA (experiment #4). Electron microscopy was
performed on 2 animals/group for vehicle, DCA or TCA treatment, with 6 randomly chosen
micrographic fields utilized for peroxisome profiles. These micrographs were analyzed without
identification as to what area of the liver lobules they were being taken from. Hence there is a
question as to whether the areas which are known to be peroxisome rich were assayed of not.
The data from all control groups were reported as pooled data in figures but statistical
comparisons were made between concurrent control and treated groups. The results for DNA
single strand breaks were reported for "13 control animals" and each experimental time point "as
at least 6 animals."
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-132 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DNA strand breaks were reported to be significantly increased over concurrent control
by a single exposure to 10 or 500 mg/kg DC A or 500 mg/kg TCA for 1, 2, or 4 hours after
administration but not at 8 or 24 hours. There did not appear to be a difference in the magnitude
of response between the 3 treatments (the fraction of unwound DNA was -2.5 times that of
control). PCO activity was reported to be not increased over control within 24 hours of either
DCA or TCA treatment. (n = 6 animals per group). The fraction of alkaline unwinding rates as
an indicator of single strand breaks were reported to not be significantly different from controls
and TCA-treated animals after 10 days of exposure (n = 5).
Relative to controls, body weights were reported to not be affected by exposures to DCA
or TCA for 10 days at 500 mg/kg (data were not shown.) (n = 6 per group). However, both DCA
and TCA were reported to significantly increase liver weight and liver/body weight ratios (i.e.,
liver weights were 1.3 ± 0.05 g, 2.1 ± 0.10 g, and 1.7 ± 0.09 g for control, 500 mg/kg DCA and
500 mg/kg TCA treatment groups, respectively while percent liver/body weights were
4.9% ± 0.14%, 7.5% ± 0.18%, and 5.7% ± 0.14% for control, 500 mg/kg DCA and 500 mg/kg
TCA treatment groups, respectively).
PCO activity ([j,mol/min/g liver) was reported to be significantly increased by DCA (500
mg/kg), TCA (500 mg/kg), and Clofibrate (250 mg/kg) treatment (i.e., levels of oxidation were
0.63 ± 0.07, 1.03 ± 0.09, 1.70 ± 0.08, and 3.26 ± 0.05 for control, 500 mg/kg DCA, 500 mg/kg
TCA and 250 mg/kg Clofibrate treatment groups, respectively). Thus, the increases were -1.63-,
2.7-, and 5-fold of control for DCA, TCA and Clofibrate treatments.
Results from randomly selected electron photomicrographs from 2 animals (6 per
animal) were reported for DCA and TCA treatment and to show an increase in peroxisomes per
unit area that was reported to be statistically significant (i.e., 9.8 ± 1.2, 25.4 ± 2.9, and 23.6 ±1.8
for control, 500 mg/kg DCA and 500 mg/kg TCA, respectively). The 2.5- and 2.4-fold of
control values for DCA and TCA gave a different pattern than that of PCO activity. The small
number of animals examined limited the power of the experiment to quantitatively determine the
magnitude of peroxisome proliferation via electron microscopy. The enzyme analyses suggested
that both DCA and TCA were weaker inducers of peroxisome proliferation that Clofibrate.
The authors reported that there was no evidence of gross hepatotoxicity in vehicle or
TCA-treated mice. Light microscopic sections from mice exposed to TCA or DCA for 10 days
were stained with H&E and PAS for glycogen. For TCA treatment, PAS staining "produced
approximately the same intensity of staining and amylase digesting revealed that the vast
majority of PAS-positive staining was glycogen." Hepatocytes were reported to be "slightly
larger in TCA-treated mice than hepatocytes from control animals throughout the liver section
with the architecture and tissue pattern of the liver intact." The histopathology after DCA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
treatment was reported to be "markedly different than that observed with either vehicle or TCA
treatments" with the "most pronounced change in the size of hepatocytes." DC A was reported to
produce marked cellular hypertrophy uniformly throughout the liver. The
hepatocytes were approximately 1.4 times larger in diameter than control liver
cells. This hypertrophy was accompanied by an increase in PAS staining;
indicating greater glycogen deposition than in TCA-treated and control liver
tissue. Multiple white streaks were grossly visible on the surface of the liver of
DCA-treated mice. The white areas corresponded with subcapsular foci of
coagulative necrosis. These localized necrotic areas were not encapsulated and
varied in size. The largest necrotic foci occupied the area of a single lobule.
These necrotic areas showed a change in staining characteristics. Often this
change consisted of increased eosinophilia. A slight inflammatory response,
characterized by neutrophil infiltration, was present. These changed were evident
in all DCA-treated mice.
The results from this experiment cannot inform as to dose-response relationships for the
parameters tested with the exception of DNA single strand breaks where 2 concentrations of
DC A were examined (10 and 500 mg/kg). For this parameter the 10 mg/kg exposure of DC A
was as effective as the 500 mg/kg dose where toxicity was observed. This effect on DNA was
also observed before evidence of induction of peroxisome proliferation. The authors did not
examine Clofibrate for effects on DNA so whether it too, would have produced this effect is
unclear. The results from this study are consistent with those of Sanchez and Bull (1990) for
induction of hepatomegaly by DC A and TCA, the lack of hepatotoxicity at this dose by TCA,
and the difference in glycogen deposition between DCA and TCA.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-134 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
E.2.4.1.8. Styles et al. (1991). In this report a similar paradigm is used as Nelson et al.
(1989) for the intention of repeating that work on single strand breakage and to study DNA
synthesis and peroxisome proliferation. In regard to the findings of single strand breaks,
Styles et al. (1991) reported for a similar paradigm of 500 mg/kg neutralized TCA
administered to male B6C3F1 mice (7-8 weeks of age) and examined at 1, 4, 8, and 24
hours after dosing. They reported no increased unwinding of DNA 1 or 24 hours after
TCA administration. In a separate experiment tritiated thymidine was administered to
mice 1 hour before sacrifice at 24, 36, 48, 72, and 96 hours after the first dose of 500 mg/kg
TCA for 3 days via gavage (n = 5 animals per group).
The hepatic DNA uptake of tritiated thymidine was reported to be similar to control
levels up to 36 hours after the first dose and then to increase to a level ~6-fold greater than
controls by 72 hours after the first dose of TCA. By 96 hours the level of tritiated thymidine
incorporation had fallen to ~4-fold greater than controls. The variation, reported by standard
deviation (SD), was very large in treated animals (e.g., SD was equal to approximately ±1.3-fold
of control for 48 hour time point). Individual hepatocytes were examined with the number of
labeled hepatocytes/1,000 cells reported for each animal.
The control level was reported to be ~1 with a SD of similar magnitude. The number of
labeled hepatocytes was reported to decrease between 24 and 36 hours and then to rise slowly
back to control levels at 48 hour and then to be significantly increased 72 hours after the first
dose of TCA (~9 cells/1,000 with a SD of 3.5) and then to decrease to a level of ~5 cells/1,000.
Thus, it appears that increases in hepatic DNA tritiated thymidine uptake preceded those of
increased labeled hepatocytes and did not capture the decrease in hepatocyte labeling at 36
hours. By either measure the population of cells undergoing DNA synthesis was small with the
peak level being less than 1% of the hepatocyte population.
The authors go on to report the zonal distribution of mean number of hepatocytes
incorporating tritiated thymidine but no variations between animals were reported. The decrease
in hepatocyte labeling at 36 hours was apparent at all zones. By 48 hours there appeared to be
slightly more perioportal than midzonal cells undergoing DNA synthesis with centrilobular cells
still below control levels. By 72 hours all zones of the liver were reported to have a similar
number of labeled cells. By 96 hours the midzonal and centrilobular regions have returned
almost to control levels while the periportal areas were still elevated. These results are consistent
with all hepatocytes showing a decrease in DNA synthesis by 36 hours and then a wave of DNA
synthesis occurring starting at the periportal zone and progressing through to the pericentral zone
until 72 hours and then the midzonal and pericentral hepatocytes completing their DNA
synthesis activity.
Peroxisome proliferation was assessed via electron photomicrographs taken in mice (4
controls and 4 treated animals) given 10 daily doses of 500 mg/kg TCA and killed 14 hours after
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
the last dose. No details were given by the authors as to methodology for peroxisome volume
estimate (e.g., how many photos per animals were examined and whether they were randomly
chosen). The mean percent cell volume occupied by peroxisome was reported to be 2.1% ±
0.386% and 3.9% ± 0.551% for control and 500 mg/kg TCA, respectively. Given there were no
time points examined before 10 days for peroxisome proliferation, correlations with DNA
synthesis activity induced by TCA cannot be made from this experiment. However, it is clear
from this study that a wave of DNA synthesis occurs throughout the liver after treatment of TCA
at this exposure concentration and that it has peaked by 72 hours even with continuous exposure
to 96 hours. Whether the DNA synthesis represents polyploidization or cell proliferation cannot
be determined from these data as neither can a dose-response.
E.2.4.1.9. Carter et al. (1995). The aim of this study was to "use correlative biochemical,
pathologic and morphometric techniques to characterize and quantify the acute, short-
term responses of hepatocytes in the male B6C3F1 mouse to drinking water containing
DCA." This report used tritiated thymidine incorporation, DNA concentration, hepatocyte
number per field (cellularity), nuclear size and binuclearity (polyploidy) parameters to
study 0, 0.5, and 5 g/L neutralized DCA exposures up to 30 days. Male B6C3F1 mice were
started on treatment at 28 days of age. Tritiated thymidine was administered by
miniosmotic pump 5 days prior to sacrifice.
E.2.4.1.10. The experiment was conduced in two phases which consisted of 5-15 days of
treatment (Phase I) and 20-30 days of treatment (Phase II) with 5 animals per group in
groups sacrificed at 5-day intervals. Liver sections were stained for H&E, PAS (for
glycogen) or methyl green pryonin stain (for RNA). DNA was extracted from liver
homogenates and the amount of tritiated thymidine determined as dpm/jig DNA.
Autoradiography was performed with the number of hepatocyte nuclei scored in
1,000 hepatocytes selected randomly to provide a labeling index of "number of labeled
cells/1000 X 100%." Changes in cellularity, nuclear size and number of multinucleate cells
were quantified in H&E sections at 40x power. Hepatocyte cellularity was determined by
counting the number of nuclei in 50 microscopic fields with multinucleate cells being
counted as one cell and nonparenchymal cells not counted. Nuclear size was also measured
in 200 nuclei with the mean area plus 2 SD was considered to be the largest possible single
nucleus. Therefore, polyploid diploid cells were identified by the authors but not cells that
had undergone polyploidy with increased DNA content in a single nucleus.
Mean body weights at the beginning of the experiment varied between 18.7 and 19.6 g in
the first 3 exposure groups of Phase I of the study. Through 15 days of exposure there did not
appear to be a change in body weight in the 0.5 g/L exposure groups but in the 5 g/L exposure
group body weight was reduced at 5, 10 and 15 days with that reduction statistically significant
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
at 5 and. 15 days. Liver weights did not appear to be increased at Day 5 but were increased at
days 10 and 15 in both treatment groups (i.e., means ± S.E.M. for Day 10; 1.36 ± 0.03,
1.46 ± 0.03, and 1.59 ± 0.08 g for control, 0.5 and 5 g/L DCA, respectively and for Day 15;
1.51 ± 0.06, 1.72 ± 0.05, and 2.08 ±0.11 g for control, 0.5 and 5 g/L DCA, respectively). The
percent liver/body weight followed a similar pattern with the exception that at Day 5 the 5 g/L
exposure group had a statistically significant increase over control (i.e., for Day 10;
6.00% ± 0.10%, 6.72% ± 0.17%, and 8.21% ± 0.10% for control, 0.5 and 5 g/L DCA,
respectively and for Day 15; 6.22 ± 0.08, 6.99 ±0.15, and 10.37 ± 0.27% g for control, 0.5 and
5 g/L DCA, respectively).
In Phase II of the study, control body weights were smaller than Phase I and varied
between 16.6 and 16.9 g in the first 3 exposure groups. Liver weights of controls were also
smaller making it difficult to quantitatively compare the two groups in terms of absolute liver
weights. However, the pattern of DCA-induced increases in liver weight and percent liver/body
weight remained. The patterns of body weight reduction only in the 5 g/L treatment groups and
increased liver weight with DCA treatment at both concentrations continued from 20 to 30 days
of exposure.
For liver weight there was a slight but statistically significant increase in liver weight for
the 0.5 g/L treatment groups over controls (i.e., for Day 20; 1.02 ±0.02, 1.18 ±0.05, and 1.98 ±
0.05 g for control, 0.5 and 5 g/L DCA, respectively, for Day 25; 1.15 ± 0.03, 1.34 ± 0.04, and
2.06 ± 0.12 g for control, 0.5 and 5 g/L DCA, respectively, for Day 30; 1.15 ± 0.03, 1.39 ± 0.08,
and 1.90 ± 0.12 g for control, 0.5 and 5 g/L DCA, respectively). For percent liver/body weight
there was a small increase at 0.5 g/L that was not statistically significant but all other treatments
induced increases in percent liver/body weight that were statistically significant (i.e., for Day 20;
4.82% ± 0.07%, 5.05% ± 0.09%, and 9.71% ± 0.11% for control, 0.5 and 5 g/L DCA,
respectively, for Day 25; 5.08%> ± 0.04%, 5.91% ± 0.09%, and 10.38%) ± 0.58% for control, 0.5
and 5 g/L DCA, respectively, for Day 30; 5.17% ± 0.09%, 6.01% ± 0.08%, and 10.28% ± 0.28%
for control, 0.5 and 5 g/L DCA, respectively).
Of note is the dramatic decrease in water consumption in the 5 g/L treatment groups that
were consistently reduced by 64% in Phase I and 46% in Phase II. The 0.5 g/L treatment groups
had no difference from controls in water consumption at any time in the study. The effects of
such water consumption decreases would affect body weight as well as dose received. Given the
differences in the size of the animals at the beginning of the study and the concurrent differences
in liver weights and percent liver/body weight in control animals between the two phases, the
changes in these parameters through time from DCA treatments cannot be accurately determined
(e.g., control liver/body weights averaged 6.32% in Phase I but 5.02% in Phase II). However,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
percent liver/body weight increase were reported to be consistently increased within and between
both phases of the study for the 0.5 g/L DCA treatment from 5 days of treatment to 30 days of
treatment (i.e., for Phase I the average increase was 9.5% and for Phase II the average increased
was 12.5% for 0.5 g/L DCA treated groups). Although increase at 5 days the nonsignificance of
the change may be resultant from the small number of animals examined. The difference in
magnitude of dose and percent liver/body weight increase is difficult to determine given that the
5 g/L dose of DCA reduced body weight and significantly reduced water consumption by -50%
in both phases of the study. Of note is that the differences in DCA-induced percent liver/body
weight were ~6-fold for the 15, 25, and 30-day data between the 0.5 and 5 g/L DCA exposures
rather than the 10-fold difference in exposure concentration in the drinking water.
The incorporation of tritiated thymidine into total hepatic DNA control treatment groups
was reported to be 73.34 ± 11.74 dpm/[j,g DNA at 5 days, 34 ± 4.12 dpm/[j,g DNA at 15 days,
and 28.48 ± 3.24 dpm/[j,g DNA at 20 days but was not reported for other treatments. The results
for 0.5 g/L treatments were not reported quantitatively but the authors stated that the results
"showed similar trends of initial inhibition followed by enhancement of labeling, the changes
relative to controls were not statistically significant." For 5 g/L treatment groups the 5-day
treated groups DNA tritiated thymidine incorporation was reported to be 42.8% of controls and
followed by a transient increase at 15 and 20 days (i.e., 2.65- and 2.45-fold of controls,
respectively) but after 25 and 30 days to not be significantly different from controls (data not
shown).
Labeling indices of hepatocytes were reported as means but variations as either SEM or
SD were not reported. Control means were reported as 5.5, 4, 2, 2, 3.2, and 3.5% of randomly
selected hepatocytes for 5, 10, 15, 20, 25, and 30 days, respectively, for 4 to 5 animals per group.
In contrast to the DNA incorporation results, no increase in labeling of hepatocytes was reported
to be observed in comparison to controls for any DCA treatment group from 5 to 30 days of
DCA exposure. The 5 g/L treatment group showed an immediate decrease in hepatocyte
labeling from Day 5 onwards that gradually increased approximately half of control levels by
Day 30 of exposure (i.e., <0.5% labeling index [LI] at Day 5, -1% LI at Day 10, -0.6% LI at
Day 20, 1% LI at Day 25 and 2% LI at Day 30). For the 0.5 g/L treatment the labeling index
was reported to not differ from controls from days 5 though 15 but to be significantly decreased
between days 20 and 30 to levels similar to those observed for the 5 g/L exposures. The
relatively higher number of hepatocytes incorporating label reported in this study than others can
be reflection of the longer times of exposure to tritiated thymidine. Here, incorporation was
shown for 1 weeks worth of exposure and reflects the percent of cell undergoing synthesis during
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
that time period. Also the higher labeling index in control animals at the 5 and 10 day exposure
periods is probably a reflection of the age of the animals at the time of study.
From the data reported by the authors, there was a correlation between the patterns of
total DNA incorporation of label and hepatocyte labeling indices in control groups (i.e., higher
level of labeling at 5 days than at 15 and 20). However, the patterns of decreased thymidine
labeling reported for hepatocytes were not correlated with a transient increase in total DNA
thymidine incorporation reported with DCA treatment, especially at the 5 g/L exposure level
with a large decrease reported for the number of labeled hepatocytes at the same time an increase
in total DNA thymidine incorporation was reported.
Although reported to be transiently increased, the total hepatic DNA labeling still
represented at most a 2.5-fold increase over control liver, which represents a small population of
cells. Given that the study examined hepatocyte labeling in random fields and did not report
quantitative zonal differences in proliferation, a more accurate determination of what hepatocytes
were undergoing proliferation cannot be made from the LI results. Also although the authors
report signs of inflammatory cells for 5-day treatment there is no reference to any inflammatory
changes that may have been observed at later time periods when cellular degeneration and loss of
nuclei were apparent. Such an increase inflammatory infiltrates can increase the DNA synthesis
measurements in the liver. The difference in LI and total DNA synthesis could reflect
differences in nonparenchymal cell proliferation or ploidy changes versus mitoses in
hepatocytes. Clearly, the increases in liver weight that were reported as early as 5 days of
exposure could not have resulted from increased hepatocyte proliferation.
The H&E sections were reported to have been fixed in an aqueous solution that reduced
glycogen content. However, residual PAS positive material (assumed to be glycogen) was
reported to be present indicating that not all of the glycogen had been dissolved. The authors
report changes in pathology between 5 and 30 days in control animals that included straightening
of hepatocyte cording, decreased mitoses, less clarity and more fine granularity of pericentral
hepatocellular cytoplasm, increased numbers of larger nuclei that were not labeled, and reported
differences between animals in the amount of glycogen present (i.e., 2 or 3 animals out of the 5
had less glycogen than other members of the group with less glycogen in the central and
midzonal areas). These changes are consistent with increased polyploidization expected for
maturing mice (see Sections E. 1.1 and E. 1.2 above).
After 5 days of treatment, 0.5 g/L exposed animals were reported to have livers with
fewer mitoses and tritiated thymidine hepatocyte labeling but by 10 days an increase in nuclear
size. Labeling was reported to be predominantly in small nuclei. Animals given 0.5 g/L DCA
for 15, 20, and 25 days were reported to have "focal cells in the middle zone with less detectable
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-13 9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
or no cell membranes and loss of the coarse granularity of the cytoplasm" with some cells not
having nuclei or cells having a loss of nuclear membrane and apparent karyolysis. "Cells
without nuclei because the plane of the section did not pass through the nuclei had the same type
of nuclei. Cells without nuclei not related to plane of section had a condensed cytoplasm."
Livers from 20-day and later sacrifice groups treated with 0.5 g/L DCA were reported to have
normal architecture. After 25 days of treatment apoptotic bodies were reported to be observed
with fewer nuclei around the central veins nuclei that were larger in central and midzonal areas.
In animals treated with 5 g/L DCA the authors report similar features as for 0.5 g/L but in
a zonal pattern. Inflammatory cells were reported to not be observed and after 5 and 10 days a
marked decrease in labeled nuclei. After 5 days of 5 g/L DCA, nuclear depletion in the central
and mid-zonal areas was reported. In methyl green pyronin-stained slides a marked loss of
cellular membranes was reported at 5 days with a loss of nuclei and formation of "lakes of liver
cell debris." After 15 days of treatment there was a reported increase in labeling in comparison
to animals sacrificed after 5 or 10 days. The cells nearest to the triads were reported to have
clearing of their cytoplasms and an increase in PAS positivity. Hepatocytes of both 0.5 and 5
g/L DCA treatment groups were reported to have "enlarged, presumably polyploidy nuclei."
Some of the nuclei were reported to be "labeled, usually in hepatocytes in the mid-zonal area."
The morphometric analyses of liver sections were reported to reveal statistically
significant changes in cellularity, nuclear size (as measured by either nuclear area or mean
diameter of the nuclear area equivalent circle), and multinucleated cells during 30 days exposure
to DCA. The authors reported that the concentration of total DNA in the liver, reported as total
j_ig nuclear DNA/g liver, ranged between 278.17 ± 16.88 and 707.00 ± 25.03 in the control
groups (i.e., 2-5-fold range). No 0.5 g/L DCA treatment groups differed from their control
group in terms of liver DNA concentration. However, for 10 though 30 days of exposure hepatic
DNA concentrations were reported to be decreased in the 5 g/L treatment groups (at 5 days there
appeared to be -30% increase over control). The number of cells per field was reported to range
between 24.28 ± 1.94 and 43.81 ± 1.93 in control livers (i.e., 1.8-fold range). From 5 to 15 days
the number of cells/field decreased with 0.5 g/L DCA treatment although only at Day 15 was the
change statistically significant. From 20 to 30 days of treatment only the 30 day treatment
showed a slight decrease in cells/field and that change was statistically significant. After 5 days
of treatment, the number of cells/field was 1.6-fold of control, by 15 days reduced by -20%, and
for 20 to 30 days continued to be reduced by as much as 40%.
Although the authors reported that the changes in cellularity and DNA concentration to
be closely correlated, the patterns in the number of cells/field varied in their consistency with
those of DNA concentration (i.e., for days 5, 20 and 25 there direction of change with dose was
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-140 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
similar between the two parameters but for days 10, 15 and 30 were not). If changes in liver
weight were due to hepatocellular hypertrophy, the increased liver size would be matched by a
decrease in liver DNA concentration and by the number of cells/field. The large increases in
liver/body weight induced by 5 g/L DCA were matched by decreases in liver DNA concentration
except for the 5 day exposure group. In general, the small increases in liver/body weight
consistently induced by 0.5 g/L treatment from Day 5 through 30 were not correlated with DNA
concentrations or cells/field.
The small number of animal examined for these parameters (i.e., n = 4-5) and the highly
variable control values limit the power to accurately detect changes. The apparent dehydration
in the animals treated at 5 g/L DCA was cited by the authors for the transient increase in
cellularity and DNA concentration in the 5-day exposure group. However, drinking water
consumption was reported to be similarly reduced at all treatment periods for 5 g/L DCA-treated
animals so that all groups would experience the same degree of dehydration.
The percentage of mononucleated cells was reported as percent of mononucleated
hepatocytes with results given as means but with no reports of variation within groups. The
mean control values were reported to range between 60 and 75% for Phase I and between 58 and
71% for Phase II of the experiment (n = 4-5 animals per group). The percent of mononucleated
hepatocytes was reported to be similar between control and DCA treatment groups at 5- and
10-day exposure. At 15 days both DCA treatments were reported to give a similar increase in
mononucleated hepatocytes (-80 vs. 60% in control) with only the 5 g/L DCA group statistically
significant. The increase in mononucleated cells reported for DCA treatment is similar in size to
the variation between control values. For Phase II of the study, DCA treatment was reported to
increase the number of mononucleated cells in at all concentrations and exposure time periods in
comparison to control values. However, only the increases for the 5 g/L treatments at days 20
and 25, and the 0.5 g/L treatment at Day 30 were reported to be statistically significant. Again,
small numbers of animals limit the ability to accurately determine a change. However, the
consistent reporting of an increasing number of mononucleated cells between 15 and 30 days
could be associated with clearance of mature hepatocytes as suggested by the report of DCA-
induced loss of cell nuclei.
Mean nuclear area was reported to range between 45 and 54 [j, in Phase I and to range
between 41 and 48 [j, in Phase II of the experiment with no variation in measurements given by
the authors. The only statistically significant differences reported between control and treated
groups in Phase I was a decrease from 54 to -42 [j, in the 0.5 g/L DCA 10 day treatment group
and a small increase from 50 to -52 [j, 15 day treatment group. Clearly the changes reported by
the authors as statistically significant did not show a dose-related pattern and were within the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-141 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
range of variation reported between control groups. For Phase II of the experiment both DCA
treatment concentrations were reported to induce a statistically significant increase the nuclear
area that was dose-related with the exception of Day 30 in which the nuclear area was similar
between the 0.5 and 5 g/L treatment groups. The largest increase in nuclear area was reported at
20 days for the 5 g/L treatment group (-72 vs. 41 [j, for control).
The patterns of increases in nuclear area were correlated with those of increased
percentage of mononucleated cells in Phase II of the study (20-30 days of treatment) as well as
the small changes seen in Phase I of the experiment. An increase in nuclear cell area is
consistent with increase polyploidization without mitosis as cells are induced towards
polyploidization. A decrease in the numbers of binucleate cells in favor of mononucleate cells is
consistent with clearance of mature binucleate hepatocyte as well induction of further
polyploidization of diploid or tetraploid binucleate cell to tetraploid or octoploid mononucleate
cells. The authors suggested that the "large hyperchromatic mononucleated hepatocytes are
tetraploid" and suggest that such increases in tetraploid cells have also been observed with
nongenotoxic carcinogens and with di(2-ethylhexyl) phthalate (DEHP).
In terms of increased cellular granularity observed by the authors with DCA treatment,
this result is also consistent with a more differentiated phenotype (Sigal et al., 1999). Thus, these
results for DCA are consistent with a DCA induced change in polyploidization of the cells
without cell proliferation.
The pattern of consistent increase in percent liver/body weight induced by 0.5 g/L DCA
treatment from days 5 though 30 was not consistent with the increased numbers of mononucleate
cells and increase nuclear area reported from Day 20 onward. The large differences in liver
weight induction between the 0.5 g/L treatment group and the 5 g/L treatment groups at all times
studied also did not correlate with changes in nuclear size and percent of mononucleate cells.
Thus, increased liver weight was not a function of cellular proliferation, but probably included
both aspects of hypertrophy associated with polyploidization and increased glycogen deposition
induced by DCA. The similar changes reported after short-term exposure for both the 0.5 and 5
g/L exposure concentration were suggested by the authors to indicate that the carcinogenic
mechanism at both concentrations would be similar. Furthermore, they suggest that although
there is evidence of cytotoxicity (e.g., loss of cell membranes and apparent apoptosis), De
Angelo et al. (1999) suggested that the present study does not support that the mechanism of
DCA-induced hepatocellular carcinogenesis is one of regenerative hyperplasia following
massive cell death nor peroxisome proliferation as the 0.5 g/L exposure concentration has been
shown to increase hepatocellular lesions after 100 weeks of treatment without concurrent
peroxisome proliferation or cytotoxicity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-142 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
E.2.4.1.11. DeAngelo et al. (1989). Various strains of rats and mice were exposed to TCA
(12 and 31 mM) or DCA (16 and 39 mM) for 14 days with S-D rats and B6C3F1 mice
exposed to an additional concentration of 6 mM TCA and 8 mM DCA. Although noting
that in a previous study that high concentrations of chloracids there was decreased water
consumption, the authors did not measure drinking water consumption in this study.
E.2.4.1.12. This study exposed several strains of male rats and mice to TCA at two
concentrations in drinking water (12 mM and 31mM neutralized TCA) for 14 days. The
conversion of mmols/L or mM TCA is 5 g/L TCA, 2 g/L TCA and 1 g/L for 31 mM, 12 mM,
and 6 mM TCA, respectively. The conversion of mmols/L of mM DCA is 5 g/L DCA, 2 g/L
DCA, and 1 g/L DCA for 39 mM, 16 mM and 8 mM DCA, respectively. The strains of mice
tested were Swiss-Webster, B6C3F1, C57BL/6, and C3H and for rats were Sprague Dawley
(SD), Osborne-Mendel, and F344. For the F344 rat and B6C3F1 mice data from two
separate experiments were reported for each. The number of animals in each group was
reported to be 6 for most experiments with the exception of the S-D rats (n = 3 at the
highest dose of TCA and n = 4 or 5 for the control and the lower TCA dose), one study in
B6C3F1 mice (n = 4 or 5 for all groups), and one study in F344 rats (n = 4 for all groups).
E.2.4.1.13. The body weight of the controls was reported to range from 269 to 341 g in the
differing strains of rats (1.27-fold) and 21 to 28 g in the differing strains of mice (1.33-fold,
age not reported). For percent liver/body weight ratios the range was 4.4 to 5.6% in
control rats (1.27-fold) and 5.1 to 6.8% in control mice (1.33-fold).
As discussed in other studies, the determination of PCO activity appears to be highly
variable. This enzyme activity is often used as a proxy for peroxisome proliferation. For PCO
activity the range of activity in controls was much greater than for either body weight or percent
liver/body weight. For rats there was a 2.8-fold difference in PCO control activity and in mice
there was a 4.6-fold difference in PCO activity. Between the two studies performed in the same
strain of rat (F344) there was a 2.83-fold difference in PCO activity between controls, and for the
two studies in the same strain of mouse (B6C3F1) there was a 3.14-fold difference in PCO
activity between controls. Not only were there differences between strains and experiments in
the same strain, but also differences in control values between species with a wider range of
values in the mice. The lowest level of PCO activity in control rats, expressed as nanomoles
NAD reduced/min/mg/protein, was 3.34 and for control mice was 1.40. The highest level
reported in control in rats was 9.46 and for control mice was 6.40.
These groups of rats and mice were exposed to 2 g/L NaCl, 2 g/L or 5 g/L TCA in
drinking water for 14 days and their PCO activity assayed. These doses of TCA did not affect
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-143 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
body weight except for the SD rats, which lost -16% of their body weight. This was also the
same group in which only 3 rats survived treatment. The Osborne-Mendel and F344 strains did
not exhibit loss of body weight or mortality due to TCA exposure.
There was a large variation in response to TCA exposure between the differing strains of
rats and mice with a much larger difference between the strains of mice. For the 3 rat strains
tested there was a range between 0% change and 2.38-fold of control for PCO activity at the 5
g/L TCA exposure. For the 2 g/L TCA exposure, there was a range of 0% change to 1.54-fold of
control for PCO activity. The Osborne-Mendel rats had 1.54-fold of control value for PCO
activity at 2 g/L TCA and 2.38-fold of control value for PCO activity reported at 5 g/L,
exhibiting the most consistent increase in PCO with increased dose of TCA. Two experiments
were reported for F344 rats with one reporting a 1.63-fold of control and the other a 1.79-fold of
control value for 5 g/L TCA. Only one of the F334 experiments also exposed rats to 2 g/L TCA
and reported no change from control values.
For the 4 strains of mice tested there was a range of 7.44- to 22.13-fold of control values
reported at the 5 g/L TCA exposures and 3.76- to 25.92-fold of control values at the 2 g/L TCA
exposures for PCO activity. For the C57BL/6 strain of mice there was little difference between
the 5 g/L and 2 g/L TCA exposures and a generally 3-fold higher induction of PCO activity by
TCA at the 5 g/L TCA exposure level than for the other mouse strains. Although there was a
2.5-fold difference between the 5 g/L and 2 g/L TCA exposure dose, the difference in magnitude
of PCO activity between these doses ranged from 0.85- to 2.23-fold for all strains of mice. For
the B6C3F1 mice there was a difference between reported increases of PCO activity in the text
(i.e., reported as 9.59-fold of control) for one of the experiments and that presented graphically
in Figure 2 (i.e., 8.70-fold of control). Nevertheless in the two studies of B6C3 F1 mice, 5 g/L
TCA was reported to induce 7.78-fold of control and 8.70-fold of control for PCO activity, and
2 g/L TCA was reported to induce 5.56-fold of control and 4.70-fold of control for PCO activity.
For the two F344 rat studies in which -200 mg/kg or 5 g/L TCA was administered for 10
or 14 days, there was 1.63-fold of control and 1.79-fold of control values reported for PCO
activity. Thus, for experiments in which the same strain and dose of TCA were administered,
there was not as large a difference in PCO response than between strains and species.
Whether increases in percent liver/body weight ratios were similar in magnitude to
increased PCO activity can be assessed by examination of the differences in magnitude of
increase over control for the 5 g/L and 2 g/L TCA treatments in the varying rat strains and mouse
strains. The relationship in exposure concentration was a 2.5:1 ratio for the 5 and 2 g/L doses.
For rats treatment of 5 g/L TCA to SD rats resulted in a significant decrease in body weight and
therefore, affected the magnitude of increase in percent liver/body weight ratio for this group.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-144 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
However, for the rest of the rat and mouse data, this dose was not reported to affect body weight
so that there is more confidence in the dose-response relationship.
For the SD rat there was no change in the percent liver/body weight ratio at 2 g/L but a
10% decrease at 5 g/L TCA exposure with no change in PCO activity for either. However, for
the Osborne-Mendel rats, there was no change in percent liver/body weight ratios for either
exposure concentration of TCA, but PCO activity was reported to be 1.54-fold of control at 2 g/L
and 2.38-fold of control at 5 g/L TCA. Thus, there was a ratio of 2.5-fold increase in PCO
activity between the 5 g/L and 2 g/L treatment groups. For the F344 rats there was a 2-fold
difference in liver weight increases (i.e., 12 vs. 6% increase over control) between the two
exposure concentrations but 1.6-fold of control value for PCO activity at the 5 g/L TCA
exposure concentration and no increase in PCO activity at the 2 g/L level. Thus, for the three
strains of rats, there did not appear to be a consistent correlation between liver weight induction
by TCA and PCO activity.
For differing strains of mice, similar concentrations of TCA were reported to vary in the
induction of liver weight increases. The range of liver weight induction was 1.26- to 1.66-fold of
control values between the 4 strains of mice at 5 g/L TCA and 1.16- to 1.63-fold at 2 g/L TCA.
In general, for mice the magnitudes of the difference in the increase in dose between the 5 g/L
and 2 g/TCA exposure concentration (2.5-fold) was generally higher than the increase percent
liver/body weight ratios at these doses. The differences in liver weight induction between the 2
and 5 g/L doses were -40% for the Swiss-Webster, C3H, and for one of the B6C3F1 mouse
experiments. For the C57BL/6 mouse there was no difference in liver weight induction between
the 2 and 5 g/L TCA exposure groups. For the other B6C3F1 mouse experiments there was a
2.5-fold greater induction of liver weight increase for the 5 g/L TCA group than for the 2 g/L
exposure group (1.39-fold of control vs. 1.16-fold of control for percent liver/body weight,
respectively).
For PCO activity the Swiss-Webster, C3H, and one of the B6C3F1 mouse experiments
were reported to have ~2-fold difference in the increase in PCO activity between the two doses.
For the other B6C3F1 mouse experiment there was only about a 50% increase and for the
C57BL/6 mouse data there was 15% less PCO activity induction reported at the 5 g/L TCA dose
that at the 2 g/L dose. None of the difference in increases in liver weight or PCO activity in mice
from the 2 or 5 g/L TCA exposures were of the same magnitude as the difference in TCA
exposure concentration (i.e., 2.5-fold) except for liver weight from the one experiment in
B6C3F1 mice. This is also the data used fore comparisons with the Sprague-Dawley rat
discussed below.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-145 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
In regard to strain differences for TCA response in mice, there did not appear to be
correlations of the magnitude of 5 g/L TCA-induced changes in percent liver/body weight ratio
or PCO activity, with the body weights reported for control mice for each strain. The control
weights between the 4 strains of mice varied from 21 to 28 grams. The strain with the greatest
response (C57B1/6) for TCA-induced changes in percent liver/body weight ratio (i.e., 1.66-fold
of control) and PCO activity (22.13-fold of control) had a mean body weight reported to be 26 g
for controls. At this dose, the range of percent liver/body weight for the other strains was
reported to be 1.26- to 1.39-fold of control and the range of PCO activity reported to be of 7.48-
to 8.71-fold of control.
Of note is that in the literature, this study has been cited as providing evidence of
differences between rats and mice for peroxisomal response to TCA and DC A. Generally the
PCO data from the Sprague Dawley rats and B6C3F1 mice at the highest dose of TCA and DCA
have been cited. However, the SD strain was reported to have greater mortality from TCA at this
exposure than the other strains tested (i.e., only 3 rats survived and provided PCO levels) and a
lower PCO response (no change in PCO activity over control) that the other two strains tested in
this study (i.e., Osborne-Mendel rats was reported to have had 2.38-fold of control and the F344-
had a 1.63- to 1.79-fold of control for PCO activity after exposure to 5 g/L TCA with no
mortality). The B6C3F1 mouse was reported to have a 7.78- or 8.71-fold of control for PCO
activity from 5 g/L TCA exposure. Certainly the male mouse is more responsive to TCA
induction of PCO activity. However, as discussed above there are large variations in control
levels of PCO activity and in the magnitude and dose-response of TCA-induction of PCO
activity between rat and mouse strains and between species. If is not correct to state that the rat
is refractory to TCA-induction of peroxisome activity.
Unfortunately, the authors chose the SD rat (i.e., the most unresponsive strain for PCO
activity and most sensitive to toxicity) for studies for comparative studies between DCA and
TCA effects. The authors also tested for carnitine acetyl CoA transferase (CAT) activity as a
marker of peroxisomal enzyme response and took morphometric analysis of peroxisome # and
cytoplasmic volume for one liver section for each of two B6C3F1 mice of SD rats from the 5 g/L
TCA and 5 g/L DCA treatment groups. Only 6 electron micrograph fields were analyzed from
each section (12 fields total) were analyzed without identification as to what area of the liver
lobules they were being taken from. Hence there is a question as to whether the areas which are
known to be peroxisome rich were assayed of not. Also as noted above, previous studies have
indicate that such high concentration of DCA and TCA inhibit drinking water consumption and
therefore, raising issues not only about toxicity but also the dose which rats and mice received.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-146 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The number of peroxisomes per 100 [j,m3 and cytoplasmic volume of peroxisomes was
reported to be 6.60 and 1.94%, respectively, for control rats, and 6.89 and 0.61% for control
mice, respectively. For 5 g/L TCA and 5 g/L DCA the numbers of peroxisomes were reported to
be increased to 7.14 and 16.75, respectively in treated Sprague Dawley rats. Thus, there was 2.5-
and 1.08-fold of control reported in peroxisome # for 5 g/L DCA and TCA, respectively. The
cytoplasmic volume of peroxisomes was reported to be 2.80% and 0.89% for 5 g/L DCA and 5
g/L TCA, respectively (i.e., a 1.44-fold of control and -60% reduction for 5 g/L DCA and 5 g/L
TCA, respectively). Thus, 5 g/L TCA was reported to slightly increase the number of
peroxisomes and but decrease the percent of the cytoplasmic volume occupied by peroxisome by
half. For DCA the reported pattern was for both to increase. PCO activity was reported to
increase by a similar magnitude as peroxisome # but not volume in the 5 g/L TCA treated S-D
rats. However, although peroxisomal volume was reported to be cut nearly in half and for
peroxisome number to be similar, 5 g/L TCA treatment was not reported to change PCO activity
in the S-D rat.
For comparisons between DCA and TCA, B6C3 F1 mice were examined at 1.0, 2.0, and
5.0 g/L concentrations. DCA was reported to induce a higher percent liver/body weight ratio
that did TCA at every concentration (i.e., 1.55-, 1.27-, and 1.21-fold of control for DCA and
1.39-, 1.16-, and 1.08-fold of control for TCA at 1.0, 2.0, and 5.0 g/L concentrations,
respectively). As noted above, for other strains of mice tested and a second experiment with
B6C3F1 mice, there was 40% or less difference in percent liver/body weight ratio between the
2.0 g/L and 5.0 g/L exposures to TCA but for this experiment there was a 2.5-fold difference.
Thus, at 5 g/L there was -40% greater induction of liver weight for DCA than TCA.
In the B6C3F1 mice, 5 g/L TCA was reported to increase peroxisome number to 30.75
and cytoplasmic volume to 4.92% (i.e., 4.4- and 8.1-fold of control, respectively). For 5 g/L
DCA treatment, the peroxisome number was reported to be 30.77 and 3.75% (i.e., 4.5- and 6.1-
fold of control, respectively). While there was no difference in peroxisome number and -40%
difference in cytoplasmic volume at the 5.0 g/L exposures of DCA and TCA, there was a greater
difference in the magnitude of PCO activity increase. The 5 g/L TCA exposure was reported to
induce 4.3-fold of control for PCO activity while 5 g/L DCA induced as 9.6-fold of control PCO
activity (although a figure in the report shows 8.7-fold of control) which is a -2.5-fold difference
between DCA and TCA at this exposure concentration. Thus, for one of the B6C3F1 mouse
studies, 5 g/L DCA and TCA treatments were reported to give a similar increase peroxisome
number, TCA to induce a 40% greater increase in peroxisomal cytoplasmic volume than DCA
and a 2.5-fold greater increase in PCO activity, but DCA to induce -40% greater liver weight
induction than TCA.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-147 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Not only were PCO activity, peroxisome number and cytoplasmic volume occupied by
peroxisomes analyzed but also CAT activity as a measure of peroxisome proliferation. For TCA
and DCA the results were opposite those reported for PCO activity. In SD rats control levels of
CAT were reported to be 1.81 nmoles of carnitine transferred/min/mg/protein. Exposure to 5 g/L
TCA was reported to increase CAT activity by 3.21-fold of control while 5 g/L DCA was
reported to induce CAT activity to 10.33-fold of control levels in S-D rats. However, while PCO
activity was reported to be the same as controls, and peroxisomal volume decreased, 5 g/L TCA
increased CAT activity 3.21-fold of control in these rats. The level of CAT induced by 5 g/L
DCA was over 10-fold of control in the rat while peroxisome # was only 2.5-fold of control and
cytoplasmic volume 1.4-fold of control. Thus, the fold increases for these three measures were
not the same for DCA treatment and for TCA in rats. Nevertheless for CAT, DCA was a
stronger inducer in rats than was TCA.
In B6C3 F1 mice 5 g/L TCA and 5 g/L DCA induced CAT activity to a similar extent
(4.50- and 5.61-fold of control, respectively). The magnitude of CAT induction was similar to
that of peroxisome # for both 5 g/L DCA and 5 g/L TCA and lower than PCO activity in DCA-
treated mice and cytoplasmic volume in TCA-treated mice by about half. Thus, using CAT as
the marker of peroxisome proliferation, the rat was more responsive than the mouse to DCA and
nearly as responsive to TCA as the mouse at this high dose in these two specific strains. These
data illustrate the difficulty of using only one measure for peroxisome proliferation and shows
that the magnitude of increased PCO activity is not necessarily predictive of the peroxisome # or
cytoplasmic volume or CAT activity. The difficulty of interpretation of the data from so few
animals and sections for the electron microscopy analysis, and the low number of animals for
PCO activity and CAT activity (n = 3 to 6), the high dose studied (5 g/L), and the selection of a
rat strain that appears to be more resistant to this activity but more susceptible to toxicity than the
others tested, should be taken into account before conclusions can be made about differences
between these chemicals for peroxisome activity between species.
Of note is that PCO activity was also shown to be increased by corn oil alone in F344 rats
and to potentiate the induction of PCO activity of TCA. After 10 days of exposure to either
water, corn oil, 200 mg/kg/d TCA in corn oil or 200 mg/kg TCA in water via gavage dosing,
there was 1.40-fold PCO activity from corn oil treatment alone in comparison to water, a
1.79-fold PCO activity from TCA in water treatment in comparison to water, and a 3.14-fold
PCO activity from TCA in corn oil treatment in comparison to water.
The authors provided data for 3 concentrations of DCA and TCA for SD and for one
experiment in the B6C3F1 mouse for examination of changes in body and percent liver/body
weight ratios (1, 2, or 5 g/L DCA or TCA) after 14 days of exposure. As noted above, not only
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-148 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
did the 5 g/L exposure concentration of DC A result in mortality in the SD strain of rat, but the
5 g/L and 2 g/L concentrations of DC A were reported to decrease body weight (-20 and 25%,
respectively). The 5 g/L dose of TCA was also reported to induce a statistically significant
decrease in body weight in the SD rat. There were no differences in final body weight in any of
the mice exposed to TCA or DCA.
As noted above no TCA or DCA exposure group of SD rats was reported to have a
statistically significant increase in percent liver/body weight ratio over control. For the B6C3F1
male mice, the percent liver/body weight ratio was 1.22-, 1.27-, and 1.55-fold of control after
exposure to 1, 2, and 5 g/L DCA, respectively, and 1.08-, 1.16-, and 1.39-fold of control after
exposure to 1, 2, and 5 g/L TCA, respectively. Thus, for DCA there was only a 20% increase in
liver weight corresponding to the 2-fold increase between the 1 and 2 g/L exposure levels of
DCA. Between the 2 and 5 g/L exposure concentrations of DCA there was a 2-fold increase in
liver weight corresponding to a 2.5-fold increase in exposure concentration. For TCA, the
magnitude of increase in dose was reported to be proportional to the magnitude of increase in
percent liver/body weight ratio in the B6C3 F1 male mouse. As stated above, the
correspondence between magnitude of dose and percent liver weight for TCA exposure in this
experiment differed from the other experiment reported for this strain of mouse and also differed
from the other 3 strains of mice examined in this study where the magnitude in liver weight gain
was much less than exposure concentration.
E.2.4.2. Subchronic and Chronic Studies of Dichloroacetic Acid (DCA) and Trichloroacetic
Acid (TCA)
Several experiments have been conducted with exposure to DCA and TCA, generally at
very high levels with a limited dose range, for less periods of time than standard carcinogenicity
bioassays, and with very limited information on any endpoints other than the liver tumor
induction. Caldwell and Keshava (2006) and Caldwell et al. (2008b) have examined these
studies for inferences of modes of action for TCE. Key studies are briefly described below for
comparative purposes of results reported in TCE studies.
E.2.4.2.1. Snyder et al. (1995). Studies of TCE have reported either no change or a slight
increase in apoptosis only after a relatively high exposure level (Channel et al., 1998; Dees
and Travis, 1993). Inhibition of apoptosis, which has been suggested to prevent removal of
"initiated" cells from the liver and lead to increased survival of precancerous cells, has
been proposed as part of the MOA for peroxisome proliferators (see Section E.3.4). The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-149 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
focus of this study was to examine whether DCA, which has been shown to inhibit DNA
synthesis after an initial transient increase (see Section E.2.3.3, below), also alters the
frequency of spontaneous apoptosis in mice. This study exposed 28-day old male B6C3F1
male mice (n = 5) to 0, 0.5 or 5.0 g/L buffered DCA in drinking water for up to 30 days
(Phase I = 5-15 days exposure and Phase II = 20-30 days treatment).
E.2.4.2.2. Portions of the left lobe of the liver were prepared for histological examination
after H&E staining. Hepatocyte number was determined by counting nuclei in 50 fields
with nonparenchymal cell nuclei excluded on the basis of nuclear size. Multinucleate cells
were counted as one cell. Apoptotic cells were visualized by in situ TDT nick end-labeling
assay from 2-4 different liver sections from each control or treated animal. The average
number of apoptotic cells was then determined for each animal in each group. The authors
reported that in none of the tissues examined were necrotic foci observed, there was no any
indication of lymphocyte or neutrophil infiltration indicative of an inflammatory response,
and suggested that no necrotic cells contributed to the responses in their analysis.
Control animals were reported to exhibit apoptotic frequencies ranging from -0.04 to
0.085% and that over the 30-day period the frequency rate declined. The authors suggested that
this result is consistent with reports of the livers of these young animals undergoing rapid
changes in cell death and proliferation. They note that animals receiving 0.5 g/L DCA also had a
similar trend of decreasing apoptosis with age, supportive of the decrease being a physiological
phenomenon. The 0.5 g/L exposure level of DCA was reported to decrease the percentage of
apoptotic hepatocytes as the earliest time point studied and to remain statistically significantly
decreased from controls from 5 to 30 days of exposure. The rate of apoptosis ranged from
-0.025 to 0.060% after 0.5 g/L DCA exposure during the 30-day period (i.e., and -30-40%)
reduction). Animals receiving the 5.0 g/L DCA dose exhibited a significant reduction at the
earliest time point that was sustained at a similar level and statistically significant throughout the
time-course of the experiment (percent apoptosis ranged from 0.015-0.030%)).
The results of this study not only provides a baseline of apoptosis in the mouse liver,
which is very low, but also to show the importance of taking into account the effects of age on
such determinations. The authors reported that the for rat liver the estimated frequency of
spontaneous apoptosis to be -0.1% and therefore, greater than that of the mouse. The
significance of the DCA-induced reduction in apoptosis, of a level that is already inherently low
in the mouse, for the MOA for induction of cancer is difficult to discern.
E.2.4.2.3. Mather et al. (1990). This 90-day study in male SD rats examined the body and
organ weight changes, liver enzyme levels, and PCO activity in livers from rats treated with
estimated concentrations of 3.9, 35.5, 345 mg/kg day DCA or 4.1, 36.5, or 355 mg/kg/d TCA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-150 DRAFT—DO NOT CITE OR QUOTE
-------
from drinking water exposures (i.e., 0, 50, 500 and 5,000 ppm or 0.05, 0.5, or 5.0 g/L DCA
or TCA in the drinking water). All dose levels of DCA and TCA were reported to result in
a dose-dependent decrease in fluid intake at 2 months of exposure. The rats were 9 (DCA)
or 10 (TCA) weeks old at the beginning of the study (n = 10/group). Animals with body
weights that varied more than 20% of mean weights were discarded from the study. The
DCA and TCA solutions were neutralized. The mean values for initial weights of the
animals in each test group varied less than 3%.
E.2.4.2.4. DCA treatment induced a dose-related decrease in body weight that was
statistically significant at the two highest levels (i.e., a 6, 9.5, and 17% decrease from
control). TCA treatment also resulted in lower body weights that were not statistically
significant (i.e., 2.1, 4.4, and 5.9%). DCA treatments were reported to result in a dose-
related increase in absolute liver weights (1.01-, 1.13-, and 1.36-fold of control that were
significantly different at the highest level) and percent liver/body weight ratios (1.07-, 1.24-,
and 1.69-fold of control that were significant at the two highest dose levels). TCA
treatments were reported to not result in changes in either absolute liver weights or percent
liver/body weight ratios with the exception of statistically significant increase in percent
liver/body weight ratios at the highest level of treatment (1.02-fold of control).
E.2.4.2.5. Total serum protein levels were reported to be significantly depressed in all
animals treated with DCA with animals in the two highest dose groups also exhibiting
elevations of alkaline phosphatase. Alanine-amino transferase levels were reported to be
elevated only in the highest treatment group. No consistent treatment-related effect on
serum chemistry was reported to be observed for the TCA-treated animals with data not
shown.
E.2.4.2.6. In terms of PCO activity, there was only a mild increase at the highest dose of
15% for TCA and a 2.5-fold level of control for DCA treatment that were statistically
significant. The difference in PCO activity between control groups for the DCA and TCA
experiments was reported to be 33%. No treatment affect was reported to be apparent for
hepatic microsomal enzymes, or measures of immunotoxicity for either DCA or TCA but
data were not shown.
E.2.4.2.7. Focal areas of hepatocellular enlargement in both DCA- and TCA-treated rats
were reported to be present with intracellular swelling more severe with the highest dose of
DCA treatment. Livers from DCA treated rats were reported to stain positively for PAS,
indicating significant amounts of glycogen with TCA treated rats reported to display "less
evidence of glycogen accumulation." Of note is that, in this study of rats, DCA was
reported to induce a greater level of PCO activity than did TCA.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-151 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
E.2.4.2.8. Parrish et al. (1996). Parrish et al. (1996) exposed male B6C3F1 mice (8 weeks
old and 20-22 g upon purchase) to TCA or DCA (0, 0.01, 0.5, and 2.0 g/L) for 3 or 10 weeks
(n = 6). Livers were excised and nuclei isolated for examination of 8-OHdG and
homogenates examined for cyanide insensitive acyl-CoA oxidase (ACO) and laurate
hydroxylase activity. The authors noted that control values between experiments varied as
much as a factor of 2-fold for PCO activity and that data were presented as percent of
concurrent controls. Initial body weights for treatment groups were not presented and
thus, differences in mean values between the groups cannot be ascertained.
Final body weights were reported to not be statistically significantly changed by DCA or
TCA treatments at 21 days or 71 days of treatment (all were within -8% of controls). The mean
percent liver/body ratios were reported to be 5.4, 5.3, 6.1, and 7.2% for control, 0.1, 0.5, and
2.0 g/L TCA, respectively and 5.4, 5.5, 6.7, and 7.9% for control, 0.1, 0.5, and 2.0 g/L DCA,
respectively after 21 days of exposure. This represents 0.98-, 1.13-, and 1.33-fold of control
levels with these exposure levels of TCA and 1.02-, 1.24-, and 1.46-fold of control levels with
DCA after 21 days of exposure. For 71 days of exposure the mean percent liver/body ratios were
reported to be 5.1, 4.6, 5.8, and 6.9% for control, 0.1, 0.5, and 2.0 g/L TCA, respectively and 5.1,
5.1, 5.9, and 8.5% for control, 0.1, 0.5, and 2.0 g/L DCA, respectively. This represents 0.90-,
1.14-, and 1.35-fold of control with TCA exposure and 1.0-, 1.15-, and 1.67-fold of control with
DCA exposure after 71 days of exposure. The magnitude of difference between the 0.1 and
0.5 g/L TCA doses is 5 and 0.5 and 2.0 g/L doses is 4-fold.
For the 21-day and 71-day exposures the magnitudes of the increases in percent
liver/body weight over control values were greater for DCA than TCA exposure at same
concentration with the exception of 0.5 g/L doses at 71 days in which both TCA and DCA
induced similar increases. For TCA, the 0.01 g/L dose produces a similar 10% decrease in
percent liver/body weight. Although there was a 4-fold increase in magnitude between the 0.5
and 2.0 g/L TCA exposure concentrations, the magnitude of increase for percent liver/body
weight increase was 2.5-fold between them at both 21 and 71 days of exposure. For DCA, the
0.1 g/L dose was reported to have a similar value as control for percent liver/body weight ratio.
Although there was a 4-fold difference in dose between the 0.5 and 2.0 g/L DCA exposure
concentrations, there was a ~2-fold increase in percent liver/body weight increase at 21 days and
~4.5-fold increase at 71 days.
As a percentage of control values, TCA was reported to induce a dose-related increase in
PCO activity at 21 days (-1.5-, 2.2-, and ~4.1-fold of control, for 0.1, 0.5, and 2 g/L TCA
exposures). Only the 2.0 g/L dose of DCA was reported to induce a statistically significant
increase at 21-days of exposure of PCO activity over control (~1.8-fold of control) with the 0.1
and 0.5 g/L exposure PCO activity to be slightly less than control values (-20% less). Thus,
although there was no increase in percent liver/body weight at 0.1 g/L TCA, the PCO activity
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-152 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
was reported to be increased by -50% after 21 days. A 13% increase in liver weight at 0.5 g/L
TCA was reported to be associated with 2.2-fold of control level of PCO activity and a 33%
increase in liver weight after 2.0 g/L TCA to be associated with 4.1-fold of control level of PCO
activity.
Thus, increases in PCO activity were not necessarily correlated with concurrent TCA-
induced increases in liver weight and the magnitudes of increase in liver weight between 0.5 and
2.0 g/L TCA (2.5-fold) was greater than the corresponding increase in PCO activity (1.8-fold of
control). Although there was a 20-fold difference in TCA dose, the magnitude of increase in
PCO activity between 0.1 and 2.0 g/L TCA was ~2.7-fold. As stated above, the 4-fold difference
in TCA dose at the two highest levels resulted in a 2.5-fold increase in liver weight. For DCA,
the increases in liver weight at 0.1 and 0.5 g/L DCA exposures were not associated with
increased PCO activity after 21 days of exposure. The 2.0 g/L DCA exposure concentration was
reported to induce 1.8-fold of control PCO activity.
After 71 days of treatment, TCA induced a dose-related increase in PCO activity that
was approximately twice the magnitude as that reported at 21 days (i.e., ~9-fold greater at 2.0
g/L level). After 71 days, for DCA the 0.1 and 0.5 g/L doses produced a statistically significant
increase in PCO activity (-1.5- and 2.5-fold of control, respectively). The administration of 1.25
g/L clofibric acid in drinking water was used as a positive control and reported to induce -6-7-
fold of control PCO activity at 21 and 71 days of exposure.
Laurate hydroxylase activity was reported to be elevated significantly only by TCA at
21 days (2.0 g/L TCA dose only) and to increased to approximately the same extent (-1.4 to
1.6-fold of control values) at all doses tested. For 0.1 g/L DCA the laurate hydroxylase activity
was reported to be similar to that of 0.1 g/L TCA (-1.4-fold of control) but to be -1.2-fold of
control at both the 0.5 and 2.0 g/L DCA exposures. At 71 days, both the 0.5 and 2.0 g/L TCA
exposures induced a statistically significant increase in laurate hydroxylase activity (i.e., 1.6- and
2.5-fold of control, respectively) with no change after DCA exposure. The actual data rather
than percent of control values were reported for laurate hydroxylase activity. The control values
for laurate hydroxylase activity varied 1.7-fold between 21 and 71 days experiments.
The results for 8-OHdG levels are discussed in Section E.3.4.2.3, below. Of note is that
the increases in PCO activity noted for DCA and TCA were not associated with 8-OHdG levels
(which were unchanged, see Section E.3.4.2.3, below) and also not with changes laurate
hydrolase activity or percent liver/body weight ratio increases observed after either DCA or TCA
exposure. A strength of this study is that is examined exposure concentrations that were lower
than those examined in many other short-term studies of DCA and TCA.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-153 DRAFT—DO NOT CITE OR QUOTE
-------
E.2.4.2.9. Bull et al. (1990). The focus of this study was the determination of "dose-
response relationships in the tumorigenic response to these chemicals [sic DCA and TCA]
in B6C3F1 mice, determine the nature of the nontumor pathology that results from the
administration of these compounds in drinking water, and test the reversibility of the
response." Male and female B6C3F1 mice (age 37 days) were treated from 15 to 52 weeks
with neutralized TCA and TCA. A highly variable number and generally low number of
animals were reported to be examined in the study with n = 5 for all time periods except for
52 weeks where in males the n = 35 for controls, n =11 for 1 g/L DCA, n = 24 for 2 g/L
DCA, n = 11 for 1 g/L TCA, and n = 24 for 2 g/L TCA exposed mice. Female mice were
only examined after 52 weeks of exposure and the number of animals examined was n = 10
for control, 2 g/L DCA, and 2 g/L TCA exposed mice.
E.2.4.2.10. "Lesions to be examined histologically for pathological examination were
selected by a random process" with lesions reported to be selected from 31 of 65 animals
with lesions at necropsy. 73 of 165 lesions identified in 41 animals were reported to be
examined histologically. All hyperplastic nodules, adenomas and carcinomas were lumped
together and characterized as hepatoproliferative lesions. Accordingly there were only
exposure concentrations available for dose-response analyses in males and only
"multiplicity of hepatoproliferative lesions" were reported from random samples. Thus,
these data cannot be compared to other studies and are unsuitable for dose-response with
inadequate analysis performed on random samples for pathological examination.
E.2.4.2.11. The authors state that some of the lesions taken at necropsy and assumed to be
proliferative were actually histologically normal, necrotic, or an abscess as well. It is also
limited by a relatively small number of animals examined in regard to adequate statistical
power to determine quantitative differences. Similar concerns were raised by Caldwell et
al. (2008b) with a subsequent study (eg., Bull et al., 2002). For example, the authors report
that 5/11 animals had "lesions" at 1 g/L TCA at 52 weeks and 19/24 animals had lesions at
2 g/L TCA at 52 weeks. However, while 7 lesions were examined in 5 mice bearing lesions
at 1 g/L TCA, only 16 of 30 lesions from 11 of the 19 animals bearing lesions examined in
the 2 g/L TCA group. Therefore, almost half of the mice with lesions were not examined
histologically in that group along with only half of the "lesions."
1	The authors reported the effects of DCA and TCA exposure on liver weight and percent
2	liver/body changes (m ± SEM) and these results gave a pattern of hepatomegaly generally
3	consistent with short-term exposure studies. The authors report "no treatment produced
4	significant changes in the body weight or kidney weight of the animals (data not shown)"
5	In male mice (n = 5) at 37 weeks of exposure, liver weights were reported to be 1.6 ± 0.1,
6	2.5 ± 0.1, and 1.9 ± 0.1 g for control, 2 g/L DCA, and 2 g/L TCA exposed mice, respectively.
7	The percent liver/body weights were reported to be 4.1% ± 0.3%, 7.3% ± 0.2%, and 5.1% ±
8	0.1% for control, 2 g/L DCA, and 2 g/L TCA exposed mice, respectively. In male mice at 52
9	weeks of exposure, liver weights were reported to be 1.7 ± 0.1, 2.5 ± 0.1, 5.1 ± 0.1, 2.2 ± 0.1,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-154 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
and 2.7 ± 0.1 g for control (n = 35), 1 g/L DCA (n = 11), 2 g/L DCA (« = 24), 1 g/L TCA (« =
11), and 2 g/L TCA (« = 24) exposed mice, respectively. In male mice at 52 weeks of exposure,
percent liver/body weights were reported to be 4.6% ± 0.1%, 6.5% ± 0.2%, 10.5% ± 0.4%, 6.0%
± 0.3%, and 7.5% ± 0.5% for control, 1 g/L DCA, 2 g/L DCA, 1 g/L TCA, and 2 g/L TCA
exposed mice, respectively. For female mice (n = 10) at 52 weeks of exposure, liver weights
were reported to be 1.3 ± 0.1, 2.6 ± 0.1, and 1.7 ± 0.1 g for control, 2 g/L DCA, and 2 g/L TCA
exposed mice, respectively. The percent liver/body weights were reported to be 4.8% ± 0.3%,
9.0% ± 0.2%), and 6.0% ± 0.3% for control, 2 g/L DCA, and 2 g/L TCA exposed mice,
respectively.
Although the number of animals examined varied 3-fold between treatment groups in
male mice, the authors reported that all DCA and TCA treatments were statistically increased
over control values for liver weight and percent body/liver weight in both genders of mice. In
terms of percent liver/body weight ratio, female mice appeared to be as responsive as males at
the exposure concentration tested. Thus, hepatomegaly reported at these exposure levels after
short-term exposures appeared to be further increased by chronic exposure with equivalent levels
of DCA inducing greater hepatomegaly than TCA.
Interestingly, after 37 weeks of treatment and then a cessation of exposure for 15 weeks
liver weights were assessed in control male mice, 2 g/L DCA treated mice, and 2 g/L TCA
treated mice (n= 11 for each group but results for controls were pooled and therefore, n = 35).
Liver weights were reported to be 1.7 ± 0.1, 2.2 ±0.1, and 1.9 ± 0.1 g for control, 2 g/L DCA,
and 2 g/L TCA exposed mice, respectively. The percent liver/body weights were reported to be
4.6% ± 0.1%), 5.7%) ± 0.3%), and 5.4%) ± 0.2% for control, 2 g/L DCA, and 2 g/L TCA exposed
mice, respectively. After 15 weeks of cessation of exposure, liver weight and percent liver/body
weight were reported to still be statistically significantly elevated after DCA or TCA treatment.
The authors partially attributed the remaining increases in liver weight to the continued
presence of hyperplastic nodules in the liver. The authors stated that because of the low
incidence of lesions in the control group and the two groups that had treatments suspended, all
the lesions from these groups were included for histological sectioning. However, the authors
presented a table indicating that, of the 23 lesions detected in 7 mice exposed to DCA for 37
weeks, 19 were examined histologically. Therefore, groups that were exposed for 52 weeks had
a different procedure for tissue examination as those at 37 weeks.
In terms of liver tumor induction, the authors stated that "statistical analysis of tumor
incidence employed a general linear model ANOVA with contrasts for linearity and deviations
from linearity to determine if results from groups in which treatments were discontinued after 37
weeks were lower than would have been predicted by the total dose consumed." The multiplicity
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-15 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of tumors observed in male mice exposed to DCA or TCA at 37 weeks and then sacrificed at 52
weeks were reported by the authors to have a response in animals that received DCA very close
to that which would be predicted from the total dose consumed by these animals. The response
to TCA was reported by the authors to deviate significantly (p = 0.022) from the linear model
predicted by the total dose consumed.
Multiplicity of lesions per mouse and not incidence was used as the measure. Most
importantly the data used to predict the dose response for "lesions" used a different methodology
at 52 weeks than those at 37 weeks. Not only were not all animal's lesions examined but foci,
adenomas, and carcinomas were combined into one measure. Therefore, foci, of which a certain
percentage have been commonly shown to spontaneously regress with time, were included in the
calculation of total "lesions." Pereira and Phelps (1996) note that in initiated mice treated with
DCA, the yield of altered hepatocytes decreases as the tumor yields increase between 31 and
51 weeks of exposure suggesting progression of foci to adenomas. Initiated and noninitiated
control mice also had fewer foci/mouse with time.
Because of differences in methodology and the lack of discernment between foci,
adenomas, and carcinomas for many of the mice exposed for 52 weeks, it is difficult to compare
differences in composition of the "lesions" after cessation of exposure. For TCA treatment the
number of animals examined for determination of which "lesions" were foci, adenomas, and
carcinomas was 11 out of the 19 mice with "lesions" at 52 weeks while all 4 mice with lesions
after 37 weeks of exposure and 15 weeks of cessation were examined.
For DCA treatment the number of animals examined was only 10 out of 23 mice with
"lesions" at 52 weeks while all 7 mice with lesions after 37 weeks of exposure and 15 weeks of
cessation were examined. Most importantly, when lesions were examined microscopically then
did not all turn out to be preneoplastic or neoplastic. Two lesions appeared "to be histologically
normal" and one necrotic. Not only were a smaller number of animals examined for the
cessation exposure than continuous exposure but only the 2 g/L exposure levels of DCA and
TCA were studied for cessation. The number of animals bearing "lesions" at 37 and then 15
week cessation weeks was 7/11 (64%) while the number of animals bearing lesions at 5 weeks
was 23/24 (96%) after 2 g/L DCA exposure. For TCA the number of animals bearing lesions at
37 weeks and then 15 weeks cessation was 4/11 (35%) while the number of animals bearing
lesions at 52 weeks was 19/24 (80%). While suggesting that cessation of exposure diminished
the number of "lesions," conclusions regarding the identity and progression of those lesion with
continuous versus noncontinuous DCA and TCA treatment are tenuous.
Macroscopically, the "livers of many mice receiving DCA in their drinking water
displayed light colored streaks on the surface" at every sacrifice period and "corresponded with
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-156 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
multi-focal areas of necrosis with frequent infiltration of lymphocytes." At the light microscopic
level, the lesions were described to also be present in the interior of the liver as well. For
TCA-treated mice, "similar necrotic lesions were also observed... but at a much lower
frequency, making it difficult to determine if they were treatment-related." Control animals were
reported not to show degenerative changes. "Marked cytomegaly" was reported for mice treated
with either 1 or 2 g/L DCA "throughout the liver" In regard to cell size the authors did not give
any description in the methods section of the paper as to how sections were selected for
morphometric analysis or what areas of the liver acinus were examined but reported after
52 weeks of treatment the long axis of hepatocytes measured (mean ± S.E.) 24.9 ± 0.3,
38.5 ± 1.0, and 29.3 ±1.4 [j,m in control, DCA- and TCA-treated mice, respectively.
Mice treated with TCA (2 g/L) for 52 weeks were reported to have livers with
"considerable dose-related accumulations of lipofuscin." However, no quantitative analyses
were presented. A series of figures representative of treatment showed photographs (l,000x) of
lipofuscin fluorescence indicating greater fluorescence in TCA treated liver than control or DCA
treated liver.
A series of photographs of H&E sections in the report (see Figures 2a, b and c) were
shown as representative histology of control mice, mice treated with 2 g/L DCA and 2 g/L TCA.
The area of the liver from which the photographs were taken did not include either portal tract or
central veins and the authors did not give the zone of the livers from which they were taken. The
figure representing TCA treatment shows only a mild increase in cell volume in comparison to
controls, while for DCA treatment the hepatocyte diameter was greatly enlarged, pale stained so
that cytoplasmic contents appear absent, nuclei often pushed to the cell perimeter, and the
sinusoids appearing to be obscured by the swollen hepatocytes. The apparent reduction of
sinusoidal volume by the enlarged hepatocytes raises the possibility of decreased blood flow
through the liver, which may have been linked to focal areas of necrosis reported for this high
exposure level.
In a second set of figures, glycogen accumulation was shown with PAS staining at the
same level of power (400x) for the same animals. In control animals PAS positive material was
not uniformly distributed between or within hepatocytes but tended to show a zonal pattern of
moderate intensity. PAS positive staining (which the authors reported to be glycogen) appeared
to be slightly less than controls but with a similar pattern in the photograph representing TCA
exposure. However, for DCA the photograph showed a uniform and heavy stain within each
hepatocyte and across all hepatocytes.
The authors stated in the results section of the paper that "the livers of TCA-treated
animals displayed less evidence of glycogen accumulation and it was more prominent in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-157 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
periportal than centrilobular portions of the liver acinus." In their abstract they state "TCA
produced small increases in cell size and a much more modest accumulation of glycogen." Thus,
the statement in the text, which is suggestive that TCA induced an increase in glycogen over
controls that was not as much as that induced by DC A, and the statement in the abstract which
concludes TCA exposure increased glycogen is not consistent with the photographs. In the
photograph shown for TCA there is less not more PAS positive staining associated with TCA
treatment in comparison to controls.
In Sanchez and Bull (1990) the authors report that "TCA exposure induced a much less
intense level of PAS staining that was confined to periportal areas" but do not compare PAS
staining to controls but only to DCA treatment. In the discussion section of the paper the authors
state "Except for a small increase in liver weight and cell size, the effects produced by DCA
were not observed with TCA." Thus, there seems to be a discrepancy with regard to what the
effects of TCA are in relation to control animals from this report that has caused confusion in the
literature. Kato-Weinstein et al. (2001) reported that in male mice exposed to DCA and TCA the
DCA increased glycogen and TCA decreased glycogen content of the liver using chemical
measurement of glycogen in liver homogenates and using ethanol-fixed sections stained with
PAS, a procedure designed to minimize glycogen loss.
E.2.4.2.12. Nelson et al. (1990). Nelson et al. (1990) reported that they used the same
exposure paradigm as Herren-Freund et al. (1987), with little description of methods used
in treatment of the animals. Male B6C3F1 mice were reported to be exposed to DCA (1 or
2 g/L) or TCA (1 or 2 g/L) for 52 weeks. The number of animals examined for nontumor
tissue was 12 for controls. The number of animals varied from 2 to 8 for examination of
nontumor tissue, hyperplastic nodules, and carcinoma tissues for c-Myc expression. There
was no description for how hyperplastic nodules were defined and whether they included
adenomas and foci. For the 52-week experiments, the results were pooled for lesions that
had been obtained by exposure to the higher or lower concentrations of DCA or TCA (i.e.,
the TCA results are for lesions induced by either 1.0 g/L or 2.0 g/L TCA).
E.2.4.2.13. A second group of mice were reported to be given either DCA or TCA for 37
weeks and then normal drinking water for the remaining time till 52 weeks with no
concentrations given for the exposures to these animals. Therefore, it is impossible to
discern what dose was used for tumors analyzed for c-Myc expression in the 37-week
treatment groups and if the same dose was used for 37 and 52 week results.
E.2.4.2.14. Autoradiography was described for 3 different sections per animal in 5
different randomly chosen high power fields per section. The number of hyperplastic
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-15 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
nodules or the number of carcinomas per animal induced by these treatments was not
reported nor the criteria for selection of lesions for c-myc expression. Apparently a second
experiment was performed to determine the expression of c-H-ras. Whereas in the first
experiment there were no hyperplastic nodules, in the second 1-control animal was
reported to have a hyperplastic nodule. The number of control animals reported to be
examined for nontumor tissue in the second group was 12. The numbers of animals in the
second group was reported to vary from 1 to 7 for examination of nontumor tissue,
hyperplastic nodules, and carcinoma tissues for c-H-ras expression. The number of
animals per group for the investigation of H-ras did not match the numbers reported for
that of c-Myc. The number of animals treated to obtain the "lesion" results was not
presented (i.e., how many animals were tested to get a specific number of animals with
tumors that were then examined). The number of lesions assessed per animal was not
reported.
At 52 weeks of exposure, hyperplastic nodules (n = 8 animals) and carcinomas
(n = 6 animals) were reported to have ~2-fold expression of c-Myc relative to nontumor tissue
(n = 6 animals) after DCA treatment. After 37 weeks of DCA treatment and cessation of
exposure, there was a -30% increase in c-Myc in hyperplastic nodules (n = 4 animals) that was
not statistically significant. There were no carcinomas reported at this time.
After 52 weeks of TCA exposure, there was ~2-fold of nontumor tissue reported for c-
Myc in hyperplastic nodules (n = 6 animals) and ~3-fold reported for carcinomas (n = 6
animals). After 37 weeks of TCA exposure there was ~2-fold c-Myc in hyperplastic nodules
(n = 2 animals) that was not statistically significant and ~2.6-fold increase in carcinomas (n = 3
animals) that was reported to be statistically significant over nontumor tissue. There was no
difference in c-Myc expression between untreated animals and nontumor tissue in the treated
animals.
The authors reported that c-Myc expression in TCA-induced carcinomas was "almost 6
times that in control tissue (corrected by subtracting nonspecific binding)," and concluded that
c-Myc in TCA-induced carcinomas was significantly greater than in hyperplastic nodules or
carcinomas and hyperplastic nodules induced by DCA. However, the c-myc expression reported
as the number of grains per cells was ~2.6-fold in TCA-induced carcinomas and ~2-fold in
DCA-induced carcinomas than control or nontumor tissue at 52 weeks. The hyperplastic nodules
from DCA- and TCA-treatments at 52 weeks gave identical ratios of ~2-fold. In 3 animals per
treatment, c-Myc expression was reported to be similar in "selected areas of high expression" for
either DCA or TCA treatments of 52 weeks.
There did not appear to be a difference in c-H-ras expression between control and
nontumor tissue from DCA- or TCA-treated mice. The levels of c-H-ras transcripts were
reported to be "slightly elevated" in hyperplastic nodules induced by DCA (-67%) or TCA
(-43%) but these elevations were not statistically significant in comparison to controls.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-159 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
However, carcinomas "derived from either DCA- or TCA-treated animals were reported to have
significantly increased c-H-ras levels relative to controls." The fold increase of nontumor tissue
at 52 weeks for DCA-induced carcinomas was ~2.5-fold and for TCA induced carcinomas
~2.0-fold. Again the authors stated that "if corrected for nonspecific hybridization, carcinomas
expressed approximately 4 times as much c-H-ras than observed in surrounding tissues" Given
that control and nontumor tissue results were given as the controls for the expression increases
observed in "lesions," it is unclear what this the usefulness of this "correction" is. The authors
reported that "focal areas of increased expression of c-H-ras were not observed within
carcinomas."
The limitations of this experiment include uncertainty as to what doses were used and
how many animals were exposed to produce animals with tumors. In addition results of differing
doses were pooled and the term hyperplastic nodule, undefined. The authors state that c-Myc
expression in itself is not sufficient for transformation and that its over expression commonly
occurs in malignancy. They also state that "Unfortunately, the limited amount of tissue available
prevented a more serious pursuit of this question in the present study." In regard to the effects of
cessation of exposure, the authors do not present data on how many animals were tested with the
cessation protocol, what doses were used, and how many lesions comprised their results and
thus, comparisons between these results and those from 52 weeks of continuous exposure are
hard to make. Quantitatively, the small number of animals, whose lesions were tested, was
n = 2-4 for the cessation groups. Bull et al. (1990) is given as the source of data for the
cessation experiment (see Section E.2.3.2.1, above).
E.2.4.2.15. DeAngelo et al. (1999). The focus of this study was to "determine a dose
response for the hepatocarcinogenicity of DCA in male mice over a lifetime exposure and to
examined several modes of action that might underlie the carcinogenic process." As
DeAngelo et al. (1999) pointed out, many studies of DCA had been conducted at high
concentrations and for less than lifetime studies, and therefore, of suspect relevance to
environmental concentrations. This study is one of the few that examined DCA at a range
of exposure concentrations to determine a dose-response in mice. The authors concluded
that DCA-induced carcinogenesis was not dependent on peroxisome proliferation or
chemically sustained proliferation. The number of hepatocellular carcinomas/animals was
reported to be significantly increased over controls at all DCA treatments including 0.05
g/L and a no-observed-effect level (NOEL) not observed. Peroxisome proliferation was
reported to be significantly increased at 3.5 g/L DCA only at 26 weeks and did not correlate
with tumor response. No significant treatment effects on labeling of hepatocytes (as a
measure of proliferation) outside proliferative lesions were also reported and thus, that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-160 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
DCA-induced liver cancer was not dependent on peroxisome proliferation or chemically
sustained cell proliferation.
Male B6C3F1 mice were 28-30 days of age at the start of study and weighed 18-21 g (or
-14% range). They were exposed to 0, 0.05, 0.5, 1.0, 2.0, and 3.5 g/L DCA via drinking water
as a neutralized solution. The time-weighted mean daily water consumption calculated over the
100-week treatment period was reported to be 147, 153, 158, 151, 147, and 124 (84% of
controls) mL/kg/day for 0, 0.05, 0.5, 1, 2, and 3.5 g/L DCA, respectively. The number of
animals reported as used for interim sacrifices were 35, 30, 30, 30 and 30 for controls, 0.5, 1.0,
2.0, and 3.5 g/L DCA treated groups respectively (i.e., 10 mice per treatment group at interim
sacrifices of 26, 52 and 78 weeks). The number of animals at final sacrifice were reported to be
50, 33, 24, 32, 14 and 8 for controls, 0.05, 0.5, 1.0, 2.0, and 3.5 g/L DCA treated groups
respectively. The number of animals with unscheduled deaths before final sacrifice were
reported to be 3, 2, 1,9, 11 and 8 for controls, 0.05, 0.5, 1.0, 2.0, and 3.5 g/L DCA treated
groups respectively. The Authors reported that early mortality tended to occur from liver cancer.
The number of animals examined for pathology were reported to be 85, 33, 55, 65, 51,
and 41 for controls, 0.05, 0.5, 1.0, 2.0, and 3.5 g/L DCA treated groups respectively. The
experiment was conducted in two parts with control, 0.5, 1.0 L, 2.0, and 3.5 g/L groups treated
and then 1 months later a second group consisting of 30 control group mice and 35 mice in a
0.05 g/L DCA exposure group studied.
The authors reported no difference in prevalence and multiplicity of hepatocellular
neoplasms in the two groups so that data were summed and reported together. The number of
animals reported as examined for tumors were n= 10 animals, with controls reported to be 35
animals split among 3 interim sacrifice times—exact number per sacrifice time is unknown. The
number of animals reported "with pathology" and assumed to be included in the tumor analyses
from Table 1, and the sum of the number of animals "scheduled for sacrifice that survived till
100 weeks" and "interim sacrifices" do not equal each other. For the 1 g/L DCA exposure
group, 30 animals were sacrificed at interim periods, 32 animals were sacrificed at 100 weeks,
9 animals were reported to have unscheduled deaths, but of those 71 animals only 65 animals
were reported to have pathology for the group. Therefore, some portion of animals with
unscheduled deaths must have been included in the tumor analyses. The exact number of
animals that may have died prematurely but included in analyses of pathology for the 100 week
group is unknown.
In Figure 3 of the study, the authors reported prevalence and multiplicity of
hepatocellular carcinomas following 79 to 100 weeks of DCA exposure in their drinking water.
The number of animals in each dose group used in the tumor analysis for 100 weeks was not
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-161 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
given by the authors. Given that the authors included animals that survived past the 78 interim
sacrifice period but died unscheduled deaths in their 100 week results, the number must have
been greater than those reported as present at final sacrifice. A comparison of the data for the
100-week data presented in Table 3a and Figure 3 shows that the data reported for 100 weeks is
actually for animals that survived from 79 to 100 weeks.
The authors report a dose-response that is statistically significant from 0.5 to 3.5 g/L
DCA for hepatocellular carcinoma incidence and a dose-response in hepatocellular carcinoma
multiplicity that is significantly increased over controls from 0.05 to 0.5 g/L DCA that survived
79 to 100 weeks of exposure (i.e., 0, 8-, 84-, 168-, 315-, and 429 mg/kg/d dose groups with
prevalences of 26, 33, 48, 71, 95, and 100%, respectively, and multiplicities of 0.28, 0.58, 0.68,
1.29, 2.47, and 2.90, respectively). Hepatocellular adenoma incidence or multiplicity was not
reported for the 0.05 g/L DCA exposure group.
In Table 3 of the report, the time course of hepatocellular carcinomas and adenoma
development are given and summarized in Table E-2, below.
The authors reported hepatocellular carcinomas and number of lesions/animal in mice
that survived 79-100 weeks of exposure. They combined exposure groups to be animals after
the Week 78 sacrifice time that did and did not make it to 100 weeks. This is the same data
reported above for the 100-week exposure with the inclusion of the 0.05 g/L DCA data. The
difference between number of animals at interim and final sacrifices and those "with pathology"
and used in the tumor analysis but most likely coming from unscheduled deaths is reported in
Table E-3 as "extra" and varied across treatment groups.
Table E-2. Prevalence and Multiplicity data from DeAngelo et al. (1999)
Prevalence
Multiplicity
(lesions/animal m ± SEM)
Carcinomas
Adenomas
52 weeks control = 0% carcinomas, 0% adenoma
0
0
0.5 g/L DCA = 0/10 carcinoma, 1/10 adenomas
0
0.10 ±0.09
1.0 g/L DCA = 0/10 carcinomas, 1/10 adenomas
0
0.10 ±0.09
2.0 g/L DCA = 2/10 carcinomas, 0/10 adenomas
0.20 ±0.13
0
3.5 g/L DCA = 5/10 carcinomas, 5/10 adenomas
0.70 ±0.25
0.80 ±0.31
78 weeks control = 10% carcinomas, 10% adenomas
0.10±0.10
0.10 ±0.09
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-162 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
0.5 g/L DCA = 0/10carcinoma, 1/10 adenomas
0
0.10 ±0.09
1.0 g/L DCA = 2/10 carcinomas, 2/10 adenomas
0.20 ±0.13
0.20 ±0.13
2.0 g/L DCA = 5/10 carcinomas, 5/10 adenomas
1.0 ±0.47
1.00 ±-0.42
3.5 g/L DCA = 7/10 carcinomas, 5/10 adenomas
1.20 ±0.37
1.00 ±0.42
100 weeks control = 26% carcinoma, 10% adenoma
0.28 ±0.07
0.12 ±0.05
0.5 g/L DCA = 48%) carcinoma, 20%> adenomas
0. 68 ±0.17
0.32 ± 0.14
1.0 g/L DCA = 71%) carcinomas, 51.4% adenomas
1.29 ± 0.17
0.80 ± 0.17
2.0 g/L DCA = 95% carcinomas, 42.9% adenomas
2.47 ±0.29
0.57 ± 0.16
3.5 g/L DCA = 100%) carcinomas, 45% adenomas
2.90 ±0.40
0.64 ±0.23
Table E-3. Difference in pathology by inclusion of unscheduled deaths from
DeAngelo et al. (1999).
Dose = Prevalence of HC
#HC/animal
n = at 100 wk
Extra added in
Control = 26%
0.28
50
0
0.05 g/L = 33%
0.58
33
0
0.5 g/L = 48%
0.68
24
1
1 g/L = 71%
1.29
32
3
2 g/L_= 95%
2.47
14
7
3.5 g/L = 100%
2.9
8
3
These data show a dose-related increase in tumor formation and decrease in time-to-
tumor associated with DCA exposure at the lowest levels examined. These findings are limited
by the small number of animals examined at 100 weeks but especially those examined at
"interim sacrifice" periods (n = 10). The data illustrate the importance of examining multiple
exposure levels at lower concentrations at longer durations of exposure and with an adequate
number of animals to determine the nature of a carcinogenic response.
Preneoplastic and non-neoplastic hepatic changes were reported to have been described
previously and summarized as large preneoplastic foci observed at 52 weeks with multiplicities
of 0.1, 0.1, 0.2 and 0.16 for 0.5, 1, 2, and 3.5 g/L DCA exposure respectively. At 100 weeks all
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-163 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
values were reported to be significant (0.03, 0.06, 0.14, 0.27 for 0.5, 1, 2, and 3.5 g/L DCA
exposure respectively). Control values were not reported by the authors.
The authors reported that the prevalence and severity of hepatocellular cytomegaly and
of cytoplasmic vacuolization with glycogen deposition to be dose-related and considered
significant in all dose groups examined when compared to control liver. However, no
quantitative data were shown.
The authors reported a severity index of 0 = none, 1 = <25%, 2 = 50-75% and 4 = 75%
of liver section for hepatocellular necrosis and report at 26 weeks scores (n= 10 animals) of 0.10
± 0.10, 0.20 ± 0.13, 1.20 ± 0.38, 1.20 ± 0.39 and 1.10 ± 0.28 for control, 0.5, 1, 2, and 3.5 g/L
DCA treatment groups, respectively. Thus, there appeared to be a treatment but not dose-related
increase in hepatocellular necrosis that is does not involve most of the liver from 1 to 3.5 g/L
DCA at this time point. At 52 weeks of exposure, the score for hepatocellular necrosis was
reported to be 0, 0, 0.20 ± 0.13, 0.40 ± 0.22 and 1.10 ± 0.43 for control, 0.5, 1, 2, and 3.5 g/L
DCA treatment groups, respectively. At 78 weeks of exposure the score for hepatocellular
necrosis was reported to be 0, 0, 0, 0.30 ± 0.21 and 0.20 ± 0.13 for control, 0.5, 1, 2, and 3.5 g/L
DCA treatment groups, respectively. Finally, at the final sacrifice time when more animals were
examined, the extent of hepatocellular necrosis was reported to be 0.20 ± 0.16, 0.20 ± 0.08,
0.42 ± 0.15, 0.38 ± 0.20 and 1.38 ± 0.42 for control, 0.5, 1, 2, and 3.5 g/L DCA treatment
groups, respectively.
Thus, there was no reported increase in hepatocellular necrosis at any exposure period for
0.5 g/L DCA treatment and the mild hepatocellular necrosis seen at the three highest exposure
concentrations at 26 weeks had diminished with further treatment except for the highest dose at
up tolOO weeks of treatment. Clearly the pattern of hepatocellular necrosis did not correlate with
the dose-related increases in hepatocellular carcinomas reported by the authors and was not
increased over control at the 0.5 g/L DCA level where there was a DCA-related tumor increase.
The authors cited previously published data and state that CN-insensitive palmitoyl CoA
oxidase activity (a marker of peroxisome proliferation) data for the 26 week time point plotted
against 100 weeks hepatocellular carcinoma prevalence of animals bearing tumors was
significantly enhanced at concentrations of DCA that failed to induce "hepatic PCO" activity.
The authors reported that neither 0.05 nor 0.5 g/L DCA had any marked effect on PCO activity
and that it was "only significantly increased after 26 weeks of exposure to 3.5 g/L DCA and
returned to control level at 52 weeks (data not shown)." In regards to hepatocyte labeling index
after treatment for 5 days with tritiated thymidine, the authors reported that animals examined in
the dose-response segment of the experiment at 26 and 52 weeks were examined but no details of
the analysis were reported. The authors commented on the results from this study and a previous
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-164 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
one that included earlier time points of study and stated that there were "no significant alterations
in the labeling indexes for hepatocytes outside of proliferative lesions at any of the DCA
concentrations when compared to the control values with the exception of 0.05 g/L DCA at
4 weeks (4.8 ± 0.6 vs. 2.7 ± 0.4 control value; data not shown)."
The effects of DCA on body weight, absolute liver weight and percent liver/body weight
were given in Table 2 of the paper for 26, 52, 78 and 100 weeks exposure. For 52 and 78 week
studies 10 animals per treatment group were examined. Liver weights were not determined for
the lowest exposure concentration (0.05 g/L DCA) except for the 100 week exposure period. At
26 weeks of exposure there was not a statistically significant change in body weight among the
exposure groups (i.e., 35.4 ± 0.7, 37.0 ± 0.8, 36.8 ± 0.8, 37.9 ± 0.6, and 34.6 ± 0.8 g for control,
0.5, 1, 2, and 3.5 g/L DCA, respectively). Absolute liver weight was reported to have a dose-
related significant increase in comparison to controls at all exposure concentrations examined
with liver weight reaching a plateau at the 2 g/L concentration (i.e., 1.86 ± 0.07, 2.27 ± 0.10,
2.74 ± 0.08, 3.53 ± 0.07, and 3.55 ± 0.1 g for control, 0.5, 1, 2, and 3.5 g/L DCA, respectively).
The percent liver/body weight ratio increases due to DCA exposure were reported to have a
similar pattern of increase (i.e., 5.25% ± 0.11%, 6.12%± 0.16%, 7.44% ± 0.12%,
9.29% ± 0.08%), and 10.24%) ± 0.12%> for control, 0.5, 1, 2, and 3.5 g/L DCA, respectively).
This represented a 1.17-, 1.41-, 1.77-, and 1.95-fold of control percent liver/body weight at these
exposures at 26 weeks.
At 52 weeks of exposure there was not a statistically significant change in body weight
among the exposure groups except for the 3.5 g/L exposed group in which there was a significant
decrease in body weight (i.e., 39.9 ± 0.8, 41.7 ± 0.8, 41.7 ± 0.9, 40.8 ± 1.0, and 35.0 ± 1.1 g for
control, 0.5, 1, 2, and 3.5 g/L DCA, respectively). Absolute liver weight was reported to have a
dose-related significant increase in comparison to controls at all exposure concentrations
examined with liver weight reaching a plateau at the 2 g/L concentration (i.e., 1.87 ± 0.13,
2.39 ± 0.04, 2.92 ± 0.12, 3.47 ± 0.13, and 3.25 ± 0.24 g for control, 0.5, 1, 2, and 3.5 g/L DCA,
respectively). The percent liver/body weight ratio increases due to DCA exposure were reported
to have a similar pattern of increase (i.e., 4.68%> ± 0.30%>, 5.76%> ± 0.12%>, 7.00%> ± 0.15%>,
8.50%o ± 0.26%o, and 9.28%> ± 0.64%> for control, 0.5, 1, 2, and 3.5 g/L DCA, respectively).
For liver weight and percent liver/body weight there was much larger variability between
animals within the treatment groups compared to controls and other treatment groups. There
were no differences reported for patterns of change in body weight, absolute liver weight, and
percent liver/body weight between animals examined at 26 weeks and those examined at 52
weeks.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-165 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
At 78 weeks of exposure there was not a statistically significant change in body weight
among the exposure groups except for the 3.5 g/L exposed group in which there was a significant
decrease in body weight (i.e., 46.7 ± 1.2, 43.8 ± 1.5, 43.4 ± 0.9, 42.3 ± 0.8, and 40.2 ± 2.2 g for
control, 0.5, 1, 2, and 3.5 g/L DCA, respectively). Absolute liver weight was reported to have a
dose-related increase in comparison to controls at all exposure concentrations examined but none
were reported to be statistically significant (i.e., 2.55 ± 0.14, 2.16 ± 0.09, 2.54 ± 0.36, 3.31 ±
0.63, and 3.93 ± 0.59 g for control, 0.5, 1, 2, and 3.5 g/L DCA, respectively). The percent
liver/body weight ratio increases due to DCA exposure were reported to have a similar pattern of
increase over control values but only the 3.5 g/L exposure level was reported to be statistically
significant (i.e., 5.50% ± 0.35%, 4.93% ± 0.09%, 5.93% ± 0.97%, 7.90% ± 1.55%, and 10.14% ±
1.73%) for control, 0.5, 1, 2, and 3.5 g/L DCA, respectively).
Finally, for the animals reported to be sacrificed between 90 and 100 weeks there was
not a statistically significant change in body weight among the exposure groups except for the
2.0 and 3.5 g/L exposed groups in which there was a significant decrease in body weight (i.e.,
43.9 ± 0.8, 43.3 ± 0.9, 42.1 ± 0.9, 43.6 ± 0.7, 36.1 ± 1.2, and 36.0 ± 1.3 g for control, 0.05, 0.5, 1,
2, and 3.5 g/L DCA, respectively). Absolute liver weight did not show a dose-response pattern
at the two lowest exposure levels but was elevated with the 3 highest doses with the two highest
being statistically significant (i.e., 2.59 ± 0.26, 2.74 ± 0.20, 2.51 ± 0.24, 3.29 ± 0.21, 4.75 ± 0.59,
and 5.52 ± 0.68 g for control, 0.05, 0.5, 1, 2, and 3.5 g/L DCA, respectively). The percent
liver/body weight ratio increases due to DCA exposure were reported to have a similar pattern of
increase over control values but only the 2.0 and 3.5 g/L exposure levels were reported to be
statistically significant (i.e., 6.03%> ± 0.73%, 6.52% ± 0.55%, 6.01% ± 0.66%, 7.65% ± 0.55%,
13.30% ± 1.62%, and 15.70% ± 2.16% for control, 0.05, 0.5, 1, 2, and 3.5 g/L DCA,
respectively).
It must be recognized that liver weight increases, especially in older mice, will reflect
increased weight due to tumor burden and thus, DCA-induced hepatomegaly will be somewhat
obscured at the longer treatment durations. However, by 100 weeks of exposure there did not
appear to be an increase in liver weight at the 0.05 and 0.5 g/L exposures while there was an
increase in tumor burden reported. Examination of the 0.5 g/L exposure group from 26 to
100 weeks shows that slight hepatomegaly, reported as either absolute liver weight increase over
control or change in percent liver/body ratio, was present by 26 weeks (i.e., 22% increase in liver
weight and 17% increase in percent liver/body weight), decreased with time, and while similar at
52 weeks, was not significantly different from control values at 78 or 100 weeks durations of
exposure. However, tumor burden was increased at this low concentration of DCA.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-166 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
The authors present a figure comparing the number of hepatocellular carcinomas per
animal at 100 weeks compared with the percent liver/body weight at 26 weeks and show a linear
correlation (r = 0.9977). Peroxisome proliferation and DNA synthesis, as measured by tritiated
thymidine, were reported to not correlate with tumor induction profiles and were also not
correlated with early liver weight changes induced by DC A exposure. Most importantly, in a
paradigm that examined tumor formation after up to 100 weeks of exposure, DCA-induced
tumor formation was reported to occur at concentrations that did not also cause cytotoxicity and
at levels 20 to 40 times lower than those used in "less than lifetime" studies reporting concurrent
cytotoxicity.
E.2.4.2.16. Carter et al. (2003). The focus of this study was to present histopathological
analyses that included classification, quantification and statistical analyses of hepatic
lesions in male B6C3F1 mice receiving DCA at doses as low as 0.05 g/L for 100 weeks and
at 0.5,1.0, 2.0, and 3.5 g/L for between 26 and 100 weeks. This analysis used tissues from
the DeAngelo et al. (1999) (two blocks from each lobe and all lesions found at autopsy).
E.2.4.2.17. This study used the following diagnostic criteria for hepatocellular changes.
Altered hepatic Foci (AHF) were defined as histologically identifiable clones that were
groups of cells smaller than a liver lobule that did not compress the adjacent liver. Large
foci of cellular alteration (LFCA) were defined as lesions larger than the liver lobule that
did not compress the adjacent architecture (previously referred to as hyperplastic nodules
by Bull et al., 1990) but had different staining. These are not non-neoplastic proliferative
lesions termed "hepatocellular hyperplasia" that occur secondary to hepatic degeneration
or necrosis. Adenomas (ADs) showed growth by expansion resulting in displacement of
portal triad and had alterations in both liver architecture and staining characteristics.
Carcinomas (CAs) were composed of cells with a high nuclear-to-cytoplasmic ration and
with nuclear pleomorphism and atypia that showed evidence of invasion into the adjacent
tissue. They frequently showed a trabecular pattern characteristic of mouse hepatocellular
CAs.
The report grouped lesions as eosinophilic, basophilic and/or clear cell, and dysplastic.
"Eosinophilic lesions included lesions that were eosinophilic but could also have clear cell,
spindle cell or hyaline cells. Basophilic lesions were grouped with clear cell and mixed cell (i.e.,
mixed basophilic, eosinophilic, hyaline, and/or clear cell) lesions." The authors reported that
this grouping was necessary because many lesions had both a basophilic and clear
cell component and a few <10 % had an eosinophilic or hyaline
component.. .Lesions with foci of cells displaying nuclear pleomorphism,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-167 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
hyperchromasia, prominent nucleoli, irregular nuclear borders and/or altered
nuclear to cytoplasmic ratios were considered dysplastic irrespective of their
tinctorial characteristics.
Therefore, Carter et al. (2003) lumped mixed phenotype lesions into the basophilic grouping so
that comparisons with the results of Bull et al. (2002) or Pereira (1996), which segregate mixed
phenotype from those without mixed phenotype, cannot be done.
This report examined type and phenotype of preneoplastic and neoplastic lesions pooled
across all time points. Therefore, conclusions regarding what lesions were evolving into other
lesions have left out the factor of time. Bannasch (1996) reported that examining the evolution
of foci through time is critical for discerning neoplastic progression and described foci evolution
from eosinophilic or basophilic lesions to more basophilic lesions. Carter et al. (2003) suggested
that size and evolution into a more malignant state are associated with increasing basophilia, a
conclusion consistent with those of Bannasch (1996). The analysis presented by Carter et al.
(2003) also suggested that there was more involvement of lesions in the portal triad, which may
give an indication where the lesions arose. Consistent with the results of DeAngelo et al. (1999),
Carter et al. (2003) reported that "DCA (0.05 - 3.5 g/L) increased the number of lesions per
animal relative to animals receiving distilled water and shortened the time to development of all
classes of hepatic lesions." They also concluded that
although this analysis could not distinguish between spontaneously arising lesions
and additional lesions of the same type induced by DCA, only lesions of the kind
that were found spontaneously in control liver were found in increased numbers in
animals receiving DCA.. .Development of eosinophilic, basophilic and/or clear
cell and dysplastic AHF was significantly related to DCA dose at 100 weeks and
overall adjusted for time.
The authors concluded that the presence of isolated, highly dysplastic hepatocytes in male
B6C3F1 mice chronically exposed to DCA suggested another direct neoplastic conversion
pathway other than through eosinophilic or basophilic foci.
It appears that the lesions being characterized as carcinomas and adenomas in
DeAngelo et al. (1999) were not the same as those by Carter et al. (2003) at 100 weeks even
though they were from the same tissues (see Table E-4). Carter et al. (2003) identified all
carcinomas as dysplastic despite tincture of lesion and subdivided adenomas by tincture. If the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-168 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
differing adenoma multiplicities are summed for Carter et al. (2003) they do not add up to the
same total multiplicity of adenoma given by DeAngelo et al. (1999).
It is unclear how many animals were included in the differing groups in both studies for
pathology. The control and high-dose groups differ in respect to "animals with pathology"
between DeAngelo et al. (1999) and the "number of animals in groups" examined for lesions in
Carter et al. Neither report gave how many animals with unscheduled deaths were treated in
regards to how the pathology data were included in presentation of results. Given that DeAngelo
et al. (1999) represents animals at 100 weeks as also animals from 79-100 weeks exposure, it is
probable that the animals that died after 79 weeks were included in the group of animals
sacrificed at 100 weeks. However, the number of animals affecting that result (which would be a
mix of exposure times) for either DeAngelo et al. (1999), or Carter et al. (2003), is unknown
from published reports.
In general, it appears that Carter et al. (2003) reported more adenomas/animal for their
100 week animals than DeAngelo et al. (1999) did, while DeAngelo et al. reported more
carcinomas/ animal.
Table E-4. Comparison of data from Carter et al. (2003) and DeAngelo et al.
(1999)
Exposure




Sum of
Sum of
level of




adenomas
adenomas
DCA at
Total
Total
Total
Total
and
and
79-100
adenoma
adenoma
carcinoma
carcinoma
carcinoma
carcinoma
wk
multiplicity
multiplicity
multiplicity
multiplicity
multiplicity
multiplicity
(g/L)
(Carter)
(DeAngelo)
(Carter)
(De Angelo)
(Carter)
(DeAngelo)
0
0.22
0.12
0.05
0.28
0.27
0.40
0.05
0.48
-
<0.025
0.58
-0.50
-
0.5
0.44
0.32
0.20
0.68
0.64
1.0
1.0
0.52
0.80
0.30
1.29
0.82
2.09
2.0
0.60
0.57
1.55
2.47
2.15
3.27
3.5
1.48
0.64
1.30
2.90
2.78
3.54
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-169 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
In order to compare these data with others (eg., Pereira and Phelps, 1996) for estimates of
multiplicity by phenotype or tincture it would be necessary to add foci and LFCA together as
foci, and adenomas and carcinomas together as tumors. It would also be necessary to lump
mixed foci together as "basophilic" from other data sets as was done for Carter et al. in
describing "basophilic lesions." If multiplicity of carcinomas and adenomas are summed from
each study to control for differences in identification between adenoma and carcinoma, there are
still differences in the two studies in multiplicity of combined lesions/animal with DeAngelo et
al. (1999) giving consistently higher estimates. However, both studies show a dose response of
tumor multiplicity with DCA and a difference between control values and the 0.05 DCA
exposure level. Error is introduced by having to transform the data presented as a graph in
Carter et al. (2003). Also no SEM is given for the Carter data.
In regard to other histopathological changes, the authors report that
necrosis was found in 11.3% of animals in the study and the least prevalent toxic
or adaptive response. No focal necrosis was found at 0.5 g/L. The incidence of
focal necrosis did not differ from controls at 52 or 78 weeks and only was greater
than controls at the highest dose of 3.5 g/L at 100 weeks. Overall necrosis was
negatively related to the length of exposure and positively related to the DCA
dose. Necrosis was an early and transitory response. There was no difference in
necrosis 0 and 0.05 g/L or 0.5 g/L. There was an increase in glycogen at 0.5 g/L
at the perioportal area. There was no increase in steatosis but a dose-related
decrease in steatosis. Dysplastic LFCA were not related to necrosis indicating
that these lesions do not represent, regenerative or reparative hyperplasia.
Nuclear atypia and glycogen accumulation were associated with dysplastic
adenomas. Necrosis was not related to occurrence of dysplastic adenomas.
Necrosis was of borderline significance in relation to presence of hepatocellular
carcinomas. Necrosis was not associated with dysplastic LFCAs or Adenomas.
They concluded that "the degree to which hepatocellular necrosis underlies the carcinogenic
response is not fully understood but could be significant at higher DCA concentrations (^lg/L) "
E.2.4.2.18. Stauber and Bull (1997). This study was designed to examine the differences in
phenotype between altered hepatic foci and tumors induced by DCA and TCA. Male
B6C3F1 mice (7 weeks old at the start of treatment) were treated with 2.0 g/L neutralized
DCA or TCA in drinking water for 38 or 50 weeks, respectively. They were then treated
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-170 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
with additional exposures (n = 12) of 0, 0.02, 0.1, 0.5,1.0, 2.0 g/L DCA or TCA for an
additional 2 weeks. Three days prior to sacrifice in DCA-treated mice or 5 days for TCA-
treated mice, animals had miniosmotic pumps implanted and administered BrdU.
E.2.4.2.19. Immunohistochemical staining of hepatocytes from randomly selected fields
(minimum of 2,000 nuclei counter per animal) from 5 animals per group were reported for
14- and 28-day treatments. It was unclear how many animals were examined for 280- and
350-day treatments from the reports. The percentage of labeled cells in control livers was
reported to vary between 0.1 and 0.4% (i.e., 4-fold).
E.2.4.2.20. There was a reported ~3.5-fold of control level for TCA labeling at 14 day time
period and a ~5.5-fold for DCA. At 28 days there was ~2.5-fold of control for TCA but a
~2.3-fold decrease of control for DCA. At 280 days there was no data reported for TCA
but for DCA there was a ~2-fold decrease in labeling over control. At 350 days there was
no data for DCA but a reported ~2.3-fold decrease in labeling of control with TCA. The
authors reported that the increases at Day 14 for TCA and DCA exposure and the decrease
at Day 28 for DCA exposure were statistically significant although a small number of
animals were examined. Thus, although there may be some uncertainty in the exact
magnitude of change, there was at most ~5-fold of control labeling for DCA within after 14
days of exposure that was followed by a decrease in DNA synthesis by Day 28 of treatment.
These data show that hepatocytes undergoing DNA synthesis represented a small
population of hepatocytes with the highest level with either treatment less than 1% of
hepatocytes. Rates of cell division were reported to be less than control for both DCA and
TCA by 40 and 52 weeks of treatment.
In this study the authors reported that there was no necrosis with the 2.0 g/L DCA dose
for 52 weeks and concluded that necrosis is a recurring but inconsistent result with chronic DCA
treatment. Histological examination of the livers involved in the present study found little or no
evidence of such damage or overt cytotoxicity. It was assumed that this effect has little bearing
on data on replication rates.
Foci and tumors were combined in reported results and therefore, cannot be compared
the results Bull et al. (2002) or to DeAngelo et al. (1999). Prevalence rates were not reported.
Data were reported in terms of "lesions" with DCA-induced "lesions" containing a number of
smaller lesions that were heterogeneous and more eosinophilic with larger "lesions" tending to
less numerous and more basophilic. For TCA results using this paradigm, the "lesions" were
reported to be less numerous, more basophilic, and larger than those induced by DCA. The
DCA-induced larger "lesions" were reported to be more "uniformly reactive to c-Jun and c-Fos
but many nuclei within the lesions displaying little reactivity to c-Jun." The authors stated that
while most DCA-induced "lesions" were homogeneously immunoreactive to c-Jen and C-Fos
(28/41 lesions), the rest were stained heterogeneously. For TCA-induced lesions, the authors
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-171 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
reported not difference in staining between "lesions" and normal hepatocytes in TCA-treated
animals. Again, of note is that not only were "lesions" comprised of foci and tumors at different
stages of progression reported in these results, but that also DCA and TCA results were reported
for different durations of exposure.
E.2.4.2.21. Pereira (1996). The focus of this study was to report the dose-response
relationship for the carcinogenic activity of DCA and TCA in female B6C3F1 mice and the
characteristics of
the lesions. Female B6C3F1 mice (7-8 weeks of age) were given drinking water with either
DCA or TCA at 2.0, 6.67, or 20 mmol/L and neutralized with sodium hydroxide to a pH or
6.5-7.5. The control received 20 mmol/L sodium chloride. Conversion of mmol/L to g/L was
as follows: 20.0 mmol/L DCA = 2.58 g/L, 6.67 mmol/L DCA = 0.86 g/L, 2.0
mmol/L = 0.26 g/L, 20.0 mmol/L TCA = 3.27 g/L, 6.67 mmol/L TCA = 1.10 g/L, 2.0 mmol/L
TCA = 0.33 g/L. The concentrations were reported to be chosen so that the high concentration
was comparable to those previously used by us to demonstrate carcinogenic activity. The mice
were exposed till sacrifice at 360 (51 weeks), or 576 days (82 weeks) of exposure.
Whole liver was reported to be cut into ~3 mm blocks and along with representative
sections of the visible lesions fixed and embedded in paraffin and stained with H&E for
histopathological evaluation of foci of altered hepatocytes, hepatocellular adenomas, and
hepatocellular carcinomas. The slides were reported to be evaluated blind. Foci of altered
hepatocytes in this study were defined as containing 6 or more cells and hepatocellular adenomas
were distinguished from foci by the occurrence of compression at greater than 80% of the border
of the lesion.
Body weights were reported to be decreased only the highest dose of DCA from
40 weeks of treatment onward. For TCA there were only 2 examination periods (Weeks 51 and
82) that had significantly different body weights from control and only at the highest dose.
Liver/body weight percentage was reported in comparison to concentration graphically and
shows a dose-response for DCA with steeper slope than that of TCA at 360 and 576 days of
exposure. The authors reported that all three concentrations of DCA resulted in increased
vacuolation of hepatocytes. Such vacuolization probably due to glycogen removal from tissue
processing. Using a score of 1-3, (with 0 indicating the absence of vacuolization, +1 indicating
vacuolated hepatocytes in the periportal zone, + 2 indicating distribution of vacuolated
hepatocytes in the midzone, and +3 indicating maximum vacuolization of hepatocytes
throughout the liver), the authors also reported "the extent of vacuolization of the hepatocytes in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-172 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
the mice administered 0, 2.0, 6.67 or 20.0 mmol/1 DCA was scored as 0.0, 0.80 ± 0.08, 2.32 ±
0.11, or 2.95 ± 0.05, respectively."
Cell proliferation was reported to be determined in treatment groups containing 10 mice
each and exposed to either DCA or TCA for 5, 12, or 33 days with animals implanted with
miniosmotic pumps 5 days prior to sacrifice and administered BrdU. Tissues were
immunohistochemically stained for BrdU incorporation. At least 2,000 hepatocytes/mouse were
reported to be evaluated for BrdU-labeled and unlabeled nuclei and the BrDU-labeling index was
calculated as the percentage of hepatocytes with labeled nuclei.
Pereira (1996) reported a dose-related increase in BrDU labeling in 2,000 hepatocytes
that was statistically significant at 6.67 and 20.mmol/L DCA at 5 days of treatment but that
labeling at all exposure concentrations decreased to control levels by Day 12 and 33 of treatment.
The largest increase in BrdU labeling was reported to be a 2-fold of controls at the highest
concentration of DCA after 5 days of exposure. For TCA all doses (2.0, 6.67 and 20 mmol/L)
gave a similar and statistically significant increase in BrDU labeling by 5 days of treatment (~3-
fold of controls) but by days 12 and 33 there were no increases above control values at any
exposure level. Given the low level of hepatocyte DNA synthesis in quiescent control liver,
these results indicate a small number of hepatocytes underwent increased DNA synthesis after
DCA or TCA treatment and that by 12 days of treatment these levels were similar to control
levels in female B6C3F1 mice.
Incidence of foci and tumors in mice administered DCA or TCA (prevalence or number
of animals with tumors of those examined at sacrifice) in this report are given below in
Tables E-5 and E-6.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-173 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table E-5. Prevalence of foci and tumors in mice administered NaCl, DCA, or TCA
2	from Pereira (1996)
3
Treatment
N
Foci
Adenomas
Carcinomas
Number
%
Number
%
Number
%
82 wks
20.0 mmol NaCl
90
10
11.1
2
2.2
2
2.2
20.0 mmol DCA
19
17
89.5*
16
84.2*
5
26.3*
6.67 mmol DCA
28
11
39.3*
7
25.0*
1
3.6
2.0 mmol DCA
50
7
14.0
3
6.0
0
0
20.0 mmol TCA
18
11
61.1*
7
38.9*
5
27.8%*
6.67 mmol TCA
27
9
33.3*
3
11.1
5
18.5*
2.0 mmol TCA
53
10
18.9
4
7.6
0
0
51 wks
20.0 mmol NaCl
40
0
0
1
2.5
0
0
20.0 mmol DCA
20
8
40.0*
7
35*
1
5
6.67 mmol DCA
20
1
5
3
15
0
0
2.0 mmol DCA
40
0
0
0
0
0
0
20.0 mmol TCA
20
0
0
2
15.8
5
25*
6.67 mmol TCA
19
0
0
3
7.5
0
0
2.0 mmol TCA
40
3
7.5
3
2.5
0
0
4
5	*p < 0.05.
6
7	NaCl = sodium chloride control.
8
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-174 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table E-6. Multiplicity of foci and tumors in mice administered NaCl, DCA,
2	or TCA from Pereira (1996)
3
Treatment
N
Foci/mouse
Adenomas/mouse
Carcinomas/mouse
82 wks
20.0 mmol NACL
90
0.11 ±0.03
0.02 ±0.02
0.02 ±0.02
20.0 mmol DCA
19
7.95 ± 2.00a
5.58 ± 1.14a
0.37 ± 0.17b
6.67 mmol DCA
28
0.39 ± 0.11b
0.32 ± 0.13b
0.04 ±0.04
2.0 mmol DCA
50
0.14 ±0.05
0.06 ±0.03
0
20.0 mmol TCA
18
1.33 ± 0.3 la
0.61 ±0.22b
0.39 ± 0.16b
6.67 mmol TCA
27
0.41 ±0.13b
0.11 ±0.06
0.22 ± 0.10b
2.0 mmol TCA
53
0.26 ±0.08
0.08 ±0.04
0
51 wks
20.0 mmol NACL
40
0
0.03 ±0.03
0
20.0 mmol DCA
20
0.60 ± 0.22a
0.45 ± 0.17a
0.10±0.10
6.67 mmol DCA
20
0.05 ±0.05
0.20 ±0.12
0
2.0 mmol DCA
40
0
0
0
20.0 mmol TCA
20
0
0.15 ± 0.11
0.50 ± 0.18b
6.67 mmol TCA
19
0
0.21 ±0.12
0
2.0 mmol TCA
40
0.08 ± 0.04
0.08 ±0.04
0
4
5	><0.01.
6	hp< 0.05.
7
8	NaCl = sodium chloride control.
9
10
11	These data show the decreased power of using fewer than 50 mice, especially at shorter
12	durations of exposure. By 82 weeks of exposure increased adenoma and carcinomas induced by
13	TCA or DCA treatment are readily apparent.
14	The foci of altered hepatocytes and the tumors obtained from this study were reported to
15	be basophilic, eosinophilic, or mixed containing both characteristics and are shown in Tables E-7
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-175 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
and E-8. DC A was reported to induce a predominance of eosinophilic foci and tumors, with over
80% of the foci and 90% of the tumors in the 6.67 and 20.0 mmol/L concentration groups being
eosinophilic. Only approximately half of the lesions were characterized as eosinophilic with the
rest being basophilic in the group administered 2.0 mmol/L DCA. The eosinophilic foci and
tumors were reported to consistently stained immunohistochemically for the presence of GST-71,
while basophilic lesions did not stain for GST-71, except for a few scattered cells or small areas
comprising less than 10% of foci.
The foci of altered hepatocytes in the TCA treatment groups were approximately equally
distributed between basophilic and eosinophilic in tincture. However, the tumors were
predominantly basophilic lacking GST-pi (21 of 28 or 75%) including all 11 hepatocellular
carcinomas. The limited numbers of lesions, i.e., 14, in the sodium chloride (vehicle control)
group were characterized as 64.3, 28.6, and 7.1% basophilic, eosinophilic, and mixed,
respectively.
Table E-7. Phenotype of foci reported in mice exposed to NaCl, DCA, or
TCA by Pereira (1996)
Treatment

% Foci
at 51 and 82 wk
N
Basophilic
Eosinophilic
Mixed
20.0 mmol NaCl
10
70
30
0
20.0 mmol DCA
150
3.3
96.7
0
6.67 DCA
11
18.2
81.8
0
2.0 mmol DCA
7
42.8
57.2
0
20.0 mmol TCA
22
36.4
54.6
9.1
6.67 mmol TCA
11
45.5
54.5
0
2.0 mmol TCA
13
38.5
61.5
0
NaCl = sodium chloride control.
Table E-8. Phenotype of tumors reported in mice exposed NaCl, DCA, or
TCA by Pereira (1996)
Treatment
N
Tumors
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-176 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
at 51 and 82 wk

Basophilic
Eosinophilic
Mixed
20.0 mmol NaCl
4
50
25
25.5
20.0 mmol DCA
105
2.9
96.1
1
6.67 DCA
10
10
90
0
2.0 mmol DCA
3
0
100
0
20.0 mmol TCA
18
61.1
22.2
16.7
6.67 mmol TCA
6
100
0
0
2.0 mmol TCA
4
100
0
0
NaCl = sodium chloride control.
These data for female B6C3F1 mice show that DC A and TCA treatment induced a
mixture of basophilic or eosinophilic foci. The pooling of the data between time and adenoma
versus carcinoma decreases the ability to ascertain the phenotype of tumor due to treatment or
the progression of phenotype with time as well as the small number of tumor examined at lower
exposure concentrations. Foci that occurred at 51 and 82 weeks were presented as one result.
Adenomas and carcinoma data were pooled as one endpoint (n = number of total foci or tumors
examined). Therefore, evolution of phenotype between less to more malignant stages of tumor
were lost.
E.2.4.2.22. Pereira and Phelps (1996). The focus of this study was to determine tumor
response and phenotype in methyl nitrosourea (MNU)-treated mice after DCA or TCA
exposure. The concentrations of DCA or TCA were the same as Pereira (1996). For
Pereira (1996) the animals were reported to be 7-8 weeks of age when started on treatment
and sacrificed after 360 or 576 days of exposure (51 or 82 weeks). For this study and Tao et
al. (2004b), animals were reported to be 6 weeks of age when exposed to DCA or TCA via
drinking water and to be 31 or 52 weeks of age at sacrifice. Thus, exposure time would be
~24 or 45 weeks. A control group of non-MNU treated animals was presented for female
B6C3F1 mice treated for 31 or 52 weeks and are discussed in Table E-9, below.
Although this paradigm appears to be the same paradigm as those reported in Pereira
(1996), fewer animals were studied. The number of animals in each group varied between
8 controls and 14 animals in the 2.0 mmol/L treatment groups. In mice that were not treated with
MNU but were treated with either DCA or TCA at 31 weeks, there were no reported statistically
significant treatment-related effect upon the yield of foci or altered hepatocytes and liver tumors
but the number of animals examined was small and therefore, of limited power to detect a
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-177 DRAFT—DO NOT CITE OR QUOTE
-------
1	response. The results below indicate a DCA-related increase in foci and percentage of mice with
2	foci.
3	See Section E.4.2.3 for further discussion of the results of coexposures to MNU and DCA
4	or TCA from this study.
5
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-178 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Table E-9. Multiplicity and incidence data (31 week treatment) from Pereira
and Phelps (1996)
Treatment
No
Foci/mouse
incidence %
Adenomas/mouse
incidence %
20.0 mmol NaCl
15
0.13 ± 0.13
6.7
0.13 ± 0.13
not reported
20.0 mmol DCA
10
0.40 ±0.16
40
0
0
6.67 DCA
10
0.10±0.10
10
0
0
2.0 mmol DCA
15
0
0
0
0
20.0 mmol TCA
10
0
0
0
0
6.67 mmol TCA
10
0
0
0
0
2.0 mmol TCA
15
0
0
0
0
NaCl = sodium chloride control.
E.2.4.2.23. Ferreira-Gonzalez et al. (1995). The focus of this study was the investigation of
differences in H-ras mutation spectra in hepatocellular carcinomas induced by TCA or
DCA in
male B6C3F1 mice. 28-day old mice were exposed for 104 weeks to 0. 1.0 g or 3.5 g/L DCA or
4.5 g/L TCA that was pH adjusted. Tumors observed from this treatment were diagnosed as
either hepatocellular adenomas or carcinomas. DNA was extracted from either spontaneous,
DCA- or TCA-induced hepatocellular carcinomas. Samples for analysis were chosen randomly
in the treatment groups of which 19% of untreated mice had spontaneous liver hepatocellular
carcinomas (0.26 carcinomas/animal).
DCA treatment induced 100% prevalence at 3.5 g/L (5.06 carcinomas/animal) and 10.6%
carcinomas at 1.0 g/L (1.29 carcinomas/animal). TCA treatment was reported to induce 73.3%
prevalence at 4.5 g/L (1.5 carcinomas/animal). The number of samples analyzed was 32 for
spontaneous carcinomas, 33 for mice treated with 3.5 g/L DCA, 13 from mice treated with
1.0 g/DCA, and 11 from mice treated with 4.5 g/L TCA.
This study has the advantage of comparison of tumor phenotype at the same stage of
progression (hepatocellular carcinoma), for allowance of the full expression of a tumor response
(i.e., 104 weeks), and an adequate number of spontaneous control lesions for comparison with
DCA or TCA treatments. However, tumor phenotype at an endstage of tumor progression
reflects of tumor progression and not earlier stages of the disease process.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-179 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
There were no ras mutations detected except at H-61 in DNA from spontaneously arising
tumors of control mice. Only 4/57 samples from carcinogen-treated mice were reported to
demonstrate mutation other than in the second exon of H-ras. In spontaneous liver carcinomas,
58% were reported to show mutations in H-61 as compared with 50% of tumor from 3.5 g/L
DCA-treated mice and 45% of tumors from 4.5.g/L TCA-treated mice. Thus, there was a
heterogeneous response for this phenotypic marker for the spontaneous, DCA-, and TCA-
treatment induced hepatocellular carcinomas.
All samples positive for mutation in the exon 2 of H-ras were sequenced for the
identification of the base change responsible for the mutation. The authors noted that H-ras
mutations occurring in spontaneously developing hepatocellular carcinomas from B6C3F1 male
mice are largely confined to codon 61 and involve a change from CAA to either AAA or CGA or
CTA in a ratio of 4:2:1. They noted that in this study, all of the H-ras second codon mutations
involved a single base substitution in H-61 changing the wild-type sequence from CAA to AAA
(80%>), CGA (20%) or CTA for the 18 hepatocellular carcinomas examined.
In the 16 hepatocellular carcinomas from 3.5 g/L DC A treatment with mutations, 21%
were AAA transversions, 50% were CGA transversions, and 29% were CTA transversions. For
the 6 hepatocellular carcinomas from 1.0 g/L DC A with mutations, 16% were an AAA
transversion, 50% were a CGA transversion, and 34% were a CTA transversion. For the 5
hepatocellular carcinomas from 4.5 g/L TCA with mutations, 80% were AAA transversions,
20%) CGA tranversions, and 0% were CTA transversions. The authors note that the differences
in frequency between DCA and TCA base substitutions did not achieve statistical significance
due to the relatively small number of tumors from TCA-treated mice. They note that the finding
of essentially equal incidence of H-ras mutations in spontaneous tumors and in tumors of
carcinogen-treated mice did not help in determining whether DCA and TCA acted as
"genotoxic" or "nongenotoxic" compounds.
E.2.4.2.24. Pereira et al. (2004b). Pereira et al. (2004b) exposed 7-8 week old female
B6C3F1 mice treated with "AIN-76A diet" to neutralized 0, or 3.2 g/L DCA in the drinking
water and 4.0
or 8.0 g/kg L-methionine added to their diet. The final concentration of methionine in the diet
was estimated to be 11.3 and 15.3 g/kg. Mice were sacrifice 8 and 44 weeks after exposure to
DCA with body and liver weights evaluated for foci, adenomas, and hepatocellular carcinomas.
No histological descriptions were given by the authors other than tinctoral phenotype of foci and
adenomas for a subset of the data. The number of mice examined was 36 for the DCA + 8.0 g/kg
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-180 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
methionine or 4.0 g/kg methionine group sacrificed at 44 weeks. However, for the DCA-only
treatment group the number of animals examined was 32 at 44 weeks and for those groups that
did not receive DCA but either methionine at 8.0 or 4.0 g/kg, there were only 16 animals
examined. All groups examined at 8 weeks had 8 animals per group.
Liver glycogen was reported to be isolated from 30-50 mg of whole liver. Peroxisomal
acyl-CoA oxidase activity was reported to be determined using lauroyl-CoA as the substrate and
was considered a marker of peroxisomal proliferation. Whole liver DNA methylation status was
analyzed using a 5-MeC antibody.
Methionine (8.0 g/kg) and DCA coexposure was reported to result in the death of 3 mice
while treatment with methionine (4.0 g/kg) and DCA or methionine (8.0 g/kg) alone was
reported to kill one mouse in each group. The authors reported that "There was an increased in
body weight during weeks 12 to 36 in the mice that received 8.0 g/kg methionine without DCA.
There was no other treatment-related alteration in body weight." However, the authors do not
present the data and initial or final body weights were not presented for the differing treatment
groups.
DCA treatment was reported to increase percent liver/body weight ratios at 8 and
44 weeks to about the same extent (i.e., ~2.4-fold of control at 8 weeks and 2.2-fold of control at
44 weeks). Methionine coexposure was reported to not affect that increase (-2.4-, 2.2-, and
2.1-fold of control after DCA treatment alone, DCA/4 g/kg methionine, and DCA/8 mg/kg
methionine treatment for 8 weeks, respectively). There was a slight increase in percent
liver/body weight ratio associated with 8.0 g/kg methionine treatment alone in comparison to
controls (-7%) at 8 weeks with no difference between the two groups at 44 weeks.
After 8 weeks of only DCA exposure, the amount of glycogen in the liver was reported to
be ~2.09-fold of the value for untreated mice (115 vs. 52.5 mg/g glycogen in treated vs. control,
respectively at 8 weeks). Both 4 g/kg and 8 g/kg methionine coexposure reduced the amount of
DCA-induced glycogen increase in the liver (~1.64-fold of control for DCA/4.0 g/kg methionine
and ~1.54-fold of control for DC A/8.0 mg/kg methionine). Thus, for treatment with DCA alone
or with the two coexposure levels of methionine, the magnitude of the increase in liver weight
was greater than that of the increase in liver glycogen (i.e., 2.42- vs. 2.09-fold of control percent
liver/body weight vs. glycogen content for DCA alone, 2.20- vs. 1.64-fold of control percent
liver/body weight vs. glycogen content for DCA/4.0 g/kg methionine, 2.10- vs. 1.54-fold of
control percent liver/body weight vs. glycogen content for DCA/8.0 g/kg methionine). Thus, the
magnitudes of treatment-related increases were higher for percent liver/body weight than for
glycogen content in these groups.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-181 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In regard to percentage of liver mass that glycogen represented, the control value for this
study is similar to that presented by Kato-Weinstein et al. (2001) in male mice (-60 mg glycogen
per gram liver) and represents -6% of liver mass. Therefore, a doubling of the amount of
glycogen is much less than the 2-fold increases in liver weight observed for DCA exposure in
this paradigm. These data suggest that DCA-related increases in liver weight gain are not only
the result of increased glycogen accumulation, and that methionine coexposure is affecting
glycogen accumulation to a much greater extent than the other underlying processes that are
contributing to DCA-induced hepatomegaly after 8 weeks of exposure. The authors reported that
8-weeks of DCA exposure alone did not result in a significant increase in cell proliferation as
measured by PCN index (neither data nor methods were shown). This is consistent with other
data showing that DCA effects on DNA synthesis were transient and had subsided by 8 weeks of
exposure.
The levels of lauroyl-CoA oxidase activity were reported to be increased (~1.33-fold of
control) by DCA treatment alone at 8 weeks and to be slightly reduced by 8 g/kg methionine
treatment alone (~0.83-fold of control). Methionine coexposure was reported to have little effect
on DCA-induced increases in lauroyl-CoA oxidase activity. The levels of DNA methylation
were reported to be increased by 8.0 g/kg methionine only treatment at 8 weeks ~1.32-fold of
control, and reduced by DCA only treatment to ~0.44-fold of control. DCA and 4.0 g/kg
methionine coexposure gave similar results as controls (within 2%). Coexposures of DCA and
8.0 g/kg methionine treatments were reported to increase DNA methylation 1.22-fold of controls
after 8 weeks of coexposure.
In the 44-week study, the authors reported that foci and hepatocellular adenomas were
found. However, the authors do not report the incidences of these lesions in their study groups
(how many of the treated animals developed lesions). As noted above, the numbers of animals in
these groups varied widely between treatments (e.g., n = 36 for DCA and coexposure to 8.0 g/kg
methionine but only n = 16 for 8 g/kg methionine treatment alone). Although reporting
unscheduled deaths in the 8.0 g/kg methionine and DCA coexposure groups, the authors did not
indicate whether these mortalities occurred in the 44-week or 8-week study groups.
Multiplicities of foci and adenoma data were presented. DCA was reported to induce
2.42 ± 0.38 foci/mouse and 1.28 ± 0.31 adenomas/mouse (m ± SE) after 44 weeks of treatment.
The DCA-induced foci and adenomas were reported to stain as eosinophilic with "relatively
large hepatocytes and nuclei." The authors did not present data on the percent of foci and
adenomas that were eosinophilic using this paradigm. The addition of 4.0 or 8.0 g/kg methionine
to the AIN-76A diet was reported to reduce the number of DCA-induced adenomas/mouse to
0.167 ± 0.093 and 0.028 ± 0.028, respectively. However, the addition of 4.0 g/kg methionine to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-182 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
the DCA treatment was reported to increase the number of foci/mouse (3.4 ± 0.46 foci/mouse).
The addition of 8.0 g/kg methionine to the DCA treatment was reported to yield
0.94 ± 0.24 foci/mouse. There were no foci or tumors in the 16 mice that received either the
control diet or the 8.0 g/kg methionine treatment without DCA. The authors did not report
whether methionine treatment had an effect on the tincture of the foci or adenomas induced by
DCA.
Therefore, a very high level of methionine supplementation to an AIN-760A diet, was
shown to affect the number of foci and adenomas, i.e., decrease them, after 44 weeks of
coexposure to very high exposure concentration of DCA. However, a lower level of methionine
coexposure increased the incidence of foci at the same concentration of DCA. Methionine
treatment alone at the 8 g/kg level was reported to increase liver weight, decrease lauroyl-CoA
activity and to increase DNA methylation.
No histopathology was given by the authors to describe the effects of methionine alone.
Coexposure of methionine with 3.2 g/L DCA was reported to decrease by -25% DCA-induced
glycogen accumulation and increase mortality, but not to have much of an effect on peroxisome
enzyme activity (which was not elevated by more than 33% over control for DCA exposure
alone). The authors suggested that their data indicate that methionine treatment slowed the
progression of foci to tumors. Whether, these results would be similar for lower concentrations
of DCA and lower concentrations of methionine that were administered to mice for longer
durations of exposure, cannot be ascertained from these data. It is possible that in a longer-term
study, the number of tumors would be similar. Whether, methionine treatment coexposure had
an effect on the phenotype of foci and tumors was not presented by the authors in this study.
Such data would have been valuable to discern if methionine coexposure at the 4.0 mg/kg level
that resulted in an increase in DCA-induced foci, resulted in foci of a differing phenotype or
resulted in a more heterogeneous composition than DCA treatment alone.
E.2.4.2.25. DeAngelo et al. (2008). In this study, neutralized TCA was administered in
drinking water to male B6C3 F1 mice (28-30 days old) in three studies. In the first, study
control animals
received 2 g/L sodium chloride while those in the second study were given 1.5 g/L neutralized
acetic acid (HAC) to account for any taste aversion to TCA dosing solutions. In a third study
deionized water served as the control.
No differences in water uptake were reported. Mean initial weights were reported to not
differ between the treatment groups (19.5 ± 2.5 g - 21.4 ± 1.6 g or -10% difference). The first
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-183 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
study was reported to be conducted at the U.S. EPA laboratory in Cincinnati, OH in which mice
were exposed to 2 g/L sodium chloride, or 0.05, 0.5, or 5 g/L TCA in drinking water for
60 weeks. There were 5 animals at each concentration that were sacrificed at 4, 15, 31, and
45 weeks with 30 animals sacrificed at 60 weeks of exposure. There were 3 unscheduled deaths
in the 0.05 g/L TCA group leaving 27 mice at final necropsy. For the other exposure groups
there were 29 or 30 animals at final necropsy.
In the second study, also conducted in the same laboratory, mice were reported to be
exposed to 1.5 g/L neutralized acetic acid or 4.5 g/L TCA for 104 weeks. Serial necropsies were
conducted (5 animals per group) at 15, 30, and 45 weeks of exposure and on, 10 animals in the
control group at 60 weeks. For this study, a total of 25 animals were sacrificed in interim
necropsies in the 1.5 g/L HAC group and 15 in the 4.5 g/L TCA group. There were 7
unscheduled deaths in the HAC group and 12 in the 4.5 g/L TCA group leaving 25 animals at
final necropsy and 30 animals in the final necropsy groups, respectively.
Study 3 was conducted at the U.S. EPA laboratory in RTP, NC. Mice were exposed to
deionized water or 0.05 or 0.5 g/L TCA in the drinking water for 104 weeks with serial
necropsies (n = 8 per group) conducted at 26, 52, and 78 weeks. There were 19-21 animals
reported at interim sacrifices and 17 unscheduled deaths in the deionized water group, 24
unscheduled deaths in the 0.05 g/L TCA group, and 24 unscheduled deaths in the 0.5 g/L TCA
group. This left 34 mice at final necropsy in the control group, 29 mice in the 0.05 g/L TCA
group, and 27 mice in the 0.5 g/L group.
At necropsy, liver, kidneys, spleen and testes weights were reported to be taken and
organs examined for gross lesions. Tissues were prepared for light microscopy and stained with
H& E. At termination of the exposure periods, a complete rodent necropsy was reported to be
performed. Representative blocks of tissue were examined only in 5 mice from the high dose
and control group with the exception of gross lesions, liver, kidney, spleen and testis at interim
and terminal sacrifices. If the number of any histopathologic lesions in a tissue was
"significantly increased above that in control animals" then that tissue was reported to be
examined in all TCA dose groups.
For Study #3 a second contract pathologist reviewed 10% of the described hepatic
lesions. No "major differences" were reported between the two pathologic diagnoses.
The prevalence and multiplicity of hepatic tumors were reported to be derived by
performing a histopathologic examination of surface lesions and four sections cut from each of
four tissue blocks excised from each liver lobe. Tumor prevalence was reported to be calculated
as the percentage of the animals with a neoplastic lesion compared to the number of animals
examined. Tumor multiplicity was reported to be calculated by dividing the number of each
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-184 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
lesion or combined adenomas and carcinomas by the number of animals examined.
Preneoplastic large foci of cellular alteration were also observed over the course of the study.
The prevalence and severity of hepatocellular cytoplasmic alterations, inflammation, and
necrosis were reported to be determined using a scale based on the amount of liver involved of
1 = minimal (occupying 25%), 2 = mild (occupying 25-50%), 3 = moderate (occupying
50-75%)) and 4 = marked (occupying >15%). The only "significant change outside of the liver"
was reported to be testicular degeneration.
LDH was determined in arterial blood collected at 30 and 60 weeks (Study 1) and 4, 30,
and 104 weeks (Study 2). Cyanide insensitive PCO was also reported to be measured. Five days
prior to sacrifice, tritiated thymidine (Studies 1 and 2) or BrdU (Study 3) was administered via
miniosmotic pumps and the number of hepatocyte nuclei with grain counts >6 were scored in
1,000 cells or chromogen pigment over nuclei (BrdU). The labeling index was calculated by
dividing the number of labeled hepatocyte nuclei by total number of hepatocytes scored.
Total neoplastic and preneoplastic lesions (multiplicity) were counted individually or
combined (adenomas and carcinomas) for each animal. The analysis of tumor prevalence data
were reported to include only those animals examined at the scheduled necropsies or animals
surviving to Week 60 (Study 1) or longer than 78 weeks (Studies 2 and 3). The data from all the
scheduled necropsies was combined for an overall test of treatment-related effect.
For Study #1 (60-week exposure) all TCA treated groups experienced a decrease in
drinking water consumption with the decreases in drinking water for the 0.5 and 5 g/L TCA
exposure groups reported as statistically significant by the authors. The water consumption in
mL/kg-day was reported to be reduced by 11, 17, and 30% in the 0.05, 0.5, and 5 g/L TCA
treated groups compared to 2 g/L NaCl control animals as measured by time-weighted mean
daily water consumption measured over the study. The control value was reported to be
171 mL/kg/day. Although the 0.05 g/L exposure concentrations were not measured, the 0.5 and
5 g/L solutions were within 4% of target concentrations. The authors estimated that the mean
daily doses were 0, 8 mg/kg, 68 mg/kg and 602 mg/kg per day.
For the 102 week studies the mean water consumption with deionized water was
reported to be 112 mL/kg/day and 132 mL/kg/day for control animals given 1.5 g/L HAC.
Therefore, there appeared to be a 35% decrease in water consumption between the controls in
Study #1 given 2 g/L NaCl and controls in a Study #3 given deionized water but conducted at a
different laboratory. There appeared to be a 23% reduction in water consumption between
animals given 2 g/L NaCl and those given 1.5 g/L HAC at the same laboratory (Study #2).
As the concentrations of TCA were increased, there would be a corresponding increase in the
amount of sodium hydroxide needed to neutralize the solutions and a corresponding increase in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-185 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
salts in the solution as well as TCA. The authors did not address nor discuss the differences in
drinking water consumption between the differing control solutions between the studies.
DeAngelo et al. (1999) reported mean drinking water consumption of 147 mL/kg/day in
control mice of over 100 weeks and that the highest dose of DC A (3.5 g/L) reduced drinking
water consumption by 26%. Carter et al. (1995) reported that DCA at 5 g/L to decrease drinking
water consumption by 64 and 46% but 0.5 g/L DCA to not affect drinking water consumption.
In this study, while reporting that Study #1 showed that increasing TCA concentration decreased
drinking water consumption, the drinking water consumption in Studies #2 and #3 were similar
between controls and TCA exposure groups with both being less than the control and low TCA
concentration values reported in Study #1 (i.e., in Study #2 the 1.5 g/L HAC and 4.5 g/L TCA
drinking water consumption was -130 mL/kg/day and in Study #3 the drinking water
consumption was ~112 mL/kg/day for the deionized water control and 0.05 g/L and 0.5 g/L TCA
exposure groups). Thus, the drinking water concentrations for Study #3 was -35% less than for
the control values for Study #1 and was also -25% less than for DeAngelo et al. (1999). The
reasons for the apparently lower drinking water averages for Study #3 and the lack of effect of
the addition of 0.5 g/L TCA that was reported in Study #1 and in other studies, was not discussed
by the authors.
In Study #1, there was little difference between exposure groups (n = 5) noted for the
final body weights (mean range of 27.6-28.1 g) in mice sacrificed after 4 weeks of exposure.
However, absolute liver weight and percent liver/body weight ratios increased with TCA dose.
The percent liver/body weight ratios were 5.7% ± 0.4%, 6.2% ± 0.3%, 6.6% ± 0.4%, and
7.7%) ± 0.6% for the 2 g/L NaCl control, 0.05, 0.5, and 5 g/L TCA exposure groups, respectively.
These represent 1.09-, 1.16-, and 1.35-fold of control levels that were statistically significant.
At 15 weeks of exposure the fold increases in percent liver/body weight ratios were 1.14-,
1.16-, and 1.47-fold of controls for 0.05, 0.5, and 5 g/L TCA. At 31 weeks of exposure the fold
increases in percent liver/body weight ratios were 0.98-, 1.09-, and 1.59-fold of controls for 0.05,
0.5, and 5 g/L TCA. At 45 weeks of exposure the fold increases in percent liver/body weight
ratios were 1.13-, 1.45-, and 1.98-fold of controls for 0.05, 0.5, and 5 g/L TCA. At 60 weeks of
exposure the percent liver/body weight ratios were 0.94-, 1.25-, 1.60-fold of controls for 0.05,
0.5, and 5 g/L TCA.
Thus, the range of increase at the lowest level of TCA exposure (i.e., 0.05 g/L) was 0.94-
to 1.14-fold of controls. These data consistently show TCA-induced increases in liver weight
from 4 to 60 weeks of the study that were dose-related. For the 0.5 g/L exposure group, the
magnitude of the increase compared to control was reported to be about the same between weeks
4 and 30 with the highest increase reported to be at Week 45 (1.45-fold of control). In regard to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-186 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
the correspondence with magnitude of difference in dose of TCA and liver weight increase, there
was ~2-fold increase in liver weight gain corresponding to 10-fold increases in TCA
concentration at 4 weeks of exposure. For the 4 and 15-week exposures there was -3.3- and 3.9-
fold difference in liver weight that corresponded to a 100-fold difference in exposure
concentration of TCA (i.e., 0.05 vs. 5.0 g/L TCA).
The small number of animals examined, n = 5, limit the power of the study to determine
the change in percent liver/body weight up to 45 weeks, especially at the lowest dose. However,
the 0.05 g/L TCA exposure groups at 4 week and 15 weeks were reported to significantly
increase percent liver/body weight ratios.
The percent liver/body weight ratios for all of the treatment groups and the ability to
detect significant changes were affected by changes in final body weight and changing numbers
of animals. After 4 to 30 weeks of exposure, the final body weights of mice increased in control
animals but were within 11% of each other between weeks 31 and 60. The percent liver/body
weight ratios in controls decreased from 4 to 31 weeks and were slightly elevated by 60 weeks
compared to the 31-week level. Although control values were changing, there appeared to be no
difference between control values and treated values in final body weight for any duration of
exposure with the exception of the 5 g/L TCA exposure group after 60 weeks of exposure, which
was decreased by -15%. At the 31-week and 60-week exposure durations, the 0.05 g/L TCA
groups did not have increased percent liver/body weight ratios over controls.
In Study #2, conducted in the same laboratory but with a 1.5 g/L HAC solution used for
control groups, there was less than 5% difference in final body weights between control mice
give HAC and those treated with 4.5 g/L TCA up to 45 weeks. However, final body weight was
reduced by TCA treatment by 104 weeks by -15%. Between the interim sacrifices of 15, 30, and
45 weeks, the percent liver/body weight ratios in control mice were similar at 15 and 45 weeks
(~4.8%>) but greater in the 30-week control group (5.3% or —10% greater than other interim
control groups). The TCA-induced increases in body weight were 1.60-, 1.40-, and 1.79-fold of
control for the 15, 30, and 45 week groups exposed to 4.5 g/L TCA in Study #2. The smaller
magnitude of TCA-induced liver weight increase at 30-weeks that that for 15 and 45 weeks, was
a reflection of the increased percent liver/body weight ratio reported for the HAC control at that
time point.
Comparisons can be made between Study #1 and Study #2 for 4.5 g/L or 5.0 g/L TCA
exposure levels and controls for 15, 30/31 and 45 weeks of exposure to ascertain the consistency
of response from the same laboratory. Although the two studies had differing control solutions
and reported different drinking water consumption overall, they were exposing the TCA groups
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-187 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
to almost the same concentration of TCA in the same buffered solutions for the same periods of
time with the same number of mice per group.
Between Study #1 and Study #2, there were consistent percent liver/body weight ratios
induced by either 5.0 g/L TCA and 4.5 g/L TCA at weeks 15 and 30/31 (i.e., within 3% of each
other). The percent liver/body ratios for these exposure groups ranged from 7.3-7.7% between
weeks 15 and 30/31 for the -5.0 g/L TCA exposure in both studies. Final body weights were
within 10%. While the percent liver/body weight ratios induced by -5.0 g/L TCA were similar,
the magnitude of increase in comparison to the controls was 1.47- and 1.59-fold of control for
Study #1, and 1.60- and 1.40-fold of control for Study #2 after 15 and 30/31 weeks of exposure,
respectively. At 45 weeks, the percent liver/body weight ratios were within 11% of each other
(9.4 vs. 8.4%>) and final body weights were within 2% of each for this exposure concentration
between the two studies giving a 1.98- and 1.79-fold of control percent liver/body weight,
respectively. Thus, the apparent magnitude of TCA-induced increase in percent liver/body
weight was affected by control values used as the basis for comparison. The percent liver/body
weights reported for either 4.5 g/L TCA or 5.0 g/L TCA exposure groups for weeks 15 and 30/31
was similar between the two studies conducted in the same laboratory.
Study #3 was conducted in a separate laboratory, interim sacrifice times were not the
same as for Study #1, the number of animals examined differed (n = 5 for Study #1 and n = 8 for
Study #3), and control animals studied for comparative purposes were given different drinking
water solutions (deionized water vs. 2 g/L NaCl). Most importantly the body weights reported at
52 weeks were much greater than that reported at 45 weeks for Studies #1 and #2.
However, a comparison of TCA-induced liver weight gain and the effects of final body
weight can be made between the 0.05 and 0.5 g/L TCA exposure groups at 30 weeks (Study #1)
and 26 weeks (Study #3), at 45 weeks and 60 weeks (Study #1), and 52 weeks (Study #3). At 31
weeks there was <2% difference in mean final body weights between control and the two TCA-
treatment groups in Study #1. There was also little difference between the TCA-treated groups
at week in Study #3 at Week 26 and the TCA treatment groups in at Week 31 in Study #1 (i.e.,
range of 42.6-43.5 g for 0.05 and 0.5 g/L TCA treatments in Studies #1 and #3). However, in
Study #3, the control value was 12%> lower than that of Study #1 for mean final body weight.
Based on final body weights, there would be an expectation of similar results between the two
studies at the 26- and 30-week time points.
At the 45 week (Study #1), and 52-week (Study #3), and 60-week (Study #1) durations of
exposure, the mean final body weights varied little between their corresponding control groups at
each sacrifice time (less than 4%> variation between control and TCA-treated groups). However,
there was variation in mean final body weights between the differing sacrifice times. Control
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-188 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
and TCA-treated groups were reported to have lower mean final body weights at 45 weeks of
exposure in Study #1 than at either 30 weeks or at 60 weeks. The 45-week mean final body
weights in Study #1 were also reported to be lower than those at 52 weeks in Study #3. Control
mean body weight values were 28% higher at 52 weeks in Study #3 than 45 weeks in Study #1
and 15% higher for 60 weeks in Study #1. In essence, for Study #1 mean final body weights
went down between 31 and 45 weeks of exposure and then went back up at 60 weeks of
exposure for control mice (-43, -40, and -44 g for 31, 45, and 60 weeks, respectively) as well as
for both TCA concentrations. However, for Study #3 final mean body weights went up between
26 and 52 weeks of exposure for control mice (-39 vs. -51 g) and for both TCA concentrations.
While for Study #1 the percent liver/body weight ratios were 0.98- and 1.09-fold of
control at 31 weeks of exposure, at Week 45 the ratios were 1.13- and 1.45-fold of control, and at
Week 60 they were 0.94- and 1.25-fold of controls for the 0.05 and 0.5 g/L TCA exposure levels,
respectively. For Study #3, the pattern differed than that of Study #1. There was a 1.07- and
1.18-fold of control percent liver/body weight for 26 weeks but a 0.92- and 1.04-fold of control
percent liver/body weight change at 52 weeks of exposure at 0.05 and 0.5 g/L TCA exposure,
respectively.
Thus, there appeared to be differences in control and the treatment groups at the 26 week
sacrifice groups in Study #3 that was not apparent at the 52-week sacrifice time. Overall, the
final body weights appeared to be similar between controls and TCA treatment groups at the
52-week sacrifice time in Study #3 and at the 31-, 45-, and 60-week sacrifice times in Study #1.
However, although consistent within sacrifice times, the final body weights differed between the
various sacrifice times in Studies #1 and #3. The patterns of percent liver/body weight at
differing and similar sacrifice times appeared to differ between the Study #1 and Study #3 at the
same concentrations of TCA. The largest difference appeared to be between Week 45 group in
Study #1 and Week 52 group in Study #3 where both concentrations of TCA were reported to
induce increases in percent liver/body weight in one study but to have little difference in the
other. The differences in mean final body weights between these two sacrifice times were also
the largest although control and TCA-treatment groups had little difference on this parameter.
Similar to the work of Kjellstrand et al with TCE (Kjellstrand et al., 1983a), the groups with the
lower body weight appeared to have the greatest response in liver weight increase.
These data illustrate the variability in findings of percent liver weight induction between
laboratories, studies, choice of controls solutions, and the affects of final body weights on this
parameter. They also illustrate the limitations for determining either the magnitude or pattern of
liver weight increases using a small number of test animals. As animals age the size of their
liver changes but also during the latter parts of the lifespan, foci and spontaneously occurring
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-189 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
liver tumors can affect liver weight. The results of Study #1 show a consistent dose-response in
TCA liver weight increases at 4 and 15 week time periods over a range of concentration from
0.05 g/L to 5 g/L TCA.
In regard to non-neoplastic pathological changes the authors reported that
Increased incidences and severity of centrilobular cytoplasmic alterations,
inflammation, and necrosis were the only nonproliferative changes seen in livers
of animals exposed to TCA for 60 weeks (Tables 7-9; Study 1. Incidences were
between 21 and 93%; severity ranged from minimal to mild; and some lesions
were transient. Centrilobular cytoplasmic alterations (Table 7) were the most
prominent nonproliferative lesion. The incidence and severity were dose related
and significantly increased at all TCA concentrations. Centrilobular alterations
are a low-grade degeneration of the hepatocytes characterized by an intense
eosinophilic cytoplasm with deep basophilic granularity (microsomes) and slight
hepatomegaly. The distribution ranged from centrilobular to diffuse. The
incidence of inflammation was increased significantly in the 5 g/L TCA treatment
group (Table 8), but was significantly lower in the 0.05- and 0.5 g/L groups
between 31 and 45 weeks, but abated by 60 weeks. There was a significant dose-
related trend, but a significant increase in severity was only found at 5 g/L. No
alteration in the severity of this lesion was observed. The occurrence and severity
of nonproliferative lesions in animals exposed to 0.5 and 4.5 g/L TCA for 104
weeks were similar to those observed at 60 weeks (data not shown). No
pathology outside the liver was observed except for a significant dose-related
trend and incidence of testicular tubular degeneration at 0.5 and 5 g/L TCA.
The results shown in Table 7 by the authors for the 60-week TCA-exposed mice did not
show a dose-response for either incidence or severity of centrilobular cytoplasmic alterations.
They reported a 7, 48, 21, and 93% incidence and a 0.10 ± 0.40, 0.70 ± 0.82, 0.34 ± 0.72 and
1.60 ± 0.62 mean severity score for control, 0.05, 0.5, and 5.0 g/L TCA exposure groups,
respectively. Thus, for control, 0.05 and 0.5 g/L TCA exposure there was less than minimal (i.e.,
score of 1 or occupying less than 25% of the microscopic field) severity of this finding for the 27
to 30 mice examined in each group. Only slight hepatomegaly is noted by the authors to be
included in their description of the centrilobular cytoplasmic alteration. Interestingly, the
elevation of this parameter for both incidence and severity in the 0.05 g/L TCA exposed group
compared to 0.5 g/L exposure group did not correspond to an increase in percent liver/body
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-190 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
weight for this same exposure group. While the percent liver/body weight ratio was 32% higher,
the incidence and severity of this lesion were reported to be half that in the 0.5 versus 0.05 g/L
exposure groups after 60 days of TCA exposure. Thus, TCA-induced hepatomegaly did not
appear to be associated with this centrilobular cytoplasmic change.
Similarly the incidence of hepatic inflammation was reported to be 10, 0, 7, and 24% and
severity, 0.11 ± 0.40, 0.09 ± 0.30, 0.12 ± 0.33, and 0.29 ± 0.48 for control, 0.05, 0.5, and 5.0 g/L
TCA exposure groups, respectively. Thus, at no TCA exposure concentration was the incidence
more than 24% and the severity was considerably less than minimal. The reported results for
hepatic necrosis were pooled from data from the 5 mice exposed for either 30 or 45 weeks (n =
10 total). No incidences of necrosis were reported for either control or 0.05 g/L TCA exposed
mice. At 0.5 g/L TCA 3/10 mice were reported to have necrosis but at a severity level of 0.50 ±
0.97. At 5.0 g/L TCA 5/10 mice were reported to have necrosis but at a severity level of 1.30 ±
1.49. The limitations of the small number of animals pooled in these data are obvious.
However, there does not appear to be much more than minimal necrosis at the highest dose of
TCA between 30 and 45 weeks and this response is reported by the authors to be transient.
Serum LDH activity was reported by the authors for 31 and 60 week TCA exposures in
Study #1. They state that
There was a dose-related trend at 31 weeks; serum LDH was significantly
increased at 0.5 and 5 g/L TCA (161 ± 39 and 190 ± 44, respectively vs. 100 ± 28
IU for the control). LDH activity returned to control levels at 60 weeks.
Similarly, elevated LDH levels were observed at early time periods for 0.5 and
4.5 g/L TCA during the 104 week exposure (data not shown: Studies 2 and 3).
The data presented by the author for Study #1 are from 5 animals/group for the 30-week results
and 30 animals/group for the 60-week results. Of interest is for the 60-week data, there appears
to be 50% decreased in LDH activity at 0.05 and -25% decrease in LDH activity at 0.5 g/L TCA
treatment with the LDH level reported to be the same as control for the 5 g/L TCA exposure
group. For the 31-week data, in which only 5 animals were tested in each treatment group, there
appeared to be a slight increase at the 0.5 g/L (60% increase over control) and 5 g/L (90%
increase over control) treatment groups. The data for necrosis detected by light microscopy and
by LDH level is consistent with no changes from control detected at the 0.05 g/L TCA treatment
group and less than minimal necrosis of on a 60% increase in LDH level over control reported
for 0.5 g/L TCA treatment. Even at the highest dose of 5.0 g/L TCA there is still little necrosis
or LDH release reported over control.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-191 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Data for testicular tubular degeneration was reported for Study #1 after 60-weeks of TCA
exposure. The incidence of testicular tubular degeneration was reported to be 7, 0, 14, and 21%
for mice exposed to 2.0 g/L NaCl, 0.05, 0.5, and 5.0 g/L TCA. The severity of the lesions was
reported to be 0.10 ± 0.40, 0, 0.17 ± 0.47, and 0.21 ± 0.41 with a significant trend with dose
reported by the authors for severity and for the 0.5 and 5 g/L treatment groups to be significantly
increased over control incidence levels. Of note, similar to the percent liver/body weight ratios
and hepatic inflammation values for this data set, the values for testicular tubular degeneration
were slightly higher in control mice than 0.05 g/L TCA exposed mice. In regard to mean
severity levels for testicular degeneration, although still minimal, there was little difference
between the results for reported for the 0.5 g/L TCA and 5.0 g/L TCA exposed mice.
In regard to peroxisome proliferation, liver PCO activity was presented for up to
60 weeks (Study #1) and 104 weeks (Study #2). Similar to the data for LDH activity, -30
animals were examined at the 60-week time point but only 5 animals per exposure group were
examined for 4-, 15-, 31-, and 45-week results. The data are presented in a figure and in some
instances it is hard to determine the magnitude of change.
Similar to other reports, the baseline level of PCO activity was variable between control
groups and ranged 2.7-fold (-1.49 to 4.06 nmol NAD reduced/min/mg protein given by the
authors). There appeared to be little change in PCO activity between the 0.05 g/L TCA exposure
and control levels for up to 45 weeks of exposure (i.e., the groups with n = 5) in Study #1. For
the 60-week group the 0.05 g/L TCA group PCO activity was -1.7-fold of control but was not
statistically significant. For the 0.5 g/L TCA treatment groups, the increase ranged from -1.3- to
2.7-fold of control after 4-, 15-, 31-, and 45-weeks of exposure with the largest differences
reported at 4 and 60 weeks (i.e., 2.2- and 2.7-fold of control, respectively). For the 5.0 g/L TCA
exposure groups, the increase ranged from -3.2- to -5.7-fold of control after 4, 15, 31, and 45
weeks of exposure.
While the data at 60-weeks had the most animals examined (-30 vs. 5) with -1.7-, 2.7-,
and 4.5-fold of control PCO activity, at this time period the authors report the occurrence of
tumors had already occurred. At the earlier time points of 4 and 15 weeks, there was a difference
in the magnitude of TCA-induced increases in PCO activity. As displayed graphically, at 4
weeks the PCO increases were -1.3-, 2.4-, and 5.3-fold of control for 0.05, 0.5, and 5.0 g/L
TCA, respectively, while at 15 weeks, the PCO levels were decreased by 5%, increased to 1.3-
fold, and increased to 3.2-fold of control with only the 5.0 g/L treatment group difference to be
statistically significant.
For Study #2 the authors present a figure (Figure #4) that states that PCO values were
given for mice given HAC or 4.5 g/L TCA for 4-60 weeks. However, the data presented in #4
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-192 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
appears to be for 15-, 30-, 45- and 104-week exposures. The number of mice is not given in the
figure but the methods section states that serial sections were conducted on 5 mice/group for
these interim sacrifice periods. The number of mice examined for PCO activity at 104 weeks
was not given by the authors but the number of mice at final sacrifice was given as 25. The
levels of PCO in the control tissues varied by -33% for weeks 15 to 45 but there was a ~5-fold
difference between the level reported at 104 weeks and that for the earlier time periods in control
mice shown in the figures (-2.23 vs. 0.41 nmol NAD reduced/min/mg protein as given by the
authors). The increase over control induced by 4.5 g/L TCA in Study #2 was shown to be -6.9-,
4.8-,	3.6-, and 19-fold of controls for 15, 30, 45 and 104 weeks, respectively.
Therefore, at a comparable level of TCA exposure (-5.0 g/L), number of mice examined
(n = 5), and durations of exposure (15, 30, and 45 weeks), the increase in PCO activity induced
by -5.0 g/L TCA varied between 3.2- to 5.7-fold of control in Study #1 and between 3.6- to
6.9-fold	of control in Study #2. There was not a consistent pattern between the two studies in
regard to level of PCO induction from -5 g/L TCA and duration of exposure. The lowest TCA-
induced PCO activity increase was recorded at 15 weeks in Study #1 (i.e., 3.2-fold of control)
and highest PCO activity increase was recorded at 15 weeks in Study #2 (i.e., 6.9-fold of
control). No PCO data were reported for data in Study #3 with the exception of the authors
stating that "PCO activity was significantly elevated for the 0.5 g/L TCA exposure over the 104
weeks (study 3). The extent of the increases was similar to those measured for 0.5 g/L TCA
(200-375%: data not shown) in Study 1." No other details are given for PCO activity in
Study #3.
Hepatocyte proliferation was reported by the authors to be assessed by either
incorporation of tritiated thymidine (Studies #1 and #2) or BrdU (Study #3) into hepatocyte
nuclei. As noted previously, these techniques measure DNA synthesis and not necessarily
hepatocyte proliferation. The authors did not report if specific areas of the liver were analyzed
by autoradiographs or how many autoradiographs were examined in the analyses they conducted.
For later time points of examination (60-104 weeks) the authors did not indicate whether
hepatocytes in foci or adenomas were excluded from DNA synthesis reports. The authors
present data for what are clearly, 31, 45, and 60 week exposure for Study #1 as the percent
tritiated thymidine labeled nuclei. An early time point that appears to be 8 weeks is also given.
However, for Study #1 only 4 week and 15 week durations were tested so it cannot be
established what time period the earlier time point represents. What is very apparent from the
data presented for Study #1 is that the baseline level of tritiated thymidine incorporation was
relatively high and highly variable for the 5 animals examined (-8% of hepatocytes were
labeled). There did not appear to be an apparent pattern of TCA treatment groups at this
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-193 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
timepoint with the 0.05 and 5.0 g/L TCA groups having a similar percentage of labeled
hepatocytes and for 0.5 g/L TCA reported to have a 60% reduction in labeled hepatocytes.
After 31 weeks of exposure the control values were reported to be 2% of hepatocytes
labeled. The authors report that only the 5.0 g/L TCA group had a statistically significant
increase of control and was elevated to -6% of hepatocytes. The two lower exposure
concentrations of TCA had similar reported incidences of labeled hepatocytes of 4.5% that were
not reported to be statistically significant.
For the 45-week exposure period in Study #1, the control value was reported to be 1.2%
with only the 5.0 g/L TCA value reported to be statistically significantly increased at 3.2% and
the other two TCA groups to be similar to control. Finally, for the 60 week group from Study
#1, the control value was reported to be 0.6% of hepatocytes labeled and the only the 0.5 g/L
TCA dose reported to be statistically significantly increased over control at 3.2%.
What is clear from this study is that the control value for the unidentified early time point
is much higher than the other values. There should not be such a large difference in mature mice
nor such a high level. The difference in control values between the earlier time point and the 31-
week time point was 4-fold. The difference between the earlier time point and the 45-week time
point was ~7-fold. There did not appear to be an increase in hepatocyte tritiated thymidine
labeling due to any concentration of TCA at the early unidentified time point (-Week 10 from
the figure) from Study #1. There was no dose-response apparent for the other study periods and
the percent of hepatocytes labeled were 3% or less. These results indicated DNA synthesis was
not increased by 10-60 week exposures to TCA exposure that induced increased liver tumor
response.
For Study #2, results were reported for tritiated thymidine incorporation into hepatocytes
in a figure that was labeled as 4.5 g/L TCA and control tissue for 104 weeks but showed data for
15, 30, and 45 weeks of exposure. Of note is that the control values for this study were much
lower than that reported for Study #1. The percent of hepatocytes labeled with tritiated
thymidine was reported to be -2% for the 15 week exposure period and less than 1% for the 30-
and 45-week exposure periods. For the 4.5 g/L TCA exposures the percent hepatocytes labeled
with tritiated thymidine were -2-4% at all time points with only the 45 week period identified
by the authors as statistically significant.
For Study #3, rather than tritiated thymidine, BrdU was used as a measure of DNA
synthesis. The results are presented in Figure #8 of the report in which the 0.5 g/L TCA
concentration is mislabeled as 0 g/L and the figure is mislabeled as having a duration of
104 weeks but the data are presented for 26, 52, and 78 weeks of exposure. The percent of
hepatocytes at 26 weeks was reported to be —1—2% for the control, 0.05 and 0.5 g/L TCA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-194 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
groups. At 52 weeks the control value was -1% the 0.05 g/L TCA value was less than 0.1% and
the 0.5 g/L TCA value was -3.5% but not statistically significant. At 78 weeks of exposure the
control value was reported to be -0.2% with only the 0.05 g/L TCA group having a statistically
significant increase over control.
From these data, the estimated control values for DNA synthesis at similar time points of
exposure ranged from 0.4 to 2% at 26-31 weeks and -0.1 to 1.2% at 45-52 weeks. The results
for Study #1 and #2 were inconsistent in regard to the magnitude of tritiated thymidine
incorporation but consistent in that there was a lot of variability in these measurements, not a
consistent pattern with time that was TCA-dose related, and, even at the highest dose of TCA,
did not indicate much of an increase in cell proliferation 15-45 weeks of exposure. Similarly the
results for Studies #1 and #3 indicate that the two lower doses of TCA there were not generally
statistically significant increases in DNA synthesis from 15-45 weeks of exposure although there
was an increase in liver tumor response at later time points.
The authors reported that "all gross and microscopic histopathological alterations were
consistent across the three studies." However, the histological descriptions that follow were
focused on the liver for both neoplastic and non-neoplastic parameters. As stated above, only a
few animals (n = 5) from the control and high TCA dose level were examined for lesions other
than liver, kidneys, spleen and testes. Thus, whether other neoplastic lesions were induced by
TCA exposure cannot be determined from this set of studies.
Study #1 was conducted for 60 weeks. Although of short duration and using 30 or less
animals, the authors reported in the text that
a significant trend with dose was found for liver cancer. The prevalence and
multiplicity of adenomas (38%; 0.55 ± 0.15) or carcinoma (38%; 0.42 ±0.11)
were statistically significant at 602 mg/kg/day TCA compared to control (7%;
0.07 ± 0.05) [sic for both adenoma and carcinoma the same value was given,
mean ± SD], When either an adenoma or a carcinoma was present, statistical
significant was seen at both 5 g/L (55%; 1.00 ± 0.19) and 0.5 g/L (38%: 0.52 ±
0.14 TCA exposure groups compared to control (13%; 0.13 ± 0.06).
No significant change in liver neoplasia were reported to be observed by the authors at
0.05 g/L TCA. Preneoplastic large foci of cellular alteration (24%) were seen in the 5
g/L TCA control compared to control.
Although not statically significant, there was an incidence of 15% adenoma in the
0.05 g/L TCA treatment group (n = 27) and a multiplicity of 0.15 ± 0.07 adenomas/mouse
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-195 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
reported with both values being twice that of the values given for the controls (n = 30). The
incidence and multiplicity for carcinomas was approximately the same for the 0.05 g/L TCA
treatment group and the control group. Given the small number of animals examined, the study
was limited in its ability to determine statistical significance for the lower TCA exposure level.
The fold increases of incidence and multiplicity of adenomas at 60 weeks was 2.1-, 3.0-,
and 5.4-fold of control incidence and 2.1-, 3.4-, and 7.9-fold of control multiplicity for 0.05, 0.5,
and 5 g/L exposure to TCA. For multiplicity of adenomas and carcinomas combined there was a
1.46-, 4.0-, and 7.68-fold of control values. Analysis of tumor prevalence data for this study
included only animals examined at scheduled necropsy. Since most animals survived until
60 weeks, most were included and a consistent time point for tumor incidence was reported.
There are significant discrepancies for reporting of data for tumor incidences in this
report for the 104 week data. While the methods section and table describing the dose
calculation and animal survival indicate that Study #3 control animals were administered
deionized water and those from Study#2 were given HAC, Table 6 of the report gives 2 g/L
NaCl as the control solution given for Study #2 and 1.5 g/L HAC for Study #3. A comparison of
the descriptions of animal survival and tumor incidence and multiplicity between the results
given in DeAngelo et al. (2008) and George et al. (2000) (see Table E-10) shows not only that
the control data presented in DeAngelo et al. (2008) for Study #3 to be the same data as that
presented by George et al. (2000) previously, but also indicates that rather than 1.5 g/L HAC, the
tumor data presented in DeAngelo et al. (2008) is for mice exposed to deionized water.
DeAngelo et al. (2008) did not report that these data were from a previous publication.
Table E-10. Comparison of descriptions of control data between George et al.
(2000) and DeAngelo et al. (2008)
Descriptor
George et al., (2000)
DeAngelo et al. (2008)
Species
Mouse
Mouse
Strain
B6C3F1
B6C3F1
Gender
Male
Male
Age
28-30 days
28-30 days
Source
Charles River, Portage
Charles River, Portage
Mean initial body wt
19.5 ±2.5 g
19.5 ±2.5 g
Water consumption
111.7 mL/kg/day
112 mL/kg/day
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-196 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Laboratory
RTP NC
RTP NC
# Animals at start
72
72
# Animals at interim sac.
22
21
# Unscheduled deaths
16
17
# Animals at final sacrifice
34
34
# Animals for pathology
65
63
Adenoma incidence
21.40%
21%
Adenoma multiplicity
0.21 ±0.06
0.21 ±0.06
Carcinoma incidence
54.80%
55%
Carcinoma multiplicity
0.74 ±0.12
0.74 ±0.12
RTP NC = Research Triangle Park, North Carolina.
For Studies #2 and #3 tumor prevalence data were reported in the methods section of the
report to include necropsies of animals that survived greater than 78 weeks and thus, included
animals that were scheduled for necropsy but also those which were moribund and sacrificed at
differing times.
Thus, for the longer times of study, there was a mixture of exposure durations that
included animals that were ill and sacrificed early and those that survived to the end of the study.
Animals that were allowed to live for longer periods or who did not die before scheduled
sacrifice times had a greater opportunity to develop tumors. However, animals that died early
may have died from tumor-related causes.
The mislabeling of the tumor data in DeAngelo et al. (2008) has effects on the
interpretation of results for if the tumor results table was not mislabeled it would indicate 17
animals were included in the liver tumor analysis that were not included in the final necropsy and
that the 7 unscheduled deaths could not account for the total number of "extra" mice included in
the tumor analysis so some of the animals had to have come from interim sacrifice times (78
weeks or less) and that for Study #3 the data from 9 animals at terminal sacrifice were not used
in the tumor analysis. Not only does it appear that the control data was mislabeled for Study #3,
but the control data were also apparently mislabeled for Study #2 as being 2.0 g/L NaCl rather
than 1.5 g/L HAC. Of the 42 animals used for the tumor analysis in Study #3, only 34 were
reported to have survived to interim sacrifice so that 8 animals were included from unscheduled
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-197 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
deaths. However, the authors report that there were 17 unscheduled deaths in the study not all
were included in the tumor analysis. The basis for the selection of the 8 animals for tumor
analysis was not give by the authors.
Not only are the numbers of control animals used in the tumor analysis different between
two studies (25 mice in Study #2 and 42 mice in Study #3), but the liver tumor results reported
for Study #2 and Study #3 were very different. Of the 42 "control" mice examined from Study
#3, the incidence and multiplicity of adenomas was reported to be 21% and 0.21 ± 0.06,
respectively. For carcinomas, the incidence and multiplicity was reported to be 55% and
0.74 ± 0.12, respectively, and for the incidence and multiplicity of adenomas and carcinomas
combined reported to be 64% and 0.93 ± 0.12, respectively. For the 25 mice reported by the
authors for Study #2 to have been treated with "2.0g/L NaCl" but were probably exposed to
1.5 g/L HAC, the incidence and multiplicity of adenomas was 0%. For carcinomas, the
incidence and multiplicity was reported to be 12% and 0.20 ± 0.12, respectively and for the
incidence and multiplicity of adenomas and carcinomas combined to be 12% and 0.20 ± 0.12,
respectively. Therefore, while -64% the 42 control mice in Study #3 were reported to have
adenomas and carcinomas, only 12% of the 25 mice were reported to have adenomas and
carcinomas in Study #2 for 104-weeks.
While the effect of using fewer mice in one study versus the other will be to reduce the
power of the study to detect a response, there are additional factors that raise questions regarding
the tumor results. Not only were the tumor incidences reported to be higher in control mice from
Study #3 than Study #2, but the number of unscheduled deaths was reported to also be 2-fold
higher. The age, gender, and strain of mouse were reported to be the same between Study #2 and
#3 with only the vehicles differing and weight of the mice to be reported to be different.
Although the study by George et al. (2000) described the same control data set as for Study #3 as
being for animals given deionized water, there is uncertainty as to the identity of the vehicle used
for the tumor results reported for Study #3 and there are some discrepancies in reporting between
the two studies. As discussed below in Section E.2.5, the differences in the weight of the mice
between Studies #1, #2, and #3 is critical to the issue of differences in background tumor rate
and hence interpretability of the study.
As noted by Leakey et al. (2003b), the greatest correlation with liver tumor incidence and
body weight appears between the ages of 20 and 60 weeks in male mice. As reported in
Section E.2.5, the mean 45-week body weight reported for control male B6C3F1 mice in the
George et al. (2000) study, which is the same control data as DeAngelo et al. (2008) was -50 g.
This is a much greater body weight than reported for Study #1 at 45 weeks (i.e., 39.6 g) and for
Study #2 at 45 weeks (i.e., 39.4 g). Using probability curves presented by Leakey et al. (2003b),
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-198 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
the large background rate of 64% of combined adenomas and carcinomas for Study #3 is in the
range predicted for such a large body weight (i.e., -65%). Such a high background incidence
compromises a 2-year bioassay as it prevents demonstration of a positive dose-response
relationship. Thus, Study #3 of DeAngelo et al. (2008) is not comparable to the results in
Study #1 and #2 for the determination of the dose-response for TCA.
The accurate determination of the background liver tumor rate is very important in
determining a treatment-related effect. The very large background level of tumor incidence
reported for Study #3 makes the detection of a TCA-related change in tumor incidence at low
exposure levels very difficult to determine. Issues also arise as to what the source of the tumor
data were in the TCA-treatment and control groups in Study #3. While 29 mice exposed to
0.05 g/L TCA were reported to have been examined at terminal sacrifice, 35 mice were used for
liver tumor analysis. Similarly, while 27 mice exposed to 0.5 g/L TCA were reported to have
been examined at terminal sacrifice, 37 mice were used for tumor analysis. Finally, for the
42 control animals examined for tumor pathology in the control group, 34 were examined at
terminal sacrifice. Clearly more animals were included in the analyses of tumor incidence and
multiplicity than were sacrificed at the end of the experiment. What effect differential addition
of the results from mice not sacrificed at 104 weeks and the selection bias that may have resulted
from their inclusion on these results cannot be determined. Not only were the background levels
of tumors reported to be increased in the control animals in Study #3 compared to Study #2 at
104 weeks, but the rate of unscheduled deaths was doubled. This is also an expected
consequence of using much larger mice (Leakey et al., 2003b).
For the 35 mice examined after 0.05 g/L TCA in Study #3, the incidence and multiplicity
of adenomas was reported to be 23% and 0.34 ± 0.12, respectively. For carcinomas, the
incidence and multiplicity was reported to be 40% and 0.71 ±0.19, respectively, and for the
incidence and multiplicity of adenomas and carcinomas combined reported to be 57% and
1.11 ± 0.21, respectively. For the 37 mice examined after 0.5 g/L TCA in Study #3, the
incidence and multiplicity of adenomas was reported to be 51% and 0.78 ±0.15, respectively.
For carcinomas, the incidence and multiplicity was reported to be 78% and 1.46 ± 0.21,
respectively, and for the incidence and multiplicity of adenomas and carcinomas combined
reported to be 87% and 2.14 ± 0.26, respectively.
Thus at 0.5 g/L TCA, the results presented for this study for the "104 week" liver tumor
data were significantly increased over the reported control values. However, these results are
identical to those reported in Study #3 for a 10-fold higher concentration of TCA (4.5 g/L TCA)
for the same 104 weeks of exposure but in the much larger mice. Of the 36 animals exposed to
4.5 g/L TCA in Study #2 and included in the tumor analysis, 30 animals were reported to be
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-199 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
examined at 104 weeks. The incidence and multiplicity of adenomas was reported to be 59%
and 0.61 ± 0.16, respectively. For carcinomas, the incidence and multiplicity was reported to be
78% and 1.50 ± 0.22, respectively, and for the incidence and multiplicity of adenomas and
carcinomas combined reported to be 89% and 2.11 ± 0.25, respectively.
The importance of selection and determination of the control values for comparative
purposes of tumor induction are obvious from these data. The very large difference in control
values between Study #2 and Study #3 is the determinant of the magnitude of the dose response
for TCA after 104 weeks of exposure. The tumor response for 0.5 and 4.5 g/L TCA exposure
between the two experiments was identical. Therefore, only the background tumor rate
determined the magnitude of the response to treatment. If a similar control values (i.e., a
historical control value) were used in these experiments, there would appear to be no difference
in TCA-tumor response between 0.5 and 4.5 g/L TCA at 104 weeks of exposure. DeAngelo et
al. (1999) report for male B6C3F1 mice exposed only water for 79 to 100 weeks the incidence of
carcinomas to be 26% and multiplicity to be 0.28 lesions/mouse. For 100-week data, the
incidence and prevalence of adenomas was reported to be 10% and 0.12 ± 0.05 and for
carcinomas to be 26% and 0.28 ± 0.07.
Issues with reporting for that study have already been discussed in Section E.2.3.2.5.
However, the data for DeAngelo et al. (1999) are more consistent with the control data for "1.5
g/L HAC" for Study #2 in which there were 0% adenomas and 12% carcinomas with a
multiplicity of 0.20 ± 0.12, than for the control data for Study #3 in which 64% of the control
mice were reported to have adenomas and carcinomas and the multiplicity was 0.93 ± 0.12. If
either the control data from DeAngelo et al. (1999) or Study #2 were used for comparative
purposes for the TCA-treatment results of Study #2 or #3, there would be a dose-response
between 0.05 and 0.5 g/L TCA but no difference between 0.5 and 4.5 g/L TCA after 100 weeks
of exposure. The tumor incidence would have peaked at -90% in the 0.5 and 4.5 g/L TCA
exposure groups. These results would be more consistent with the 60-week results in Study #1
in which 0.5 and 5 g/L TCA exposure groups already had incidences of 38 and 55% of adenomas
and carcinomas combined, respectively, compared to the 13% control level. With increased time
of exposure the differences between the two highest TCA exposure concentrations may diminish
as tumor progression is allowed to proceed further. However, the use of the larger and more
tumor prone mice in Study #3 also increases the tumor incidence at the longer period of study.
The authors also presented data for multiplicity of combined adenomas or carcinomas for
mice sacrificed at weeks 26, 52, and 78 for Study #3 (n = 8 per group). No indication of
variability of response, incidence data, statistical significance, or data for adenomas versus
carcinomas, or the incidence of adenomas was reported. The authors reported that "neoplastic
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-200 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
lesions were first found in the control and 0.05 g/L TCA groups at 52 weeks. At 78 weeks,
adenomas or carcinomas were found in all groups (0.29, 0.20, and 0.57 tumors/animals for
control, 0.05 g/L TCA, and 0.5 g/L TCA, respectively)." Because no other data were presented
at the 52 and 78 week time points in this study, these results cannot be compared to those
presented for Study #1, which was conducted for 60 weeks. Of note, the results presented from
Study #1 for 60 weeks of exposure to control, 0.05 g/L or 0.5 g/L TCA exposure in 27-30 mice
show a 13, 15, and 38% incidence of hepatocellular adenomas and carcinomas and a multiplicity
of 0.13 ± 0.06, 0.19 ± 0.09, and 0.52 ± 0.14, respectively. Both the incidence and multiplicity of
adenomas were 2-fold higher in the 0.05 g/L TCA treatment group than for the control.
However, the interim data presented by the authors from Study #3 for 52 weeks of exposure in
only 8 mice per group gives a higher multiplicity of adenomas and carcinomas for control
animals (-0.25) than for either 0.05 or 0.5 g/L TCA treatments. Again, comparisons between
Study #2 and #3 are difficult due to difference in mouse weight.
Of note, there are no descriptions given in this report in regard to the phenotype of the
tumors induced by TCA or for the liver tumors reported to occur spontaneously in control mice.
Such information would have been of value as this study reports results for a range of TCA
concentration and for 60 and 100 weeks of exposure. Insight could have been gained as to the
effects of differing concentrations of TCA exposure, whether TCA-induced liver tumors had a
similar phenotype as those occurring spontaneously, as well as information in regard to effects
on tumor progression and heterogeneity.
Although only examining tissues from 5 mice from the control and high-dose groups only
at 104 weeks at organ sites other than the liver, the authors report that
neoplastic lesions at 104 weeks (Studies #2 and #3) at organ sites other than the
liver were found in the lung, spleen, lymph nodes, duodenum (lymphosarcoma),
seminal vesicles, skin, and thoracic cavity of control and treated animals. All
were considered spontaneous for the male B6C3F1 mouse and did not exceed the
tumor incidences when compared to a historical control database (Haseman et al.,
1984; NIEHS, 1998).
No data were shown. The limitations involved in examining only 5 animals in the control and
high-dose groups, and the need to examine the concurrent control data in each experiment,
especially given the large variation in liver tumor response between long-term studies carried out
in the two different laboratories used for Study #2 and Study #3 using the same strain and gender
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-201 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of mouse, make assertions regarding extrahepatic carcinogenicity of TCA from this study
impossible to support.
A key issue raised from this study is whether changes in any of the parameters measured
in interim sacrifice periods before the appearance of liver tumors (i.e., 4-15 weeks)
corresponded to the induction of liver tumors. The first obstacle for determining such a
relationship is the experimental design of these studies in which only a full range of TCA
concentrations is treated for 60 weeks of exposure with a small number of animals available for
determination of a carcinogenic response (i.e., 30 animals or less in Study #1) and a very small
number of animals (n = 5 group) examined for other parameters. Also as stated above, PCO
activity was highly variable between controls and between treatment groups (e.g., the PCO
activity for Study #1 and #2 at ~5 g/L exposure for 15 weeks).
On the other hand, most of the animals that were examined at terminal sacrifice were also
utilized for the tumor results without the differential deletion or addition of "extra" animals for
the tumor analysis. For the 60-week data in Study #1 there appeared to be a consistent dose-
related increase in the incidence and multiplicity of tumors after TCA exposure (Table E-l 1).
The TCA-induced increases in liver tumor responses can be compared with both increased liver
weight and PCO activity that were also reported to be increased with TCA dose as earlier events.
Although the limitations of determining the exact magnitude of responses has already been
discussed, as shown below, the incidence and multiplicity of adenomas show a dose-related
increase at 60 weeks. However, the magnitude of differences in TCA concentrations was not
similar to the magnitude of increased liver tumor induction by TCA after 60 weeks of exposure.
First of all, the greater occurrence of TCA-induced increases in adenomas than
carcinomas reported after 60 weeks of exposure would be expected for this abbreviated duration
of exposure as they would be expected to occur earlier than carcinomas. For adenoma induction,
there was a ~2-fold increase between the 0.05 g/L dose of TCA and the control group for
incidence (7 vs. 15%) and multiplicity (0.07 vs. 0.15 tumors/animals). However, an additional
10-fold increase in TCA dose (0.5 g/L) only resulted in a reported 1.8-fold greater incidence
(15 vs. 21%) and 2.2-fold increase in multiplicity (0.15 vs. 0.24 tumors/animal) of control
adenoma levels. An additional 10-fold increase in dose (5.0 vs. 0.5 g/L TCA) resulted in a
2.2-fold increase in incidence (21 vs. 38%) and 2.9-fold increase in multiplicity (0.24 vs.
0.55 tumors/animal) of control adenoma levels.
Thus, a 100-fold difference in TCA exposure concentration resulted in differences of 4-
fold of control incidence and 6-fold of control multiplicity for adenomas. For adenomas or
carcinomas combined (a parameter that included carcinomas for which only the two highest
exposure levels of TCA were reported to increase incidence and multiplicity) the incidences
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-202 DRAFT—DO NOT CITE OR QUOTE
-------
1	were reported to be 13, 15, 38, and 55%, and the multiplicity reported to be 0.13, 0.19, 0.52, and
2	1.00 for control, 0.05, 0.5, and 5.0 g/L TCA at 60 weeks. For multiplicity of adenomas or
3	carcinomas, the 0.05 g/L TCA exposure induced a 1.5-fold increase over control. An additional
4	10-fold increase in TCA (0.5 g/L) induced a 6-fold increase in tumors/animal. An additional 10-
5	fold increase in TCA (5.0 vs. 0.5 g/L) induced an additional 2.2-fold increase in tumors/animal.
6	Therefore, using combinations of adenomas or carcinomas, there was a 13-fold increase in
7	multiplicity that corresponded with a 100-fold increase in dose.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-203 DRAFT—DO NOT CITE OR QUOTE
-------
Table E-ll. TCA-induced increases in liver tumor occurrence and other parameter over control after 60 weeks
(Study #1)
Dose TCA
g/L
Adenomas
Adenomas or carcinomas
% liver/body weight
PCO activity
NaCl
Incidence 7%
Multiplicity 0.07
Incidence 13%
Multiplicity
0.13
4-week
15-week
4-week
15-week
0.05
15% (2.1-fold)
0.15 (2.1-fold)
15% (1.2-fold)
0.19(1.5-fold)
1.09-fold
1.14-fold
1.3-fold
1.0 -fold
0.5
21% (3.0-fold)
0.24 (3.4-fold)
38% (2.9-fold)
0.52 (4.0-fold)
1.16-fold
1.16-fold
2.4-fold
1.3-fold
5.0
38% (5.4-fold)
0.55 (7.9-fold)
55% (4.2-fold)
1.00 (7.7-fold)
1.35-fold
1.47-fold
5.3-fold
3.2-fold
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The results for adenoma induction at 60 weeks of TCA exposure (i.e., ~2-fold increased
incidences and 2- to 3-fold increases in multiplicity with 10-fold increases in TCA dose) are
similar to the ~2-fold increase in liver weight gain resulting from 10-fold differences in dose
reported at 4-weeks of exposure. For PCO activity there was a -30% increase in PCO activity
from control at 0.05 g/L TCA. A 10-fold increase in TCA exposure concentration (0.5 g/L)
resulted in an additional ~5-fold increase in PCO activity. However, another 10-fold increase in
TCA concentration (0.5 vs. 5 g/L) resulted in a 3-fold increase in PCO activity. The 100-fold
increase in TCA dose (0.05 vs. 5 g/L TCA) was correlated with a 14-fold increase in PCO
activity. For 15 weeks of TCA exposure there was no difference in 0.05 and control PCO
activity and only a 30% difference between the 0.05 and 0.5 g/L TCA exposures. There was a
7-fold difference in PCO activity between the 0.5 and 5.0 g/L TCA exposure concentrations.
The increases in PCO activity and liver weight data at 15-weeks did not fit the magnitude of
increases in tumor multiplicity or incidence data at 60 weeks as well as did the 4-week data.
However, the TCA-induced increase in tumors at 60 weeks (especially adenomas) seemed to
correlate more closely with the magnitude of liver weight increase than for PCO activity at both
4 and 15 weeks.
In regard to Studies #1 and #2 there were consistent periods of study for percent
liver/body weight with the consistency of the control values being a large factor in the magnitude
of TCA-induced liver weight increases. As discussed above, there were differences in the
magnitude of percent liver/body weight increase at the same concentration between the two
studies (e.g., a 1.47-fold of control percent liver/body weight in the 5 g/L TCA exposed group in
Study #1 and 1.60-fold of control in Study #2 at 15 weeks). For the two studies that had
extended durations of exposure (Studies #2 and #3) the earliest time period for comparison of
percent liver/body weight is 26 weeks (Study #3) and 30 weeks (Study #2). If those data sets (26
weeks for Study #3 and 30 weeks for Study #2) are combined, 0.05, 05, and 4.5 g/L TCA gives a
percent liver body/weight increase of 1.07-, 1.18-, and 1.40-fold over concurrent control levels.
Using this parameter, there appears to be a generally consistent pattern as that reported for Study
#1 at weeks 4 and 15. Generally, a 10-fold increase in TCA exposure concentration resulted in
~2.5-fold increased in additional liver weight observed at -30 weeks of exposure which
correlated more closely with adenoma induction at 60 weeks than did changes in PCO activity.
A similar comparison between Studies of longer duration (Studies #2 and #3) could not be made
for PCO activity as data were not reported for Study #3.
For 104-week studies of TCA-tumor induction (Studies #2 and #3) the lower TCA
exposure levels (0.05 and 0.5 g/L TCA) were assayed in a separate experiment and by a separate
laboratory than the high dose (5.0 g/L TCA) and most importantly in larger more tumor prone
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-205 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
mice. The total lack of similarity in background levels of tumors in Study #2 and #3, the
differences in the number of animals included in the tumor analyses, and the low number of
animals examined in the tumor analysis at 104 weeks (less than 30 for the TCA treatment
groups) makes the determination of a dose-response TCA-induced liver tumor formation after
104-weeks of exposure problematic. The correlation of percent liver/body weight increases with
incidence and multiplicity of liver tumors in Study #1 and the similarity of dose-response for
early induction of percent liver/body weight gain between Study #1 suggest that there should be
a similarity in tumor response. However, as noted above, the 104-week studies had very
difference background rates of spontaneous tumors reported in the control mice between
Study #2 and #3.
Table E-12, below, shows the incidence and multiplicity data for Studies #2 and #3 along
with the control data for DeAngelo et al. (1999) for the same paradigm. It also provides an
estimate of the magnitude of increase in liver tumor induction by TCA treatments if the control
values from the DeAngelo et al. (1999) data set were used as the background tumor rate. As
shown below, the background rates for Study #2 are more consistent with those of DeAngelo et
al. (1999). Whereas there was a 2:1 ratio of multiplicity for adenomas and adenomas and
carcinomas between 0.5 and 5.0 g/L TCA after 60 weeks of exposure, there was no difference in
any of the data (i.e., adenoma, carcinoma, and combinations of adenoma and carcinoma
incidence and multiplicity) for these exposure levels in Study #2 and #3 for 104 weeks. The
difference in the incidences and multiplicities for all tumors was 2-fold between the 0.05 and
0.5 g/L TCA exposure groups in Study #2. These results are consistent with the two highest
exposure levels reaching a plateau of response with a long enough duration of exposure (-90%
of animals having liver tumors) and with the 2-fold difference in liver tumor induction between
concentrations of TCA that differed by 10-fold, reported in Study #1.
If either the control values for Study #2 or the control values from DeAngelo et al. (1999)
were used for as the background rate of spontaneous liver tumor formation, the magnitude of
liver tumor induction by the 0.05 g/L TCA over control levels differs dramatically from that
reported as control tumor rates in Study #3. To put the 64% incidence data for carcinomas and
adenomas reported in DeAngelo et al. (DeAngelo et al., 2008) for the control group of Study #3
in context, other studies cited in this review for B6C3F1 mice show a much lower incidence in
liver tumors in that: (1) the National Cancer Institute (NCI, 1976) study of TCE reports a colony
control level of 6.5% for vehicle and 7.1% incidence of hepatocellular carcinomas for untreated
male B6C3F1 mice (n = 70-77) at 78 weeks, (2) Herren-Freund et al. (1987) report a 9%
incidence of adenomas in control male B6C3F1 mice with a multiplicity of 0.09 ± 0.06 and no
carcinomas (n = 22) at 61 weeks, (3) NTP (1990) report an incidence of 14.6% adenomas and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-206 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
16.6% carcinomas in male B6C3F1 mice after 103 weeks (n = 48), and (4) Maltoni et al. (1986)
report that B6C3F1 male mice from the "NCI source" had a 1.1% incidence of "hepatoma"
(carcinomas and adenomas) and those from "Charles River Co." had a 18.9% incidence of
"hepatoma" during the entire lifetime of the mice (n = 90 per group). The importance of
examining an adequate number of control or treated animals before confidence can be placed in
those results in illustrated by Anna et al. (1994) in which at 76 weeks 3/10 control male B6C3F1
mice that were untreated and 2/10 control animals given corn oil were reported to have
adenomas but from 76 to 134 weeks, 4/32 mice were reported to have adenomas (multiplicity of
0.13 ± 0.06) and 4/32 mice were reported to have carcinomas (multiplicity of 0.12 ± 0.06).
Using concurrent control values reported in Study #3, there is no increase in incidence of
multiplicity of adenomas and carcinomas for the 0.05 g/L exposure group. However, compared
to either the control data from DeAngelo et al. (1999) or the control data from Study #3, there is
a -2-3- or ~5-fold increased in incidence or multiplicity of liver tumors, respectively. Thus,
trying to determine a correspondence with either liver weight increases or increases in PCO
activity at earlier time points will be depend on the confidence placed in the concurrent control
data reported in Study #3 in the 104 week studies. As noted previously, the use of larger tumor
prone mice in Study #3 limits its usefulness to determine the dose-response for TCA.
The authors provided a regression analysis for "tumors/animal" or multiplicity as a
percent of control values and PCO activity for the 60-week and 104-week data. Whether
adenomas and carcinomas combined or individual tumor type were used was not stated. Also
comparing PCO activity at the end of the experiments, when there was already a significant
tumor response rather than at earlier time points, may not be useful as an indicator of PCO
activity as a key event in tumorigenesis. A regression analysis of these data are difficult to
interpret because of the dose spacing of these experiments as the control and 5 g/L exposure
levels will basically determine the shape of the dose-response curve. The 0.05 and 0.5 g/L
exposure groups in the regression were so close to the control value in comparison to the 5 g/L
exposure, that the dose response will appear linear between control and the 5.0 g/L value with
the two lowest doses not affecting the slope of the line (i.e., "leveraging" the regression). The
value of this analysis is limited by (1) the use of tumor prone larger mice in Study #3 that had
large background rates of tumors which make inappropriate the apparent combination of results
from Studies #2 and #3 for the multiplicity as percentages of control values (2) the low and
varying number of animals analyzed for PCO values and the variability in PCO control values
(3) the appropriateness of using PCO values from later time points, and (4) the dose-spacing of
the experiment.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-207 DRAFT—DO NOT CITE OR QUOTE
-------
IS*
Table E-12. TCA-induced increases in liver tumor occurrence after 104 wks (Studies #2 and #3)
Dose TCA
Adenomas
Carcinomas
Adenomas or carcinomas
Incidence
Multiplicity
Incidence
Multiplicity
Incidence
Multiplicity
Study #3
1.5 g/LHAC (H20?)
21%
0.21
55%
0.74
64%
0.93
0.05 g/L TCA
23%
0.34
40%
0.71
57%
1.11
(1.1-fold)
(1.6-fold)
(0.7-fold)
(1.0-fold)
(0.9-fold)
(1.2-fold)
0.5 g/L TCA
51%
0.78
78%
1.46
87%
2.14
(2.4-fold)
(3.7-fold)
(1.4-fold)
(2.0-fold)
(1.4-fold)
(2.3-fold)
Study #2
2.0 g/L NaCl (HAC?)
0%
0
12%
0.20
12%
0.20
4.5 g/L TCA
59%
0.61
78%
1.50
89%
2.14
(?)
(?)
(6.5-fold)
(7.5-fold)
(7.4-fold)
(11-fold)
DeAngelo etal., 1999
h2o
10%
0.12
26%
0.28


0.05 g/TCA (S #3)
(2.3-fold)
(2.8-fold)
(1.5-fold)
(2.5-fold)


0.5 g/L TCA (S #3)
(5.1-fold)
(6.5-fold)
(3.0-fold)
(5.2-fold)


5.0 g/L TCA (S #2)
(5.9-fold)
(6.5-fold)
(3.0-fold)
(5.4-fold)


H20 = water.
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
Oq ^
53
^S
§*
Co
S5
S*
<§>
>
to
*
^s
«
s
TO
Co
0
s
1
a.
VO
to
o
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Similarly, the authors reported a regression analysis that compares "percent of
hepatocellular neoplasia" which again is indicated by tumor multiplicity with TCA dose as
represented by mg/kg/d. This regression analysis also is of limited value for the same reasons as
that for PCO with added uncertainty as the exposure concentrations in drinking water have been
converted to an internal dose and each study gave different levels of drinking water with one
study showing a reduction of drinking water at the 5 g/L level. The authors attempted to identify
a NOEL for tumorigenicity using tumor multiplicity and TCA dose. However, it is not an
appropriate descriptor for these data, especially given that "statistical significance" of the tumor
response is the determinant of the conclusions regarding a dose in which there is no TCA-
induced effect. Only the 60-week experiment (i.e., Study #1) is useful for the determination of
tumor dose-response due to the issues related to appropriateness of control in Study #3. A power
calculation of the 60-week study shows that the type II error, which should be >50% and thus,
greater than the chances of "flipping a coin," was 41 and 71% for incidence and 7 and 15% for
multiplicity of adenomas for the 0.05 and 0.5 g/L TCA exposure groups. For the combination of
adenomas and carcinomas, the power was 8 and 92% for incidence and 6 and 56% for
multiplicity at 0.05 and 0.5 g/L TCA exposure. Therefore, the designed experiment could accept
a false null hypothesis, especially in terms of tumor multiplicity, at the lower exposure doses and
erroneously conclude that there is no response due to TCA treatment.
E.2.4.2.26. DeAngelo et al. (1997). The design of this study appears to be similar to that of
DeAngelo et al. (2008) but to have been conducted in F344 rats. 28-30 day old rats,
reported to be of similar weights, were exposed to 2.0 g/L NaCl, 0.05, 0.5, or 5.0 g/L TCA in
drinking water for 104 weeks. There were groups of animals sacrificed at 15, 30, 45 and
60 weeks (n = 6) for PCO analysis. There were 23, 24, 19, and 22, animals reported to be
examined at terminal sacrifice at 104 weeks and 23, 24, 20, and 22 animals reported to be used in
the liver tumor analysis reported by the authors for the control, 0.05, 0.5, and 5.0 g/L treatment
groups, respectively. Complete pathological exams were reported to be performed for all tissues
from animals in the high dose TCA group at 104 weeks. No indication was given as to whether a
complete necropsy and pathological exam was performed for controls at terminal sacrifice.
Tritiated thymidine was reported to be administered at interim sacrifices five days prior to
sacrifice and to be examined with autoradiography. The 5 g/L TCA treatment group was reported
to have a reduction in growth to 89.3% of controls.
For water consumption TCA versus reported to slightly decrease water consumption at all
doses with a 7, 8, and 4% decrease in water consumption reported for 0.05, 0.5 and 5.0 g/L TCA,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-210 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
respectively. Body weight was decreased by 5.0 g/L TCA dose only through 78 weeks of
exposure to 89.3% of the control value. All of the percent liver/body weight ratios were reported
to be slightly decreased (1-4%) by all of the exposure concentrations of TCA but the data shown
does not indicate if the liver weight data were taken at interim sacrifice times and appears to be
only for animals at terminal sacrifice of 104 weeks.
No data were shown for hepatocyte proliferation but the authors reported no TCA
treatment effects. For PCO there was a 2.3-fold difference between control values between the
15-week and 104-week data. For the 0.05 and 0.5 g/L TCA treatment groups there was not a
statistically significant difference reported between control and treated group PCO levels. At
15 weeks the PCO activity was reduced by 55%, increased to 1.02-fold, and increased 2.12-fold
of control for 0.05, 0.5 and 5.0 g/L TCA exposures, respectively. For the 30 week exposure
groups, the 0.05 and 0.5 g/L TCA groups were reported to have PCO levels within 5% of the
control level. However, for the 5.0 g/L TCA treatment groups there was ~2-fold of control PCO
activity at the 15, 30, 45 and 60 weeks and at 104 weeks there was a 4-fold of control PCO
activity. Of note is that the control PCO value was lowest at 104 weeks while the TCA treatment
group was similar to interim values.
For analysis of liver tumors, there were 20-24 animals examined in each group. Unlike
the study of DeAngelo et al. (2008), it appeared that most of the animals that were sacrificed at
104 weeks were used in the tumor analysis without addition of "extra" animals or deletion of
animal data. The incidence of adenomas was reported to be 4.4, 4.2, 15, and 4.6% and the
incidence of hepatocellular carcinomas was reported to be 0, 0, 0, and 4.6% for the control, 0.05,
0.5, and 5.0 g/L TCA exposure groups. The multiplicity or tumors/animal was reported to be
0.04, 0.08, 0.15, and 0.05 for adenomas and 0, 0, 0, and 0.05 for carcinomas for the control, 0.05,
0.5, and 5.0 g/L TCA exposure groups.
Although there was an increase in the incidence of adenomas at 0.5 g/L and an increase in
carcinomas at 5.0 g/L TCA, they were not reported to be statistically significant by the authors.
Neither were the increases in adenoma multiplicity at the 0.05 and 0.5 g/L exposures. However,
using such a low number of animals per treatment group (n = 20-24) limits the ability of this
study to determine a statistically significant increase in tumor response and to be able to
determine that there was no treatment-related effect. A power calculation of the study shows that
the type II error, which should be >50%) and thus, greater than the chances of "flipping a coin,"
was less than 6% for incidence and multiplicity of tumors at all exposure DCA concentrations
with the exception of the incidence of adenomas for 0.5 g/L treatment group (58.7%). Therefore,
the designed experiment could accept a false null hypothesis, especially in terms of tumor
multiplicity, at the lower exposure doses and erroneously conclude that there is no response due
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-211 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
to TCA treatment. Thus, while suggesting a lower response than for mice for TCA-induced liver
tumors, the study is inconclusive for determination of whether TCA induces a carcinogenic
response in the liver of rats. The experimental design is such that extrahepatic carcinogenicity of
TCA in the male rat cannot be determined.
E.2.4.2.27. DeAngelo et al. (1996). In this study, 28-day-old male F344 rats were given
drinking water containing DCA at concentrations of 0, 0.05, 0.5, or 5.0 g/L with another
group
was provided water containing 2.0 g/L NaCl for 100 weeks. This experiment modified its
exposure protocol due to toxicity (peripheral neuropathy) such that the 5.0 g/L group was lowered
to 2.5 g/L at 9 weeks and then 2.0 g/L at 23 weeks and finally to 1.0 g/L at 52 weeks. When the
neuropathy did not reverse or diminish, the animals were sacrificed at 60 weeks and excluded
from the results. Based on measured water intake in the 0, 0.05, and 0.5 g/L groups, the time-
weighted average doses were reported to be 0, 3.6, and 40.2 mg/kg/d respectively. This
experiment was conducted at a U.S. EPA laboratory in Cincinnati and the controls for this group
were given 2.0 g/L NaCl (Study #1). In a second study rats were given either deionized water or
2.5 g/L DCA, which was also lowered to 1.5 g/L at 8 weeks and to 1.0 g/L at 26 weeks of
exposure (Study #2).
Although 23 animals were reported to be sacrificed at terminal sacrifice that had been
given 2 g/L NaCl, the number of animals reported to be examined in this group for hepatocellular
lesions was 3. The incidence data for this group for adenomas was 4.4% so this is obviously a
typographical error. The number of rats included in the water controls for tumor analysis was
reported to be 33 which was the same number as those at final sacrifice. The number of animals
at final sacrifice was reported to be 23 for 2 g/L NaCl, 21 for 0.05 g/L DCA, 23 for 0.5 g/L DCA
in experiment #1 and 33 for deionized water and 28 for the initial dose of 2.5 g/L DCA in
experiment #2.
Although these were of the same strain, the initial body weight was 59.1 g versus 76 g for
the 2.0 g/L control group versus deionized water group. The treatment groups in both studies
were similar to the deionized water group. The percent liver/body weights were greater (4.4 vs.
3.7% in the NaCl vs. deionized water control groups (-20%). The number of unscheduled deaths
was greater in Study #2 (22%) than in Study #1 (12%). Interim sacrifice periods were conducted.
As with the DeAngelo et al. (DeAngelo et al., 2008) study in mice, the number of animals
reported at final sacrifice was not the same as the number examined for liver tumors in Study #1
(5 more animals examined than sacrificed at the 0.05 g/L DCA and 6 more animals examined
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-212 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
than sacrificed at the 0.5 g/L DCA exposure groups) with n = 23, n = 26, and n = 29 for the 2 g/L
NaCl, 0.05 g/L DCA and 0.5 g/L DCA groups utilized in the tumor analysis. For Study #2 the
same number of rats was reported to be sacrificed as examined. The source of the extra animals
for tumor analysis in Study #1, whether from interim sacrifice or unscheduled deaths, was not
given by the authors and is unknown. Carcinomas prevalence data were not reported for the
control group or 0.05 g/L DCA group in Study #1 and multiplicity data were not reported to the
control group, or 0.05 g/L DCA group. Multiplicity was not reported for adenomas in the 0.05
g/L DCA group in Study #1.
There was a lack of hepatocyte DNA synthesis and necrosis reported at any dose group
carried out to final sacrifice at 100 weeks. The authors reported that the incidence of adenomas to
be 4.4% in 2 g/L NaCl control, 0 in 0.05 g/L DCA, and 17.2% in the 0.5 g/L DCA exposure
groups. For carcinomas no data were reported for the control or 0.05 g/L DCA group but an
incidence of 10.3% was reported for the 0.5 g/L DCA group. The authors reported increased
hepatocellular adenomas and carcinomas in male F344 rats although no data were reported for
carcinomas in the control and 0.05 g/L exposure groups. They reported that for 0.5 g/L DCA,
24.1 versus 4.4% adenomas and carcinomas combined (Study #1) and 28.6 versus 3.0%
(Study #2) at what was initially 2.5 g/L DCA but continuously reduced. Tumor multiplicity was
reported to be significantly increased in the 0.5 g/L DCA group (0.04 adenomas and
carcinomas/animal in control vs. 0.31 in 0.5 g/L DCA in Study #1 and 0.03 in control vs. 0.36 in
what was initially 2.5 g/L DCA in Study #2). The issues of use of a small number of animals,
additional animals for tumor analysis in Study #1, and most of all the lack of a consistent dose for
the 2.5 g/L animals in Study #2, are obvious limitations for establishment of a dose-response for
DCA in rats.
E.2.4.2.28. Richmond et al. (1995). This study was conducted by the same authors as
DeAngelo et al. (1996) and appears to report results for the same data set for the 2 g/L
NaCl control,
0.05 g/L DCA and 0.5 g/L DCA exposed groups. Of note is that while DeAngelo et al. (1996)
refer to the 28-day old rats as "weanlings" the same aged rats are referred to as "adults" in this
study. Male Fischer 344 rats were administered time-weighted average concentrations of 0, 0.05,
0.5, or 2.4 g/L DCA in drinking water. Concentrations were kept constant but due to hind-limb
paralysis all 2.4 g/L DCA exposed rats had been sacrificed by 60 weeks of exposure. In the
104-week sacrifice time, there were 23 rats reported to be analyzed for incidence of hepatocellular
adenomas and carcinomas in the control group, 26 rats in the 0.05 g/L DCA group and 29 rats in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-213 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
the 0.5 g/L DCA exposed group. This is the same number of animals included in the tumor
analysis reported in DeAngelo et al. (1996). Tumor multiplicity was not given.
Richmond et al. (1995) reported that there was a 4% incidence of adenomas reported in
the 2.0 g/L NaCl control animals, 0% at 0.05 g/L DCA, and 21% in the 0.5 DCA group at 104
weeks. These figures are similar to those reported by DeAngelo et al. (1996) for the same data set
with the exception of a 17.2% incidence of adenomas reported for the 0.5 g/L DCA group.
There were no hepatocellular carcinomas reported in the control or 0.05 g/L exposure
groups but a 10% incidence reported in the 0.5 g/L DCA exposure group at 104 weeks of
exposure. While carcinomas were not reported by DeAngelo et al. (1996) for the control and 0.05
g/L groups they are assumed to be zero in the summary data for carcinomas and adenomas
combined. The 10% incidence at 0.5 g/L DCA is similar to the 10.4% incidence reported for this
group by DeAngelo et al. (1996).
At 60 weeks at 2.4 g/L DCA, the incidences of hepatocellular adenomas were reported to
be 26%) and hepatocellular carcinomas to be 4%. This is not similar to the values reported by
DeAngelo for 2.5 g/L DCA that was continuously decreased so that the estimated final
concentration was 1.6 g/L DCA for 100 weeks. For those animals, the incidence of adenomas
was reported by DeAngelo et al. (1996) to be 10.7% and carcinomas 21.4%, probably more a
reflection of longer exposure time allowing for adenoma to carcinoma progression. The authors
did not report any of the results of DCA-induced increases of adenomas and carcinomas to be
statistically significant. As it appears the same data set was used for the 2.0g/L NaCl control,
0.05 g/L DCA and 0.5 g/L DCA exposure groups as was reported in DeAngelo et al. (1996), the
same issues arise as regarding the differences in numbers of animals were included in tumor
analysis than were reported to have been present at final sacrifice. As stated previously for the
DeAngelo et al. (1997) study of TCA in rats, the use of small numbers of rats limits the detection
of and ability to determine whether there was no treatment-related effects, especially at the low
concentrations of DCA exposure.
E.2.5. Summaries and Comparisons Between Trichloroethylene (TCE), Dichloroacetic
Acid (DCA), and Trichloroacetic Acid (TCA) Studies
There are a number of studies to TCE that have reported effects on the liver. However,
the study of this compound is difficult as its concentration does not remain stable in drinking
water, some studies have been carried out using TCE with small quantities of a carcinogenic
stabilizing agent, some studies have been carried out in whole body inhalation chambers that
resulted in additional oral administration and for which individual animal data were not recorded
throughout the experiment, and the results of gavage studies have been limited by gavage related
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-214 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
deaths and vehicle effects. In addition some studies have been conducted using the i.p. route of
administration, which results in route-related toxicity and inflammation. For many studies, liver
effects consisted of measured increases in liver weight with little or no description of attendant
histological changes induced by TCE treatment. A number of studies were conducted at a few
relatively high doses with attendant effects on body weight, indicative of systemic toxicity and
affecting TCE-induced liver weight gain. Although, many studies have been performed in male
mice, the inhalation studies of Kjellstrand et al. indicate that male mice, regardless of strain
appear to have a greater variability in response, as measured by TCE-induced liver weight gain,
and susceptibility to TCE-induced decreases in body weight than female mice. However, the
body of the TCE literature is consistent in identifying the liver as a target of TCE-induced affects
and with the most commonly reported change to be a dose-related TCE-induced increase in liver
weight in multiple species, strains, and genders from both inhalation and oral routes of exposure.
The following sections will not only summarize results for studies of TCE reported in
Sections E.2. l-E.2.2, but provide comparison of studies of either TCA or DCA that have used
similar paradigms or investigated similar parameters described in Sections E.2.3.1 and E.2.3.2. A
synopsis of the results from studies of CH and in comparison with TCE results is presented in
Section E.2.5. While the study of Bull et al. (2002), described in Section E.2.2.21, presents data
for combinations of DCA or TCA exposure for comparisons of tumor phenotype with those
induced by TCE, the examination of co-exposure studies of TCE metabolites in rodents that are
also exposed to a number of other carcinogens, and descriptions of the toxicity data for
brominated haloacetates that also occur with TCE in the environment, are presented in Section
E.4.3.3.
E.2.5.1. Summary of Results For Short-term Effects of Trichloroethylene (TCE)
In regard to early changes in DNA synthesis, the data for TCE are very limited. The study
by Mirsalis et al. (1989) used an in vivo-in vitro hepatocyte DNA repair and S-phase DNA
synthesis in primary hepatocytes from male Fischer-344 rats (180-300 g) and male and female
B6C3F1 mice (20-29 g for male mice and 18-25 g female mice) administered TCE by gavage in
corn oil. They reported negative results 2-12 hours after treatment from 50-1,000 mg/kg TCE in
rats and mice (male and female) for unscheduled DNA synthesis and repair using 3 animals per
group. After 24 and 48 hours of 200 or 1,000 mg/kg TCE in male mice (n = 3) and after 48 hours
of 200 (n = 3) or 1,000 (n = 4) mg/kg TCE in female mice, similar values of 0.30 to 0.69% of
hepatocytes were reported as undergoing DNA synthesis in those hepatocytes in primary culture
with only the 1,000 mg/kg TCE dose in male mice at 48 hours giving a result considered to be
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-215 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
positive (-2.2%). No statistical analyses were performed on these measurements, which were
obviously limited by both the number of animals examined and the relevance of the paradigm.
TCE-induced increases in liver weight have been reported to occur quickly. The
inhalation study of Okino et al. (1991) in male rats demonstrates that liver weight and
metabolism were increased with as little as 8 hours of TCE exposure (500 and 2,000 ppm) and as
early as 22 hours after cessation of such exposures with little concurrent hepatic necrosis.
Laughter reported increase liver weight in SV129 mice in their 3-days study (see below). Tao et
al. (2000a) reported a 1.26-fold of control percent liver/body weight in female B6C3F1 mice fed
1,000 mg/kg TCE in corn oil for 5 days. Elcombe et al. (1985) and Dees and Travis (1993)
reported gavage results in mice and rats after 10 days exposure to TCE which showed TCE-
induced increases in liver weight (see below for more detail on dose-response). Tucker et al.
(1982) reported that 14 days of exposure to 24 mg/kg and 240 mg/kg TCE via gavage to induce a
dose-related increase in liver weight in male CD-I mice but did not show the data.
TCE-induced increases in percent liver/body weight ratios have been studied most
extensively in B6C3F1 and Swiss mice. Both strains have been shown to have a TCE-induced
increase in liver tumors from long-term exposure as well (see Section E.2.4.2, below). A number
of studies have provided dose-response information for TCE-induced increases in liver weight
from 10 days to 13 weeks of exposure in mice. Most studies have reported that the magnitude of
increase in TCE exposure concentration is similar to the magnitude increase of percent liver/body
weight increase. For example a 2-fold increase in TCE exposure has often resulted in a 2-fold
increase in the percent change in liver/body weight over control (i.e., 500 mg/kg TCE induces a
20% increase in liver weight and 1,000 mg/kg TCE induces a 50% increase in liver weight as
reported by Elcombe et al. (1985). The range in which this relationship is valid has been reported
to vary from 100 mg/kg TCE at 10 days (Dees and Travis, 1993) to 1,600 mg/kg (Buben and
O'Flaherty, 1985) at 6 weeks and up to 1,500 mg/kg TCE for 13 weeks (NTP, 1990). The
consistency in the relationship between magnitude of liver weight increase and TCE exposure
concentration has been reported for both genders of mice, across oral and inhalation routes of
exposure, and across differing strains of mice tested. For rats, there are fewer studies with fewer
exposure levels tested, but both Berman et al. (1995) and Melnick et al. (1987) report that short-
term TCE exposures from 150 mg/kg to -2,000 mg/kg induced percent liver/body weight that
increased proportionally with the magnitude of TCE exposure concentration.
Dependence of PPARa activation for TCE-liver weight gain has been investigated in
PPARa null mice by both Nakajima et al. (2000) and Laughter et al. (2004). After 2 weeks of
750 mg/kg TCE exposure to carefully matched SV129 wild-type or PPARa-null male and female
mice (n = 6 group), there was a reported 1.50-fold of control in wild-type and 1.26-fold of control
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-216 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
percent liver/body weight in PPARa-null male mice by Nakajima et al. (2000). For female mice,
there was ~1.25-fold of control percent liver/body weight ratios for both wild-type and PPARa-
null mice. Ramdhan et al. (2010) also reported increased liver weight in male PPARa-null mice
after a high levels of inhalation exposure that were comparable to that in wild type mice after 7
days of exposure (up to 40-50% increases at the highest dose). Thus, TCE-induced liver weight
gain was not dependent on a functional PPARa receptor in female mice and data indicate that a
significant portion of it may have also not have been PPARa receptor-dependent in male mice.
Nakajima et al. (2000) report that both wild-type male and female mice were reported to
have similar increases in the number of peroxisome in the pericentral area of the liver after TCE
exposure and, although increased 2-fold, were still only -4% of cytoplasmic volume. Female
wild-type mice were reported to have less TCE-induced elevation of very long chain acyl-CoA
synthetase, D-type peroxisomal bifunctional protein, mitochondrial trifunctional protein a
subunits a and P, and cytochrome P450 4A1 than males mice, even though peroxisomal volume
was similarly elevated in male and female mice. The induction of PPARa protein by TCE
treatment was also reported to be slightly less in female than male wild-type mice (2.17- vs. 1.44-
fold of control, respectively).
Ramdhan et al. (2010) examined TCE-induced hepatice steatosis and toxicity using male
wild type, PPARa-null, and human PPARa-inserted mice (humanized) exposed to high inhalation
concentrations of TCE for 7 days. Significant differences were observed among control mice for
each genotype with reduced body weight in untreated humanized mice. Liver/body weight ratios
were 11% higher in untreated PPARa- null mice than wild type mice. Higher levels of liver
triglycerides and hepatic steatosis were reported in the untreated humanized mice and PPARa-
null mice than wild type mice. Background expression of a number of genes and protein
expression levels were significantly different between the untreated strains. In particular human
PPARa protein levels were >10-fold greater in the humanized mice than mouse PPARa in
untreated wild type mice. Insertion of human PPARa in the null mice did not return the mice to a
normal state. Both PPARa-null and humanized mice were more susceptible to TCE toxicity as
evidenced by serum AST and ALT (liver injury biomarkers), hepatic triglyceride levels, and
hepatic steatosis. Hepatomegally was induced in all strains to a similar extent after TCE
exposure. However, urinary TCA concentrations were reported to be significantly lower and
trichloroethanol levels significantly higher in TCE-treated PPARalpha-null mice in comparison to
treated wild type mice. This difference was not realted to changes in expression of metabolic
enzymes. Thus, TCE-induced liver toxicity was not dependent on PPARa with dysregulation of
the receptor in null or humanized mice rendering them more susceptible to TCE-induced toxicity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-217 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Laughter et al. (2004) also studied SV129 wild-type and PPARa-null male mice treated
with 3 daily doses of TCE in 0.1% methyl cellulose for either 3 days (1,500 mg/kg TCE) or
3 weeks (0, 10, 50, 125, 500, 1,000, or 1,500 mg/kg TCE 5 days a week). However, not only is
the paradigm not comparable to other gavage paradigms, but no initial or final body weights of
the mice were reported and thus, the influence of differences in initial body weight on percent
liver/body weight determinations could not be ascertained. In the 3-day study, while control
wild-type and PPARa-null mice were reported to have similar percent liver/body weight ratios
(-4.5%), at the end of the 3-week experiment the percent liver/body weight ratios were reported
to be increased in the PPARa-null male mice (5.1%).
TCE treatment for 3 days was reported to increase the percent liver/body weight ratio 1.4-
fold of control in the wild-type mice and 1.07-fold of control in the null mice. In the 3-week
study, wild-type mice exposed to various concentrations of TCE had percent liver/body weights
that were reported to be within -2% of control values except for the 1,000 mg/kg and 1,500
mg/kg groups (-1.18- and 1.30-fold of control levels, respectively). For the PPARa-null mice the
variability in percent liver/body weight was reported to be greater than that of the wild-type mice
in most of the groups and the baseline level of percent liver/body weight ratio also 1.16-fold
greater. TCE exposure was apparently more toxic in the null mice with death at the 1,500 mg/kg
TCE exposure level resulting in the prevention of recording of percent liver/body weights. At
1,000 mg/kg TCE exposure level there was a reported 1.10-fold of control percent liver/body
weight in the PPARa-null mice.
None of the increases in percent liver/body weight in the null mice were reported to be
statistically significant by Laughter et al. (2004). However, the statistical power of the study was
limited due to low numbers of animals and increased variability in the null mice groups. The
percent liver/body weight after TCE treatment that was reported in this study was actually greater
in the null mice than the wild-type male mice at the 1,000 mg/kg TCE exposure level
(5.6%) ± 0.4%) vs. 5.2% ± 0.5%o, for null and wild-type mice, respectively). At 1-weeks and at
3-weeks, TCE appeared to induce increases in liver weight in PPARa-null mice, although not
reaching statistical significance in this study. At a 1,000 mg/kg TCE exposure for 3 weeks
percent liver/body weights were reported to be 1.18-fold of control in wild-type and 1.10-fold of
control in null mice. Although the experiments in Laughter et al. for DCA and TCA were not
conducted using the same paradigm, the TCE-induced increase in percent liver/body weight more
closely resembled the dose-response pattern for DCA than for DCA wild-type SV129 and
PPARa-null mice.
Many studies have used cyanide-insensitive PCO as a surrogate for peroxisome
proliferation. Of note is that several studies have shown that this activity is not correlated with
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-218 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
the volume or number of peroxisomes that are increased as a result of exposure to TCE or it
metabolites (Elcombe et al., 1985; Nakajima et al., 2000; Nelson et al., 1989). This activity
appears to be highly variable both as a baseline measure and in response to chemical exposures.
Laughter et al. (2004) presented data showing that WY -14,643 induced increases in PCO activity
varied up to 6-fold between experiments in wild-type mice. They also showed that PCO activity,
in some instances, was up to 6-fold of wild-type mice values in untreated PPARa-null mice.
Parrish et al. (1996) noted that control values between experiments varied as much as a factor of
2-fold for PCO activity and thus, their data were presented as percent of concurrent controls.
Goldsworthy and Popp (1987) reported that 1,000 mg/kg TCE induced a 6.25-fold of control PCO
activity in B6C3F1 mice in two 10-day experiments. However, for F344 rats, the increases over
control between two experiments conducted at the same dose were reported to vary by >30%.
Finally, Melnick et al. (1987) have reported that corn oil administration alone can elevate PCO
activity as well as catalase activity.
For TCE there are two key 10-days studies (Dees and Travis, 1993; Elcombe et al., 1985)
that examine the effects of short-term exposure in mice and rats via gavage exposure and attempt
to determine the nature of the dose- response in a range of exposure concentrations that include
levels below which there is concurrent decreased body weights. Although they have limitations,
they reported generally consistent results. In regard to liver weight in mice, gavage exposure to
TCE at concentrations ranging from 100 to 1,500 mg/kg TCE produced increases in liver/body
weight that was dose-related (Dees and Travis, 1993; Elcombe et al., 1985).
Elcombe et al. (1985) reported a small decrease in DNA content with TCE treatment
(consistent with hepatocellular hypertrophy) that was not dose-related, increased tritiated
thymidine incorporation in whole mouse liver DNA that was that was treatment but not dose-
related (i.e., a 2-, 2-, and 5-fold of control values in mice treated with 500, 1,000, and
1,500 mg/kg TCE), and slightly increased numbers of mitotic figures that were treatment but not
dose-related and not correlated with DNA synthesis as measured by thymidine incorporation.
Elcombe et al. (1985) reported an increase in peroxisome volume after TCE exposure that was
correlated with the magnitude of increase in peroxisomal-associated enzyme activity at the only
dose in which both were tested. Peroxisome increases after TCE treatment in mice livers were
identified as being pericentral in location. After TCE treatment, increased peroxisomal volumes
in B6C3F1 mice were reported to be not dose-related (i.e., there was little difference between 500
to 1,500 mg/kg TCE exposures). The TCE-induced increases in peroxisomal volumes were also
not correlated with the reported increases in thymidine incorporation or mitotic activity in mice.
Neither TCE-induction of peroxisomes or hepatocellular proliferation, as measured by
either mitotic index or thymidine incorporation, was correlated with TCE-induced liver weight
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-219 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
increases. Elcombe et al. (1985) only measured PCO activity in a subset of B6C3F1 mice at the
1,000 mg/kg TCE exposure level for 10 days of exposure and reported an 8-fold of control PCO
activity and a 1.5-fold of control catalase activity. This result was similar to that of Goldsworthy
and Popp (1987) who reported 6.25-fold of control PCO activity in male B6C3F1 mice exposed
to 1,000 mg/kg/d TCE for 10 days in two separate experiments.
Similar to Elcombe et al., who reported no difference in response between 500 and
1,000 mg/kg TCE treatments, (Dees and Travis, 1993) reported that incorporation of tritiated
thymidine in DNA from mouse liver was elevated after TCE treatment and the mean peak level of
tritiated thymidine incorporation occurred at 250 mg/kg TCE treatment level remaining constant
for the 500 and 1,000 mg/kg treated groups. (Dees and Travis, 1993) specifically report that
mitotic figures, although very rare, were more frequently observed after TCE treatment, found
most often in the intermediate zone, and found in cells resembling mature hepatocytes. They
reported that there was little tritiated thymidine incorporation in areas near the bile duct epithelia
or close to the portal triad in liver sections from both male and female mice. They also reported
no evidence of increased lipofuscin and that increased apoptoses from TCE exposure "did not
appear to be in proportion to the applied TCE dose given to male or female mice" (i.e., the mean
number of apopotosis 0, 0, 0, 1 and 8 for control, 100, 250, 500, and 1,000 mg/kg TCE treated
groups, respectively). Both Elcombe et al. (1985) and (Dees and Travis, 1993) reported no
changes in apoptosis other than increased apoptosis only at a treatment level of 1,000 mg/kg TCE.
Elcombe et al. (1985) reported increased in percent liver/body weight after TCE treatment
in both the Osborne-Mendel and Alderly Park rat strain, although to a smaller extent than in mice.
For both strains, Elcombe et al. (1985) reported no TCE-induced changes in body weight at doses
ranging from 500 to 1,500 mg/kg. For male Osborne-Mendel rats administration of TCE in corn
oil gavage resulted in a 1.18-, 1.26-, and 1.30-fold of control percent liver/body weight at
500 mg/kd/day, 1,000 mg/kg/d, and 1,500 mg/kg/d exposures, respectively. For Alderly Park rats
those increases were 1.14-, 1.17-, and 1.17-fold of control at the same respective exposure levels
for 10 days of exposure.
In regard to liver weight increases, Melnick et al. (1987) reported a 1.13- and 1.23-fold of
control percent liver/body weight in male Fischer 344 rats fed 600 mg/kg/d and 1,300 mg/kg/d
TCE in capsules, respectively. There was no difference in the extent of TCE-induced liver
increase between the two lowest dosed group administered TCE in corn oil gavage (-20%
increase in percent liver/body weight at 600 mg/kd and 1,300 mg/kg TCE) for 14 days. However,
the magnitude of increases in percent liver/body weight in these groups was affected by difference
between control groups in liver weight although initial and final body weights appeared to be
similar. By either type of vehicle, Melnick et al. (1987) reported decreases in body weights in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-220 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
rats treated with concentrations of TCE 2,200 mg/kg/d or greater for 14 days. Similarly, Nunes et
al. (2001) reported decreased body weight in S-D rats administered 2,000 mg/kg/d for 7 days in
corn oil. Melnick et al. (1987) reported that both exposures to either 600 or 1,300 mg/kg/d TCE
in capsules did not result in decreased body weight and caused less than minimal focal necrosis
randomly distributed in the liver. At 2,200 and 4,800 mg/kg TCE fed via capsule, Melnick et al.
(1987) reported that although there was decreased body weight in rats treated at these exposures,
there was little TCE-induced necrosis, and no evidence of inflammation, cellular hypertrophy or
edema with TCE exposure. Similarly, Berman et al. (1995) reported increases in liver weight
gain at doses as low as 50 mg/kg TCE, no necrosis up to doses of 1,500 mg/kg, and hepatocellular
hyper trophy only at the 1,500 mg/kg level in female Fischer 344 rats.
For rats, Elcombe et al. (1985) reported an increase over untreated rats of 1.13-fold of
control PCO activity in Alderly Park rats after 1,000 mg/kg/d TCE exposure for 10 days, while
Goldsworthy and Popp (1987) reported a 1.8- and 2.39-fold of control in male Fischer 344 rats at
the same exposure in two separate experiments. Melnick et al. (1987) reported PCO activity of
1.23- and 1.75-fold of control in male Fischer 344 rats fed 600 mg/kg/d and 1,300 mg/kg/d TCE
for 14 days in capsules. For rats treated by gavage with 600 mg/kg/d or 1,200 mg/kg d TCE corn
oil, they reported 1.16- and 1.29-fold of control values. However, control levels of PCO were
16% higher in corn oil controls than in untreated controls. In addition Melnick et al. (1987)
reported little catalase increases in rats fed TCE via capsules in food (less than 6% increase) but a
1.18- and 1.49-fold of control catalase activity in rats fed 600 mg/kg/d or 1,200 mg/kg/TCE via
corn oil gavage, indicative of a vehicle effect.
The data from Elcombe et al. (1985) included reports of TCE-induced pericentral
hypertrophy and eosinophilia for both rats and mice but with "fewer animals affected at lower
doses." In terms of glycogen deposition, Elcombe report "somewhat" less glycogen pericentrally
in the livers of rats treated with TCE at 1,500 mg/kg than controls with less marked changes at
lower doses restricted to fewer animals. They do not comment on changes in glycogen in mice.
Dees and Travis (1993) reported TCE-induced changes to "include an increase in eosinophilic
cytoplasmic staining of hepatocytes located near central veins, accompanied by loss of
cytoplasmic vacuolization." Since glycogen is removed using conventional tissue processing and
staining techniques, an increase in glycogen deposition would be expected to increase
vacuolization and thus, the report from Dees and Travis is consistent with less not more glycogen
deposition. Neither study produced a quantitative analysis of glycogen deposition changes from
TCE exposure. Although not explicitly discussing liver glycogen content or examining it
quantitatively in mice, these studies suggest that TCE-induced liver weight increases did not
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-221 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
appear to be due to glycogen deposition after 10 days of exposure and any decreases in glycogen
were not necessarily correlated with the magnitude of liver weight gain either.
For both rats and mice the data from Elcombe et al. (1985) showed that tritiated thymidine
incorporation in total liver DNA observed after TCE exposure did not correlate with mitotic index
activity in hepatocytes with both Elcombe et al. (1985) and Dees and Travis (1993) reporting a
small mitotic indexes and evidence of periportal hepatocellular hypertrophy from TCE exposure.
Neither mitotic index or tritiated thymidine incorporation data support a correlation with TCE-
induced liver weight increase in the mouse. If higher levels of hepatocyte replication had
occurred earlier, such levels were not sustained by 10 days of TCE exposure. Both Elcombe et al.
(1985) and Dees and Travis (1993) present data that represent "a snapshot in time" which does
not show whether increased cell proliferation may have happened at an earlier time point and then
subsided by 10 days. These data suggest that increased tritiated thymidine levels were targeted to
mature hepatocytes and in areas of the liver where greater levels of polyploidization occur. Both
Elcombe et al. (1985) and Dees and Travis (1993) show that tritiated thymidine incorporation in
the liver was ~2-fold of controls between 250-1,000 mg/kg TCE, a result consistent with a
doubling of DNA. Thus, given the normally quiescent state of the liver, the magnitude of this
increase over control levels, even if a result of proliferation rather than polyploidization, would be
confined to a very small population of cells in the liver after 10 days of TCE exposure.
Laughter et al. (2004) reported that there was an increase in DNA synthesis after aqueous
gavage exposure to 500 and 1,000 mg/kg TCE given as 3 boluses a day for 3 weeks with BrdU
given for the last week of treatment in mice. An examination of DNA synthesis in individual
hepatocytes was reported to show that 1 and 4.5% of hepatocytes had undergone DNA synthesis
in the last week of treatment for the 500 and 1,000 mg/kg doses, respectively. Both Elcombe et
al. (1985) and Dees and Travis (1993) show TCE-induced changes for several parameters at the
lowest level tested without toxicity and without evidence of regenerative hyperplasia or sustained
hepatocellular proliferation.
In regards to susceptibility to liver cancer induction, the more susceptible (B6C3F1)
versus less susceptible (Alderly Park/Swiss) strains of mice to TCE-induced liver tumors (Maltoni
et al., 1988), the "less susceptible" strain was reported by Elcombe et al. (1985) to have, a greater
baseline level of liver weight/body weight ratio, a greater baseline level of thymidine
incorporation as well as greater responses for those endpoints due to TCE exposure. However,
both strains showed a hepatocarcinogenic response after TCE exposure, although there are
limitations regarding determination of the exact magnitude of response for these experiments as
previously discussed.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-222 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E.2.5.2. Summary of Results For Short-Term Effects of Dichloroacetic Acid (DCA) and
Trichloroacetic Acid (TCA): Comparisons With Trichloroethylene (TCE)
Short-term exposures from DCA and TCA have been studied either through gavage or in
drinking water. Palatability became an issue at the highest level of DCA tested in drinking water
experiments (5 g/L) which caused a significant reduction of drinking water intake in mice of 46 to
64% (Carter et al., 1995). Decreases in drinking water consumption have also been reported for a
range of concentrations of DCA and TCA from 0.05 g/L to 5.0 g/L, in both mice and rats, and
with generally the higher concentrations producing the highest decrease in drinking water (Carter
et al., 1995; DeAngelo et al., 1997; DeAngelo et al., 1999; Mather et al., 1990) (DeAngelo et al.,
2008). However, results within studies (e.g., (DeAngelo et al., 2008) and between studies have
been reported to vary as to the extent of the reduction in drinking water from the presence of TCA
or DCA. Some drinking water studies of DCA or TCA have not reported drinking water
consumption as well. Therefore, although in general DCA and TCA studies have do not include
vehicle effects, such as those posed by corn oil, they have been affected by differences in drinking
water consumption not only changing the dose received by the rodents and therefore, potentially
the shape of the dose-response curve, but also the effects of dehydration are potentially added to
any chemically-related reported effects.
Studies have attempted to determine short-term effects on DNA by TCE and its
metabolites. Nelson and Bull (1988) administered TCE male to Sprague Dawley rats and male
B6C3F1 mice and measured the rate of DNA unwinding under alkaline conditions 4 hours later.
For rats there was a significantly increased rate of unwinding at the two highest dose and for mice
there was a significantly increased level of DNA unwinding at a lower dose. In this same study,
DCA was reported to be most potent in this assay with TCA being the lowest, while CH closely
approximated the dose-response curve of TCE in the rat. In the mouse the most potent metabolite
in the assay was reported to be TCA followed by DCA with CH considerably less potent. Nelson
and Bull (1988) and Nelson et al. (1989) have reported increases in single strand breaks after
DCA and TCA exposure. However, Styles et al. (1991) (for mice) and Chang et al. (1992) (for
mice and rats) did not. Austin et al. (1996) note that the alkaline unwinding assay, a variant of the
alkaline elution procedure, is noted for its variability and inconsistency depending on the
techniques used while performing the procedure. In regard to oxidative damage as measured by
TBARS for lipid peroxidation and 8-OHdG levels in DNA, increases appear to be small (less than
50% greater than control levels) and transient after DCA and TCA treatment in mice (see Section
E.3.4.2.3) with TCE results confounded by vehicle or route of administration effects.
Although there is no comparative data for TCE, the study of Styles et al. (1991) is
particularly useful for determining effects of TCA from 1 to 4 days of exposure in mice. Styles et
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-223 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
al. (1991) reported no change in "hepatic" DNA uptake of tritiated thymidine up to 36 hours, a
peak at 72 hours (~6-fold of control), and falling levels by 96 hours (~4-fold of controls) after
500 mg/kg TCA gavage exposure. Incorporation of tritiated thymidine observed for individual
hepatocytes decreased between 24 and 36 hours, rose slowly back to control levels at 48 hours,
significantly increased by 72 hours, and then decreased by 96 hours. Thus, increases in "hepatic"
DNA tritiated thymidine uptake did not capture the decrease observed in individual hepatocytes at
36 hours. By either measure, the population of cells undergoing DNA synthesis was small with
the peak level being less than 1% of the hepatocyte population. Zonal distribution of labeled
hepatocytes were decreased at 36 hours in all zones, appeared to be slightly greater in perioportal
than midzonal cells with centrilobular cells still below control levels by 48 hours, similarly
elevated over controls in all zones by 72 hours, and to have returned to near control levels in the
midzonal and centrilobular regions but with periportal areas still elevated by 96 hours. These
results are consistent with all hepatocytes showing a decrease in DNA synthesis by 36 hours and
then a wave of DNA synthesis to occur, starting at the periportal zone and progressing through the
liver acinus that is decreased by 4 days after exposure.
Along with changes in liver weight, DNA synthesis, and glycogen accumulation, several
studies of DCA and TCA have focused on the extent of peroxisome proliferation as measured by
changes in peroxisome number, cytoplasmic volume and enzyme activity induction as potential
"key events" occurring from shorter-term exposures that may be linked to chronic effects such as
liver tumorigenicity. As noted above in Section E.2.4.1, TCE-induced liver weight gain has been
reported to not be dependent on a functional PPARa receptor in female mice while as well as a
significant portion of it not dependent on functional PPARa receptor in male mice. Also as noted,
cyanide-insensitive PCO has also been reported to not be correlated with the volume or number of
peroxisomes that are increased as a result of exposure to TCE or it metabolites (Elcombe et al.,
1985; Nakajima et al., 2000; Nelson et al., 1989) and to be highly variable both as a baseline
measure and in response to chemical exposures (e.g., variation of up to 6-fold between after WY-
14,643 exposure in mice). Also as noted above, the vehicle used in many TCE gavage
experiments, com oil, has been reported to elevate PCO activity as well as catalase activity.
A number of short-term studies have examined the effects of TCA and DCA on liver
weight increases and evidence of peroxisome proliferation and changes in DNA synthesis. In
particular two studies of DCA and TCA used a similar paradigm presented by Elcombe et al.
(1985) and Dees and Travis (1993) for TCE effects in mice. Nelson et al. (1989) report findings
from gavage doses of unbuffered TCA (500 mg/kg) and DCA (500 mg/kg) in male B6C3F1 mice
and Styles et al. (1991) also providing data on peroxisome proliferation using the same paradigm.
Nelson et al. (1989) reported levels of PCO activity in mice administered 500 mg/kg DCA or
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-224 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
TCA for 10 days with 250 mg/kg Clofibrate administration serving as a positive control. DCA
and TCA exposure were reported to not affect body weight, but both to significantly increase liver
weight (1.63-fold of control for DCA and 1.30-fold of control for TCA treatments), and percent
liver/body weight ratios (1.53-fold of control for DCA and 1.16-fold of control for DCA
treatments). PCO activity was reported to be significantly increased by -1.63-, 2.7-, and 5-fold of
control for DCA, TCA and Clofibrate treatments, respectively and indicated that both DCA and
TCA were weaker inducers of this activity than Clofibrate.
Results from randomly selected electron photomicrographs showed an increase in
peroxisomes per unit area but gave a different pattern than PCO enzyme activity (i.e., 2.5- and
2.4-fold of control peroxisome volume for DCA and TCA, respectively). Evidence of gross
hepatotoxicity was reported to not occur in vehicle or TCA-treated mice. Light microscopic
sections were reported to show TCA and control hepatocytes to have the same intensity of PAS
staining, but with slightly larger hepatocytes occurring in TCA-treated mice throughout the liver
section with architecture and tissue pattern of the liver intact. For DCA, the histopathology was
reported to be markedly different than control mice or TCA treated mice. DCA was reported to
induce a marked increase in the size of hepatocytes throughout the liver with an approximately
1.4-fold of control diameter that was accompanied by increased PAS staining (indicative of
glycogen deposition). All DCA-treated mice were reported to have multiple white streaks grossly
visible on the surface of the liver corresponding with subcapsular foci of coagulative necrosis that
were not encapsulated, varied in size, and accompanied by a slight inflammatory response
characterized by neutrophil infiltration.
A quantitative comparison of effects from equivalent exposures of TCE, TCA, and DCA
(500 mg/kg for 10 days in mice via corn oil gavage for TCE) shown in Table E-13 can be drawn
between the Elcombe et al. (1985), Dees and Travis (1993), Styles et al. (1991), and Nelson et al.
(1989) data for relationship to control values for percent liver/body weight, PCO, and
qualitatively for glycogen deposition.
Table E-13. Comparison of liver effects from TCE, TCA, and DCA (10-day
exposures in mice)
Model
Expo-
sure
% Liver/body
wt.
Peroxisome
volume
Peroxisome
enzyme
activity
Glycogen
deposition
Nelson et al., (1989)a
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-225 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
B6C3F1 male
TCA
1.16-fold
2.4-fold
2.7-fold
No change
DCA
1.53-fold
2.5-fold
1.63-fold
Increased
Styles et al. (1991)
B6C3F1 male
TCA
NR
1.9-fold
NR
NR
Elcombe et al. (1985)
B6C3F1 male
TCE
1.20-fold
8-fold
NR
NR
Alderly Park male (Swiss)
TCE
1.43-fold
4-fold
NR
NR
Dees and Travis (1993)
B6C3F1 male
TCE
1.05-foldb
NR
NR
NR
B6C3F1 female
TCE
1.18-fold
NR
NR
NR
aUnbuffered. NR = not reported as no analysis was performed for this dose or the authors did not report this finding
(i.e., did not note a change in glycogen in description of exposure-related changes).
Statistically significant although small increase.
Although using a similar species, route of exposure, and dose, the comparison of
responses for TCE and its metabolites shown above are in male mice and also are reflective of
variability in strain, and variability and uncertainty of initial body weights. As described in more
detail in Section E.2.2, initial age and body weight have an impact on TCE-related increases in
liver weight. Male mice have been reported to have greater variability in response than female
mice within and between studies and most of the comparative data for the 10-day 500 mg/kg
doses of TCE or its metabolites were from studies in male mice. Corn oil, used as the vehicle for
TCE gavage studies but not those of its metabolites, has been noted to specifically affect
peroxisomal enzyme induction, body weight gain, and hepatic necrosis, specifically, in male mice
(Merrick et al., 1989). Corn oil alone has also been reported to increase PCO activity in F344 rats
and to potentiate the induction of PCO activity of TCA (DeAngelo et al., 1989). Thus,
quantitative inferences regarding the magnitude of response in these studies are limited by a
number of factors.
The variability in the magnitude of TCE-induced increases in percent liver/body weight
across studies in readily apparent but for TCE, TCA and DCA there is an increase in liver weight
in mice at this dose after 10 days of exposure. The volume of the peroxisomal compartment in
hepatocytes was reported to be more greatly increased from TCE-treatment by Elcombe et al.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-226 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
(1985) than for either TCA or DCA by Nelson et al. (1989) or Styles et al. (1991). However, the
control values for the B6C3F1 mice were half that of the other strain reported by Elcombe et al.
(1985)and this parameter in general did not match the pattern of PCO activity values reported for
TCA and DCA (Nelson et al., 1989). There is no PCO activity data at this dose for TCE but
Elcombe et al. (1985) reported that the magnitude of TCE-induced increase in peroxisome
volume was similar to that of PCO activity at the only dose where both were tested (1,000 mg/kg
TCE).
However, Elcombe et al. (1985) reported increased peroxisomal volumes in B6C3F1 mice
after 10 days of TCE treatment were not dose-related (i.e., there was little difference between 500,
1,000, and 1,500 mg/kg TCE exposures in the magnitude of TCE-induced increases in
peroxisomal volume). The lack of dose-response for TCE-induced peroxisomal volume increases
was not consistent with increases in percent liver/body weight that increased with increasing TCE
exposure concentration. Also as noted above, PCO activity appears to be highly variable in
untreated and treated rodents and to vary between experiments and between studies.
From the above comparison it is clear that TCE, DCA, and TCA exposures were
associated with increased liver weight in mice but a question arises as to what changes account
for the liver weight increases. For TCE and TCA 500 mg/kg treatments, changes in glycogen
were not reported in the general descriptions of histopathological changes (Dees and Travis,
1993; Elcombe et al., 1985; Styles et al., 1991) or were specifically described by the authors as
being similar to controls (Nelson et al., 1989). However, for DCA, glycogen deposition was
specifically noted to be increased with treatment, although no quantitative analyses was presented
that could give information as to the nature of the dose-response (Nelson et al., 1989). Issues in
regard to not only whether TCE and its metabolites each gives a similar response for a number of
parameters, but what potential changes may be associated with carcinogenicity from long-term
exposures can be examined by a comparison of the dose-response curves for these parameters
from a range of exposure concentrations and durations of exposure. In addition, if glycogen
accumulation results from DCA exposure, what proportion of DCA-induced liver weight
increases result from such accumulation or other events that may be similar to those occurring
with TCE exposure (see Section E.4.2.4, below)?
As noted above in Section E.2.4.I., TCE-induced changes in liver weight appear to be
proportional to the exposure concentration across route of administration, gender and rodent
species. As an indication of the potential contribution of TCE metabolites to this effect, a
comparison of the shape of the dose-response curves for liver weight induction for TCE and its
metabolites is informative. A number of studies of TCA and DCA in drinking water, conducted
from 10-days to 4 weeks, have attempted to measure changes in liver weight induction,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-227 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
peroxisomal enzyme activity, and changes in DNA synthesis predominantly in mice to provide
insight into the MOA(s) for liver cancer induction (Carter et al., 1995; DeAngelo et al., 1989;
DeAngelo et al., 2008; Parrish et al., 1996; Sanchez and Bull, 1990).
Direct comparisons are harder to make between the drinking water studies of DCA and
TCA and the gavage studies of TCE (Tables E-14, E-15, and E-16). Similar to 10-day gavage
exposures to TCE, 14-day exposures to TCA or DCA via drinking water were reported to induce
dose-related increases in liver weight in male B6C3F1 mice (0.3, 1.0, and 2.0 g/L TCA or DCA)
with a greater increase in liver weight from DCA than TCA at 2 g/L and a difference in the shape
of the dose-response curve (Sanchez and Bull, 1990). They reported a 1.08-, 1.31-, and 1.62-fold
of control liver weight for DCA and a 1.15-, 1.22-, and 1.38-fold of control values for TCA at 0.3
g/L, 1.0 g/L and 2.0 g/L concentrations, respectively (n = 12-14 mice). While the magnitude of
difference between the exposures was ~6.7-fold between the lowest and highest dose, the
differences between TCA exposure groups for change in percent of liver weight was -2.5, but for
DCA the slope of the dose-response curve for liver weight increases appeared to be closer to the
magnitude of difference in exposure concentrations between the groups (i.e., a difference of
7.7-fold between the highest and lowest dose for liver weight induction).
DeAngelo et al. (1989) reported that after 14 days of exposure to 5 g/L or 2 g/L TCA in
male mice, the magnitudes of the difference in the increase in exposure concentration (2.5-fold)
was generally higher than the increase percent liver/body weight ratios at these doses (i.e., -40%
for the Swiss-Webster, C3H, and for one of the B6C3F1 mouse experiments, and for the C57BL/6
mouse there was no difference in liver weight induction between the 2 and 5 g/L TCA exposure
groups). There was a range in the magnitude of percent liver/body weight ratio increases between
the strains of mice with liver weight induction reported to range between 1.26- to 1.66-fold of
control values for the 4 strains of mice at 5 g/L TCA and to range between 1.16- to 1.63-fold of
control values at 2 g/L TCA. One strain, B6C3F1, was chosen to compare responses between
DCA and TCA. At 1 g/L, 2 g/L and 5 g/L TCA or DCA, DCA was reported to induce a greater
increase in liver weight that TCA (i.e., 1.55- vs. 1.39-fold of control percent liver/body weight
ratio for 5.0 g/L DCA vs. TCA, respectively). At the 5 g/L exposures DCA induced -40% greater
percent liver/body weight than TCA. Although as noted above, the majority of the data from this
study in mice did not indicate that the magnitude of difference in exposure concentration was the
same as that of liver weight induction for TCA, in the particular experiment that examined both
DCA and TCA, the increase in percent liver/body weight ratios were similar to the magnitude of
difference in dose between the 2 g/L and 5 g/L exposure concentrations for both DCA and TCA
(i.e., 2- to2.5-fold increase in liver weight change corresponding to a 2.5-fold difference in
exposure concentration).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-228 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
Carter et al. (1995) examined 0.5 and 5.0 g/L exposures to DCA in B6C3F1 male mice
and reported that percent liver/body weights were increased consistently from 0.5 g/L DCA
treatment from 5 days to 30 days of treatment (i.e., a range of 1.05- to 1.16-fold of control). For
5.0 g/L DCA exposure the range of increase in percent liver/body weight was reported to be 1.37-
to 2.04-fold of control for the same time period. At the 15 days of exposures the percent
liver/body weight ratios were 1.67- and 1.12-fold of control for 5.0 and 0.5 g/L DCA and at
30 days were 1.99- and 1.16-fold, respectively. The difference in magnitude of dose and percent
liver/body weight increase is difficult to determine given that the 5 g/L dose of DCA reduced
body weight and significantly reduced water consumption by -50%. The differences in DCA-
induced percent liver/body weights were ~6-fold for the 15, 25, and 30-day data between the 0.5
and 5 g/L DCA exposures rather than the 10-fold difference in exposure concentration in the
drinking water.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-229 DRAFT—DO NOT CITE OR QUOTE
-------
Table E-14. Liver weight induction as percent liver/body weight fold-of-control in male B6C3F1 mice from
DCA or TCA drinking water studies
Concentration
(g/L)
Duration of exposure
Mean for
average of days
14-30
14 or 15 days
20 or 21 days
25 days
28 or 30 days
DCA
0.1

1.02-fold


1.02-fold
0.3
1.08-fold



1.08-fold
0.5
1.12-fold
1.24-fold, 1.05-fold
1.16-fold
1.16-fold
1.15-fold
1.0
1.31-fold



1.31-fold
2.0
1.62-fold
1.46-fold, 2.01-fold
2.04-fold
1.99-fold, 1.42-fold
1.83-fold
5.0
1.67-fold



1.67-fold
TCA
0.05



1.09-fold
1.09-fold
0.1

0.98-fold


0.98-fold
0.3
1.15-fold



1.15-fold
0.5

1.13-fold

1.16-fold
1.15-fold
1.0
1.23-fold, 1.08-fold



1.16-fold
2.0
1.38-fold, 1.16-fold, 1.26-fold
1.33-fold


1.30-fold
3.0



1.33-fold
1.33-fold
-------
°s
TO
s
&
TO
Co
s
o
TO
O
s
Co
3"
rS"
TO
S
Co
a,
Sr
<§>

TO
*
^S
«
s
TO
Co
0
S
1
a.
1.39-fold, 1.35-fold
vo
to
o
1.37-fold
-------
Table E-15. Liver weight induction as percent liver/body weight fold-of-control in male B6C3F1 or Swiss mice
from TCE gavage studies
Concentration
(mg/kg/d)
10 days
28 days
42 days
Mean for average of
days 10-42
B6C3F1
100
1.00-fold


1.00-fold
250
1.00-fold


1.00-fold
500
1.20-fold, 1.06-fold


1.13-fold
600

1.36-fold

1.36-fold
1,000
1.50-fold, 1.17-fold, 1.50-fold


1.39-fold
1,200

1.64-fold

1.64-fold
1,500
1.47-fold


1.47-fold
2,400

1.81-fold

1.81-fold
Swiss
100


1.12-fold
1.12-fold
200


1.15-fold
1.15-fold
400


1.25-fold
1.25-fold
500
1.43-fold
1.32-fold

1.38-fold
800


1.36-fold
1.36-fold
1,000
1.56-fold
1.41-fold

1.49-fold
1,500
1.75-fold


1.75-fold
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
yq s
"tS
§*
3'
Co
a
s^
Sf
<§>
>S
TO
*
^s
«
s
TO
>;
0
s
1
a.
VO
o
to
1,600


1.63-fold
1.63-fold
2,000

1.38-fold

1.38-fold
2,400

1.69-fold

1.69-fold
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Table E-16. B6C3F1 and Swiss (data sets combined)
Concentration (mg/kg/d)
Mean for average of days 10-42
100
1.06-fold
200
1.15-fold
250
1.00-fold
400
1.25-fold
500
1.26-fold
600
1.36-fold
800
1.36-fold
1,000
1.49-fold
1,200
1.64-fold
1,500
1.61-fold
1,600
1.63-fold
2,000
1.38-fold
2,400
1.75-fold
Parrish et al. (1996) reported that for male B6C3F1 mice exposed to TCA or DCA (0,
0.01, 0.5, and 2.0 g/L) for 3 or 10 weeks, the 4- to 5-fold magnitude of difference in doses
resulted in increases in percent liver/body weight for the 21-day and 71-day exposures that were
greater for DCA than TCA. The percent liver/body weight ratio were 0.98-, 1.13-, and 1.33-fold
of control levels at 0.1, 0.5, and 2.0 g/L TCA and for DCA were 1.02-, 1.24-, and 1.46-fold of
control levels, respectively, after 21 days of exposure. Both TCA and DCA exposures at 0.1 g/L
resulted in difference in percent liver/body weight change of 2% or less. For TCA, although there
was a 4-fold increase in magnitude between the 0.5 and 2.0 g/L TCA exposure concentrations, the
magnitude of increase for percent liver/body weight increase was 2.5-fold between them at both
21 and 71 days of exposure. For DCA, the 4-fold difference in dose between the 0.5 and 2.0 g/L
DCA exposure concentrations were reported to result in a ~2-fold increase in percent liver/body
weight increase at 21 days and ~4.5-fold increase at 71 days.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-234 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DeAngelo et al. (2008) studied 3 exposure concentrations of TCA in male B6C3F1 mice,
which were an order of magnitude apart, for 4 weeks of exposure. The percent liver/body weight
ratios were 1.09-, 1.16-, and 1.35-fold of control levels, for 0.05, 0.5, and 5.0 g/L TCA exposures,
respectively. The 10-fold differences in exposure concentration of TCA resulted in ~2-fold
differences in percent liver/body weight increases. No dose-response inferences can be drawn
from the 4-week study of DCA and TCA in B6C3F1 male mice by Kato-Weinstein et al. (2001)
but 2 g/L DCA and 3 g/L TCA in drinking water were reported to induce percent liver/body
weights of 1.42- and 1.33-fold of control, respectively (n = 5).
The majority of short-term studies of DCA and TCA in mice have been conducted in the
B6C3F1 strain and in males. Studies conducted from 14 to 30 days show a consistent increase in
percent liver/body weight induction by TCA or DCA. Analyses of this information regarding
inferences for attribution and comparisons of dose-response have been published by Evans et al.
(2009). Chiu et al., (2004). Chiu (In Press), and is discussed in Section 4 of the TCE assessment
document and in Appendix A. A broader discussion of primarily issues and data related to Evans
(2009) is contained below.
An examination of all of the data from Parrish et al. (1996), Sanchez and Bull (1990),
Carter et al. (1995), Kato-Weinstein et al. (2001), and DeAngelo et al. (1989; DeAngelo et al.,
2008) from 14 to 30 days of exposure in male B6C3F1 mice can give an approximation of the
dose-response differences between DCA and TCA for liver weight induction as shown in Table
E-14 and Figure E-l, below. Although the data for B6C3F1 mice from Sanchez and Bull (1990)
is reported as the fold of liver weight rather that percent liver/body weight increase, it is included
in the comparison as both reflect increase in liver weight. Similar data can be assessed for TCE
for comparative purposes. Short duration studies (10-42 days) were selected because (1) in
chronic studies, liver weight increases are confounded by tumor burden, (2) multiple studies are
available, and (3) in this duration range, Kjellstrand et al. (1981a) reported that TCE-induced
increases in liver weight plateau, and (4) TCA studies do not show significant duration-dependent
differences in this duration range. These comparisons are presented in Table E-14.
DeAngelo et al. (1989) and Carter et al. (1995) used up to 5 g/L DCA and TCA in their
experiments with Carter et al. (1995) noting a dramatic decrease in water consumption in the
5 g/L DCA treatment groups (46-64% reduction) which can affect body weight as well as dose
received. DeAngelo et al. (1989) did not report drinking water consumption. The drinking water
consumption was reported by DeAngelo et al. (Chiu and White, 2004) to be reduced by 11, 17,
and 30% in the 0.05, 0.5, and 5 g/L TCA treated groups compared to 2 g/L NaCl control animals
over 60 weeks. DeAngelo et al. (1999) reported mean drinking water consumption to be reduced
by 26%) in mice exposed to 3.5 g/L DCA over 100 weeks. Carter et al. (1995) reported that DCA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
at 5 g/L to decrease drinking water consumption by 64 and 46% but 0.5 g/L DCA to not affect
drinking water consumption. Thus, it appears that the 5 g/L concentrations of either DCA or
TCA can significantly affect drinking water consumption as well as inducing reductions in body
weight. Accordingly, an estimation of the shape of the dose-response curve for comparative
purposes between DCA or TCA drinking water studies is best examined at concentrations at 2 g/L
or less, especially for DCA.
Male B6C3F1 mice liver weight for TCA and DCA in drinking water - days 14-30
2.0
DCA
TCA
1.8
T3 1.4
1.2
1.0
0.0
0.5
2.0
2.5
Concentration of DCA or TCA (g/l)
Figure E-l. Comparison of average fold-changes in relative liver weight to
control and exposure concentrations of 2 g/L or less in drinking water for
TCA and DCA in male B6C3F1 mice for 14-30 days (Carter et al., 1995;
DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-Weinstein et al., 2001;
Parrish et al., 1996; Sanchez and Bull, 1990)\. (Reproduced from Section
4.5.)
The dose-response curves for similar concentrations of DCA and TCA are presented in
Figure E-l for durations of exposure from 14-28 days in the male B6C3F1 mouse, which was the
most common sex and strain used. For this comparative analysis an average is provided between
two values for a given concentration and duration of exposure for comparison with other doses
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
and time points. As noted in the discussion of individual experiments, there appears to be a linear
correlation between dose in drinking water and liver weight induction up to 2 g/L of DCA.
However, the shape of the dose-response curve for TCA appears to be quite different (i.e., lower
concentrations of TCA inducing larger increase that does DCA but then the response reaching an
apparent plateau for TCA at higher doses while that of DCA continues to increase). As shown by
DeAngelo et al. (2008), 10-fold differences in the magnitude of exposure concentration to TCA
corresponded to ~2-fold differences in liver weight induction increases. In addition, TCA studies
did not show significant duration-dependent difference in liver weight induction in this duration
range as shown in Table E-14.
Of interest is the issue of how the dose-response curves for TCA and DCA compare to
that of TCE in a similar model and dose range. Since TCA and DCA have strikingly different
dose-response curves, which one if either best fits that of TCE and thus, can give insight as to
which is causative agent for TCE's effects in the liver? In the case of the TCE database in the
mouse two strains have been predominantly studied, Swiss and B6C3F1, and both have been
reported to get liver tumors in response to chronic TCE exposure.
Rather than administered in drinking water, oral TCE studies have been conducted via oral
gavage and generally in corn oil for 5 days of exposure per week. The study by Goel et al. (1992)
was conducted in ground-nut oil. Vehicle effects, the difference between daily and weekly
exposures, the dependence of TCE effects in the liver on its metabolism to a variety of agents
capable inducing effects in the liver, differences in response between strains, and the inherent
increased variability in use of the male mouse model all add to increased difficulty in establishing
the dose-response relationship for TCE across studies and for comparisons to the DCA and TCA
database. Despite difference in exposure route, etc., a consistent pattern of dose-response
emerges from combining the available TCE data. The effects of oral exposure to TCE from
10-42 days on liver weight induction is shown in Figure E-2 using the data of Elcombe et al.
(1985), Dees and Travis (1993), Goel et al. (1992), Merrick et al. (1989), Goldsworthy and Popp
(1987), and Buben and O'Flaherty (1985). More detailed discussion of the 4- to 6-week studies is
presented in Section E.2.4.3, below (e.g., for (Buben and O'Flaherty, 1985; Goel et al., 1992;
Merrick et al., 1989)). For this comparative analysis an average is provided between two values
per concentration and duration of exposure for comparison with other doses and time points. As
shown by the 10-day data in B6C3 F1 mice, there are significant differences in response between
studies of male B6C3F1 mice at the same dose of TCE. This variability is similar to findings
from inhalation studies of TCE in male mice (Kjellstrand et al., 1983b).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 7 DRAFT—DO NOT CITE OR QUOTE
-------
Male mice liver weight for TCE oral gavage - days 10-42
2.0
• B6C3F1
	 Regression
O)
0
500
1000
1500
2000
2500
3000
Concentration of TCE (mg/kg/day)
Male mice liver weight for TCE oral gavage - days 10-42
2.0
• B6C3F1 and Swiss
	 Plot 2 Regr
0
500
1000
1500
2000
2500
3000
Concentration of TCE (mg/kg/day)
1	Figure E-2. Comparisons of fold-changes in average relative liver weight and
2	gavage dose of (top panel) male B6C3F1 mice for 10-28 days of exposure
3	(Dees and Travis, 1993; Elcombe et al., 1985; Goldsworthy and Popp, 1987;
4	Merrick et al., 1989) and (bottom panel) in male B6C3F1 and Swiss mice.
5	(Reproduced from Section 4.5.)
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
As shown in Figure E-2, oral TCE administration in male B6C3F1 and Swiss mice
appeared to induce a dose-related increase in percent liver/body weight that was generally
proportional to the increase in magnitude of dose, though as expected, with more variability than
observed for a similar exercise for DCA or TCA in drinking water. Common exposure
concentrations between B6C3F1 and Swiss mice were 100, 500, 1,000, 1,500 and 2,400 mg/kg/d
TCE which corresponded to a 5-, 2-, 1.5-, and 1.6-fold difference in the magnitude of dose. For
the data from studies in B6C3 F1 mice, there was no increase reported at 100 mg/kg/d TCE but
between 500 and 1,000, 1,000 and 1,500, and 1,500 and 2,400 mg/kg/d TCE the magnitude of
difference in doses matched that of the magnitude of increase in percent liver/body weight (i.e., a
2.6-, 1.4-, and 1.7-fold increase in liver weight was matched by a 2-, 1.5-, and 1.6-fold increase in
TCE exposure concentration at these exposure intervals).
However, only 10-day was available for doses between 100 and 500 mg/kg in B6C3F1
mice and at the lower doses, a 10-day interval may have been too short for the increase in liver
weight to have been fully expressed. The database for the Swiss mice, which has more data from
28 and 42 days of exposure, support this conclusion. At 28-42 days of exposure there was a
much greater increase in liver weight from TCE exposure in Swiss mice than the 10-day data in
B6C3F1 mice.
In Figure E-2, the 10-day data are included for comparative purpose for the B6C3F1 data
set and the Swiss and B6C3F1 data sets combined. Both the combined TCE data and that for only
B6C3F1 mice shows a correlation with the magnitude of dose and magnitude of percent
liver/body weight increase. The slope of the dose-response curves are both closer to that of DCA
than TCA. The correlation coefficients for the linear regressions presented for the B6C3F1 data
2	2
are R = 0.861 and for the combined data sets is R = 0.712. Comparisons of the slopes of the
dose-response curves indicate that TCA is not responsible for TCE-induced liver effects. In this
regression all data points were treated equally although some came from several sets of data and
others did not. Of note is that the 2,000 mg/kg TCE data point in the combined data set, which is
much lower in liver weight response than the other data, is from one experiment (Goel et al.,
1992), from 6 mice, at one time point (28 days), and one strain (Swiss). Deletion of these data
point from the rest of the 23 used in the study results in a better fit to the data of the regression
analysis.
A more direct comparison would be on the basis of dose rather than drinking water
concentration. The estimations of internal dose of DCA or TCA from drinking water studies have
been reported to vary with DeAngelo et al. reporting DCA drinking water concentrations of 1.0,
2.0, and 5.0 g/L to result in 90, 166, and 346 mg/kg/d, respectively. For TCA, 0.05, 0.5, 1.0, 2.0,
and 5 g/L drinking water exposures were reported to result in 5.8 (range 3.6-8.0), 50 (range of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-23 9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
32.5 to 68), 131, 261, and 469 (range 364 to 602) mg/kg/d doses. The estimations of internal dose
of DCA or TCA from drinking water studies, while varying considerably (DeAngelo et al., 1989;
2008), nonetheless suggest that the doses of TCE used in the gavage experiments were much
higher than those of DCA or TCA. However, only a fraction of ingested TCE is metabolized to
DCA or TCA, as, in addition to oxidative metabolism, TCE is also cleared by glutathione (GSH)
conjugation and by exhalation.
While DCA dosimetry is highly uncertain (see Sections E.3.3 and E.3.5), the mouse
physiologically based pharmacokinetic (PBPK) model, described in Section E.3.5 was calibrated
using extensive in vivo data on TCA blood, plasma, liver, and urinary excretion data from
inhalation and gavage TCE exposures, and makes robust predictions of the rate of TCA
production. If TCA were predominantly responsible for TCE-induced liver weight increases, then
replacing administered TCE dose (e.g., mg TCE/kg/day) by the rate of TCA produced from TCE
(mg TCA/kg/day) should lead to dose-response curves for increased liver weight consistent with
those from directly administered TCA.
Figure E-3 shows this comparison using the PBPK model-based estimates of TCA
production for 4 TCE studies from 28-42 days in the male NMRI, Swiss, and B6C3F1 mice
(Buben and O'Flaherty, 1985; Goel et al., 1992; Kjellstrand et al., 1983a; Merrick et al., 1989) and
4 oral TCA studies in B6C3F1 male mice at 2 g/L or lower drinking water exposure (DeAngelo et
al., 1989; DeAngelo et al., 2008; Kato-Weinstein et al., 2001; Parrish et al., 1996) from 14-28
days of exposure. The selection of the 28-42 day data for TCE was intended to address the
decreased opportunity for full expression of response at 10 days. PBPK modeling predictions of
daily internal doses of TCA in terms of mg/kg/d via produced via TCE metabolism would be are
indeed lower than the TCE concentrations in terms of mg/kg/d given orally by gavage. The
predicted internal dose of TCA from TCE exposure studies are of a comparable range to those
predicted from TCA drinking water studies at exposure concentrations in which palability has not
been an issue for estimation of internal dose. Thus, although the TCE data are for higher
exposure concentrations, they are predicted to produce comparable levels of TCA internal dose
estimated from direct TCA administration in drinking water.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-240 DRAFT—DO NOT CITE OR QUOTE
-------
2.5
~ TCE Studies [28-42 d]
o TCA Studies [14-28 d]
— - Linear (TCA Studies [14-28 d])
	Linear (TCE Studies [28-42 d])
~ ~
.0-
u_
0	100	200	300	400	500
mg TCA/kg-d
(produced [TCE studies] or administered [TCA studies])
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-241 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
Figure E-3. Comparison of fold-changes in relative liver weight for data sets
in male B6C3F1, Swiss, and NRMI mice between TCE studies (Buben and
O'Flaherty, 1985; Goel et al., 1992; Kjellstrand et al., 1983a; Merrick et al.,
1989) [duration 28-42 days] and studies of direct oral TCA administration to
B6C3 F1 mice (DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-Weinstein
et al., 2001; Parrish et al., 1996) [duration 14-28 days]. Abscissa for TCE
studies consists of the median estimates of the internal dose of TCA predicted
from metabolism of TCE using the PBPK model described in Section 3.5 of
the TCE risk assessment. Lines show linear regression with intercept fixed
at 1. All data were reported fold-change in mean liver weight/body weight
ratios, except for Kjellstrand et al. (1983a), with were the fold-change in the
ratio of mean liver weight to mean body weight. In addition, in Kjellstrand
et al. (1983a), some systemic toxicity as evidence by decreased total body
weight was reported in the highest dose group. (Reproduced from Section
4.5.)
Figure E-3 clearly shows that for a given amount of TCA produced from TCE, but going
through intermediate metabolic pathways, the liver weight increases are substantially greater than,
and highly inconsistent with, that expected based on direct TCA administration. In particular, the
response from direct TCA administration appears to "saturate" with increasing TCA dose at a
level of about 1.4-fold, while the response from TCE administration continues to increase with
dose to 1.75-fold at the highest dose administered orally in Buben and O'Flaherty (1985) and over
2-fold in the inhalation study of Kjellstrand et al. (1983a). For this analysis is unlikely that strain
differences can account for this inconsistency in the dose-response curves.
TCE-induced increases in liver weight appear to be generally similar between B6C3F1
and Swiss male mice (see Table E-14) via oral exposure and between NMRI male and female
mice after inhalation, although the NMRI strain appeared to be more prone to TCE-induced
toxicity in male mice and for females to have a smaller TCE-induced liver weight increase than
other strains (Kjellstrand et al., 1983a). As noted previously, the difference in response between
strains and between studies in the same strain for TCE liver weight increases can be highly
variable. Little data exist to examine this issue for TCA studies although DeAngelo et al. (1989)
report a range of 1.16- to 1.63-fold of control percent liver/body weight increase after 14 days
exposure at 2 g/L TCA in the Swiss-Webster, C3H, C57BL/6, and B6C3F1 strains, with
differences also noted between 2 studies of the B6C3F1 mouse.
Furthermore, while as noted previously, oral studies appear to report a linear relationship
between TCE exposure concentration and liver weight induction, the inclusion of inhalation
studies on the basis of internal dose led to a highly consistent dose-response curve for among
TCE study. Therefore, it is unlikely that differing routes of exposure can explain the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-242 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
inconsistencies in dose-response. The PBPK model predicted that matching average TCA
production by TCE with the equivalent average dose from drinking water-administered TCA also
led to an equivalent area-under-the-curve (AUC) of TCA in the liver.
Moreover, Dees and Travis (1993) administered 100 to 1,000 mg/kg/d TCA by gavage to
male and female B6C3F1 mice for 11 days, and did not observe increases in liver/body weight
ratios more than 1.28-fold, no higher than those observed with drinking water exposures. Finally,
the dose-response consistency between TCE inhalation and gavage studies argues against route of
exposure significantly impacting liver weight increases. Thus, no level of TCA administration
appears able account for the continuing increase in liver weights observed with TCE,
quantitatively inconsistent with TCA being the predominant metabolite responsible for TCE-
induced liver weight changes. Involvement of other metabolites, besides TCA, is implicated as
the causes of TCE-induced liver effects.
Additional analyses do, however, support a role for oxidative metabolism in TCE-induced
liver weight increases, and that the parent compound TCE is not the likely active moeity (as
suggested previously by Buben and O'Flaherty, 1985). In particular, the same studies are shown
in Figure E-4 using PBPK-model based predictions of the AUC of TCE in blood and total
oxidative metabolism, which produces chloral, trichloroethanol, DC A, and other metabolites in
addition to TCA. The dose-response relationship between TCE blood levels and liver weight
increase, while still having a significant trend, shows substantial scatter and a low R of 0.43. On
the other hand, using total oxidative metabolism as the dose metric leads to substantially more
consistency dose-response across studies, and a much tighter linear trend with an R of 0.90 (see
Figure E-4). A similar consistency is observed using liver-only oxidative metabolism as the dose
metric, with R of 0.86 (not shown). Thus, while the slope is similar between liver weight
increase and TCE concentration in the blood and liver weight increase and rate of total oxidative
metabolism, the data are a much better fit for total oxidative metabolism.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-243 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
R =0.426
0
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
calculation shows that the type II error, which should be >50% and thus, greater than the chances
of "flipping a coin," was only a 6 and 7% and therefore, the designed experiment could accept a
false null hypothesis.
Although the qualitative similarity to the linear dose-response relationship between DCA
and liver weight increases is suggestive of DCA being the predominant metabolite responsible for
TCE liver weight increases, due to the highly uncertain dosimetry of DCA derived from TCE, this
hypothesis cannot be tested on the basis of internal dose. Similarly, another TCE metabolite, CH,
has also been reported to induce liver tumors in mice, however, there are no adequate comparative
data to assess the nature of liver weight increases induced by this TCE metabolite (see Section
E.2.5, below). Whether its formation in the liver after TCE exposure correlates with TCE-
induced liver weight changes cannot be determined. Of note is the high variability in total
oxidative metabolism reported in mice and humans of Section 3.3, which suggests that the
correlation of total TCE oxidative metabolism with TCE-induced liver effects should lead not
only to a high degree of variability in response in rodent bioassays which is the case (see Section
E.2.4.4, below) but also make detection of liver effects more difficult in human epidemiological
studies (see Section 4.3.2).
The bioavailability of TCA has been assumed to be 100% in the analyses in Figure E-3.
Further analyses are presented in Appendix A and in Chiu (In Press) regarding the assertions by
Sweeney, et al. (Sweeney et al.. 2009) that previously unpublished kinetic data for mice exposed
to TCA in drinking water indicates much lower absorption. The conclusions of Sweeney et al.
(2009) were based on the TCE PBPK model of Hack et al. (2006) and not that of Evans et al.
(2009) and Chiu et al. (2009). The analyses by Chiu (In Press) show that while there is some
decreased absorption of TCA at higher doses, it was not as low as that estimated by Sweeney et
al. (2009) and as discussed in Appendix A, it may be more accurate to characterize the fractional
absorption as an empirical parameter reflecting unaccounted-for biological processes as well as
experimental variation. The Chiu (In Press)re-analyses the data on TCE- and TCA-induced
hepatomegaly, using the central estimates of the fractional absorption of TCA, showed that while
reduced fractional absorption inferred from drinking water data reported by Sweeney et al. (2009)
accounts for part of the difference in dose-responses between TCE- and TCA-induced
hepatomegaly reported by Evans et al. (2009), it does not appear to be able to account for the
entire difference. The inability of TCA to account for TCE-induced hepatomegaly was
confirmed statistically by ANOVA and even with an assumption of reduced TCA bioavailability,
the available data are inconsistent with the toxicological hypothesis that TCA can fully account
for TCE-induced hepatomegaly.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-245 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
What mechanisms or events are leading to liver weight increases for DCA, TCA and TCE
can be examined by correlations between changes in glycogen content, hepatocyte volume, and
evidence of polyploidization noted in short-term assays. Data have been reported regarding the
nature of changes the TCE and its metabolites induce in the liver and are responsible for the
reported increases in liver weight. Increased liver weight may result from increased size or
hypertrophy of hepatocytes through changes in glycogen deposition, but also through increased
polyploidization. Increased cell number may also contribute to increased liver weight. As noted
above in Section E.2.4.1, hepatocellular hypertrophy appeared to be related to TCE-induced liver
weight changes after short-term exposures. However, neither glycogen deposition, DNA
synthesis, or increases in mitosis appear to be correlated with liver weight increases. In particular
DNA synthesis increases were similar from 250-1,000 mg/kg and peroxisomal volume was
similar between 500 and 1,500 mg/kg TCE exposures after 10 days. Autoradiographs identified
hepatocytes undergoing DNA synthesis in "mature" hepatocytes that were in areas where
polyploidization typically takes place in the liver.
By 14 days of exposure, Sanchez and Bull (1990) reported that both dose-related TCA-
and DCA-induced increases in liver weight were generally consistent with changing cell size
increases, but were not correlated with patterns of change in hepatic DNA content, incorporation
of tritiated thymidine in DNA extracts from whole liver, or incorporation of tritiated thymidine in
hepatocytes. There are conflicting reports of DNA synthesis induction in individual hepatocytes
for up to 14 days of DCA or TCA exposure and a lack of correlation with patterns observed for
this endpoint and those of whole liver thymidine incorporation. The inconsistency of whole liver
DNA tritiated thymidine incorporation with that reported for hepatocytes was noted by the
Sanchez and Bull (1990) to be unexplained. Carter et al. (1995) also report a lack of correlation
between hepatic DNA tritiated thymidine incorporation and labeling in individual hepatocytes in
male mice. Carter et al. (1995) reported no increase in labeling of hepatocytes in comparison to
controls for any DCA treatment group from 5 to 30 days of DCA exposure. Rather than increase
hepatocyte labeling, DCA induced a decrease with no change reported from days 5 though 15 but
significantly decreased levels between days 20 and 30 for 0.5 g/L that were similar to those
observed for the 5 g/L exposures.
The most comparable time period between TCE, TCA and DCA results for whole liver
thymidine incorporation is the 10- and 14-day durations of exposure when peak tritiated
thymidine incorporation into individual hepatocytes and whole liver for TCA and DCA have been
reported to have already passed (Carter et al., 1995; Pereira, 1996; Sanchez and Bull, 1990; Styles
et al., 1991). Whole liver DNA synthesis was elevated over control levels by ~2-fold after from
250 to 1,000 mg/kg TCE exposure after 10 days of exposure but did not correlate with mitosis
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-246 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
(Dees and Travis, 1993; Elcombe et al., 1985). After 3 weeks of exposure to TCE, Laughter et al.
(2004) reported in individual hepatocytes that 1 and 4.5% of hepatocytes had undergone DNA
synthesis in the last week of treatment for the 500 and 1,000 mg/kg TCE levels, respectively.
More importantly, these data show that hepatocyte proliferation in TCE-exposed mice at 10 days
of exposure or for DCA- or TCA-exposed mice for up to 14 days of exposure is confined to a
very small population of cells in the liver.
In regard to cell size, although increased glycogen deposition with DCA exposure was
noted by Sanchez and Bull (1990), lack of quantitative analyses of that accumulation in this study
precludes comparison with DCA-induced liver weight gain. Although not presenting a
quantitative analysis, Sanchez and Bull (1990) reported DCA-treated B6C3F1 mice to have large
amounts of PAS staining material and Swiss-Webster mice to have similar increase despite
reporting differences of DCA-induced liver weight gain between the two strains. The lack of
concordance of the DCA-induced magnitude of increase in liver weight with that of glycogen
deposition is consistent with the findings for longer-term exposures to DCA reported by
Kato-Weinstein et al. (2001) and Pereira et al. (2004b) in mice (see Section E.2.4.4, below).
Carter et al. (1995) reported that in control mice there was a large variation in apparent glycogen
content and also did not perform a quantitative analysis of glycogen deposition. The variability
of this parameter in untreated animals and the extraction of glycogen during normal tissue
processing for light microscopy makes quantitative analyses for dose-response difficult unless
specific methodologies are employed to quantitatively assess liver glycogen levels as was done
by Kato-Weinstein et al. (2001) and Pereira et al. (2004b).
Although suggested by their data, polyploidization was not examined for DCA or TCA
exposure in the study of Sanchez and Bull (1990). Carter et al. (1995) reported that hepatocytes
from both 0.5 and 5 g/L DCA treatment groups were reported to have enlarged, presumably
polyploidy nuclei with some hepatocyte nuclei labeled in the mid-zonal area. There were
statistically significant changes in cellularity, nuclear size, and multinucleated cells during
30 days exposure to DCA. The percentage of mononucleated cells hepatocytes was reported to
be similar between control and DCA treatment groups at 5- and 10-day exposure.
However, at 15 days and beyond, DCA treatments were reported to induce increases in
mononucleated hepatocytes. At later time periods there were also reports of DCA-induced
increases nuclear area, consistent with increased polyploidization without mitosis. The consistent
reporting of an increasing number of mononucleated cells between 15 and 30 days could be
associated with clearance of mature hepatocytes as suggested by the report of DCA-induced loss
of cell nuclei. The reported decrease in the numbers of binucleate cells in favor of mononucleate
cells is not typical of any stage of normal liver growth (Brodsky and Uryvaeva, 1977). The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-247 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
linear dose-response in DCA-induced liver weight increase was not consistent with the increased
numbers of mononucleate cells and increase nuclear area reported from Day 20 onward by Carter
et al. (1995). Specifically, the large differences in liver weight induction between the 0.5 g/L
treatment group and the 5 g/L treatment groups at all times studied also did not correlate with
changes in nuclear size and percent of mononucleate cells. Thus, DCA-induced increases in liver
weight were not a function of cellular proliferation, but probably included hypertrophy associated
with polyploidization, increased glycogen deposition and other factors.
In regard to necrosis, Elcombe et al. (1985) reported only small incidence of focal
necrosis in 1,500 mg/kg TCE-exposed mice and no necrosis at exposures up to 1,000 mg/kg for
10 days as did Dees and Travis (1993). Sanchez and Bull (1990) report DCA-induced localized
areas of coagulative necrosis both for B6C3F1 and Swiss-Webster mice at higher exposure
levels (1 or 2 g/L) by 14 days but not at the 0.3 g/L level or earlier time points. For TCA
treatment, necrosis was reported to not be associated with TCA treatment for up to 2 g/L and up
to 14 days of exposure. Carter et al. (1995) reported that mice given 0.5 g/L DCA for 15, 20,
and 25 days had midzonal focal cells with less detectable or no cell membranes, loss of the
coarse granularity of the cytoplasm, with some cells having apparent karyolysis, but for liver
architecture to be normal.
As for apoptosis, Both Elcombe et al. (1985) and Dees and Travis (1993) reported no
changes in apoptosis other than increased apoptosis only at a treatment level of 1,000 mg/kg
TCE. Rather than increases in apoptosis, peroxisome proliferators have been suggested to
inhibit apoptosis as part of their carcinogenic MOA (see Section E.3.4.1). However, the age and
species studied appear to greatly affect background rates of apoptosis. Snyder et al. (1995)
report that control mice were reported to exhibit apoptotic frequencies ranging from -0.04 to
0.085%, that over the 30-day period of their study the frequency rate of apoptosis declined, and
suggest that this pattern is consistent with reports of the livers of young animals undergoing
rapid changes in cell death and proliferation. They reported rat liver to have a greater the
estimated frequency of spontaneous apoptosis (~0.1%) and therefore, greater than that of the
mouse.
Carter et al. (!!! INVALID CITATION !!!) reported that after 25 days of 0.5 g/L DCA
treatment apoptotic bodies were reported as well as fewer nuclei in the pericentral zone and
larger nuclei in central and midzonal areas. This would indicate an increase in the apoptosis
associated potential increases in polyploidization and cell maturation. However, Snyder et al.
(1995) report that mice treated with 0.5 g/L DCA over a 30-day period had a similar trend as
control mice of decreasing apoptosis with age. The percentage of apoptotic hepatocytes
decreased in DCA-treated mice at the earliest time point studied and remained statistically
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-248 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
significantly decreased from controls from 5 to 30 days of exposure. Although the rate of
apoptosis was very low in controls, treatment with 0.5 g/L DCA reduced it further (-30-40%
reduction) during the 30-day study period. The results of this study not only provide a baseline
of apoptosis in the mouse liver, which is very low, but also to show the importance of taking
into account the effects of age on such determinations. The significance of the DCA-induced
reduction in apoptosis reported in this study, from a level that is already inherently low in the
mouse, to account for the MOA for induction of DCA-induced liver cancer is difficult to
discern.
Finally, short-term inhalation studies by Ramdhan et al. (2010) indicate that in wild type,
PPARa-null, and humanized null mice, relatively high exposures to TCE induced increased liver
size after 7 days of inhalation exposure. At the same highest concentration of TCE, although
urinary TCA concentrations were lower in PPARa-null mice than wild type mice, the sum of
urinary trichloroethanol and TCA concentrations were the same, increases in % liver/body
weight were the same, and liver triglyceride content was much greater in the PPARa-null mice
than wild type mice after TCE exposure. Hepatic steatosis was also greater as a baseline
condition along with hepatic triglyceride content in the PPARa-null mice than wild type mice.
These parameters were more elevated in humanized mice as a background dysregulation and
even more elevated after treatment with TCE. Therefore, the nature of hepatomegally induced
by TCE is complex and dependent on baseline lipid dysregulation states.
E.2.5.3. Summary Trichloroethylene (TCE) Subchronic and Chronic Studies
The results of longer-term (Channel et al., 1998; Parrish et al., 1996; Toraason et al.,
1999) studies of "oxidative stress" for TCE and its metabolites are discussed in
Section E.3.4.2.3. Of note are the findings that the extent of increased enzyme activities
associated with peroxisome proliferation do not appear to correlate with measures of oxidative
stress after longer term exposures (Parrish et al., 1996) and single strand breaks (Chang et al.,
1992).
Similar to the reports of Melnick et al. (1987) in rats, Merrick et al. (1989) report that
vehicle (aqueous or gavage) affects TCE-induced toxicity in mice. Vehicle type made a large
difference in mortality, extent of liver necrosis, and liver weight gain in male and female
B6C3F1 mice after 4 weeks of exposure. The lowest dose used in this experiment was
600 mg/kg/d in males and 450 mg/kg/d in females. Administration of TCE via gavage using
Emulphor resulted in mortality of all of the male mice and most of the female mice at a dose in
corn oil that resulted in few deaths. However, use of Emulphor vehicle induced little if any
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-249 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
focal necrosis in males at concentrations of TCE in corn oil gavage that caused significant focal
necrosis, indicating vehicle effects.
As discussed above in Section E.2.4.2, the extent of TCE-induced liver weight increases
was consistent between 4 and 6 weeks of exposure and between 10-day and 4-week exposure at
higher dose levels. In general, the reported elevations of enzymatic markers of liver toxicity and
results for focal hepatocellular necrosis were not consistent and did not reflect TCE dose-
responses observed for induction of liver weight increases (Merrick et al., 1989). Female mice
given corn oil and male and female mice given TCE in Emulphor were reported to have "no to
negligible necrosis" although they had increased liver weight from TCE exposure.
Using a different type of oil vehicle, Goel et al. (1992) exposed male Swiss mice to TCE
in groundnut oil at concentrations ranging from 500 to 2,000 mg/kg for 4 weeks and reported no
changes in body weight up to 2,000 mg/kg. There was a 15% decrease at the highest dose and
increased TCE-induced percent liver/body weight ratio. At a dose of 1,000 and 2,000 mg/kg,
liver swelling, vacuolization, and widespread degenerative necrosis of hepatocytes was reported
along with marked proliferation of "endothelial cells" but no quantitation regarding the extent or
location of hepatocellular necrosis was reported, nor whether there was a dose-response
relationship in these events. They reported a TCE-related dose-response in catalase, liver
protein but decreased induction at the 2,000 mg/kg level where body weight had decreased.
Three studies were published by Kjellstrand et al. that examined effects of TCE
inhalation primarily in mice using whole body inhalation chambers (Kjellstrand et al., 1981a;
Kjellstrand et al., 1983a; Kjellstrand et al., 1983b). Liver weight changes were used as the
indication of TCE-induced effects. The quantitative results from these experiments had many
limitations due to their experimental design including failure to determine body weight changes
for individual animals and inability to determine the exact magnitude of TCE due to concurrent
oral TCE ingestion from food and grooming behavior. An advantage of this route of exposure
was that there were not confounding vehicle effects. The results from Kjellstrand et al. (1981a)
were particularly limited by experimental design errors showed similar increases in liver weight
gain in gerbils and rats exposed at 150 ppm TCE. For rats, Kjellstrand et al. (1981a) reported
increases in liver/body weight ratios of 1.26- and 1.21-fold of control in male and female rat 30
days of continuous TCE inhalation exposure.
The unpublished report of Woolhiser et al. (2006) reports 1.05-, 1.07-, and 1.13-fold of
control percent liver/body weight changes in 100-, 300- and 1,000-ppm-exposure groups that are
exposed for 6 hours/day, 5 days/week for 4 weeks in groups of 8 female SD rats. At the two
highest exposure levels, body weight was reduced by TCE exposure. The 150 ppm continuous
exposure concentrations of Kjellstrand were analogous to 750-ppm-exposures using the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-250 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
paradigm of Woolhiser et al. (2006) in terms of total daily dose. Therefore, the very limited
inhalation database for rats does indicate TCE-related increases in liver weight.
The study of Kjellstrand et al. (1983b) employed a more successful experimental design
that recorded liver weight changes in carefully matched control and treatment groups to
determine TCE-treatment related effects on liver weight in 7 strains of mice after 30 days of
continuous inhalation exposure at 150 ppm TCE. Individual animal body weight changes were
not recorded so that such an approach cannot take into account the effects of body weight
changes and determine a relative percent liver/body weight ratio. The data presented in this
report were for absolute liver weight changes between treated and nontreated groups with
carefully matched average body weights at the initiation of exposure. A strength of the
experimental design is its presentation of results between duplicate experiments and thus, to
show the differences in results between similar exposed groups that were conducted at different
times. This information gives a measure of variability in response with time. Mouse strain
groups, that did not experience TCE-induced decreased body weight gain in comparison to
untreated groups (i.e., DBA and wild-type mice), represented the most accurate determination of
TCE-induced liver weight changes given that systemic toxicity that affects body weight can also
affect liver weight.
The C57BL, B6CBA, and NZB groups all had at least one group out of two of male mice
with changes in final body weight due to TCE exposure. Only one group of NMRI mice were
reported in this study and that group had TCE-induced decreases in final body weight. The A/sn
group not only had both male groups with decreased final body weight after TCE exposure
(along with differences between exposed and control groups at the initiation of exposure) but
also a decrease in body weight in one of the female groups and thus, appears to be the strain
with the greatest susceptibility to TCE-induced systemic toxicity. In strains of male mice in
which there were no TCE-induced affects on final body weight (wild-type and DBA), the
influence of gender on liver weight induction and variability of the response could be more
readily assessed. In wild-type mice there was a 1.76- and 1.80-fold of control liver weight in
groups 1 and 2 for female mice, and for males a 1.84- and 1.62-fold of control liver weight for
groups 1 and 2, respectively. For DBA mice there was a 1.87- and 1.88-fold of control liver
weight in groups 1 and 2 for female mice, and for males a 1.45- and 2.00-fold of control liver
weight for groups 1 and 2, respectively. Of note, as described previously, the size of the liver is
under strict control in relation to body size. An essential doubling of the size of the liver is a
profound effect with the magnitude of liver weight size increase physiologically limited.
Overall, the consistency between groups of female mice of the same strain for TCE-
induced liver weight gain, regardless of strain examined, was striking as was the lack of body
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-251 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
weight changes at TCE exposure levels that induced body weight changes in male mice. In the
absence of body weight changes, the difference in TCE-response in female mice appeared to be
reflective of strain and initial weight differences. Groups of female mice with higher body
weights, regardless of strain, generally had higher increases in TCE-induced liver weight
increases. For the C57BL and As/n strains, female mice starting weights were averaged 17.5
and 15.5 g, while the average liver weights were 1.63- and 1.64-fold of control after TCE
exposure, respectively. For the B6CBA, wild-type, DBA, and NZB female groups the starting
body weights averaged 22.5, 21.0, 23.0, and 21.0 g, while the average liver weights were 1.70-,
1.78-, 1.88-, and 2.09-fold of control after TCE exposure, respectively. The NMRI group of
female mice, did not follow this general pattern and had the highest initial body weight for the
single group of 10 mice reported (i.e., 27 g) associated with 1.66-fold of control liver weight.
The results of Kjellstrand et al. (1983b) suggested that there was more variability
between male mice than female mice in relation to TCE-induced liver weight gain. More strains
exhibited TCE-induced body weight changes in male mice than female mice suggesting
increased susceptibility of male mice to TCE toxicity as well as more variability in response.
Initial body weight also appeared to be a factor in the magnitude of TCE-induced liver weight
induction rather than just strain. In general, the strains and groups within strain that had TCE-
induced body weight decreases had smaller TCE-induced increase in liver weight. Therefore,
only examining liver weight in males as an indication of TCE treatment effects would not be an
accurate predictor of strain sensitivity nor the magnitude or response at doses that also affect
body weight. The results from this study show that comparison of the magnitude of TCE
response, as measured by liver weight increases, should take into account, strain, gender, initial
body weight and systemic toxicity. It shows a consistent pattern of increased liver weight in
both male and female mice after TCE exposure of 150 ppm for 30 days.
Kjellstrand et al. (1983a) presented data in the NMRI strain of mice (a strain that
appeared to be more prone to TCE-induced toxicity in male mice and a smaller TCE-induced
increase in liver weight in female mice) after inhalation exposure of 37 to 300 ppm TCE. They
used the same experimental paradigm as that reported in Kjellstrand et al. (1983b) except for
exposure concentration.
For female mice exposed to concentrations of TCE ranging from 37 to 300 ppm TCE
continuously for 30 days, only the 300 pm group experienced a 16% decrease in body weight
between control and exposed animals. Therefore, changes in TCE-induced liver weight
increases were affected by changes in body weight only for that group. Initial body weights in
the TCE-exposed female mice were similar in each of these groups (i.e., range of 29.2-31.6 g,
or 8%), with the exception of the females exposed to 150 ppm TCE for 30 days (i.e., initial body
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-252 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
weight of 27.3 g), reducing the effects of differences in initial body weight on TCE-induced
liver weight induction. Exposure to TCE continuously for 30 days was reported to result in a
linear dose-dependent increase in liver weight in female mice with 1.06-, 1.27-, 1.66-, and 2.14-
fold of control liver weights reported at 37 ppm, 75 ppm, 150 ppm, and 300 ppm TCE,
respectively.
In male mice there were more factors affecting reported liver weight increases from TCE
exposure. For male mice both the 150- and 300-ppm-exposed groups experienced a 10 and 18%
decrease in final body weight after TCE exposure, respectively. The 37- and 75-ppm groups did
not have decreased final body weight due to TCE exposure but varied by 12% in initial body
weight. TCE-induced increases in liver weight were reported to be 1.15-, 1.50-, 1.69-, and 1.90-
fold of control for 37, 75, 150, and 300 ppm TCE exposure in male mice, respectively. The
flattening of the dose-response curve at the two highest doses is consistent with the effects of
toxicity on final body weight.
Kjellstrand et al. (1983a) noted that liver mass increased and the changes in liver cell
morphology were similar in TCE-exposed male and female mice. They report that after 150
ppm exposure for 30 days, liver cells were generally larger and often displayed a fine
vacuolization of the cytoplasm, changes in nucleoli appearance. Kupffer cells of the sinusoid
were reported to be increased in cellular and nuclear size. The intralobular connective tissue
was infiltrated by inflammatory cells. Exposure to TCE in higher or lower concentrations
during the 30 days was reported to produce a similar morphologic picture.
For mice that were exposed to 150 ppm TCE for 30 days and then examined 120 days
after the cessation of exposure, liver weights were 1.09-fold of control for TCE-exposed female
mice and the same as controls for TCE-exposed male mice. However, the livers were not the
same as untreated liver in terms of histopathology. The authors reported that "after exposure to
150 ppm for 30 days, followed by 120 days of rehabilitation, the morphological picture was
similar to that of the air-exposure controls except for changes in cellular and nuclear sizes." The
authors did not present any quantitative data on the lesions they describe, especially in terms of
dose-response, and most of the qualitative description is for the 150-ppm-exposure level in
which there are consistent reports of TCE induced body weight decreases in male mice.
Although stating that Kupffer cells were increased in cellular and nuclear size, no
differential staining was applied to light microscopy sections and used to distinguish Kupffer
from endothelial cells lining the hepatic sinusoid in this study. Without differential staining
such a determination is difficult at the light microscopic level and a question remains as to
whether theses are the same cells as described by Goel et al. (1992) as a proliferation of
sinusoidal endothelial cells after exposures of 1,000 and 2,000 mg/kg/d TCE exposure for 28
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-253 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
days in male Swiss mice. As noted in Section E.2.4.2, the discrepancy in DNA synthesis
measures between hepatocyte examinations of individual hepatocytes and whole liver measures
in several reports of TCE metabolite exposure, is suggestive of increased DNA synthesis in the
nonparenchymal cell compartment of the liver. Thus, nonparenchymal cell proliferation is
suggested as an effect of subchronic TCE exposures in mice without concurrent focal necrosis
via inhalation studies (Kjellstrand et al., 1983a) and with focal necrosis in the presence of TCE
in a groundnut oil vehicle (Goel et al., 1992).
Although Kjellstrand et al. (1983a) did not discuss polyploidization, the changes in cell
size and especially the continued change in cell size and nuclear staining characteristics after
120 days of cessation of exposure are consistent with changes in polyploidization induced by
TCE that were suggested in studies from shorter durations of exposure (Dees and Travis, 1993;
Elcombe et al., 1985) and of longer durations (e.g., Buben and O'Flaherty, 1985). Of note is that
in the histological descriptions provided by Kjellstrand et al. (1983a) there was no mention of
focal necrosis or apoptosis resulting from these exposures to TCE to mice. Vacuolization is
reported and consistent with hepatotoxicity or lipid accumulation, which is lost during routine
histological slide preparation. The lack of reported focal necrosis in mice exposed through
inhalation is consistent with reports of gavage experiments of TCE in mice that do not use corn
oil as the vehicle (Merrick et al., 1989).
Buben and O'Flaherty (1985) reported the effects of TCE via corn oil gavage after six
weeks of exposure at concentrations ranging from 100 to 3,200 mg/kg d. This study was
conducted with older mice than those generally used in chronic exposure assays (Male Swiss-
Cox outbred mice between 3 and 5 months of age). Liver weight increases, decreases in liver
G6P activity, increases in liver triglycerides, and increases in SGPT activity were examined as
parameters of liver toxicity. Few deaths were reported during the 6-week exposure period
except at the highest dose and related to central nervous system depression. TCE exposure
caused dose-related increases in percent liver/body weight with a dose as low as 100 mg/kg/d
reported to cause a statistically significant increase (i.e., 112% of control).
The increases in liver size were attributed to hepatocyte hypertrophy, as revealed by
histological examination and by a decrease in the liver DNA concentration, and although
enlarged, were reported to appear normal. A dose-related trend toward triglyceride
concentration was also noted. A dose-related decrease in glucose-6-phophatase activity was
reported with similar small decreases (-10%) observed in the TCE exposed groups that did not
reach statistical significance until the dose reached 800 mg/kg TCE exposure. SGPT activity
was not observed to be increased in TCE-treated mice except at the two highest doses and even
at the 2,400 mg/kg dose half of the mice had normal values. The large variability in SGPT
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-254 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
activity was indicative of heterogeneity of this response between mice at the higher exposure
levels for this indicator of liver toxicity. Such variability of response in male mice is consistent
with the work of Kjellstrand et al. Thus, the results from Buben and O'Flaherty (1985) suggest
that hepatomegaly is a robust response that was reported to be observed at the lowest dose
tested, dose-related, and not accompanied by overt toxicity.
In terms of histopathology, Buben and O'Flaherty (1985) reported swollen hepatocytes
with indistinct borders; their cytoplasm was clumped and a vesicular pattern was apparent and
not simply due to edema in TCE-treated male mice. Karyorhexis (the disintegration of the
nucleus) was reported to be present in nearly all specimens from TCE-treated animals and
suggestive of impending cell death. It was not present in controls, appeared at a low level at
400 mg/kg TCE exposure level, and appeared to be slightly higher at 1,600 mg/kg TCE
exposure level. Central lobular necrosis was present only at the 1,600 mg/kg TCE exposure
level and at a very low level. Buben and O'Flaherty report increased polyploidy in the central
lobular region for both 400 mg/kg and 1,600 mg/kg TCE and described it as hepatic cells having
two or more nuclei or enlarged nuclei containing increased amounts of chromatin, but at the
lowest level of severity or occurrence. Thus, the results of this study are consistent with those of
shorter-term studies via gavage which report hepatocellular hypertrophy in the centralobular
region, increased liver weight induced at the lowest exposure level tested and at a level much
lower than those inducing overt toxicity, and that TCE exposure is associated with changes in
ploidy.
The National Toxicology Program 13-week study of TCE gavage exposure in 10 F344/N
rats [125 to 2,000 mg/kg (males) and 62.5 to 1,000 mg/kg (females)] and in B6C3Flmice (375
to 6,000 mg/kg) reported all rats survived the 13-week study. However male rat receiving
2,000 mg/kg exhibited a 24% difference in final body weight. The study descriptions of
pathology in rats and mice were not very detailed and included only mean liver weights. The
rats had increased pulmonary vasculitis at the highest concentration of TCE and viral titers were
positive for Sendai virus. No liver effects were noted for them in the study.
For mice, liver weights (both absolute and percent liver/body weight) were reported to
increase in a dose-related fashion with TCE exposure and to be increased by more than 10% in
750 mg/kg TCE-exposed males and 1,500 mg/kg or more TCE-exposed females.
Hepatotoxicity was reported as centrilobular necrosis in 6/10 males and 1/10 females exposed to
6,000 mg/kg TCE and multifocal areas of calcifications scattered throughout 3,000 mg/kg TCE
exposed male mice and only a single female 6,000 mg/kg dose, considered to be evidence of
earlier hepatocellular necrosis. One female mouse exposed to 3,000 mg/kg TCE also had a
hepatocellular adenoma, an extremely rare lesion in female mice of this age (20 weeks). At the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-255 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
lowest dose of exposure there was a consistent decrease in liver weight in female and male mice
after 13 weeks of TCE exposure.
Kawamoto et al. (1988b) exposed rats to 2 g/kg TCE subcutaneously for 15 weeks and
reported TCE-induced increases in liver weight. They also reported increase in cytochrome
P450, cytochrome b-5, and NADPH cytochrome c reductase. The difficulties in relating this
route of exposure to more environmentally relevant ones is discussed in Section E.2.2.11.
For 2-year or lifetime studies of TCE exposure a consistent hepatocarcinogenic response
has been observed in mice of differing strains and genders and from differing routes of
exposure. However, for rat studies some studies have been confounded by mortality from
gavage error or the toxicity of the dose of TCE administered. In some studies, a relative
insensitive strain of rat has been used. However, in general it appears that the mouse is more
sensitive than the rat to TCE-induced liver cancer. Three studies give results the authors
consider to be negative for TCE-induced liver cancer in mice, but have either design and/or
reporting limitations, or are in strains and paradigms with apparent low ability for liver cancer
induction or detection.
Fukuda et al. (1983) reported a 104-week inhalation bioassay in female Crj:CD-l (ICR)
mice and female Crj :CD (S-D) rats exposed to 0, 50, 150 and 450 ppm TCE (n = 50). There
were no reported incidences of mice or rats with liver tumors for controls indicative of relatively
insensitive strains used in the study for liver effects. While TCE was reported to induce a
number of other tumors in mice and rats in this study, the incidence of liver tumors was less than
2% after TCE exposure. Of note is the report of cystic cholangioma reported in 1 group of rats.
Henschler et al. (1980) exposed NMRI mice and WIST random bred rats to 0, 100, and
500 ppm TCE for 18 months (n = 30). This study is limited by short duration of exposure, low
number of animals, and low survival in rats. Control male mice were reported to have one
hepatocellular carcinoma and 1 hepatocellular adenoma with the incidence rate unknown. In the
100 ppm TCE exposed group, 2 hepatocellular adenomas and 1 mesenchymal liver tumor were
reported. No liver tumors were reported at any dose of TCE in female mice or controls. For
male rats, only 1 hepatocellular adenomas at 100 ppm was reported. For female rats no liver
tumors were reported in controls, but 1 adenoma and 1 cholangiocarcinoma was reported at
100 ppm TCE and at 500 ppm TCE, 2 cholangioadenomas, a relatively rare biliary tumor, was
reported. The difference in survival in mice, did not affect the power to detect a response, as
was the case for rats. However, the low number of animals studied, abbreviated exposure
duration, and apparently low sensitivity of this paradigm (i.e., no background response in
controls) suggests a study of limited ability to detect a TCE carcinogenic liver response. Of note
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-256 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
is that both Fukuda et al. (1983) and Henschler et al. (1980) report rare biliary cell derived
tumors in rats in relatively insensitive assays.
Van Duuren et al. (1979), exposed mice to 0.5 mg/mouse to TCE via gavage once a
week in 0.1 mL trioctanion (n = 30). Inadequate design and reporting of this study limit that
ability to use the results as an indicator of TCE carcinogenicity.
The NCI (1976) study of TCE was initiated in 1972 and involved the exposure of
Osborn-Mendel rats and B6C3F1 mice to varying concentrations of TCE. The animals were co-
exposed to a number of other carcinogens as exhalation as multiples studies and control animals
all shared the same laboratory space. Treatment duration was 78 weeks and animals received
TCE via gavage in corn oil at 2 doses (n = 20 for controls, but n = 50 for treatment groups). For
rats, the high dose was reported to result in significant mortality (i.e., 47/50 high-dose rats died
before scheduled termination of the study). A low incidence of liver tumors was reported for
controls and carbon tetrachloride positive controls in rats from this study. In B6C3F1 mice,
TCE was reported to increase incidence of hepatocellular carcinomas in both doses and both
genders of mice (-1,170 and 2,340 mg/kg for males and 870 and 1,740 mg/kg for female mice).
Hepatocellular carcinoma diagnosis was based on histologic appearance and metastasis to the
lung. The tumors were described in detail and to be heterogeneous "as described in the
literature" and similar in appearance to tumors generated by carbon tetrachloride. The
description of liver tumors in this study and tendency to metastasize to the lung are similar to
descriptions provided by Maltoni et al. (1986) for TCE-induced liver tumors in mice via
inhalation exposure.
For male rats, noncancer pathology in the NCI (1976) study was reported to include
increased fatty metamorphosis after TCE exposure and angiectasis or abnormally enlarged blood
vessels. Angiectasis can be manifested by hyperproliferation of endothelial cells and dilatation
of sinusoidal spaces. The authors conclude that due to mortality, "the test is inconclusive in
rats." They note the insensitivity of the rat strain used from their data on the positive control of
carbon tetrachloride exposure.
The NTP (1990) study of TCE exposure in male and female F344/N rats, and B6C3F1
mice (500 and 1,000 mg/kg for rats and 1,000 mg/kg for mice) was limited in the ability to
demonstrate a dose-response for hepatocarcinogenicity. There was also little reporting of
non-neoplastic pathology or toxicity and no report of liver weight at termination of the study.
However, by the end of a 2-year cancer bioassay, liver tumor induction can be a significant
factor in any changes in liver weight. No treatment-related increases in necrosis in the liver
were observed in mice. A slight increase in the incidence of focal necrosis was noted for TCE-
exposed male mice (8 vs. 2% in control) with a slight reduction in fatty metamorphosis in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-257 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
treated male mice (0 treated vs. 2 control animals). In female mice there was a slight increase in
focal inflammation (29 vs. 19% of animals) and no other changes. Therefore, this study did not
show concurrent evidence of liver toxicity but did show TCE-induced neoplasia after 2 years of
TCE exposure in mice. The administration of TCE was reported to cause earlier expression of
tumors as the first animals with carcinomas were reported to have them 57 weeks for TCE-
exposed animals and 75 weeks for control male mice.
The NTP (1990) study reported that TCE exposure was associated with increased
incidence of hepatocellular carcinoma (tumors with markedly abnormal cytology and
architecture) in male and female mice. Hepatocellular adenomas were described as
circumscribed areas of distinctive hepatic parenchymal cells with a perimeter of normal
appearing parenchyma in which there were areas that appeared to be undergoing compression
from expansion of the tumor. Mitotic figures were sparse or absent but the tumors lacked
typical lobular organization. Hepatocellular carcinomas had markedly abnormal cytology and
architecture with abnormalities in cytology cited as including increased cell size, decreased cell
size, cytoplasmic eosinophilia, cytoplasmic basophilia, cytoplasmic vacuolization, cytoplasmic
hyaline bodies and variations in nuclear appearance. Furthermore, in many instances several or
all of the abnormalities were present in different areas of the tumor and variations in architecture
with some of the hepatocellular carcinomas having areas of trabecular organization. Mitosis
was variable in amount and location. Therefore, the phenotype of tumors reported from TCE
exposure was heterogeneous in appearance between and within tumors.
For rats, the NTP (1990) study reported no treatment-related non-neoplastic liver lesions
in males and a decrease in basophilic cytological change reported from TCE-exposure in female
rats. The results for detecting a carcinogenic response in rats were considered to be equivocal
because both groups receiving TCE showed significantly reduced survival compared to vehicle
controls and because of a high rate (e.g., 20% of the animals in the high-dose group) of death by
gavage error.
The NTP (1988) study of TCE exposure in four strains of rats to "diisopropylamine-
stabilized TCE" was also considered inadequate for either comparing or assessing TCE-induced
carcinogenesis in these strains of rats because of chemically induced toxicity, reduced survival,
and incomplete documentation of experimental data. TCE gavage exposures of 0, 500 or
1,000 mg/kg per day (5 days per week, for 103 weeks) male and female rats was also marked by
a large number of accidental deaths (e.g., for high-dose male Marshal rats 25 animals were
accidentally killed).
Results from a 13-week study were briefly mentioned in the report and indicated
exposure levels of 62.5-2,000 mg/kg TCE were not associated with decreased survival (with the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-258 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
exception of 3 male August rats receiving 2,000 mg/kg TCE). Administration of the chemical
for 13 weeks was not associated with histopathological changes.
In regard to evidence of liver toxicity, the 2-year study of TCE exposure reported no
evidence of TCE-induced liver toxicity described as non-neoplastic changes ACI, August,
Marshal, and Osborne-Mendel rats. Interestingly, for the control animals of these four strains
there was, in general, a low background level of focal necrosis in the liver of both genders. In
summary, the negative results in this bioassay are confounded by the killing of a large portion of
the animals accidently by experimental error but TCE-induced overt liver toxicity was not
reported.
Maltoni et al. (1986) reported the results of several studies of TCE via inhalation and
gavage in mice and rats. A large number of animals were used in the treatment groups but the
focus of the study was detection of a neoplastic response with only a generalized description of
tumor pathology phenotype given and limited reporting of non-neoplastic changes in the liver.
Accidental death by gavage error was reported not to occur in this study. In regards to effects of
TCE exposure on survival, "a nonsignificant excess in mortality" correlated to TCE treatment
was observed only in female rats (treated by ingestion with the compound) and in male B6C3F1
mice.
TCE-induced effects on body weight were reported to be absent in mice except for one
experiment (BT 306 bis) in which a slight nondose correlated decrease was found in exposed
animals. "Hepatoma" was the term used to describe all malignant tumors of hepatic cells, of
different subhistotypes, and of various degrees of malignancy and were reported to be unique or
multiple, and have different sizes (usually detected grossly at necropsy) from TCE exposure. In
regard to phenotype tumors were described as usual type observed in Swiss and B6C3F1 mice,
as well as in other mouse strains, either untreated or treated with hepatocarcinogens and to
frequently have medullary (solid), trabecular, and pleomorphic (usually anaplastic) patterns.
Swiss mice from this laboratory were reported to have a low incidence of hepatomas without
treatment (1%). The relatively larger number of animals used in this bioassay (n = 90 to 100), in
comparison to NTP standard assays, allows for a greater power to detect a response.
TCE exposure for 8 weeks via inhalation at 100 ppm or 600 ppm may have been
associated with a small increase in liver tumors in male mice in comparison to concurrent
controls during the life span of the animals. In Swiss mice exposed to TCE via inhalation for
78 weeks there a reported increase in hepatomas associated with TCE treatment that was dose-
related in male but not female Swiss mice. In B6C3F1 mice exposed via inhalation to TCE for
78 weeks, the results from one experiment indicated a greater increase in liver cancer in females
than male mice but in a second experiment in males there was a TCE-exposure associated
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-259 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
increase in hepatomas. Although the mice were supposed to be of the same strain, the
background level of liver cancer was significantly different in male mice. The finding of
differences in response in animals of the same strain but from differing sources has also been
reported in other studies for other endpoints (see Section E.3.1.2). However, for both groups of
male B6C3F1 mice the background rate of liver tumors over the lifetime of the mice was less
than 20%.
For rats, there were 4 liver angiosarcomas reported (1 in a control male rat, 1 both in a
TCE-exposed male and female at 600 ppm TCE for 8 weeks, and 1 in a female rat exposed to
600 ppm TCE for 104 weeks) but the specific results for incidences of hepatocellular
"hepatomas" in treated and control rats were not given. Although Maltoni et al. (1986)
concluded that the small number of these tumors was not treatment-related, the findings were
brought forward because of the extreme rarity of this tumor in control SD rats, untreated or
treated with vehicle materials. In rats treated for 104 weeks, there was no report of a TCE
treatment-related increase in liver cancer in rats. This study only presented data for positive
findings so it did not give the background or treatment-related findings in rats for liver tumors in
this study. Thus, the extent of background tumors and sensitivity for this endpoint cannot be
determined.
Of note is that the SD strain used in this study was also noted in the Fukuda et al. (1983)
study to be relatively insensitive for spontaneous liver cancer and to also be negative for TCE-
induced hepatocellular liver cancer induction in rats. However, like Fukuda et al. (1983) and
Henschler et al. (1980), that reported rare biliary tumors in insensitive strains of rat for
hepatocellular tumors, Maltoni et al. (1986) reported a relatively rare tumor type, angiosarcoma,
after TCE exposure in a relatively insensitive strain for "hepatomas." As noted above, many of
the rat studies were limited by premature mortality due to gavage error or premature mortality
(Henschler et al., 1980; NCI, 1976; NTP, 1988, 1990), which was reported not occur in Maltoni
et al. (1986).
There were other reports of TCE carcinogenicity in mice from chronic exposures that
were focused primarily on detection of liver tumors with limited reporting of tumor phenotype
or non-neoplastic pathology. Herren-Freund et al. (1987) reported that male B6C3 F1 mice
given 40 mg/L TCE in drinking water had increased tumor response after 61 weeks of exposure.
However, concentrations of TCE fell by about '/2 at this dose of TCE during the twice a week
change in drinking water solution so the actual dose of TCE the animals received was less than
40 mg/L. The percent liver/body weight was reported to be similar for control and TCE-exposed
mice at the end of treatment. Despite difficulties in establishing accurately the dose received, an
increase in adenomas per animal and an increase in the number of animals with hepatocellular
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-260 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
carcinomas were reported to be associated with TCE exposure after 61 weeks of exposure and
without apparent hepatomegaly.
Anna et al. (1994) reported tumor incidences for male B6C3F1 mice receiving 800
mg/kg/d TCE via gavage (5 days/week for 76 weeks). All TCE-treated mice were reported to be
alive after 76 weeks of treatment. Although the control group contained a mixture of exposure
durations (76-134 weeks) and concurrent controls had a very small number of animals, TCE-
treatment appeared to increase the number of animals with adenomas, the mean number of
adenomas and carcinomas, but with no concurrent TCE-induced cytotoxicity.
E.2.5.4. Summary of Results For Subchronic and Chronic Effects of Dichloroacetic Acid
(DCA) and Trichloroacetic Acid (TCA): Comparisons With
Trichloroethylene (TCE)
There are no similar studies for TCA and DCA conduced at 6 weeks and with the range
of concentrations examined in Buben and O'Flaherty (1985) for TCE. In general, many studies
of DCA and TCA have been conducted at few and high concentrations, with shortened durations
of exposure, and varying and low numbers of animals to examine primarily a liver tumor
response in mice. However, the analyses presented in Section E.2.4.2 gives comparisons of
administered TCA and DCA dose-responses for liver weight increases for a number of studies in
combination as well as comparing such dose-responses to that of TCE and its oxidative
metabolism. As stated above, many subchronic studies of DCA and TCA have focused on
elucidating a relationship between dose and hypothesized events that may be indicators of
carcinogenic potential that have been described in chronic studies with a focus on indicators of
peroxisome proliferation and DNA synthesis. Many chronic studies have focused on the nature
of the DCA and TCA carcinogenic response in mouse liver through examination of the tumors
induced.
Most all of the chronic studies for DCA and TCA have been carried out in mice. As the
database for examination of the ability of TCE to induce liver tumors in rats includes several
studies that have been limited in ability determine a carcinogenic response in the liver, the
database for DCA and TCA in rats is even more limited. For TCA, the only available study in
rats (DeAngelo et al., 1997) has been frequently cited in the literature to indicate a lack of
response in this species for TCA-induced liver tumors. Although reporting an apparent dose-
related increase in multiplicity of adenomas and an increase in carcinomas over control at the
highest dose, DeAngelo et al. (1997) use such a low number of animals per treatment group
(n = 20-24) that the abilities of this study to determine a statistically significant increase in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-261 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
tumor response and to be able to determine that there was no treatment-related effect were
limited. A power calculation of the study shows that the type II error, which should be >50%,
was less than 8% probability for incidence and multiplicity of all tumors at all exposure TCA
concentrations with the exception of the incidence of adenomas and adenomas and carcinomas
for 0.5 g/L treatment group (58%) in which there was an increased in adenomas reported over
control (15 vs. 4%) that was the same for adenomas and carcinomas combined. Therefore, the
designed experiment could accept a false null hypothesis and erroneously conclude that there is
no response due to TCA treatment. While suggesting a lower response than for mice for liver
tumor induction, it is inconclusive for determination of whether TCA induces a carcinogenic
response in the liver of rats.
For DC A, there are two reported long-term studies in rats (DeAngelo et al., 1996;
Richmond et al., 1995) that appear to have reported the majority of their results from the same
data set and which consequently were subject to similar design limitations and DCA-induced
neurotoxicity in this species. DeAngelo et al. (1996) reported increased hepatocellular
adenomas and carcinomas in male F344 rats exposed for 2 years. However, the data from
exposure concentrations at a 5 g/L dose had to be discarded and the 2.5 g/L DCA dose had to be
continuously lowered during the study due to neurotoxicity. There was a DCA-induced
increased in adenomas and carcinomas combined reported for the 0.5 g/L DCA (24.1 vs. 4.4%
adenomas and carcinomas combined in treated vs. controls) and an increase at a variable dose
started at 2.5 g/L DCA and continuously lowered (28.6 vs. 3.0% adenomas and carcinomas
combined in treated vs. controls). Only combined incidences of adenomas and carcinomas for
the 0.5 g/L DCA exposure group was reported to be statistically significant by the authors
although the incidence of adenomas was 17.2 versus 4% in treated versus control rats.
Hepatocellular tumor multiplicity was reported to be increased in the 0.5 g/L DCA
group (0.31 adenomas and carcinomas/animal in treated vs. 0.04 in control rats) but was
reported by the authors to not be statistically significant. At the starting dose of 2.5 g/L,
continuously lowered due to neurotoxicity, the increased multiplicity of hepatocellular
carcinomas was reported by the authors to be to be statistically significant
(0.25 carcinomas/animals vs. 0.03 in control) as well as the multiplicity of combined adenomas
and carcinomas (0.36 adenomas and carcinomas/animals vs. 0.03 in control rats).
Issues that affected the ability to determine the nature of the dose-response for this study
include: (1) the use of a small number of animals (n = 23, n =21 and n = 23 at final sacrifice for
the 2.0 g/L NaCl control, 0.05 and 0.5 g/L treatment groups) that limit the power of the study to
both determine statistically significant responses and to determine that there are not treatment-
related effects (i.e., power) (2) apparent addition of animals for tumor analysis not present at
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-262 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
final sacrifice (i.e., 0.05 and 0.5 g/L treatment groups), and (3) most of all, the lack of a
consistent dose for the 2.5 g/L DCA exposed animals.
Similar issues were present for the study of Richmond et al. (1995) that was conducted
by the same authors as DeAngelo et al. (1996) and appeared to be from the same data set. The
Richmond et al. (1995) data for the 2 g/L NaCl. 0.05 g/L DCA and 0.5 g/L DCA exposure
groups were the same data set reported by DeAngelo et al. (1996) for these groups. Additional
data was reported for F344 rats administered and 2.5 g/L DCA that, due to hind-limb paralysis,
were sacrificed 60 weeks (DeAngelo et al., 1996). Tumor multiplicity was not reported by the
authors. There was a small difference in reports of the results between the two studies for the
same data for the 0.5 g/L DCA group in which Richmond et al. (1995) reported a 21%
incidence of adenomas and DeAngelo et al. (1996) reported a 17.2% incidence. The authors did
not report any of the results of DCA-induced increases of adenomas and carcinomas to be
statistically significant. The same issues discussed above for DeAngelo et al. (1996) apply to
this study. Similar to the DeAngelo study of TCA in rats (DeAngelo et al., 1997) the study of
DCA exposure in rats reported by DeAngelo et al. (1996) and Richmond et al. (1995), the use of
small numbers of rats limits the detection of treatment-related effects and the ability to
determine whether there was no treatment related effects (Type II error), especially at the low
concentrations of DCA exposure.
For mice, the data for both DCA and TCA is much more extensive and has shown that
both DCA and TCA induced liver tumors in mice. Many of the studies are for relatively high
concentrations of DCA or TCA, have been conducted for a year or less, and have focused on the
nature of tumors induced to ascertain potential MO As and to make inferences as to whether
TCE-induced tumors in mice are similar. As shown previously in Section E.2.4.2, the dose-
response curves for increased liver weight for TCE administration in male mice are more similar
to those for DCA administration and TCE oxidative metabolism than for direct TCA
administration. There are two studies in male B6C3F1 mice that attempt to examine multiple
concentrations of DCA and TCA for 2-year studies (DeAngelo et al., 2008; DeAngelo et al.,
1999) at doses that do not induce cytotoxicity and attempt to relate them to subchronic changes
and peroxisomal enzyme induction. However, the DeAngelo et al. (2008) study was carried out
in B6C3F1 mice that were of large size and prone to liver cancer and premature mortality
limiting its use for the determination of TCA-dose response in a 2-year bioassay. One study in
female B6C3F1 mice describes the dose-response for liver tumor induction at a range of DCA
and TCA concentrations after 51 or 82 weeks (Pereira, 1996) with a focus on the type of tumor
each compound produced.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-263 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DeAngelo et al. (1999) conducted a study of DC A exposure to determine a dose
response for the hepatocarcinogenicity of DC A in male B6C3F1 mice over a lifetime exposure
and especially at concentrations that did not illicit cytotoxicity or were for abbreviated exposure
durations. DeAngelo et al. (1999) used 0.05, 0.5, 1.0, 2.0, and 3.5 g/L exposure concentrations
of DC A in their 100-week drinking water study. The number of animals at final sacrifice was
generally low in the DCA treatment groups and variable (i.e., n = 50, n = 33, n = 24, n = 32,
n = 14, and n = 8 for control, 0.05, 0.5, 1, 2.0, and 3.5 g/L DCA exposure groups). It was
apparent that animals that died unscheduled deaths between weeks 79 and 100 were included in
data reported for 100 weeks. Although the authors did not report how many animals were
included in the 100-week results, it appeared that the number was no greater than 1 for the
control, 0.05, and 0.5 exposure groups and varied between 3 and 7 for the higher DCA exposure
groups.
The multiplicity or number of hepatocellular carcinomas/animals was reported to be
significantly increased over controls in a dose-related manner at all DCA treatments including
0.05 g/L DCA, and a NOEL reported not to be observed by the authors (i.e., 0.28, 0.58, 0.68,
1.29, 2.47, and 2.90 hepatocellular carcinomas/animal for control, 0.05, 0.5, 1.0, 2.0, and 3.5 g/L
DCA). Between the 0.5 and 3.5 g/L exposure concentrations of DCA the magnitude of increase
in multiplicity was similar to the increases in magnitude in dose. The incidence of
hepatocellular carcinomas were reported to be increased at all doses as well but not reported to
be statistically significant at the 0.05 g/L exposure concentration. However, given that the
number of mice examined for this response (n = 33), the power of the experiment at this dose
was only 16.9% to be able to determine that there was not a treatment related effect. The
authors did not report the incidence or multiplicity of adenomas for the 0.05 g/L exposure group
in the study and neither did they report the incidence or multiplicity of adenomas and
carcinomas in combination. For the animals surviving from 79 to 100 weeks of exposure, the
incidence and multiplicity of adenomas peaked at 1 g/L while hepatocellular carcinomas
continued to increase at the higher doses. This would be expected where some portion of the
adenomas would either regress or progress to carcinomas at the higher doses.
DeAngelo et al. (1999) reported that peroxisome proliferation was significantly
increased at 3.5 g/L DCA only at 26 weeks, not correlated with tumor response, and to not be
increased at either 0.05 or 0.5 g/L treatments. The authors concluded that DCA-induced
carcinogenesis was not dependent on peroxisome proliferation or chemically sustained
proliferation, as measured by DNA synthesis. DeAngelo et al. (1999) reported not only a dose-
related increase in DCA-induced liver tumors but also a decrease in time-to-tumor associated
with DCA exposure at the lowest levels examined. In regards to cytotoxicity there appeared to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-264 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
be a treatment but not dose-related increase in hepatocellular necrosis that did not involve most
of the liver from 1 to 3.5 g/L DCA exposures for 26 weeks of exposure. By 52 this effect was
diminished with no necrosis observed at the 0.5 g/L DCA treatment for any exposure period.
Hepatomegaly was reported to be absent by 100 weeks of exposure at the 0.05 and
0.5 g/L exposures while there was an increase in tumor burden reported. However, slight
hepatomegaly was present by 26 weeks in the 0.5 g/L group and decreased with time. Not only
did the increase in multiplicity of hepatocellular carcinomas increase proportionally with DCA
exposure concentration after 79-100 weeks of exposure, but so did the increases in percent
liver/body weight.
DeAngelo et al. (1999) presented a figure comparing the number of hepatocellular
carcinomas/animal at 100 weeks compared with the percent liver/body weight at 26 weeks that
showed a linear correlation (r = 0.9977) while peroxisome proliferation and DNA synthesis did
not correlate with tumor induction profiles. The proportional increase in liver weight with DCA
exposure was also reported for shorter durations of exposure as noted in Section E.2.4.2. The
findings of the study illustrate the importance of examining multiple exposure levels at lower
concentrations, at longer durations of exposure, and with an adequate number of animals to
determine the nature of a carcinogenic response. Although Carter et al. (1995) suggested that
there is evidence of DCA-induced cytotoxicity (e.g., loss of cell membranes and apparent
apoptosis) at higher levels, the 0.5 g/L exposure concentration was shown by DeAngelo et al.
(1999) to increase hepatocellular tumors after 100 weeks of treatment without concurrent
peroxisome proliferation or cytotoxicity in mice.
As noted in detail in E. 2.3.2.13, DeAngelo et al. (2008) exposed male B6C3F1 mice to
neutralized TCA in drinking water to male B6C3 F1 mice in three studies. Rather than using
5 exposure levels that were generally 2-fold apart, as was done in DeAngelo et al. (1999) for
DCA, DeAngelo et al. (2008) studied only 3 doses of TCA that were an order of magnitude
apart which limits the elucidation of the shape of the dose-response curve. In addition
DeAngelo et al. (2008) contained 2 studies, each conducted in a separate laboratories, for the
104-week data so that the two lower doses were studied in one study and the highest dose in
another. The first study was conducted using 2 g/L NaCl, or 0.05, 0.5, or 5 g/L TCA in drinking
water for 60 weeks (Study #1) while the other two were conducted for a period of 104 weeks
(Study #2 with 2.5 g/L neutralized acetic acid or 4.5 g/L TCA exposure groups and Study #3
with deionized water, 0.05 and 0.5 g/L TCA exposure groups). In the studies reported in
DeAngelo et al. (2008), a small number of animals has been used for the determination of a
tumor response (~n = 30 at final necropsy), but for the data for liver weight or PCO activity at
interim sacrifices the number was even smaller (n = 5).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-265 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
The percent liver/body weight changes at 4 weeks in Study #1 have been included in the
analysis for all TCA data in Section E.2.4.2, and are consistent with that data. Although there
was a 10-fold difference in TCA exposure concentration, there was a 9, 16, and 35% increase in
liver weight over control for the 0.05, 0.5, and 5 g/L TCA exposures. PCO activity varied 2.7-
fold as baseline controls but the increase in PCO activity at 4 weeks was 1.3-, 2.4-, and 5.3-fold
of control for the 0.05, 0.5, and 5 g/L TCA exposure groups in Study #1. The incidence data for
adenomas observed at 60 weeks was 2.1-, 3.0-, and 5.4-fold of control values and the fold
increases in multiplicity were similar after 0.05, 0.5, and 5.0 g/L TCA. Thus, in general the
dose-response for TCA-induced liver weight increases at 4 weeks was similar to the magnitude
of induction of adenomas at 60 weeks. Such a result is more consistent with the ability of TCA
to induce tumors and increases in liver weight at low doses with little change with increasing
dose as shown by this study and the combined data for TCA liver weight induction by
administered TCA presented in Section E.2.4.2.
While the 104-week data from Study's #2 and #3 could have been more valuable for
determination of the dose-response, as it would have allowed enough time for full tumor
expression, serious issues were apparent for Study #3, which was reported to have a 64%
incidence rate of adenomas and carcinomas for controls while that of Study #2 was 12%. As
stated in Section E.2.3.2.13, the mice in Study #3 were of larger size than those of either Study
#1 or #2 and the large background rate of tumors reported is consistent with mice of these size
(Leakey et al., 2003b). However, the large background rate and increased mortality for these
mice limit their use for determining the nature of the dose-response for TCA liver
carcinogenicity.
Examination of the data for treatment groups shows that there was no difference in any
of the results between the 0.5 g/L (Study #3) and 5 g/L (Study #2) TCA exposure groups (i.e.,
adenoma, carcinoma, and combinations of adenoma and carcinoma incidence and multiplicity)
for 104 weeks of exposure. For these same exposure groups, but at 60 weeks of exposure
(Study #1), there was a 2-fold increase in multiplicity for adenomas, and for adenomas and
carcinomas combined between the 0.5 and 5.0 g/L TCA exposure groups. At the two lowest
doses of 0.05 and 0.5 g/L TCA from Study #3 in the large tumor prone mice, the differences in
the incidences and multiplicities for all tumors were 2-fold at 104 weeks. These results are
consistent with (1) the two highest exposure levels reaching a plateau of response after a long
enough duration of exposure for full expression of the tumors (i.e., -90% of animals having
liver tumors at the 0.5 and 5 g/L exposures) with the additional tumors observed in a tumor-
prone paradigm. Thus, without use of the 0.05 and 0.5 g/L TCA data from Study #3, only the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-266 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
4.5 g/L TCA data from Study #2 can be used for determination of the TCA cancer response in a
2-year bioassay.
To put the 64% incidence data for carcinomas and adenomas reported in DeAngelo et al.
(2008) for the control group of Study #3 in context, other studies cited in this review for male
B6C3F1 mice show a much lower incidence in liver tumors with: (1) NCI (1976) study of TCE
reporting a colony control level of 6.5% for vehicle and 7.1%> incidence of hepatocellular
carcinomas for untreated male B6C3F1 mice (n = 70-77) at 78 weeks, (2) Herren-Freund et al.
(1987) reporting a 9%> incidence of adenomas in control male B6C3F1 mice with a multiplicity
of 0.09 ± 0.06 and no carcinomas (n = 22) at 61 weeks, (3) NTP (1990) reporting an incidence
of 14.6% adenomas and 16.6%> carcinomas in male B6C3F1 mice after 103 weeks (n = 48), and
(4) Maltoni et al. (1986) reporting that B6C3F1 male mice from the "NCI source" had a 1.1%
incidence of "hepatoma" (carcinomas and adenomas) and those from "Charles River Co." had a
18.9%o incidence of "hepatoma" during the entire lifetime of the mice (n = 90 per group).
The importance of examining an adequate number of control or treated animals before
confidence can be placed in those results in illustrated by Anna et al. (1994) in which at 76
weeks 3/10 control male B6C3F1 mice that were untreated and 2/10 control animals given corn
oil were reported to have adenomas but from 76 to 134 weeks, 4/32 mice were reported to have
adenomas (multiplicity of 0.13 ± 0.06) and 4/32 mice were reported to have carcinomas
(multiplicity of 0.12 ± 0.06). Thus, the reported combined incidence of carcinomas and
adenomas of 64%> reported by DeAngelo et al. (2008) for the control mice of Study #3, not only
is inconsistent and much higher than those reported in Studies #1 and #2, but also much higher
than reported in a number of other studies of TCE.
Trying to determine a correspondence with either liver weight increases or increases in
PCO activity after shorter periods of exposure will be depend whether data reported in Study #3
in the 104 week studies can be used. DeAngelo et al. (2008) reported a regression analyses that
compared "percent of hepatocellular neoplasia," indicated by tumor multiplicity, with TCA
dose, represented by estimations of the TCA dose in mg/kg/d, and with PCO activity for the 60-
week and 104-week data. Whether adenomas and carcinomas combined or individual tumor
type were used in these analysis was not reported by the authors. Concerns arise also from
comparing PCO activity at the end of the experiments, when there was already a significant
tumor response, rather than at earlier time points. Such PCO data may not be useful as an
indicator key event in tumorigenesis when tumors are already present.
In addition, regression analyses of these data are difficult to interpret because of the
dose spacing of these experiments as the control and 5 g/L exposure levels will basically
determine the shape of the dose-response curve. The 0.05 and 0.5 g/L exposure levels are close
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-267 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
to the control value in comparison to the 5 g/L exposure level, the dose response appears to be
linear between control and the 5.0 g/L value with the two lowest doses not affectly changing the
slope of the line (i.e., "leveraging" the regression). Thus, the value of these analyses is limited
by (1) use of data from Study #3 in a tumor prone mouse that is not comparable to those used in
Studies #1 and #2, (2) the appropriateness of using PCO values from later time points and the
variability in PCO control values (3) the uncertainty of the effects of palatability on the 5 g/L
TCA results which were reported in one study to reduce drinking water consumption, and (4)
the dose-spacing of the experiment.
DeAngelo et al. (2008) attempted to identify a NOEL for turnorigenicity using tumor
multiplicity data and estimated TCA dose. However, it is not an appropriate descriptor for these
data, especially given that "statistical significance" of the tumor response is the determinant
used by the authors to support the conclusions regarding a dose in which there is no TCA-
induced effect. Due to issues related to the appropriateness of use of the concurrent control in
Study #3, only the 60-week experiment (i.e., Study #1) is useful for the determination of tumor
dose-response. However, there no allowance for full expression of a tumor response at the
60-week time point. In addition, a power calculation of the 60-week study shows that the type II
error, which should be >50% and thus, greater than the chances of "flipping a coin," was 41 and
71% for incidence and 7 and 15% for multiplicity of adenomas for the 0.05 and 0.5 g/L TCA
exposure groups. For the combination of adenomas and carcinomas, the power calculation was
8 and 92% for incidence and 6 and 56% for multiplicity at 0.05 and 0.5 g/L TCA exposure.
Therefore, the designed experiment could accept a false null hypothesis, especially in terms of
tumor multiplicity, at the lower exposure doses and erroneously conclude that there is no
response due to TCA treatment.
Pereira (1996) examined the tumor induction in female B6C3 F1 mice and demonstrated
that foci, adenoma, and carcinoma development in mice are dependent on duration of exposure,
(or period of observation in the case of controls) for full expression of a carcinogenic response.
In control female mice a 360- versus 576-day observation period showed that at 360 days no
foci or carcinomas and only 2.5% of animals had adenomas whereas by 576 days of observation,
11% had foci, 2% adenomas, and 2% had carcinomas. For DCA and TCA treatments, foci,
adenomas, and carcinoma incidence and multiplicity did not reach full expression until
82 weeks at the 3 doses employed (2.58 g/L DCA, 0.86 g/L DCA, 0.26 g/L DCA, 3.27 g/L
TCA, 1.1.0 g/L TCA, and 0.33 g/L TCA). Although the numbers of animals were relatively low
and variable at the two highest doses (18-28 mice) there were 50-53 mice studied at the lowest
dose level and 90 animals studied in the control group.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-268 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The results of Pereira (1996) showed that not only were the incidences of mice with
foci, adenoma, and carcinomas greatly increased with duration of exposure, but concentration
also affected the nature and magnitude of the response in female mice. At 2.86 g/L, 0.86 g/L,
0.26 g/L DCA exposures and controls, after 82 weeks the incidence of adenomas in female
B6C3 F1 mice was reported to be 84.2, 25.0, 6.0, and 2.2%, respectively, and carcinomas to be
26.3, 3.6, 0, and 2.2%, respectively. For the multiplicity or number of tumors/animal at these
same exposure levels of DCA, the multiplicity was reported to be 5.58, 0.32, 0.06, and 0.02
adenomas/animal, and 0.37, 0.04, 0, and 0.02 carcinomas/animal. Thus, for DCA exposure in
female mice, for ~3-fold increases in DCA exposure concentration, after 82 weeks of exposure
there was a similar magnitude of increase in adenomas incidence with much greater increases in
multiplicity. For hepatocellular carcinoma induction, there was no increase in the incidence or
multiplicity or carcinomas between the control and 0.33 g/L DCA dose.
At 3.27, 1.10, and 0.33 g/L TCA and controls, after 82 weeks the incidence of adenomas
in female B6C3F1 mice was reported to be 38.9, 11.1, 7.6, and 2.2%, respectively, and
carcinomas to be 27.8, 18.5, 0, and 2.2%, respectively. At these same exposure levels of TCA,
the multiplicity was reported to be 0.61, 0.11, 0.08, and 0.02 adenomas/animal, and 0.39, 0.22,
0, and 0.02 carcinomas/animal, respectively. Thus, for TCA, the incidences of adenomas were
lower at the two highest doses than DCA and the ~3-fold differences in dose between the two
lowest doses only resulted in -50% increase in incidences of adenomas. For incidence of
carcinomas the ~3-fold difference in dose between the two highest doses only resulted in -50%
increase in carcinoma incidence. A similar pattern was reported for multiplicity after TCA
exposure. Foci were also examined and, in general., were similar to adenomas regarding
incidence and multiplicity. Thus, the dose-response curve for tumor induction in female mice
differed between DCA and TCA after 82 weeks of exposure with TCA having a much less steep
dose-response curve than DCA. This is consistent with the pattern of liver weight increases
reported for male B6C3F1 mice in Section E.2.4.2.
DeAngelo et al. (1999) reported a linear increase in incidence and multiplicity of
hepatocellular carcinomas that was proportional to dose and as well as proportional to the
magnitude of liver weight increase from subchronic exposure to DCA. However, the studies of
DeAngelo et al. (2008) and Pereira (1996) are suggestive that TCA induced increase in tumor
incidence are less proportional to increases in dose as are liver weight increases from subchronic
exposure.
Given that TCE subchronic exposure also induced an increase in liver weight that was
proportional to dose (i.e., similar to DCA but not TCA), it is of interest as to whether the dose-
response for TCE induced liver cancer in mice was similar. The database for TCE, while
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-269 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
consistently showing a induction of liver tumors in mice, is very limited for making inferences
regarding the shape of the dose-response curve. For many of these experiments multiplicity was
not given but only liver tumor incidence. NTP (1990), Bull et al. (2002), and Anna et al. (1994)
conducted gavage experiments in which they only tested one dose of-1,000 mg/kg/d TCE. NCI
(1976) tested 2 doses that were adjusted during exposure to an average of 1,169 mg/kg/d and
2,339 mg/kg/d in male mice with only 2-fold dose spacing in only 2 doses tested. Maltoni et al.
(1988) conducted inhalation experiments in 2 sets of B6C3F1 mice and one set of Swiss mice at
3 exposure concentrations that were 3-fold apart in magnitude between the low and mid-dose
and 2-fold apart in magnitude between the mid- and high-dose. However, for one experiment in
male B6C3F1 mice, the mice fought and suffered premature mortality and for two the
experiments in B6C3F1 mice, although using the same strain, the mice were obtained from
differing sources with very different background liver tumor levels.
For the Maltoni et al. (1988) study a general descriptor of "hepatoma" was used for liver
neoplasia rather than describing hepatocellular adenomas and carcinomas so that comparison of
that data with those from other experiments is difficult. More importantly, while the number of
adenomas and carcinomas may be the same between treatments or durations of exposure, the
number of adenomas may decrease as the number of carcinomas increase during the course of
tumor progression. Such information is lost by using only a hepatoma descriptor.
Maltoni et al. (1988) did not report an increase over control for 100 ppm TCE for the
Swiss group and one of the B6C3F1 groups and only a slight increase (1.12-fold) in the second
B6C3F1 group. At 300 ppm TCE exposure, the incidences of hepatoma were 2-fold of control
values for the Swiss, 4-fold of control for group of B6C3F1 mice, and 1.6-fold of control for the
other group of B6C3F1 mice. At 600 ppm TCE the incidences of hepatoma were 3.3-fold of
control for the Swiss group, 6.1-fold of control for one group of B6C3F1 mice, and 1.2-fold for
the other group of B6C3F1 mice. Thus, for each group of TCE exposed mice in the Maltoni et
al. (1988) inhalation study, the background levels of hepatomas and the shape of the dose-
response curve for TCE-hepatoma induction were variable. However, an average of the
increases, in terms of fold of control, between the 3 experiments gives a ~2.9-fold increase
between the low- and mid-dose (100 ppm and 300 ppm) and ~1.4-fold increase between the
mid- and high-dose (300 ppm and 600 pm) groups.
Although such a comparison obviously has a high degree of uncertainty associated with
it, it suggests that the magnitude of TCE-induced hepatoma increases over control is similar to
the 3- and 2-fold difference in the magnitude of exposure concentrations between these doses.
Therefore, the increase in TCE-induced liver tumors would roughly proportional to the
magnitude of exposure dose. This result would be similar to the result for the concordance of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-270 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
the increases in liver weight and exposure concentration observed 28-42 day exposures to TCE
(see Section E.2.4.2) using oral data from B6C3F1 and Swiss mice, and inhalation data from
NMRI mice.
The available inhalation data for TCE induced liver weight dose-response is from one
study in a strain derived from Swiss mice (Kjellstrand et al., 1983a) and was conducted in male
and female mice with comparable doses of 75 ppm and 300 ppm TCE. However, male mice of
this strain exhibited decreased body weight at the 300 ppm level, which can affect percent
liver/body weight increases. The magnitude of TCE-induced increases in liver weight between
the 75 ppm and 300 ppm exposures were ~1.80-fold for males (1.50 vs. 1.90-fold of control
liver weights) and 4.2-fold for females (1.27- vs. 2.14-fold of control liver weight) in this strain.
Female mice were examined in one study each of Swiss and B6C3F1 mice by Maltoni et
al. (1988). Both the Swiss and B6C3F1 studies reported increases in incidences of hepatomas
over controls only at the 600 ppm TCE level in female mice indicating less of a response than
males. Similarly, the Kjellstrand et al. (1983a) data also showed less of a response in females
compared to males in terms TCE induction of liver weight at the 37 to 150 ppm range of
exposure in NMRI strain. While the data for TCE dose-response of liver tumor induction is
very limited, it is suggestive of a correlation of TCE-induced increases in liver weight
correlating liver tumor induction with a pattern that is dissimilar to that of TCA.
Of those experiments conducted at -1,000 mg/kg/d gavage dose of TCE in male
B6C3F1 mice for at least 79 weeks (Anna et al., 1994; Bull et al., 2002; NCI, 1976; NTP, 1990),
the control values were conducted in varying numbers of animals [some as low as n= 15, i.e.,
(Bull et al., 2002)] and with varying results. The incidence of hepatocellular carcinomas ranged
from 1.2 to 16.7% (Anna et al., 1994; NCI, 1976; NTP, 1990) and the incidence of adenomas
ranged from 1.2 to 14.6% (Anna et al., 1994; NTP, 1990) in control B6C3F1 mice. After
-1,000 mg/kg/d TCE treatment, the incidence of carcinomas ranged from 19.4 to 62% (Anna et
al., 1994; Bull et al., 2002; NCI, 1976; NTP, 1990) with 3 of the studies (Anna et al., 1994; NCI,
1976; NTP, 1990) reporting a range of incidences between 42.8 to 62.0%). The incidence of
adenomas ranged from 28 to 66.7% (Anna et al., 1994; Bull et al., 2002; NTP, 1990). These
data are illustrative of the variability between experiments to determine the magnitude and
nature of the TCE response in the same gender (male), strain (B6C3F1), time of exposure (3/4
studies were for 76-79 weeks and 1 for 2 years duration), and roughly the same dose
(800-1,163 mg/kg/d TCE).
Given, that the TCE-induced liver response, as measured by liver weight increase, is
highly correlated with total oxidative metabolism to a number of agents that are hepatoactive
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-271 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
agents and hepatocarcinogens, the variability in response from TCE exposure would be expected
to be greater than studies of exposure to a single metabolite such as TCA or DCA.
Caldwell et al. (2008a; 2008b) have commented on the limitations of experimental
paradigms used to study liver tumor induction by TCE metabolites and show that 51-week
exposure duration has consistently produced a tumor response for these chemicals, but with
greater lesion incidence and multiplicity at 82 weeks. As reported by DeAngelo et al. (1999)
and Pereira (1996), full expression of tumor induction in the mouse does not occur until 78 to
100 weeks of DCA or TCA exposure, especially at lower concentrations. Thus, use of
abbreviated exposure durations and concurrently high exposure concentrations limits the ability
of such experiments to detect a treatment-related effect with the occurrence of additional
toxicity not necessarily associated with tumor-induction. Caldwell et al. (2008a) present a table
that shows that the differences in the ability of the studies to detect treatment-related effects
could also be attributed to a varying and low number of animals in some exposure groups and
that because of the low numbers of animals tested at higher exposures, the power to detect a
statistically significant change is very low and in fact for many of the endpoints is considerably
less than "50% chance." Table E-17 from Caldwell et al. (2008a) illustrates the importance of
experimental design and the limitations in many of the studies in the TCE metabolite database.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-272 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Table E-17. Power calculations" for experimental design described in text,
using Pereira et al. as an example
Exposure concentrationb in female
B6C3F1 mice (1996) (1996)
Number
of
animals
Power
calculation
for foci
Power
calculation for
adenomas
Power
calculation for
carcinomas
20.0 mmol/L NaCl (control) (82 wks)
90
Null
hypothesis
Null
hypothesis
Null hypothesis
2.58 g/L DCA (82 wks)
19
0.03
0.03
0.13
0.86 g/L DCA (82 wks)
28
0.74
0.20
0.91
0.26 g/L DCA (82 wks)
50
0.99
0.98
-
3.27 g/L TCA (82 wks)
18
0.15
0.09
0.14
1.10 g/L TCA (82 wks)
27
0.60
0.64
0.3
0.33 g/L TCA (82 wks)
53
0.93
0.91
-
"The power calculations represent the probability of rejecting the null hypothesis when in fact the alternate
hypothesis is true for tumor multiplicity (i.e., the total number of lesions/number of animals). The higher the
power number calculated, the more confidence we have in the null hypothesis. Assumptions made included:
normal distribution for the fraction of tumors reported, null hypothesis represents what we expected the control
tumor fraction to be, the probability of a Type I error was set to 0.05, and the alternate hypothesis was set to four
times the null hypothesis value.
bConversion of mmol/L to g/L from the original reports of Pereira (1996) and Pereira and Phelps
(1996) is as follows: 20.0 mmol/L DCA = 2.58 g/L, 6.67 mmol/L DCA = 0.86 g/L, 2.0 mmol/L
= 0.26 g/L, 20.0 mmol/L TCA = 3.27 g/L, 6.67 mmol/L TCA =1.10 g/L, 2.0 mmol/L TCA =
0.33 g/L.
Bull et al. (1990) examined male and female B6C3F1 mice (age 37 days) exposed from
15 to 52 weeks to neutralized DCA and TCA (1 or 2 g/L) but tumor data were not suitable for
dose response. They reported effects of DCA and TCA exposure on liver weight and percent
liver/body changes that gave a pattern of hepatomegaly generally consistent with short-term
exposure studies. Only 10 female mice were examined at 52 weeks but the female mice were
reported to be as responsive as males at the exposure concentration tested. After 37 weeks of
treatment and then a cessation of exposure for 15 weeks, liver weights percent liver/body weight
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-273 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
were reported to be elevated over controls which Bull et al. (1990) partially attribute the
remaining increases in liver weight to the continued presence of hyperplastic nodules in the liver.
Macroscopically, livers treated with DCA were reported to have multifocal areas of
necrosis and frequent infiltration of lymphocytes on the surface and an interior of the liver. For
TCA-treated mice, similar necrotic lesions were reported but at such a low frequency that they
were similar to controls. Marked cytomegaly was reported from exposure to either 1 or 2 g/L
DCA throughout the liver. Cell size was reported to be increased from TCA and DCA treatment
with DCA producing the greatest change. The 2 g/L TCA exposures were observed to have
increased accumulations of lipofuscin but no quantitative analysis was done. Photographs of
light microscopic sections, that were supposed to be representative of DCA and TCA treated
livers at 2 g/L, showed such great hepatocellular hypertrophy from DCA treatment that sinusoids
were obscured. Such a degree of cytomegaly could have resulted in reduction of blood flow and
contributed to focal necrosis observed at this level of exposure.
As discussed in Sections E.3.2 and E.3.4.2.1, glycogen accumulation has been described
to be present in foci in both humans and animals as a result from exposure to a wide variety of
carcinogenic agents and predisposing conditions in animals and humans. Bull et al (1990)
reported that glycogen deposition was uniformly increased from 2 g/L DCA exposure with
photographs of TCA exposure showing slightly less glycogen staining than controls. However,
the abstract and statements in the paper suggest that there was increased PAS positive material
from TCA treatment that has caused confusion in the literature in this regard. Kato-Weinstein et
al. (2001) reported that in male B6C3F1 mice exposed to DCA and TCA, the DCA treatment
increased glycogen and TCA decreased glycogen content of the liver by using both chemical
measurement of glycogen in liver homogenates and by using ethanol-fixed sections stained with
PAS, a procedure designed to minimize glycogen loss. Kato-Weinstein et al. (2001) reported
that glycogen rich and poor cells were scattered without zonal distribution in male B6C3F1 mice
exposed to 2 g/L DCA for 8 weeks. For TCA treatments they reported centrilobular decreases in
glycogen and -25% decreases in whole liver by 3 g/L TCA.
Kato-Weinstein et al. (2001) reported whole liver glycogen to be increased ~1.50-fold of
control (90 vs. 60 mg glycogen/g liver) by 2 g/L DCA after 8 weeks exposure male B6C3F1
mice with a maximal level of glycogen accumulation occurring after 4 weeks of DCA exposure.
Pereira et al. (2004b) reported that after 8 weeks of exposure to 3.2 g/L DCA liver glycogen
content was 2.20-fold of control levels (155.7 vs. 52.4 mg glycogen/g liver) in female B6C3F1
mice. Thus, the baseline level of glycogen content reported by (-60 mg/g) and the increase in
glycogen after DCA exposure was consistent between Kato-Weinstein et al. (2001) and Pereira
et al. (2004b). However, the increase in liver weight reported by Kato-Weinstein et al. (2001) of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-274 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
1.60-fold of control percent liver/body weight cannot be accounted for by the 1.50-fold of
control glycogen content. Glycogen content only accounts for 5% of liver mass so that 50%
increase in glycogen cannot account for the 60% increase liver mass induced by 2 g/L DCA
exposure for 8 weeks reported by Kato-Weinstein (2001). Thus, DCA-induced increases in liver
weight are occurring from other processes as well.
Carter et al. (2003) and DeAngelo et al. (1999) reported increased glycogen after DCA
treatment at much lower doses after longer periods of exposure (100 weeks). Carter reported
increased glycogen at 0.5 g/L DCA and DeAngelo et al. (1999) at 0.03 g/L DCA in mice.
However, there was no quantitation of that increase.
The issues involving identification of MO A through tumor phenotype analysis are
discussed in detail below for the more general case of liver cancer as well as for specific
hypothesized MOAs (see Sections E.3.1.4, E.3.1.8, E.3.2.1, and E.3.4.1.5). For TCE and its
metabolites, c-Jun staining, H-rats mutation, tincture, heterogeneity in dysplacity have been used
to describe and differentiate liver tumors in the mouse.
Bull et al. (2002) reported 1,000 mg/kg TCE administered via gavage daily for 79 weeks
in male B6C3F1 mice to produce liver tumors and also reported deaths by gavage error (6 out of
40 animals). The limitations of the experiment are discussed in Caldwell et al. (2008a).
Specifically, for the DCA and TCA exposed animals, the experiment was limited by low
statistical power, a relatively short duration of exposure, and uncertainty in reports of lesion
prevalence and multiplicity due to inappropriate lesions grouping (i.e., grouping of hyperplastic
nodules, adenomas, and carcinomas together as "tumors"), and incomplete histopatholology
determinations (i.e., random selection of gross lesions for histopathology examination).
For the TCE results, a high prevalence (23/36 B6C3F1 male mice) of adenomas and
hepatocellular carcinoma (7/36) was reported. For determinations of immunoreactivity to c-Jun,
as a marker of differences in "tumor" phenotype, Bull et al. (2002) included all lesions in most of
their treatment groups, decreasing the uncertainty of his findings. However, for
immunoreactivity results hyperplastic nodules, adenomas, and carcinomas were grouped and
thus, changes in c-Jun expression between the differing types of lesions were not determined.
Bull et al. (2002) reported lesion reactivity to c-Jun antibody to be dependent on the
proportion of the DCA and TCA administered after 52 weeks of exposure. Given alone, DCA
was reported to produce lesions in mouse liver for which approximately half displayed a diffuse
immunoreactivity to a c-Jun antibody, half did not, and none exhibited a mixture of the two.
After TCA exposure alone, no lesions were reported to be stained with this antibody. When
given in various combinations, DCA and TCA coexposure induced a few lesions that were only
c-Jun+, many that were only c-Jun-, and a number with a mixed phenotype whose frequency
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-275 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
increased with the dose of DCA. For TCE exposure of 79 weeks, TCE-induced lesions were
reported to also have a mixture of phenotypes (42% c-Jun+, 34% c-Jun-, and 24% mixed) and to
be most consistent with those resulting from DCA and TCA coexposure but not either metabolite
alone.
Stauber and Bull (1997) exposed male B6C3F1 mice (7 weeks old at the start of
treatment) to 2.0 g/L neutralized DCA or TCA in drinking water for 38 or 50 weeks, respectively
and then exposed (n = 12) to 0, 0.02, 0.1, 0.5, 1.0, 2.0 g/L DCA or TCA for an additional 2
weeks. Foci and tumors were combined in reported results as "lesions" and prevalence rates
were not reported. The DCA-induced larger "lesions" were reported to be more "uniformly
reactive to c-Jun and c-Fos" but many nuclei within the lesions displaying little reactivity to c-
Jun. Stauber and Bull (1997) stated that while most DCA-induced "lesions" were
homogeneously immunoreactive to c-Jun and C-Fos (28/41 lesions), the rest were stained
heterogeneously. For TCA-induced lesions, the authors reported no difference in staining
between "lesions" and normal hepatocytes in TCA-treated animals. These results are slightly
different that those reported by Bull et al. (2002) for DCA, who report c-Jun positive and
negative foci in DCA-induced liver tumors but no mixed lesions. Because "lesions" comprised
of foci and tumors, different stages of progression reported in these results. The duration of
exposures also differed between DCA and TCA treatment groups that can affect phenotype. The
shorter duration of exposure can also prevent full expression of the tumor response.
Stauber et al. (1998) presented a comparison of in vitro results with "tumors" from
Stauber and Bull (1997) and note that 97.5% of DCA-induced "tumors" were c-Jun + while none
of the TCA-induced "tumors" were c-Jun +. However, the concentrations used to give tumors in
vivo for comparison with in vitro results were not reported. This appears to differ from the
heterogeneity of result for c-Jun staining reported by Bull et al. (2002) and Stauber and Bull
(1997). There was no comparison of c-Jun phenotype for spontaneous tumors with the authors
stating that because of such short time, no control tumors results were given. However, the
results of Bull et al. (2002) and Stauber and Bull (1997), do show TCA-induced lesions to be
uniformly c-Jun negative and thus, the phenotypic marker was able to show that TCE-induced
tumors were more like those induced by DCA than TCA.
The premise that DCA induced c-Jun positive lesions and TCA-induced c-Jun negative
lesions in mouse liver was used as the rationale to study induction of "transformed" hepatocytes
by DCA and TCE treatment in vitro. Stauber et al. (1998) isolated primary hepatocytes from
5-8 week old male B6C3F1 mice (n = 3) and subsequently cultured them in the presence of
DCA or TCA. In a separate experiment 0.5 g/L DCA was given to mice as pretreatment for
2 weeks prior to isolation. The authors assumed that the anchorage-independent growth of these
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-276 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
hepatocytes was an indication of an "initiated cell." DCA and TCA solutions were neutralized
before use.
After 10 days in culture with DCA or TCA (0, 0.2, 0.5 and 2.0 mM), concentrations of
0.5 mM or more DCA and TCA both induced an increase in the number of colonies that was
statistically significant, increased with dose with DCA, and slightly greater for DCA. In a time
course experiment the number of colonies from DCA treatment in vitro peaked by 10 days and
did not change through days 15-25 at the highest dose and, at lower concentrations of DCA,
increased time in culture induced similar peak levels of colony formation by days 20-25 as that
reached by 10 days at the higher dose. Therefore, the number of colonies formed was
independent of dose if the cells were treated long enough in vitro.
However, not only did treatment with DCA or TCA induce anchorage independent
growth but untreated hepatocytes also formed larger numbers of colonies with time, although at a
lower rate than those treated with DCA. The level reached by untreated cells in tissue culture at
20 days was similar to the level induced by 10 days of exposure to 0.5 mM DCA. The time
course of TCA exposure was not tested to see if it had a similar effect with time as did DCA.
The colonies observed at 10 days were tested for c-Jun expression with the authors noting that
"colonies promoted by DCA were primarily c-Jun positive in contrast to TCA promoted colonies
that were predominantly c-Jun negative." Of the colonies that arose spontaneously from tissue
culture conditions, 10/13 (76.9%) were reported to be c-Jun +, those treated with DCA 28/34
(82.3%) were c-Jun +, and those treated with TCA 5/22 (22.1%) were c-Jun +. Thus, these data
show heterogeneity in cell in colonies but with more c-Jun + colonies occurring by tissue culture
conditions alone and in the presence of DCA, rather than in the presence of TCA.
The authors reported that with time (24, 48, 72, and 96 hours) of culture conditioning the
number of c-Jun+ colonies was increased in untreated controls. The authors reported that DCA
treatment delayed the increase in c-Jun+ expression induced by tissue culture conditions alone in
untreated controls while TCA treatment was reported to not affect the increasing c-Jun+
expression that increased with time in tissue culture. These results seems paradoxical given that
DCA induced a higher number of colonies at 10 days of tissue culture than TCA and that most of
the colonies were c-Jun positive. The number of colonies was greater for pretreatment with
DCA, but the magnitude of difference over the control level was the same after DCA treatment
in vitro without and without pretreatment. As to the relationship of c-Jun staining and
peroxisome proliferators as a class, as pointed out by Caldwell and Keshava (2006), although
Bull et al. (2004) have suggested that the negative expression of c-jun in TCA-induced tumors
may be consistent with a characteristic phenotype shown in general by peroxisome proliferators
as a class, there is no supporting evidence of this.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-277 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
An approach to determine the potential MO As of DCA and TCA through examination of
the types of tumors each "induced" or "selected" was to examine H-ras activation (Anna et al.,
1994; Bull et al., 2002; Ferreira-Gonzalez et al., 1995; Nelson et al., 1990). This approach has
also been used to try to establish an H-ras activation pattern for "genotoxic" and "nongenotoxic"
liver carcinogens compounds and to make inferences concerning peroxisome proliferator-
induced liver tumors.
However, as noted by Stanley et al. (1994), the genetic background of the mice used and
the dose of carcinogen may affect the number of activated H-ras containing tumors that develop.
In addition, the stage of progression of "lesions" (i.e., foci vs. adenomas vs. carcinomas) also has
been linked the observance of H-ras mutations.
Fox et al. (1990) note that tumors induced by phenobarbital (0.05% drinking water (H20),
1 year), chloroform (200 mg/kg corn oil gavage, 2 times weekly for 1 year) or Ciprofibrate
(0.0125% diet, 2 years) had a much lower frequency of H-ras gene activation than those that
arose spontaneously (2-year bioassays of control animals) or induced with the "genotoxic"
carcinogen benzidine-2 hydrochloric acid (HC1; 120 ppm, drinking H20, 1 year) in mice. In that
study, the term "tumor" was not specifically defined but a correlation between the incidence of
H-ras gene activation and development of either a hepatocellular adenoma or hepatocellular
carcinoma was reported to be made with no statistically significant difference between the
frequency of H-ras gene activation in the hepatocellular adenomas and carcinomas.
Histopathological examination of the spontaneous tumors, tumors induced with benzidine-
2HCL, Phenobarbital, and chloroform was not reported to reveal any significant changes in
morphology or staining characteristics.
Spontaneous tumors were reported to have 64% point mutation in codon 61 (n = 50
tumors examined) with a similar response for Benzidine of 59% (n = 22 tumors examined),
whereas for Phenobarbital the mutation rate was 7% (n= 15 tumors examined), chloroform 21%
(n = 24 tumors examined) and Ciprofibrate 21% (n = 39 tumors examined). The Ciprofibrate-
induced tumors were reported to be more eosinophilic as were the surrounding normal
hepatocytes.
Hegi et al. (1993) tested Ciprofibrate-induced tumors in the NIH3T3 cotransfection-nude
mouse tumorigenicity assay, which the authors stated is capable of detecting a variety of
activated proto-oncogenes. The tumors examined (Ciprofibrate-induced or spontaneously
arising) were taken from the Fox et al. study (1990), screened previously, and found to be
negative for H-ras activation. With the limited number of samples examined, Hegi et al.
concluded that ras proto-oncogene activation or activation of other proto-oncogenes using the
nude mouse assay were not frequent events in Ciprofibrate-induced tumors and that spontaneous
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-278 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
tumors were not promoted with it. Using the more sensitive methods, the H-ras activation rate
was reported to be raised from 21 to 31% for Ciprofibrate-induced tumors and from 64 to 66%
for spontaneous tumors.
Stanley et al. (1994) studied the effect of methylclofenapate (MCP) (25 mg/kg for up to 2
years), a peroxisome proliferator, in B6C3F1 (relatively sensitive) and C57BL/10J (relatively
resistant) mice for H-ras codon 61 point mutations in MCP-induced liver tumors (hepatocellular
adenomas and carcinomas). In the B6C3F1 mice the number of tumors with codon 61 mutations
was 11/46 and for C57BL/10J mice 4/31. Unlike the findings of Fox et al. (1990), Stanley etal.
(1994) reported an increase in the frequency of mutation in carcinomas, which was reported to be
twice that of adenomas in both strains of mice, indicating that stage of progression was related to
the number of mutations in those tumors, although most tumors induced by MCP did not have
this mutation.
In terms of liver tumor phenotype, Anna et al. (1994) reported that the H-ras codon 61
mutation frequency was not statistically different in liver tumors from DCA and TCE-treated
mice from a highly variable number of tumors examined. In regard to mutation spectra in H-ras
oncogenes in control or spontaneous tumors, the patterns were slightly different but mostly
similar to that of DCA-induced tumors (0.5% in drinking water). From their concurrent controls
they reported that H-ras codon 61 mutations in 17% (n = 6) of adenomas and 100% (n = 5) of
carcinomas. For historical controls (published and unpublished) they reported mutations in 73%
(n = 33) of adenomas and mutations in 70% (n = 30) of carcinomas. For tumors from TCE
treated animals they reported mutations in 35% (n = 40) of adenomas and 69% (n = 36) of
carcinomas, while for DCA treated animals they reported mutations in 54% (n = 24) of
adenomas and in 68% (n = 40) of carcinomas. Anna et al. (1994) reported more mutations in
TCE-induced carcinomas than adenomas.
The study of Ferreira-Gonzalez et al. (1995) in male B6C3 F1 mice has the advantage of
comparison of tumor phenotype at the same stage of progression (hepatocellular carcinoma), for
allowance of the full expression of a tumor response (i.e., 104 weeks), and an adequate number
of spontaneous control lesions for comparison with DCA or TCA treatments. However, tumor
phenotype at an endstage of tumor progression reflects of tumor progression and not earlier
stages of the disease process. In spontaneous liver carcinomas, 58% were reported to show
mutations in H-61 as compared with 50% of tumor from 3.5 g/L DCA-treated mice and 45% of
tumors from 4.5.g/L TCA-treated mice. Thus, there was a heterogeneous response for this
phenotypic marker for the spontaneous, DCA-, and TCA-treatment induced hepatocellular
carcinomas and not a pattern of reduced H-ras mutation reported for a number of peroxisome
proliferators.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-279 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
A number of peroxisome proliferators have been reported to have a much smaller
mutation frequency that spontaneous tumors (e.g., 13-24% H-ras codon 61 mutations after
Methylclofenopate depending on mouse strain, Stanley et al. (1994): 21 to 31% for Ciprofibrate-
induced tumors and from 64 to 66% for spontaneous tumors, Fox et al. (1990) and Hegi et al.
(1993).
Bull (2000) suggested that "the report by Anna et al. (1994) indicated that TCE-induced
tumors possessed a different mutation spectra in codon 61 of the H-ras oncogene than those
observed in spontaneous tumors of control mice." Bull (2000) stated that "results of this type
have been interpreted as suggesting that a chemical is acting by a mutagenic mechanism" but
went on to suggest that it is not possible to a priori rule out a role for selection in this process
and that differences in mutation frequency and spectra in this gene provide some insight into the
relative contribution of different metabolites to TCE-induced liver tumors. Bull (2000) noted
that data from Anna et al. (1994), Ferreira-Gonzalez et al. (1995), and Maronpot et al. (1995a)
indicated that mutation frequency in DCA-induced tumors did not differ significantly from that
observed in spontaneous tumors. Bull (2000) also noted that the mutation spectra found in DCA-
induced tumors has a striking similarity to that observed in TCE-induced tumors, and DCA-
induced tumors were significantly different than that of TCA-induced liver tumors.
Bull et al. (2002) reported that mutation frequency spectra for the H-ras codon 61 in
mouse liver "tumors" induced by TCE (n = 37 tumors examined) to be significantly different
than that for TCA (n = 41 tumors examined), with DCA-treated mice tumors giving an
intermediate result (n = 64 tumors examined). In this experiment, TCA-induced "tumors" were
reported to have more mutations in codon 61 (44%) than those from TCE (21%) and DC A
(33%>). This frequency of mutation in the H-ras codon 61 for TCA is the opposite pattern as that
observed for a number of peroxisome proliferators in which the number of mutations at H-ras 61
in tumors has been reported to be much lower than spontaneously arising tumors (see Section
E.3.4.1.5). Bull et al. (2002) noted that the mutation frequency for all TCE,TCA or DCA tumors
was lower in this experiment than for spontaneous tumors reported in other studies (they had too
few spontaneous tumors to analyze in this study), but that this study utilized lower doses and was
of shorter duration than that of Ferreira-Gonzalez et al. (1995). These are additional concerns in
addition to the effects of lesion grouping in which a lower stage of progression is grouped with
more advanced stages. In a limited subset of tumors that were both sequenced and characterized
histologically, only 8 of 34 (24%) TCE-induced adenomas but 9/15 (60%) of TCE-induced
carcinomas were reported to have mutated H-ras at codon 61, which the authors suggest is
evidence that this mutation is a late event.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-280 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Thus, in terms of H-ras mutation, the phenotype of TCE-induced tumors appears to be
more like DCA-induced tumors (which are consistent with spontaneous tumors), or those
resulting from a coexposure to both DCA and TCA (Bull et al., 2002), than from those induced
by TCA. As noted above, Bull et al. (2002) reported the mutation frequency spectra for the H-
ras codon 61 in mouse liver tumors induced by TCE to be significantly different than that for
TCA, with DCA-treated mice tumors giving an intermediate result and for TCA-induced tumors
to have a H-ras profile that is the opposite than those of a number of other peroxisome
proliferators. More importantly, these data suggest that using measures, other than dysplasticity
and tincture, mouse liver tumors induced by TCE are heterogeneous in phenotype.
With regard to tincture, Stauber and Bull (1997) reported the for male B6C3F1 mice,
DCA-induced "lesions" contained a number of smaller lesions that were heterogeneous and more
eosinophilic with larger "lesions" tending to less numerous and more basophilic. For TCA
results using this paradigm, the "lesions" were reported to be less numerous, more basophilic,
and larger than those induced by DCA.
Carter et al. (2003) used tissues from the DeAngelo et al. (1999) and examined the
heterogeneity of the DCA-induced lesions and the type and phenotype of preneoplastic and
neoplastic lesions pooled across all time points. Carter et al. (2003) examined the phenotype of
liver tumors induced by DCA in male B6C3 F1 mice and the shape of the dose-response curve
for insight into its MOA. They reported a dose-response of histopathologic changes (all classes
of premalignant lesions and carcinomas) occurring in the livers of mice from 0.05-3.5 g/L DCA
for 26-100 weeks and suggest foci and adenomas demonstrated neoplastic progression with time
at lower doses than observed DCA genotoxicity. Preneoplastic lesions were identified as
eosinophilic, basophilic and/or clear cell (grouped with clear cell and mixed cell) and dysplastic.
Altered foci were 50% eosinophilic with about 30% basophilic. As foci became larger
and evolved into carcinomas they became increasingly basophilic. The pattern held true through
out the exposure range. There was also a dose and length of exposure related increase in atypical
nuclei in "noninvolved" liver. Glycogen deposition was also reported to be dose-dependent with
periportal accumulation at the 0.5 g/L exposure level. Carter et al. (2003) suggested that size and
evolution into a more malignant state are associated with increasing basophilia, a conclusion
consistent with those of Bannasch (1996) and that there a greater periportal location of lesions
suggestive as the location from which they arose.
Consistent with the results of DeAngelo et al. (1999), Carter et al. (2003) reported that
DCA (0.05-3.5 g/L) increased the number of lesions per animal relative to animals receiving
distilled water, shortened the time to development of all classes of hepatic lesions, and that the
phenotype of the lesions were similar to those spontaneously arising in controls. Along with
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-281 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
basophilic and eosinophilic lesions or foci, Carter et al. (2003) concluded that DCA-induced
tumors also arose from isolated, highly dysplastic hepatocytes in male B6C3F1 mice chronically
exposed to DCA suggesting another direct neoplastic conversion pathway other than through
eosinophilic or basophilic foci.
Rather than male B6C3F1 mice, Pereira (1996) studied the dose-response relationship for
the carcinogenic activity of DCA and TCA and characterized their lesions (foci, adenomas and
carcinomas) by tincture in females (the generally less sensitive gender). Like the studies of TCE
by Maltoni et al. (1986), female mice were also reported to have increased liver tumors after
TCA and DCA exposures. Pereira (1996) pool lesions for phenotype analyses so the affect of
duration of exposure could not be determined nor adenomas separated from carcinomas for
"tumors."
However, as the concentration of DCA was decreased the number of foci was reported by
Pereira (1996) to be decreased but the phenotype of the foci to go from primarily eosinophilic
foci (i.e., -95% eosinophilic at 2.58 g/L DCA) to basophilic foci (-57% eosinophilic at 0.26
g/L). For TCA the number of foci was reported to -40 basophilic and -60 eosinophilic
regardless of dose. Spontaneously occurring foci were more basophilic by a ratio of 7/3. Pereira
(1996) described the foci of altered hepatocytes and tumors induced by DCA in female B6C3F1
mice to be eosinophilic at higher exposure levels but at lower or intermittent exposures to be half
eosinophilic and half basophilic. Regardless of exposure level, half of the TCA-induced foci
were reported to be half eosinophilic and half basophilic with tumors 75% basophilic. In control
female mice, the limited numbers of lesions were mostly basophilic, with most of the rest being
eosinophilic with the exception of a few mixed tumors. The limitations of descriptions tincture
and especially for inferences regarding peroxisome proliferator from the description of
"basophilia" is discussed in Section E.3.4.1.5.
The results appear to differ between male and female B6C3F1 mice in regard to tincture
for DCA and TCA at differing doses. What is apparent is that the tincture of the lesions is
dependent on the stage of tumor progression, agent (DCA or TCA), gender, and dose. Also what
is apparent from these studies is the both DCA and TCA are heterogeneous in their tinctoral
characteristics as well as phenotypic markers such as mutation spectra or expression of c-Jun.
The descriptions of TCE-induced tumors in mice reported by the NCI, NTP, and Maltoni
et al. studies are also consistent with phenotypic heterogeneity as well as consistency with
spontaneous tumor morphology (see Section E.3.4.1.5). As noted in Section E.3.1,
hepatocellular carcinomas observed in humans are also heterogeneous. For mice, Maltoni et al.
(1986) described malignant tumors of hepatic cells to be of different subhistotypes, and of
various degrees of malignancy and were reported to be unique or multiple, and have different
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-282 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
sizes (usually detected grossly at necropsy) from TCE exposure. In regard to phenotype, tumors
were described as usual type observed in Swiss and B6C3F1 mice, as well as in other mouse
strains, either untreated or treated with hepatocarcinogens and to frequently have medullary
(solid), trabecular, and pleomorphic (usually anaplastic) patterns.
For the NCI (1976) study, the mouse liver tumors were described in detail and to be
heterogeneous "as described in the literature" and similar in appearance to tumors generated by
carbon tetrachloride. The description of liver tumors in this study and tendency to metastasize to
the lung are similar to descriptions provided by Maltoni et al. (1986) for TCE-induced liver
tumors in mice via inhalation exposure.
The NTP (1990) study reported TCE exposure to be associated with increased incidence
of hepatocellular carcinoma (tumors with markedly abnormal cytology and architecture) in male
and female mice. Hepatocellular adenomas were described as circumscribed areas of distinctive
hepatic parenchymal cells with a perimeter of normal appearing parenchyma in which there were
areas that appeared to be undergoing compression from expansion of the tumor. Mitotic figures
were sparse or absent but the tumors lacked typical lobular organization. Hepatocellular
carcinomas were reported to have markedly abnormal cytology and architecture with
abnormalities in cytology cited as including increased cell size, decreased cell size, cytoplasmic
eosinophilia, cytoplasmic basophilia, cytoplasmic vacuolization, cytoplasmic hyaline bodies and
variations in nuclear appearance. Furthermore, in many instance several or all of the
abnormalities were reported to be present in different areas of the tumor and variations in
architecture with some of the hepatocellular carcinomas having areas of trabecular organization.
Mitosis was variable in amount and location. Therefore, the phenotype of tumors reported from
TCE exposure was heterogeneous in appearance between and within tumors from all 3 of these
studies.
Caldwell and Keshava (2006) reported
that Bannasch (2001) and Bannasch et al. (2001) describe the early phenotypes of
preneoplastic foci induced by many oncogenic agents (DNA-reactive chemicals,
radiation, viruses, transgenic oncogenes and local hyperinsulinism) as
insulinomimetic. These foci and tumors have been described by tincture as
eosinophilic and basophilic and to be heterogeneous. The tumors derived from
them after TCE exposure are consistent with the description for the main tumor
lines of development described by Bannasch et al. (2001) (see Section 3.4.1.5).
Thus, the response of liver to DC A (glycogenosis with emergence of glycogen
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-283 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
poor tumors) is similar to the progression of preneoplastic foci to tumors induced
from a variety of agents and conditions associated with increased cancer risk.
Furthermore Caldwell and Keshava (2006) noted that Bull et al. (2002) reported expression of
insulin receptor (IR) to be elevated in tumors of control mice or mice treated with TCE, TCA and
DCA but not in nontumor areas suggesting that this effect is not specific to DCA.
There is a body of literature that has focused on the effects of TCE and its metabolites
after rats or mice have been exposed to "mutagenic" agents to "initiate" hepatocarcinogenesis
and this is discussed in Section E.4.2, below. TCE and its metabolites were reported to affect
tumor incidence, multiplicity, and phenotype when given to mice as a coexposure with a variety
of "initiating" agents and with other carcinogens. Pereira and Phelps (1996) reported that MNU
alone induced basophilic foci and adenomas. MNU and low concentrations of DCA or TCA in
female mice were reported to induce heterogeneous for foci and tumor with a higher
concentration of DCA inducing more eosinophilic and a higher concentration of TCA inducing
more tumors that were basophilic. Pereira et al. (2001) reported that not only dose, but gender
also affected phenotype in mice that had already been exposed to MNU and were then exposed
to DCA. As for other phenotypic markers, Lantendresse and Pereira (1997) reported that
exposure to MNU and TCA or DCA induced tumors that had some commonalities, were
heterogeneous, but for female mice were overall different between DCA and TCA as
coexposures with MNU.
Stop experiments which attempt to ascertain the whether progression differences exist
between TCA and DCA have used higher concentrations at much lower durations of exposure.
A question arises as to whether the differences in results between those animals in which
treatment was suspended in comparison to those in which had not had been conducted so that full
expression of response had not been allowed rather than "progression" as well as the effects of
using large doses.
After 37 weeks of treatment and then a cessation of exposure for 15 weeks Bull et al.
(1990) reported that after 15 weeks of cessation of exposure, liver weight and percent liver/body
weight were reported to still be statistically significantly elevated after DCA or TCA treatment.
The authors partially attribute the remaining increases in liver weight to the continued presence
of hyperplastic nodules in the liver. In terms of liver tumor induction, the authors stated that
"statistical analysis of tumor incidence employed a general linear model ANOVA with contrasts
for linearity and deviations from linearity to determine if results from groups in which treatments
were discontinued after 37 weeks were lower than would have been predicted by the total dose
consumed."
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-284 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
The multiplicity of tumors observed in male mice exposed to DCA or TCA at 37 weeks
and then sacrificed at 52 weeks were reported by the authors to have a response in animals that
received DCA very close to that which would be predicted from the total dose consumed by
these animals. The response to TCA was reported by the authors to deviate significantly (p =
0.022) from the linear model predicted by the total dose consumed. Multiplicity of lesions per
mouse and not incidence was used as the measure. Most importantly the data used to predict the
dose response for "lesions" used a different methodology at 52 weeks than those at 37 weeks.
Not only were not all animal's lesions examined, but foci, adenomas, and carcinomas were
combined into one measure. Therefore, foci, of which a certain percentage have been commonly
shown to spontaneously regress with time, were included in the calculation of total "lesions."
Pereira and Phelps (1996) note that in MNU-treated mice that were then treated with
DCA, the yield of altered hepatocytes decreases as the tumor yields increase between 31 and 51
weeks of exposure suggesting progression of foci to adenomas. Initiated and noninitiated control
mice were reported to also have fewer foci/mouse with time. Because of differences in
methodology and the lack of discernment between foci, adenomas, and carcinomas for many of
the mice exposed for 52 weeks, it is difficult to compare differences in composition of the
"lesions" after cessation of exposure in the Bull et al. (1990) study.
For TCA treatment the number of animals examined for determination of which
"lesions" were foci, adenomas, and carcinomas was 11 out of the 19 mice with "lesions" at 52
weeks while all 4 mice with lesions after 37 weeks of exposure and 15 weeks of cessation were
examined. For DCA treatment the number of animals examined was only 10 out of 23 mice with
"lesions" at 52 weeks while all 7 mice with lesions after 37 weeks of exposure and 15 weeks of
cessation were examined. Most importantly, when lesions were examined microscopically they
did not all turn out to be preneoplastic or neoplastic. Two lesions appeared "to be histologically
normal" and one necrotic. Not only were a smaller number of animals examined for the
cessation exposure than continuous exposure but only the 2 g/L exposure levels of DCA and
TCA were studied for cessation. The number of animals bearing "lesions" at 37 and then 15
week cessation weeks was 7/11 (64%) while the number of animals bearing lesions at 52 weeks
was 23/24 (96%) after 2 g/L DCA exposure. For TCA the number of animals bearing lesions at
37 weeks and then 15 weeks cessation was 4/11 (35%) while the number of animals bearing
lesions at 52 weeks was 19/24 (80%). While suggesting that cessation of exposure diminished
the number of "lesions," conclusions regarding the identity and progression of those lesion with
continuous versus noncontinuous DCA and TCA treatment are tenuous.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-285 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E.2.6. Studies of Chloral Hydrate (CH)
Given that total oxidative metabolism appears to be highly correlated with TCE-induced
increases in liver weight in the mouse rather than merely the presence of TCA, other metabolites
are of interest as potential agents mediating the effects observed for TCE. Recently Caldwell
and Keshava provided a synopsis of the results of more recent studies involving CH (Caldwell
and Keshava, 2006). A large fraction of TCE oxidative metabolism appears to go through CH,
with subsequent metabolism to TCA and trichloroethanol (Chiu et al., 2006b). Merdink et al.
(2008) demonstrated that CH administered to humans can be extremely variable and complex in
its pharmacokinetic behavior with a peak plasma concentration of CH in plasma 40-50 times
higher than observed at the same time interval for other subjects. Studies of CH toxicity in
rodents are consistent with the general presumption that oxidative metabolites are important for
TCE-induced liver tumors, but whether CH and its metabolites are sufficient to explain all of
TCE liver tumorigenesis remains unclear, particularly because of uncertainties regarding how
DC A may be formed (Chiu et al., 2006b). Studies of CH may enable a comparison between
toxicity of TCE and CH and may help elucidate its role in TCE effects. As with other TCE
metabolites, the majority of the studies have focused on the mouse liver tumor response. For
rats, while the limited data suggests that there is less of a response than mice to CH, those studies
are limited in power or reporting.
Daniel et al. (1992) exposed adult male B6C3F1 (C57B1/6JC male mice bred to
C3Heb/Fej female mice) 28-day old mice to CH, 2-chloroacetaldehyde, or DC A in 2 different
phases (1 and II) with initial weights ranging from 9.4 to 13.6 g. The test compounds were
buffered and administered in drinking water for 30 and 60 weeks (n = 5 for interim sacrifice),
and for 104 weeks (n = 40). The concentration of CH was 1 g/L and for DC A 0.5 g/L and the
estimated doses of DC A were 85, 93, and 166 mg/kg/d for the DC A group I, DC A group II, and
CH exposed group, respectively. Microscopic examination of tissues was conducted for all
tissues for five animals of the CH groups with liver, kidneys, testes, and spleen, in addition to all
gross lesions, reported to be examined microscopically in all of the 104-week survivors.
The initial body weight for drinking water controls was reported to be 12.99 ± 3.04 g for
group I (n = 23) and 10.48 ± 1.70 for group II (n = 10). For DC A treated animals, initial body
weights were 13.44 ± 2.57 g for group I (// = 23) and 9.65 ± 2.72 g for group II (n = 10). For the
CH treated group the initial body weights were reported to be 10.42 ± 2.49 g (n = 40). It is not
clear from the report what control group best matched, if any, the CH group. Thus, the mean
initial body weights of the groups as well as the number of animals varied considerably in each
group (i.e., -40% difference in mean body weights at the beginning of the study).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-286 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The number of animals surviving till the termination of the experiment was 10, 10, 16, 8,
and 24 for the control group I, control group II, DC A group I, DC A group II, and CH groups,
respectively. An increase in absolute and relative liver weight versus reported to be observed at
30 weeks for DC A and CH groups and at 60 weeks for CH but data were not shown in the study.
At 104 weeks, the data for the surviving control groups were combined as was that for the 2
DC A treatment groups. Of note was that for CH treated survivors (n = 24), water consumption
was significantly reduced in comparison to controls. Absolute liver weight was reported to be
2.09 ± 0.6 g, 3.17 ± 1.3 g and 2.87 ± 1.1 g for control, DC A and CH treatment groups,
respectively. The % liver to body weight was reported to be similarly elevated (1 57-fold of
control for DC A and 1.41-fold of control for CH) at 104 weeks.
At 104 weeks the treatment-related liver lesions in histological sections were reported to
be most prominently hepatocytornegaly and vacuolization in DCA-treated animals. Cytomegaly
was also reported to be in 5, 92, and 79% of control, DC A and CH treatment groups,
respectively. Cytomegaly in CH treated mice was described as minimal and associated with an
increased number of basophilic granules (rough endoplasmic reticulum). Hepatocellular necrosis
and chronic active inflammation were reported to be mildly increased in both prevalence and
severity in all treated groups. The histological findings, from interim sacrifices (n = 5), were
considered by the authors to be unremarkable and were not reported.
Liver tumors were increased by DC A and CH treatment. The percent incidence of liver
carcinomas and adenomas combined in the surviving animals was 15, 75, and 71% in control,
DC A and CH treated mice, respectively. In the CH treated group, the incidence of hepatocellular
carcinoma was 46%. The number of tumors/animals was also significantly increased with CH
treatment. Most importantly, morphologically the authors noted that there did not appear to be
any discernable differences in the visual appearance of the DC A- and CH-induced tumors.
George et al. (2000) exposed male B6C3F1 mice and male F344/N rats to CH in drinking
water for 2 years (up to 162.6 mg/kg/d). Target drinking water concentrations were 0, 0.05, 0.5,
and 2 g/L CH in rats and 0, 0.05, 0.5 and 1.0 g/L CH in mice. Groups of animals (n = 6/group)
were sacrificed at 13 (rats only), 26, 52 and 78 weeks following the initiation of dosing with
terminal sacrifices at Week 104. A complete pathological examination was performed on 5 rats
and mice from the high-dose group, with examination primarily of gross lesions except for liver,
kidney, spleen and testes. BrdU incorporation was measured in the interim sacrifice groups in
rats and mice with PCO examined at 26 weeks in mice. In rats, the number of animals surviving
>78 weeks and examined for hepatocellular proliferative lesions was 42, 44, 44, and 42 for the
control, 7.4, 37.4 and 163.6 mg/kg/d CH treatment groups, respectively. Only 32, 36, 35, and
32 animals were examined at the final sacrifice time.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-287 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Only the lowest treatment group had increased liver tumors, which were marginally
significantly increased by treatment. The percent of animals with hepatocellular adenomas and
carcinomas was reported to be 2.4, 14.3, 2.3 and 6.8% in male rats. In mice, preneoplastic foci
and adenomas were reported to be increased in the livers of all CH treatment groups (13.5—146.6
mg/kg/d) at 104 weeks. The incidences of adenomas were reported to be statistically increased
at all dose levels, the incidences of carcinomas significantly increased at the highest dose, and
time-to-tumor decreased in all CH-treatment groups. The percent incidence of hepatocellular
adenomas was reported to be 21.4, 43.5, 5 1.3, and 50% in control, 13.5, 65.0, and 146.6 mg/kg
day treatment groups, respectively. The percent incidence of hepatocellular carcinomas was
reported to be 54.8, 54.3, 59.0, and 84.4% in these same groups. The resulting percent incidence
of hepatocellular adenomas and carcinomas was reported to be 64.3, 78.3, 79.5, and 90.6%.
The number of mice surviving >78 weeks was reported to be 42, 46, 39, and 32 and the
number surviving to final sacrifice to be 34, 42, 3 1, and 25 for control, 13.5, 65.0 and 146.56
mg/kg/d, respectively. CH exposure was reported to not alter serum chemistry, hepatocyte
proliferation (i.e., DNA synthesis), or hepatic PCO activity (an enzyme associated with
PPARa agonism) in rats and mice at any of the time periods monitored (all interim sacrifice
periods for BrdU incorporation, 52 or 78 weeks for serum enzymes, and 26 weeks for PCO) with
the exception of 0.58 g/L CH at 26 weeks slightly increasing hepatocyte labeling ( -2-3-fold
increase over controls) in rats and mice but the percent labeling still represented 3% or less of
hepatocytes.
With regard to other carcinogenic endpoints only five animals were examined at the high
dose, thereby limiting the study's power to determine an effect. Control mice were reported to
have a high spontaneous carcinoma rate (54%), thereby limiting the ability to detect a treatment-
related response. No descriptions of the foci or tumor phenotype were given. However, of note
is the lack of induction of PCO response with CH at 26 weeks of administration in either rats or
mice.
Leakey et al. (2003a) studied the effects of CH exposure (0, 25, 50, and 100 mg/kg,
5 days/week, 104-105 weeks via gavage) in male B6C3F1 mice with dietary control used to
manipulate body growth (n = 48 for 2 year study and n = 12 for the 15-month interim study).
Dietary control was reported to decrease background liver tumor rates (incidence of 15-20%)
and was reported to be associated with decreased variation in liver-to-body weight ratios, thereby
potentially increasing assay sensitivity. In dietary-controlled groups and groups fed ad libitum,
liver adenomas and carcinomas (combined) were reported to be increased with CH treatment.
With dietary restriction there was a more discernable CH tumor-response with overall tumor
incidence reduced, and time-to-tumor increased by dietary control in comparison to ad libitum
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-288 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
fed mice. Incidences of hepatocellular adenoma and carcinoma overall rates were reported to be
33, 52, 49, and 46% for control, 25, 50, and 100 mg/kg ad libitum-fed mice, respectively. For
dietary controlled mice the incidence rates were reported to be 22.9, 22.9, 29.2, and 37.5% for
controls, 25, 50, and 100 mg/kg CH, respectively. Body weights were matched and carefully
controlled in this study.
After 2 years of CH treatment the heart weights of ad libitum-fed male mice administered
100 mg/kg CH were reported to be significantly less and kidney weights of the 50 and 100
mg/kg less than vehicle controls. No other significant organ weight changes due to CH treatment
were reported to be observed in either diet group except for liver. The liver weights of CH
treated groups for by dietary groups were reported to be increased at 2 years and the absolute
liver weights of dosed groups to be generally increased at 15 months with percent liver/body
weight ratios increased in CH treated dietary-controlled mice at 15 months. There was 1.0-,
0.87-, and 1 08-fold of control percent liver/body weight for ad libitum fed mice exposed to 25,
50, and 100 mg/kg CH, respectively. For dietary controlled mice, there was 1.05-, 1.08-, and
1.11 -fold of control percent liver/body weight for the same dose groups at 15 months. Thus,
there was no corresponding dose-response for percent liver/body weight in the ad libitum-fed
mice, which were reported to show a much larger variation in 1 iver-to-body-vveight ratios (i.e.,
the standard deviation and standard errors were 2- to 17-fold lower in dietary controlled groups
than for ad libitum-fed groups).
Liver weight increases at 15-months did not correlate with 2-year tumor incidences with
this group. However, for dietary controlled groups the increase in percent liver/body weights at
15 months were generally correlated with increases in liver tumors at 2 years.
The incidences of peripheral or focal fatty change were reported to be increased in all
CH-treated groups of ad libitum-fed mice at 15 months (approximately half the animals showed
these changes for all dose groups, with no apparent dose-response). Of the enzymes associated
with PPARa agon ism (total CYP, CYP2B isoform, CYP4A, or 1 auric acid P-hydroxylase
activity), only CYP4A and 1 auric acid P-hydroxylase activity were significantly increased at
15 months of exposure in the dietary-restricted group administered 100 mg/kg CH with no other
groups reported showing a statistically significant increased response (n = 12/group). Although
not statistically significant, the 100 mg/kg CH exposure group of ad libitum-fed mice also had an
increase in CYP4A and 1 auric acid P-hydroxylase activity.
The authors reported that the increase in magnitude of CYP4A and 1 auric acid P-
hydroxylase activity at 100 mg/kg CH at 15 months in dietary controlled mice correlated with
the increase incidence of mice with tumors. However, there was no correlation of tumor
incidence and the increased enzyme activity associated with peroxisome proliferation in the ad
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-289 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
libitum-fed mice. No descriptions of liver pathology were given other than incidence of mice
with fatty liver changes. Hepatic malondialdehyde concentration in ad libitum fed and dietary
controlled mice did not change with CH exposure at 15 months but the dietary controlled groups
were all approximately half that of the ad libitum-fed mice. Thus, while overall increased
tumors observed in the ad libitum diet correlated with increased malondialdehyde concentration,
there was no association between CH dose and malondialdehyde induction for either diet.
Induction of peroxisome-associated enzyme activities was also reported for shorter times
of CH exposure. Seng et al. (2003) described CH toxicokinetics in mice at doses up to
1,000 mg/kg/d for 2 weeks with dietary control and caloric restriction slightly reducing acute
toxicity. Laurie acid P-hydroxylase and PCO activities were reported to be induced only at doses
>100 mg/kg in all groups, with dietary-restricted mice showing the greatest induction.
Differences in serum levels of TCA, the major metabolite remaining 24 hr after dosing, were
reported not to correlate with hepatic 1 auric acid P-hydroxylase activities across groups.
Leuschner and Beuscher (1998) examined the carcinogenic effects of CH in male and
female S-D rats (69-79 g, 25-29 days old at initiation of the experiment) administered 0, 15, 45,
and 135 mg/kg CH in unbuffered drinking water 7 days/week (n = 50/group) for 124 weeks in
males and 128 weeks in females. Two control groups were noted in the methods section without
explanation as to why they were conducted as two groups.
The mean survival for males was similar in treated and control groups with 20, 24, 20,
24, and 20% of Ccontrol I, Control II, 15, 45, and 135 mg/kg CH-treated groups, respectively,
surviving till the end of the study. For female rats, the percent survival was 12, 30, 24, 28, and
16% for of Control I, Control II, 15, 45, and 135 mg/kg CH-treated groups, respectively. The
authors reported no substance-related influence on organ weights and no macroscopic evidence
of tumors or lesions in male or female rats treated with CH for 124 or 128 weeks. However, no
data were presented on the incidence of tumors using this paradigm, especially background rates.
The authors reported a statistically significant increase in the incidence of hepatocellular
hypertrophy in male rats at the 135 mg/kg dose (14/50 animals vs. 4/50 and 7/50 in controls I
and II). For female rats, the incidence of hepatocellular hypertrophy was reported to be 10/50
rats (Control I) and 16/50 (Control II) rats with 18/50, 13/50 and 12/50 female rats having
hepatocellular hypertrophy after 15, 45, and 135 mg/kg CH, respectively. The lack or reporting
in regard to final body weights, histology, and especially background and treatment group data
for tumor incidences, limit the interpretation of this study. Whether this paradigm was sensitive
for induction of liver cancer cannot be determined.
From the CH studies in mice, there is an apparent increase in liver adenomas and
carcinomas induced by CH treatment by either drinking water or gavage with all available
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-290 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
studies performed in male B6C3F1 mice. However, the background levels of hepatocellular
adenomas and carcinomas in the mice in George et al. (2000) and body weight data from this
study show it is from a tumor prone mouse model.
Comparisons with concurrent studies of mice exposed to DC A revealed that while both
CH and DC A induced hepatomegaly and cytomegaly, DCA-induced cytomegaly was
accompanied by vacuolization while that of CH to be associated with increased number of
basophilic granules (rough endoplasmic reticulum) which would suggest separate effects.
However, the morphology of the CH-induced tumors was reported to be similar between DC A
and CH-induced tumors (Daniel et al., 1992).
Using a similar paradigm (2-year study of B6C3F1 male mice), De Angelo et al. (1999)
and Carter et al. (2003) described DCA-induced tumors to be heterogeneous. This is the same
description given for TCE-induced tumors in the studies by NTP, NCI, and Maltoni et al. and to
be a common description for tumors caused by a variety of carcinogenic agents. Similar to the
studies cited above for CH, De Angel o et al. (1999) reported that PCO levels were only elevated
at 26 weeks at 3.5 g/L DC A and had returned to control levels by 52 weeks. Similar to CH, no
increased tritiated thymidine was reported for DC A at 26 and 52 weeks with only 2-fold of
control values reported at 0.05 g/L at 4 weeks.
Leakey et al. (2003a) reported that ad libitum fed male mice exhibited a similar degree of
increased incidence of peripheral or focal fatty change at 15 months for all CH doses but not
enzymes associated with peroxisome proliferation. While dietary restriction seemed to have
decreased background levels of tumors and increased time-to-tumor, CH-gave a clear dose-
response in dietary restricted animals. However, while the overall level of tumor induction was
reduced there was a greater induction of PPARa enzymes by CH. Induction of liver tumors by
CH observed in ad libitum fed mice were not correlated with PPARa induction, with dietary
restriction alone appearing to have greater levels of 1 auric acid co-hydrolase activity in control
mice at 15 months. Seng et al. (2003) report that 1 auric acid |3-hydroxylase and PCO were
induced only at exposure levels >100 mg/kg CH, again with dietary restricted groups showing
the greatest induction. Such data argues against the role of peroxisome proliferation in CH-liver
tumor induction in mice.
E.2.7. Serum Bile Acid Assays
Serum bile acids (SBA) have been suggested as a sensitive indicator of hepatotoxicity to
a variety of halogenated solvents with an advantage of increased sensitivity and specificity over
conventional liver enzyme tests that primarily reflect the acute perturbation of hepatocyte
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-291 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
membrane integrity and "cell leakage" rather than liver functional capacity (i.e., uptake,
metabolism, storage, and excretion functions of the liver) (Bai et al., 1992a; Neghab et al., 1997).
While some studies have reported negative results, a number of studies have reported elevated
SB A in organic solvent-exposed workers in the absence of any alterations in normal liver
function tests. These variations in results have been suggested to arise from failure of some
methods to detect some of the more significantly elevated SB A and the short-lived and reversible
nature of the effect (Neghab et al., 1997).
Neghab et al. (1997) have reported that occupational exposure to 1,1,2-trichloro-1,2,2-
trifluoroethane and trichloroethylene has resulted in elevated SB A and that several studies have
reported elevated SB A in experimental animals to chlorinated solvents such as carbon
tetrachloride, chloroform, hexachlorobutadiene, tetrachloroethylene, 1,1,1-trichloroethane, and
trichloroethylene at levels that do not induce hepatotoxicity (Bai et al., 1992a; Hatndan and
Stacey, 1993; Wang and Stacey, 1990). Toluene, a nonhalogenated solvent, has also been
reported to increase SB A in the absence of changes in other hepatobiliary functions (Neghab and
Stacey, 1997). Thus, disturbance in S AB appears to be a generalized effect of exposure to
chlorinated solvents and nonchlorinated solvents and not specific to TCE exposure.
Neghab et al. (1997) reported that 8 hour time-weighted averages exposure to TCE of
8.9 ppm, measured in the breathing zone using a charcoal tube personal sampler for the whole
mean duration of exposure of 3 .4 years, to have not significant changes in albumin, bilirubin,
alkaline phosphatase, alanine aminotransferase, 5"-nucleosidase, y-glutamyltransferase, but to
have significantly increased total serum bile acids. Not only were total bile acids significantly
increased in these TCE-exposed workers compared to controls ( -2-fold of control), but,
specifically, deoxycholic acid and subtotal of free bile acids were increased. Neghab et al.
(1997) did not show the data, but also reported that "despite the apparent overall low level of
exposure, there was a very good correlations (r = 0.94) between the degree of increase in serum
concentration of total bile acids and level of TCE." Neghab et al. (1997) noted that while a
sensitive indicator or exposure to such solvents in asymptomatic workers, there is no indication
that actual liver injury occurs in conjunction with SAB increases.
Wang and Stacey (1990) administered TCE in corn oil via i.p. injection to male S-D rats
(300-500 g) at concentrations of 0.01, 0.1, 1, 5, and 10 mmol/kg on 3 consecutive days (n = 4, 5,
or 6) with liver enzymes and SB A examined 4 hours after the last TCE treatment. At these dose,
there were not differences between treated and control animals in regard to alkaline phosphatase
and sorbitol dehydrogenase concentrations, and an elevation of alanine aminotransferase only at
the highest dose. However, there was generally a reported dose-related increase in cholic acid,
chenodeoxycholic acid, deoxycholic acid, taurocholic acid, tauroursodeoxycholic acid, with
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-292 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
cholic acid and taurochlolic acid increased at the lowest dose. The authors reported that
"examination of liver sections under light microscopy yielded no consistent effects that could be
ascribed to trichloroethylene."
In the same study a rats were also exposed to TCE via inhalation (n = 4) at 200 ppm for
28 days, and 1,000 ppm for 6 hours/day. Using this paradigm, cholic acid and taurocholic acid
were significantly elevated at the 200 ppm level, (-10- and -5-fold of control, respectively) with
very large standard errors of the mean. At the 1,000 ppm level (6 hours, day) cholic acid and
taurocholic acid were elevated to -2-fold of control but neither was statistically significant. The
large variability in responses between rats and the low number of rats tested in this paradigm
limit its ability to determine quantitative differences between groups. Nevertheless, without the
complications associated with i.p. exposure (see Section E.2.2.1, above), both inhalation
exposure of TCE at a relative low exposure level was also associated with increased SB A levels.
The authors stated that "no increases in alanine amino transferase levels were observed in the
rats exposed to trichloroethylene via inhalation." No histopathology results were reported for
rats exposed via inhalation.
As stated by Wang and Stacey (1990), "intraperitoneal injection is not particularly
relevant to humans" which was the rationale given for the inhalation exposure experiments in the
study. They point out that intestinal interactions require consideration because a major
determinant of SB A is their absorption from the gut and intestinal flora may play a role in bile
acid metabolism. They also noted that grooming done by the experimental rats would probably
give small exposure via ingestion of TCE as well. However, Wang and Stacey (1990) reported
consistent results in terms of TCE-induced changes in SB A at relatively low concentrations by
either inhalation or i .p. routes of exposure that were not associated with other measures of
toxicity.
Ham dan and Stacey (1993) administered TCE in com oil (1 mmol/kg) in male Sprague
Davvley rats (300-400 g) and followed the time-course of SB A elevation, TCE concentration and
trichloroethanol in the blood at 2, 4, 8, and 16 hours after dosing (n = 4,5, or 6 per group). Liver
and blood concentration of TCE were reported to peak at 4 hours while those of trichloroethanol
peaked at 8 hours after dosing. TCE levels were not detectable by 16 hours in either blood or
liver while those of trichloroethanol were still elevated. Elevations of SB A were reported to
parallel those of TCE with cholic acid and taurochloate acid reported to show the highest levels
of bile acids. The dose given was based on that reported by Wang and Stacey (1990) to give no
hepatotoxicity but an increase in SB A. The authors stated that liver injury parameters were
checked and found unaffected by TCE exposure but do not show the data. Thus, it was TCE
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-293 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
concentration and not that of its metabolite that was most closely related to changes in SB A and
after a single exposure, the effect was reversible.
In an in vitro study by Bai and Stacey (1993), TCE was studied in isolated rat hepatocytes
with TCE reported to cause a dose-related suppression of initial rates of cholic acid and
taurocholic acid but with no significant effects on enzyme leakage and intracellular calcium
contents, further supporting a role for the parent compound in this effect. The authors noted that
the changes in SB A result from interference with a physiological process rather "than an event
associated with significant pathological consequences "
E.3. STATE OF SCIENCE OF LIVER CANCER MODES OF ACTION (MOAs)
The experimental evidence in mice shows that TCE and its metabolites induce foci,
hepatocellular adenomas, and carcinomas that are heterogeneous in nature as indicated by
phenotypic differences in tincture, mutational markers, or gene expression markers. The tumors
induced by TCE are reflective of phenotypes that are either similar to those induced by mixtures
of DC A and TCA exposure, or more like those induced by DC A. These tumors have been
described to be similar also to those arising spontaneously in mice or from chemically induced
hepatocarcinogenesis and to arise from preneoplastic foci, and in the case of DCA, single
dysplastic hepatocytes as well as foci. HCC observed in humans also has been described to be
heterogeneous and to be associated with formation of preneoplastic nodules. Although several
conditions have been associated with increased risk of liver cancer in humans, the mechanism of
HCC is unknown at this time. A great deal of attention has been focused on predicting which
cellular targets (e.g., "stem-cell" or mature hepatocyte) are associated with HCC as well as on
phenotypic markers in HCC that can provide insight not only into MOA and origin of tumor, but
also for prediction of clinical course. Examination of pathways and epigenetic changes
associated with cancer, and the relationship of these changes to liver cancer are also discussed
below.
The field of cancer research has been transformed by the recent discoveries of epigenetic
changes and their role in cancer and chronic disease states. The following discussion describes
these advances but also the issues involved with the technologies that have emerged to describe
them (see Section E.3.1.2, below). Exposure to TCE and its metabolites, like many others,
induces a heterogeneous response, even in a relatively homogeneous genetic paradigm as the
experimental laboratory rodent model. The importance of phenotypic anchoring is a major issue
in the study of any MOAs using these new technologies of gene expression pattern. Although a
large amount of information is now available using microarray technologies and transgenic
mouse models, specifically for TCE and in study of suggested MOAs for TCE and its
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-294 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
metabolites, use of these approaches has limitations that need to be considered in the
interpretation of data and conclusions derived from such data, especially quantitative
conclusions.
For TCE and its metabolites, the extent of acute to subchronic induction of hepatomegaly
correlated with hepatocellular carcinogenicity, although each had differing factors contributing
to that hepatomegaly from periportal glycogen deposition to hepatocellular hypertrophy and
increased polyploidy. The extent of transient DNA synthesis, peroxisome proliferation, or
cytotoxicity was not correlated with carcinogenicity. Hepatomegaly is also a predictor of
carcinogenicity for a number of other compounds in mice and rats. Allen et al. (2004) examined
the NTP database (87 compounds for rat and 83 for mice) and tried to correlate specific
hepatocellular pathology in prechronic studies with carcinogenic endpoints in the chronic 2-year
assays. The best single predictor of liver cancer in mice was hepatocellular hypertrophy.
Hepatocellular cytomegaly and hepatocyte necrosis also contributed, although the numbers of
positive findings were less than hypertrophy.
With regard to genotoxicity studies, there was no evidence of a correlation between
mouse liver tumor chemicals and Salmonella or micronucleus assay outcome. None of the
prechronic liver lesions examined were correlated with either Salmonella or Micronucleus
assays. In rats, no single prechronic liver lesions (when considered individually) was a strong
predictor of liver cancer in rats. The most predictive lesions was hepatocellular hypertrophy.
There was not significant correlation between liver turnors/toxicity and the 2 mutagenicity
measures.
Although the lack of correlation with the mutagenicity assays could be interpreted as
rodent assays predominantly identifying nongenotoxic liver carcinogens, this conclusion could
be questioned because it is solely dependent on Salmonella mutagenicity and additional
genotoxic endpoints could conceivably shift the association between liver cancer and
genotoxicity towards a more positive correlation. As to questions of the usefulness of the mouse
bioassay, the two mutagenicity assays did not correlate with rat results either and an important
indicator for carcinogenicity would be lost.
Examination of tumor phenotype from TCE, DCA and TCA exposures in mice shows a
large heterogeneity, which is also consistent with the heterogeneity observed in human HCC (see
Section E.3.1.8, below). The heterogeneity of tumor phenotype has been correlated with survival
outcome and tumor aggressiveness in humans and in transgenic mouse models that share some of
the same perturbations in gene pathway expression (see Sections E.3.1.8 and E.3.2.1, below).
An examination of common pathway disturbances that may be common to all cancers and those
of liver tumors shows that there are pathways in common, but that there is greater heterogeneity
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-295 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
in disturbance of hepatic pathways in cancer that may make is useful as a marker of disturbances
indicative of different targets of carcinogenicity depending on the cellular context and target.
Thus, although primate and human liver may not be as susceptible to HCC as the rodent liver,
the pathways leading to HCC in rodents and humans appear to be similar and heterogeneous,
with some indicative of other susceptible cellular targets for neoplasia in a differing context.
E.3.1. State of Science for Cancer and Specifically Human Liver Cancer
E.3 .1.1. Epigenetics and Disease States (Transgenerational Effects, Effects of Aging and
Background Changes)
Wood et al. (2007) published their work on "genomic landscapes" of human breast and
colorectal cancers that significantly forwards the understanding of "key events" involved with
induction of cancer. They state that there are -80 DNA mutations that alter amino acid in a
typical cancer but that examination of the overall distribution these mutations in different cancers
of the same type leads to a new view of cancer genome landscapes: they are composed of a
handful of commonly mutated genes "mountains" but are dominated by a much larger number of
infrequently mutated gene "hills."
Statistical analyses suggested that most of the ~ 80 mutation in an individual
tumor were harmless and that <15 were likely to be responsible for driving the
initiation, progression, or maintenance of the tumor.. .Historically the focus of
cancer research has been on the gene mountains, in part because they were the
only alterations that could be identified with available technologies. However,
our data show that vast majority of mutations in cancers do not occur in such
mountains. This new view of cancer is consistent with the idea that a large
number of mutations, each associated with a small fitness advantage, drive tumor
progression. It is the "hills" and not the "mountains" that dominate the cancer
genomic landscape.
The large number of "hills" actually reflects alterations in a much smaller number of cell
signaling pathways. Indeed, pathways rather than individual genes appear to govern the course
of tumorigenesis.
It is becoming increasingly clear that pathways rather than individual genes
govern the course of tumorigenesis. Mutations in any of several genes of a single
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-296 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
pathway can thereby cause equivalent increases in net cell proliferation.... This
new view of cancer is consistent with the idea that a large number of mutations,
each associated with a small fitness advantage, drive tumor progression.
Thus, when pathways are altered the same phenotype can arise from alterations in any of several
genes.
Consistent with the arguments put forth by Wood et al. (2007) for mutations in cancer is
the additional insight into pathway alterations by epigenomic mechanisms, which can act
similarly as mutation. Weidman et al. (2007) report that
cell phenotype is not only dependent on its genotype but also on its unique
epigenotype, which is shaped by developmental history and environmental
exposures. The human and mouse genome projects identified approximately
15,500 and 29,000 CpG islands, respectively. Hypermethylation of CpG-rich
regions of gene promoters inhibit expression by blocking the initiation of
transcription. DNA methylation is also involved in the allelic inactivation of
imprinted genes, the silencing of genes on the inactive X chromosome, and the
reduction of expression of transposable elements. Because epigenomic
modifications are copied after DNA synthesis by DNMT1, they are inherited
during somatic cell replication.. .Inherited and spontaneous or environmentally
induced epigenetic alterations are increasingly being recognized as early
molecular events in cancer formation. Furthermore, such epigenetic alterations
are potentially more adverse than nucleotide mutations because their effects on
regional chromatin structure can spread, thereby affecting multiple genetic loci.
Although tumor suppressor gene silencing by DNA methylation occurs frequently
in cancer, genome-wide hypomethylation is one of the earliest events to occur in
the genesis of cancer. Demethylation of the genome can lead to the reactivation
of transposable elements, thereby altering the transcription of adjacent genes, the
activation of oncogenes such as H-Ras, and biallelic expression of imprinted loci
(e.g., loss of IGF2 imprinting).
Thus, epigenetic modification may be worse than mutation in terms of cancer induction.
Dolinoy et al. (2007) report on the role of environmental exposures on the epigenome,
especially during critical periods of development and their role in adult disease susceptibility.
They report that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-297 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
aberrant epigenetic gene regulation has been proposed as a mechanism of action
for nongenotoxic carcinogenesis, imprinting disorders, and complex disorders
including Alzheimer's disease, schizophrenia, asthma, and autism. Epigenetic
modifications are inherited not only during mitosis but also can be transmitted
transgenerationally (Anway et al., 2005; Rakyan et al., 2002; Rakyan et al.,
2003)). The influence on environmental factors on epigenetic gene regulation
may also persist transgenerationally despite lack of continued exposure in second,
third, and fourth generations (Anway et al., 2005). Therefore if the genome is
compared to the hardware in a computer, the epigenome is the software that
directs the computer's operation.. .The epigenome is particularly susceptible to
deregulation during gestation, neonatal development, puberty and old age.
Nevertheless, it is most vulnerable to environmental factors during embryogenesis
because DNA synthetic rate is high, and the elaborate DNA methylation pattern
and chromatin structure required for normal tissue development is established
during early development... 83 imprinted genes have been identified in mice and
humans with 29 or about one third being imprinted in both species. Since
imprinted genes are functionally haploid, they are denied the protection from
recessive mutations that diploidy would normally afford. Imprinted genes that
have been linked to carcinogenesis include IGF2 (bladder, lung, ovarian and
others), IGF2R (breast, colon, lung, and others), and Neuronatin (pediatric
leukemia).
Bjornsson et al. (2008) recently reported that not only were there time-dependent changes
in global DNA methylation within the same individuals in 2 separate populations in widely
separated geographic locations, these changes showed familial clustering in both increased and
decreased methylation. These results were not only suggested to support the relationship of age-
related loss of normal epigenetic patterns as a mechanism for late onset of common human
diseases but also that losses and gains of DNA methylation observed over time in different
individuals could contribute to disease with the example provided of cancer which is associated
with both hypomethylation and hypermethylation through activation of oncogenes and silencing
of tumor suppressor genes. The study also showed considerable interindividual age variation,
with differences accruing over time within individuals that would be missed by studies that
employed group averaging.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-298 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The review by Reamone-Buettner and Borlak (2007) provide insight into the role of
noncoding RNAs in diseases such as cancer. They report that
a large number of noncoding RNAs (ncRNAs) play important role in regulating
gene expressions, and advances in the identification and function of eukaryotic
ncRNAs, e.g., microRNAs and their function in chromatin organization, gene
expression, disease etiology have been recently reviewed. The regulatory
pathways mediated by small RNAs are usually collectively referred to as RNA
interference (RNAi) or RNA-mediated silencing. RNAi can be triggered by small
double-stranded RNA (dsRNA) either introduced exogenously into cells as small
interfering siRNAs or that have been produced endogenously from small non-
coding RNAs known as microRNAs (miRNAs). The dsRNAs are
characteristically cleaved by the ribonuclease Ill-enzyme Dicer into 21- to 23 nt
duplexes and the resulting fragments base-pair with complementary mRNA to
target cleavage or to repress translation.. .Two mechanisms exist of miRNA-
mediated gene regulation, degradation of the target mRNA, and translational
repression. Whether one or the other of these mechanisms is used depends on the
degree of the complementary between the miRNA and target mRNA. For a near
perfect match, the Argonaute protein in the RNA-induced silencing complex
(RISC) cleaves the mRNA target, which is destined for subsequent degradation by
ribonucleases. In the situation of a less degree of complimentarity, commonly
occurring in humans, the translational repression mechanism is used to control
gene expression. However, the exact mechanism for translational inhibition is
unclear.
The varying degrees in complimentarity would help explain the large number of genes that could
be affected by miRNA and pleiotropic response.
The review by Feinberg et al. (2006) specifically addresses the epigenetic progenitor
origin of human cancer. They conclude that epigenetic alterations are ubiquitous and serve as
surrogate alterations for genetic change (oncogene activation, tumor-suppressor-gene silencing),
by mimicking the effect of genetic change. They report that:
Advances in characterizing epigenetic alterations in cancer include global
alterations, such as hypomethylation of DNA and hypoacetylation of chromatin,
as well as gene-specific hypomethylation and hypermethylation. Global DNA
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-299 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
hypomethylation leads to chromosomal instability and increased tumour
frequency, which has been shown in vitro and in vivo in mouse models, as well as
gene-specific oncogene activation, such as R-ras in gastric cancer, and cyclin D2
and maspin in pancreatic cancer. In addition, the silencing of tumour-suppressor
genes is associated with promoter DNA hypermethylation and chromatin
hypoacetylation, which affect divergent genes such as retinoblastoma 1 (RBI),
pl6 (also known as cyclin-dependent kinase inhibitor 2A (CDKN2A), von
Hippel-Lindau tumor suppressor (VHL), and MutL protein homologue (MLH1).
Genetic mechanisms are not the only path to gene disruption in cancer.
Pathological epigenetic changes - non-sequence-based alteration that are inherited
through cell division - are increasingly being considered as alternatives to
mutations and chromosomal alterations in disrupting gene function. These
include global DNA hypomethylation, hypermethylation and hypomethylation of
specific genes, chromatin alterations and loss of imprinting. All of these can lead
to aberrant activation of growth-promoting genes and aberrant silencing of
tumour-suppressor genes.
Most CG dinucleotides are methylated on cytosine residues in vertebrate
genomes. CG methylation is heritable, because after DNA replication the DNA
methyltransferase 1, DNMT1, methylates unmethylated CG on the base-paired
strand. CG dinucleotides within promoters within promoters tend to be protected
from methylation. Although individual genes vary in hypomethylation, all
tumours have shown global reduction of DNA methylation. This is a striking
feature of neoplasia.
In addition to global hypomethylation, promoters of individual genes show
increased DNA methylation levels. Hypermethylation of tumour-suppressor
genes can be tumour-type specific. An increasing number of genes are found to
be normally methylated at promoters but hypomethylated and activated in the
corresponding tumours. These include R-RAs in gastric cancer, melanoma
antigen family A, l(MAGEl) in melanoma, maspin in gastric cancer, S100A4 in
colon cancer, and various genes in pancreatic cancer.
Our genetic material is complexed with proteins in the form of histones in a one-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 00 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
to-one weight ratio. Core histones H2A, H2B, H3 and H4 form nucleosome
particles that package 147 bp of DNA, and the linker histone HI packages more
DNA between core particles, forming chromatin. It is chromatin and not just
DNA, that is the substrate for all processes that affect genes and chromosomes. In
recent years, it has become increasingly evident that chromatin, like DNA
methylation, can impart memory to genetic activity. There are dozens of post-
translational histone modifications. Studies in many model systems have shown
that particular histone modifications are enriched at sites of active chromatin
(histone H3 and H4 hyperacetylation, lysing at 4 and H3 (H3-K4) dimethylation
and trimethylation, and H3-K79 methylation) and others are enriched at sites of
silent chromatin (H3-K9 and H3-K27 methylation). These and other histone
modifications survive mitosis and have been implicated in chromatin memory.
Overproduction of key histone methyltransferases that catalyze the methylation of
either H3-K4 or H3-K27 residues are frequent events in neoplasia. Global
reductions in monoacetylated H4-K16 and trimethylated H4-K20 are general
features of cancer cells.
Genomic imprinting is parent-of -origin-specific gene silencing. It results from a
germ-line mark that causes reduced or absent expression of a specific allele of a
gene in somatic cells of the offspring. Imprinting is a feature of all mammals
affecting genes that regulate cell growth, behaviour, signaling, cell cycle and
transport; moreover, imprinting is necessary for normal development. Imprinting
is important in neoplasia because both gynogenotes (embryos derived only from
the maternal genetic complement) and androgenotes (embryos derived only from
the paternal genetic complement) form tumours - ovarian teratomas, and
hydtidiform moles/ choriocarcinomas, respectively. Loss of imprinting (LOI)
refers to activation of the normally silenced allele, or silencing of the normally
active allele, of an imprinted gene. LOI of the insulin-like growth factor 2 gene
(IGF2) accounts for half of Wilms tumours in children. LOI of IGF2 is also a
common epigenetic variant in adults and is associated with a fivefold increased
frequency of colorectal neoplasia. LOI of IGF2 might cause cancer by increasing
the progenitor cell population in the kidney in Wilm's tumor and in the
gastrointestinal tract in colorectal cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-301 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Feinberg et al. (2006) propose that epigenetic changes can provide mechanistic unity to
understanding cancer, they can occur earlier and set the stage for genetic alterations, and have
been linked to the pluripotent precursor cells from which cancers arise. "To integrate the idea of
these early epigenetic events, we propose that cancer arises in three steps; an epigenetic
disruption of progenitor cells, an initiating mutation and genetic and epigenetic plasticity."
The first step involves an epigenetic disruption of progenitor cells in a given
organ or system, which leads to a polyclonal precursor population of neoplasia-
ready cells. These cells represent a main target of environmental, genetic and
age-dependent exposure that largely accounts for the long latency period of
cancer. Epigenetic disruption might perturb the normal balance between
undifferentiated progenitor cells and differentiated committed cells within a given
anatomical compartment, either in number or in their capacity for aberrant
differentiation, which provides a common mechanism of neoplasia.
All tumours show global changes in DNA methylation, and DNA methylation is
clonally inherited through cell division. Because the conventional genetic
changes in cancer are also clonal, global hypomethylation would have to occur
universally, at the same moment as the mutational changes, which seems unlikely.
This suggests that global DNA hypomethylation (and global reductions of specific
histone modifications) precedes genetic change in cancer. Similarly,
hypermethylation of tumour-suppressor genes has been observed in the normal
tissue of patients in which the same gene is hypermethylated in the tumour tissue.
Recent data demonstrate LOI of IGF2 throughout the normal colonic epithelium
of patients who have LOI-associated colorectal cancer. LOI is associated with
increased risk of intestinal cancers in both humans and mice. A specific change
in the epithelium is seen in mice that are engineered to have biallelic expression
of IGF2 - a shift in the proportion of progenitor to differentiated cells throughout
the epithelium; a similar abnormality was observed in humans with LOI of IGF2.
The proposed existence of the epigenetically disrupted progenitors of cancer
implies that the earliest stages in neoplastic progression occur even before what a
pathologist would recognize as a benign pre-neoplastic lesion. Such alterations
are inherently polyclonal. This is in contrast with the widely accepted model of
cancer as a monoclonal disorder that arises from an initiating mutation- a model
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 02 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
that was proposed and accepted when little was known about epigenetic
phenomena in cancer.
Thus, Feinberg et al. (2006) provide a hypothesis for the latency period of cancer and
suggest that epigenetic changes predate mutational ones in cancer. Tissues that look
phenotypically "normal" may harbor epigenetic changes and predispositions toward neoplasia.
In regard to what cells may be targets or epigenetic changes that can be "progenitor cells" in the
case of cancer, Feinberg et al. (2006) define such cell having "capacity for self-renewal and
pluripotency - over their tendency toward limited replicative potential and differentiation."
Within the liver, there are multiple cell types that would fit such a definition including those who
are considered "mature" (see Section E.3.1.4, below). Feinberg et al. (2006) also note that
epigenetic states can be continuously modified to become heterogeneous at all states of the
neoplastic process.
Telomere erosion results in chromosome shortening and uncapped ends that begin
to fuse and the resulting dicentric chromosomes break at anaphase. DNA
palindromes have recently been found to form at high levels in cancer cells. Like
telomere erosion, DNA palindrome formation can lead to genetic instability by
initiating bridge-breakage-fusion cycles. However, it is not known how or
exactly when palindromes form, although they appear early in cancer progression.
Epigenetic instability can also promote cancer through pleiotropic alterations in
the expression of genes that modify chromatin.
Epigenetic changes are reversible but the changes can initiate irreversible genetic
changes. Permanent epigenetic changes can have an epigenetic basis. On a
background of cancer-associated epigenetic instability, the effects of mutations in
oncogenes and tumour -suppressor genes might be exacerbated. Therefore the
risk of developing malignancy would be much higher for a given mutations event
if it occurred on the background of epigenetic disruption.
The environmental dependence of cancer fits an epigenetic model generally for
human disease - the environment might influence disease onset not simply
through mutational mechanisms but in epigenetically modifying genes that are
targets for either germline or acquired mutation; that is, by allowing genetic
variates to be expressed. Little is known about epigenetic predispositions to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 03 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
cancer, but a recent twin study indicates that, similar to cancer risk, global
epigenetic changes show striking increase with age.
Environmental insults might affect the expression of tumour-progenitor genes,
leading to both genetic and epigenetic alterations. Liver regeneration after tissue
injury leads to widespread hypomethylation and hypermethylation of individual
genes; both of these epigenetic changes occur in cancer.
In regard to the implications of epigenomic changes and human susceptibility to toxic
insult, the review by Szyf (2007) provided additional insights.
The basic supposition in the field has been that the interindividual variations in
response to xenobiotic are defined by genetic differences and that the main hazard
anticipated at the genomic level from xenobiotic is mutagenesis or physical
damage to DNA. In accordance with this basic hypothesis, the main focus of
attention in pharmacogenetics has been on identifying polymorphisms in genes
encoding drug metabolizing enzymes and receptors. New xenobiotics were
traditionally tested for their genotoxic effects. However, it is becoming clear that
epigenetic programming plays an equally important role in generating
interindividual phenotypic differences, which could affect drug response.
Moreover, the emerging notion of the dynamic nature of the epigenome and its
responsibility to multiple cellular signaling pathways suggest that it is potentially
vulnerable to the effects of xenobiotics not only during critical period in
development but also later in life as well. Thus, non-genotoxic agents might
affect gene function through epigenetic mechanisms in a stable and long-term
fashion with consequences, which might be indistinguishable form the effects of
physical damage to the DNA. Epigenetic programming has the potential to
persist and even being transgenerationally transmitted (Anway et al., 2005) and
this possibility creates a special challenge for toxicological assessment of safety
of xenobiotics. Any analysis of interindividual phenotype diversity should
therefore take into account epigenetic variations in addition to genetic sequence
polymorphisms. Whereas, a germ-line polymorphism is a static property of an
individual and might be mapped in any tissue at any point in life, epigenetic
differences must be examined at different time points and at diverse cell types.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 04 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Karpinets and Foy (2005) proposed that epigenetic alterations precede mutations and that
succeeding mutations are not random but in response to specific types of epigenetic changes the
environment has encouraged. This mechanism was also suggested as to both explain the delayed
effects of toxicant exposure and the bystander effect of radiation on tumor development, which
are inconsistent with the accepted mechanism of direct DNA damage.
In a study of ionizing radiation, non-irradiated cells acquired mutagenesis through
direct contact with cells whose nuclei had previously been irradiated with alpha-
particles (Zhou et al., 2003). Molecular mechanisms underlying these
experimental findings are not known but it is believed that it may be a
consequence of bystander interactions involving intercellular signaling and
production of cytokines (Lorimore et al., 2003).
Caldwell and Keshava (2006) reported that
aberrant DNA methylation has emerged in recent years as a common hallmark of
all types of cancers with hypermethylation of the promoter region of specific
tumor suppressor genes and DNA repair genes leading to their silencing (an effect
similar to their mutation), and genomic hypomethylation (Ballestar and Esteller,
2002; Berger and Daxenbichler, 2002; Herman et al., 1998; Pereira et al., 2004b;
Rhee et al., 2002). Whether DNA methylation is a consequence or cause of cancer
is a long-standing issue(Ballestar and Esteller, 2002). Fraga et. al. (2005; 2004)
report global loss of monoacetylation and trimethylation of histone H4 as
common a hallmark of human tumor cells but suggest genomone-wide loss of 5-
methylcytosine (associated with the acquisition of a transformed phenotype) does
not exist as a static predefined value throughout the process of carcinogenesis but
as a dynamic parameter (i.e., decreases are seen early and become more marked in
later stages).
E.3.1.2. Emerging Technologies, DNA and siRNA, miRNA Microarrays—Promise and
Limitations for Modes of Action (MOAs)
Currently new approaches are emerging for the study of changes in gene expression and
protein production induced by chemical exposure that could be related to their toxicity and serve
as an anchor for determining similar patterns between rodent models and human diseases or risks
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-305 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of chemically-induced health impacts. Such approaches have the promise to extend the
definitions of "genotoxic" and "nongenotoxic" effects which with the advent of epigenomic
study have become obsolete as they assume only alteration of the DNA sequence is important in
cancer induction and progression. However, not only is phenotypic anchoring an issue in regard
to the differing cell types, regions, and lobes of the liver (see Section E.1.2, above), it is also an
issue for overall variability of response between animals and is critical for interpretation of
microarray and other genomic database approaches.
As shown in the discussions of TCE effects in animal models, TCE treatment resulted in
a large variability in response between what are supposed to be relatively homogeneous
genetically similar animals and there was an apparent difference in response between studies
using the same paradigm. It is important that as varying microarray approaches and analyses of
TCE toxicity or of potential MO As are published, the issue of phenotypic anchoring at the
cellular to animal level is addressed. Several studies of TCE microarray results and those of
PPARa agonists have been reported in the literature in an attempt to discern MO As. Issues
related to conduct of these experiments and interpretation of their results are listed below.
Perhaps one of the most important studies of this issue has been reported by Baker et al.
(2004). The ILSIHESI formed a hepatotoxicity working group to evaluate and compare
biological and gene expression responses in rats exposed to well-studied hepatotoxins (Clofibrate
and methapyrilene), using standard experimental protocol and to address the following issues: (a)
how comparable are the biological and gene expression data from different laboratories running
identical in vivo studies (b) how reproducible are the data generated across laboratories using the
same microarray platform (c) how do data compare using different microarray platforms; (d)
how do data compare using RNA from pooled and individual animals; (e) do the gene expression
changes demonstrate time- and dose-dependent responses that correlate with known biological
markers of toxicity? (Baker et al., 2004).
The rat model studied was the male S-D rat (57 or 60-66 days of age) exposed to 250 or
25 mg/kg/d Clofibrate for 1, 3 or 7 days. Two separate in vivo studies were conducted: one at
Abbott Laboratories and on at GlaxoSmithKline [GSK, in United Kingdom (UK)]. There was a
difference in biological response between the two laboratories. The high dose (250 mg/kg/d)
group at Day 3 had a 15% increase in liver weight relative to body weight in the GSK study,
compared with a 3% liver weight increase in the Abbott study. At 7-days, there was a 31% liver
weight increase in the GSK study and 15% in the Abbott study. Observed changes in clinical
chemistry parameters also indicated difference in the biological response of the in vivo study
concordant with difference in liver weight. A significant reduction in total cholesterol levels was
seen in the GSK study at the high dose for all time points. However, the Abbott study
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 06 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
demonstrated a significant reduction only at one dose and time point. The incidence of mitotic
figures also differed between the labs. In both studies there was a 2-3 times greater Acyl-CoA
enzyme (ACOX) activity at the high dose but no difference from control in the low dose. Again
the GSK lab gave greater response. For microarrays, GSK and ULR pooled samples from each
treatment group of four animals. U.S. EPA did some of the microarray analyses as well as GSK
and ULR (GSK in UK). It is apparent that although the changes in genes were demonstrated by
both laboratories, there were quantitative differences in the fold change values observed between
the two sites.
The U.S. EPA analyzed gene expression in individual RNA samples obtained from Day 7
high and low-dose animals that had been treated at Abbot. GSK (U.S.) and ULR analyzed gene
expression in pooled RNA from Day 7 high and low dose animals treated at GSK (UK). Gene
expression data from individual animal samples indicated that 7 genes were significantly
upregulated (maximum of 7.2-fold) and 12 were down regulated (maximum of 4.3-fold decrease)
in the high-dose group. The low-dose group generated only one statistically significant gene
expression change, namely heat shock protein 70 (HSP70). In comparison, expression changes
in the 7-day pooled high-dose samples analyzed by GSK (U.S.) ranged from 43.3-fold to a
3.5-fold decrease. Changes in these same samples analyzed by ULR ranged from a 4.9-fold
increase to a 4.3-fold decrease. As an example, the microarray fold change at 7-day 250 mg/kg/d
Clofibrate showed a 3.8-fold increase for U.S. EPA individual animals sampled, and 2.2-fold
increase for pooled samples by ULR, and a 20.3-fold increase in pooled samples by GSK (U.S.)
for CYP4A1 (Baker et al., 2004). Thus, these results show a very large difference not only
between treatment groups but between pooled an nonpooled data and between labs analyzing the
same RNA.
Not only was there a difference in DNA microarray results but a comparison of gene
expression data from Day 7 high-dose samples obtained using quantitative realtime PCR versus
data generated using cDNA microarrays has shown a quantitative difference but qualitative
similar patterns. Although both methods of quantitative real time PCR on the pooled sample
showed the PPARa gene to be down regulated, the GSK (U.S.) pooled sample microarray
analysis indicated upregulation; the URL pooled and U.S. EPA individual microarray analyses
showed no change. The microarray for PPARa at 7-day 250 mg/kg/d Clofibrate showed no
change for individual animals (U.S. EPA), no change for pooled samples (ULR) and
upregulation of 1.8-fold value for pooled samples for GSK(U.S.). The quantitative real time
PCR on the pooled sample using Taqman gave a 4.5-fold down regulation and using SYBR
Green gave a 1.2-fold down regulation of PPARa.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 07 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Baker et al. (2004) reported that the pooling of samples for microarray analysis has been
used in the past to defray the cost of microarray experiments, reduce the effect of biological
variation, and in some cases overcome availability of limiting amounts of tissues. Unfortunately
this approach essentially produced a sample size (n) of one animal. Repeated microarray
experiments with such pooled RNA produces technical replicates as opposed to true biological
replicates and thus, does not allow calculation of biologically significant changes in gene
expression between different dose groups or time points. Another possible consequence of
pooling is to mask individual gene changes and leave open the possibility of introducing error
due to individual outlier responses.
Woods et al. (2007b) note that
because toxicogenomics is a relatively novel technology, there are a number of
limitations that must be resolved before array data is widely accepted. Microarray
studies have been touted as being highly sensitive for detecting toxic responses at
much earlier time points and/or lower doses than histopathology, clinical
chemistry or other traditional toxicological assays can detect. However, based on
the nature of the assay, measurements of extreme levels of gene expression - low
or high -are thought to be unreliable. Also the reproducibility of microarray
experiments has raised concerns. "Batch effects" based on the day, user, and
laboratory environment have been observed in array datasets. To address these
concerns, confirmation of microarray-derived gene expression profiles is typically
performed using quantitative real time polymerase chain reaction (RT-PCR) or
Northern blot analysis.
In addition to the issues raised above, Waxman and Wurmbach (2007) raise issues
regarding how quantitative realtime PCR experiments are conducted. They state that cancer
development affects almost all pathways and genes including the "housekeeping" genes, which
are involved in the cell's common basic functions (e.g., glyceraldehyde-3-phosphate
dehydrogenase [GADPH], beta actin [ACTB], TATA-binding protein, ribosomal proteins, and
many more). However, "many of these genes are often used to normalize quantitative real-time
RT-PCR (qPCR) data to account for experimental differences, such as differences in RNA
quantity and quality, the overall transcriptional activity and differences in cDNA synthesis.
GADPH and ACTB are most commonly used for normalization, including studies of cancer."
Waxman and Wurmbach (2007) suggest that despite the fact that it has been shown that these
genes are differentially expressed in cancers, including colorectal-, prostate-, and bladder-cancer,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-308 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
some qPCR studies on hepatocellular carcinoma used GAPDH or ACTB for normalization.
Since many investigations on cancer include multiple comparisons, and analyze different stages
of the disease, such as normal tissue, preneoplasm, and consecutive stages of cancer, "it crucial
to find an appropriate gene for normalization" whose expression is constant throughout all
disease stage and not response to treatment.
For liver cancers associated with exposure to hepatitis C virus (HCV), Waxman and
Wurmbach (2007) reported that differing states, including preneoplastic lesions (cirrhosis and
dysplasia) and consecutive stages of hepatocellular carcinoma, had differential expression of
"housekeeping" genes and that using them for normalization had an effect on the fold change of
qPCR data and on the general direction (up or down) of differentially expressed genes. For
example, GAPDH was strongly upregulated in advanced and very advanced stages of
hepatocellular carcinoma (in some samples up to 7-fold) and ACTB was up-regulated 2- to
3-fold in many advanced and very advanced tumor samples. Waxman and Wurmbach (2007)
concluded that
microarray data are known to be highly variable. Due to its higher dynamic range
qPCR is thought to be more accurate and therefore is often used to corroborate
microarray results. Mostly, general direction (up and down-regulation) and rank
order of the fold-changes are similar, but the levels of the fold changes of
microarray experiments differ compared to qPCR data and show a marked
tendency of being smaller. This effect is more pronounced as the fold change is
very high.
In relation to use of gene expression and indicators of cancer causation, Volgelstein and
Kinzler (2004) made important points regarding their use:
Levels of gene expression are unreliable indicators of causation because
disturbance of any network invariably leads to a multitude of such changes only
peripherally related to the phenotype. Without better ways to determine whether
an unmutated but interesting candidate gene has a causal role in neoplasia, cancer
researchers will likely be spending precious time working on genes only
peripherally related to the disease they wish to study.
This is important caveat for gene expression studies for MOA that are "snapshots in time"
without phenotypic anchoring and even more applicable to experimental paradigms where there
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 09 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
is ongoing necrosis or toxicity in addition to gene changes that may or may not be associated
with neoplasia.
For an endpoint that is not as complex as neoplasia, there are issues regarding uses of
microarray data. In regard to the determination of acute liver toxicity caused by one of the most
studied hepatotoxins, acetaminophen, and its correlation with microarray data, Beyer et al.
(2007) also have reported the results of a landmark study examining issues regarding use of this
approach.
The biology of liver and other tissues in normal and disease states increasingly is
being probed using global approaches such as microarray transcriptional profiling.
Acceptance of this technology is based principally on a satisfactory level of
reproducibility of data among laboratories and across platforms. The issue of
reproducibility and reliability of genomics data obtained from similar
(standardized) biological experiments performed in different laboratories is
crucial to the generation and utility of large databases of microarray results.
While several recent studies uncovered important limitation of expression
profiling of chemical injury to cells and tissues (Baker et al., 2004; Beekman et
al., 2006; Ulrich et al., 2004), determining the effects of intralaboratory variables
on the reproducibility, validity, and general applicability of the results that are
generated by different laboratories and deposited into publicly available databases
remains a gap.. .The National Institutes of Environmental Health Sciences
(NIEHS) established the Toxicogenomics Research Consortium to apply the
collective and specialized expertise from academic institutions to address issues in
integrating gene expression profiling, bioinformatics, and general toxicology.
Key elements include developing standardized practices for gene expression
studies and conducting systematic assessments of the reproducibility of traditional
toxicity endpoints and microarray data within and among laboratories. To this
end the consortium selected the classical hepatotoxicant acetaminophen (APAP)
for its proof of concept experiments. Despite more than 30 years of research on
APAP, we are far from a complete understanding of the mechanisms of liver
injury, risk factors, and molecular markers that predict clinical outcome after
poisoning. APAP-induced hepatotoxicity was performed at seven geographically
dispersed Centers. Parallel studies with N-acetyl-m-aminophenol (AMAP), the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-310 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
non-hepatotoxic isomer of APAP, provided a method to isolate transcripts
associated with hepatotoxicity (Beyer et al., 2007).
Beyer et al. (2007) identified potential sources of interlaboratory variability when
microarray analyses were conducted by one laboratory on RNA samples generated in different
laboratories but using the same experimental paradigm and source of animals. Toxic injury by
APAP showed variability across Centers and between animals (e.g., percent liver affected by
necrosis [<20 to 80% at one time period and 0 to 60% at another], control animal serum ALT [3-
fold difference], and in glutathione depletion [<5 to >60%] between centers). There was
concordance between APAP toxicity as measured in individual animals (rather than expressed as
just a mean with SE) and transcriptional response. Of course the variability between gene
platforms and processing of the microarrays had been reduced by using the same facility to do all
of the microarray analyses. However, the results show that phenotypic anchoring of gene
expression data are required for biologically meaningful meta-analysis of genomic experiments.
Woods et al. (2007b) noted that
improvements should continue to be made on statistical analysis and presentation
of microarray data such that it is easy to interpret. Prior to the current advances in
bioinformatics, the most common way of reporting results of microarray studies
involved listing differentially expressed genes, with little information about the
statistical significance or biological pathways with which the genes are
associated.
However, there are issues with the use of "Classifiers" or predictive genomic computer programs
based on genes showing altered expression in association with the observed toxicities.
Although these metrics built on different machine learning algorithms could be
useful in estimating the severity of potential toxicities induced by compounds, the
applications of these classifiers in understanding the mechanisms of drug-induced
toxicity are not straightforward. In particular this approach is unlikely to
distinguish the upstream causal genes from the downstream responsive genes
among all the genes associated with an induced toxicity. Without knowledge of
the causal sufficiency order, designing experiments to test predicted toxicity in
animal models remains difficult" (Dai et al., 2007).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-311 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Ulrich (2003) stated that limitation of microarray analysis to study nuclear receptors (e.g.,
PPARa).
Nuclear receptors comprise a large group of ligand-activated transcription factors
that control much of cellular metabolism. Toxicogenomics is the study of the
structure and output of the entire genome as it related and responds to adverse
xenobiotic exposure. Traditionally, the genes regulated by nuclear receptors in
cells exposed to toxins have been explored at the mRNA and protein levels using
northern and western blotting techniques. Though effective when studying the
expression of individual genes, these approaches do not enable the understanding
of the myriad of genes regulated by individual receptors or of the crosstalk
between receptors.. .Discovery of the multiple genes regulated by each receptor
type has thus been driven by technological advances in gene expressional
analysis, most commonly including differential display, RT-PCR and DNA
microarrays., and in the development or receptor transgenic and knockout animal
models. There is much cross talk between receptors and many agonists interact
with multiple receptors. Off target effects cannot be predicted by target
specificity. Though RCR can affect transcription directly, much of its effects are
exerted through heterodimeric binging with other nuclear receptors (PXR, CAR,
PPARa, PPARy, FXR, LXR, TR) (Ulrich, 2003).
Another tool recent developed is gene silencing by introduction of siRNA. Dai et al.
(2007) noted issues involved in the siRNA to change gene expression for exploration of MO A
etc. to include the potential of off-target effects, incomplete knockdown, and nontargeting of
splice variants by the selected siRNA sequence. Using knockdown of PPARa in mice, Dai et al.
(2007) report "PPARa knockdown was variable between mice ranging from ~ 80 % knockdown
to little or no knockdown and that differing siRNAs gave different patterns of gene expression
with some grouped with PPARa -/- null mice but others grouped with expression patterns of
mice injected with control siRNA or Ringers buffer alone and showing no PPARa knockdown."
Dai et al. concluded that it is possible that it is the change in PPARa levels that is important for
perturbing expression of genes modulated by PPARa rather than the absolute levels of PPARa.
Not only is the finding of variability in knockdowns by siRNA technologies important
but the finding that level of PPAR is not necessarily correlated with function and that it could be
the change and not absolute level that matters in modulation in gene expression by PPARa is of
importance as well. How an animal responds to decreased PPARa function may also depend on
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-312 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
its gender. Dai et al. (2007) observed more dramatic phenotypes in female vs. male mice treated
with siRNA and noted that in aged PPARa -/- mice. Costet et al. (1998) have reported sexually
dimorphic phenotypes including obesity and increased serum triglyceride levels in females, and
steatosis and increased hepatic triglyceride levels in male PPARa-null mice. Ramdhan et al
(2010) , provided extensive date regarding lipid dysregulation in male PPARa-null mice and
humanized mice.
In regard to the emerging science and preliminary reports of the effects of microRNA as
oncogenes and tumor suppressors and of possible importance to hypothesized MO As for liver
cancer, the same caveats as described for DNA microarray analyses all apply along with
additional uncertainties. miRNAs repress their targeted mRNAs by complementary base pairing
and induction of the RNA interference pathway. Zhang et al. (2007) reported Northern blot
detection of gene expression at the mRNA level and its correlation with miRNA expression in
cancer cells as well as realtime PCR. These PCR-based analyses quantify miRNA precursors
and not the active mature miRNAs. However, they reported that the relationship between
pri-miRNA and mature miRNA expression has not been thoroughly addressed and is critical in
order to use real time PCR analysis to study the function of miRNAs in cancers. They go on to
state that
although Northern Blotting is a widely used method for miRNA analysis, it has
some limitations, such as unequal hybridization efficiency of individual probes
and difficulty in detecting multiple miRNAs simultaneously. For cancer studies,
it is important to be able to compare the expression pattern of all known miRNAs
between cancer cells and normal cells. Thus, it is better to have methods which
detect all miRNA expression at a single time.. .Although Northern blot analysis,
real-time PCR, and miRNA microarray can detect the expression of certain
miRNAs and determine which miRNAs may be associated with cancer formation,
it is difficult to determine whether or not miRNAs play a unique role in cancers.
Also these techniques cannot directly determine the correlation between mRNA
expression levels and whether the up-regulation or down-regulation of certain
miRNAs is the cause of cancer or a downstream effect of the disease.. .Many
miRNA genes have been found that are significantly overexpressed in different
cancers. All of them appear to function as oncogenes; however, only a few of
them have been well characterized.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-313 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Zhang et al. (2007) suggested that bioinformatic studies indicate that numerous genes are the
targets of miR-17-92: more than 600 for miR-19a and miR-20, two members of the miR-17-92
cluster.
Cho (2007) stated that
though more than 530 miRNAs have been identified in human, much remains to
be understood about their precise cellular function and role in the development of
diseases.. .Although each miRNA can control hundreds of target genes, it remains
a great challenge to identify the accurate miRNA targets for cancer research.
Thus, miRNAs have multiple targets so, like other transcription factors, may have pleotropic
effects that are cell, timing, and context specific.
Vogelstein and Kinzler (2004) stated "in the last decade many important gene responsible
for the genesis of various cancers have been discovered." Most importantly they and others
suggest that pathways rather than individual gene expression should be the focus of study. As a
specific example, Volgelstein and Kinzler noted
another example of the reason for focusing on pathways rather than individual
genes has been provided by studies of TP53 tumor-suppressor gene. The p53
protein is a transcription factor that normally inhibits cell growth and stimulates
cell death when induced by cellular stress. The most common way to disrupt the
p53 pathway is through a point mutation that inactivates its capacity to bind
specifically to its cognate recognition sequence. However, there are several other
ways to achieve the same effects, including amplification of the MDM2 gene and
infection with DNA tumor viruses whose products bind to p53 and functionally
inactivate it.
In regard to cellular anchoring for gene expression or pathway alterations associated with
cancer and the importance of "context" of gene expression changes, Vogelstein and Kinzler
(2004) gave several examples.
In solid tumors the important of the interactions between stroma and epithelium is
becoming increasingly recognized (e.g., the importance of the endothelial
cell).. .One might expect that a specific mutation of a widely expressed gene
would have identical or at least similar effects in different mammalian cell types.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-314 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
But this is not in general what is observed. Different effects of the same mutation
are not only found in distinct cell types; difference can even be observed in the
same cell types, depending on when the mutation occurred during the tumorigenic
process. The RAS gene mutations provide informative examples of these
complexities. KRAS2 gene mutation in normal pancreatic duct cells seem to
initiate the neoplastic process, eventually leading to the development of
pancreatic cancer. The same mutations occurring in normal colonic or ovarian
epithelial cells lead to self-limiting hyperplastic or borderline lesions that do not
progress to malignancy. In many human and experimental cancers, RAS genes
seem to function as oncogenes. But RAS genes can function as suppressor genes
under other circumstances, inhibiting tumorigenesis after administration of
carcinogens to mice. These and similar observation on other cancer genes are
consistent with the emerging notion that signaling molecules play multiple roles
at multiple time, even in the same cell type. However, the biochemical bases for
such variations among cancer cells are almost unknown.
In regard to the major pathways and mediators involved in cancer, several investigators
have reported a coherent set that are involved in many types of cancers. Vogelstein and Kinzler
(2004) noted that major pathways and mediators include p53, RB, WNT, E-cadherin, GL1,
APC, ERK, RAS:GTP, P13K,SMAD, RTK BAD, BAX, and H1F1. In regard to coherence and
site concordance between animal and human data, the disturbance of a pathway in one species
may result in the different expression of tumor pattern in another but both linked to a common
endpoint of cancer. Thus, pathways rather than a single mutation should be the focus of MO A
and cancer as several actions can be manifested by one pathway or change at one time that lead
to cancer.
Vogelstein and Kinzler (2004) also noted that pathways that are common to "cancer" are
also operative in liver cancer where, as a heterogeneous disease, multiple pathways have been
implicated in differing manifestations of this disease. Thus, liver cancer may be an example in
its multiple forms that are analogous to differing sites being affected by common pathways
leading to "cancer." Pathway concordance may not always show up as site concordance as
expression of cancer between species. Liver cancer may be the example where many pathways
can lead a cancer that is characterized by its heterogeneity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-315 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.3 .1.3. Etiology, Incidence and Risk Factors for Hepatocellular Carcinoma (HCC)
The review article of Farazi and DePinho (2006) provides an excellent summary of the
current state of human liver cancer in terms of etiology and incidence. The 5-year survival rate
of individuals with liver cancer in the United States is only 8.9% despite aggressive conventional
therapy with lethality of liver cancer due in part from its resistance to existing anticancer agents,
a lack of biomarkers that can detect surgically respectable incipient disease, and underlying liver
disease that limits the use of chemotherapeutic drugs. Chen et al. (2002) reported that surgical
resection is considered the only "curative treatment" but >80 of patients have widespread HCC at
the time of diagnosis and are not candidates for surgical treatment. Among patients with
localized HCC who undergo surgery, 50% suffer a recurrence. Primary liver cancer is the fifth
most common cancer worldwide and the third most common cause of cancer mortality. HCC
accounts for between 85 and 90% of primary liver cancers (El-Serag and Rudolph, 2007). Seitz
and Stickel (2006) report that epidemiological data from the year 2000 indicate that more than
560,000 new cases of HCC occurred worldwide, accounting for 5.6% of all human cancers and
that HCC is the fifth most common malignancy in men and the eighth in women.
Overall, incidence rates of HCC are higher in males compared to females. In almost all
populations, males have higher liver cancer rates than females, with male:female ratios usually
averaging between 2:1 and 4:1 and the largest discrepancies in rates (>4:1) found in medium-risk
European populations (El-Serag and Rudolph, 2007). Experiments showed a 2- to 8-fold of
control HCC development in male mice as well supporting the hypothesis that androgens
influence HCC progression rather than sex-specific exposure to risk factors (El-Serag and
Rudolph, 2007). El-Serag and Rudolph (2007) also reported that
in almost all areas, female rates peak in the age group 5 years older than the peak
age group for males. In low risk population (e.g., U.S.) the highest age-specific
rates occur among persons aged 75 and older. A similar pattern is seen among
most high-risk Asian populations. In contrast male rats in high-risk African
populations (e.g., Gambia) ten to peak between ages 60 and 65 before declining,
whereas female rates peak between 65 and 70 before declining.
Age adjusted incidence rates for HCC are extremely high in East and Southeast Asia and
in Africa but in Europe, there is a gradually decreasing prevalence from South to North. HCC
incidence rates also vary greatly among different populations living in the same region and vary
by race (e.g., for all ages and sexes in the United States, HCC rates are 2 times higher in Asian
than in African Americans, whose rates are 2 times higher than those in whites) ethnic variability
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-316 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
likely to include differences in the prevalence and acquisition time of major risk factors for liver
disease and HCC (El-Serag and Rudolph, 2007).
Worldwide HCC incidence rate doubled during the last two decades and younger age
groups are increasingly affected (El-Serag, 2004). The high prevalence of HCC in Asia and
Africa may be associated with widespread infection with hepatitis B virus (HBV) and HCV but
other risk factors include chronic alcohol misuse, non alcoholic fatty liver disease (NAFLD),
tobacco, oral contraceptives, and food contamination with aflatoxins (Seitz and Stickel, 2006).
El-Serag and Rudolph (2007) reported HCC to be the fastest growing cause of cancer-related
death in men in the United States with age-adjusted HCC incidence rates increasing more than 2-
fold between 1985 and 2002 and that, overall, 15-50% of HCC patients in the United States have
no established risk factors.
Although liver cirrhosis is present in a large portion of patients with HCC, it is not always
present. Fattovich et al. (2004) reported that
differences of geographic area, method of recruitment of the HCC cases (medical
or surgical) and the type of material studied (liver biopsy specimens, autopsy, or
partial hepatectomies) may account for the variable prevalence of HCC without
underlying cirrhosis (7% to 54%) quoted in a series of studies. Percutaneous liver
biopsy specimens are subject to sampling error. However, only a small
proportion of patients with HCC without cirrhosis have absolutely normal liver
histology, the majority of them showing a range of fibrosis intensity from no
fibrosis are all to septal and bridging fibrosis, necroinflammation, steatosis, and
liver cell dysplasia.
Farazi and DePinho (2006) noted that for diabetes, a higher indices of HCC has been
described in diabetic patients with no previous history of liver disease associated with other
factors. El-Serag and Rudolph (2007) reported that in their study of VA patients (173,643
patients with and 650,620 patients without diabetes), that HCC incidence doubled among
patients with diabetes and was higher among those with a longer follow-up of evaluation.
"Although most studies have been conducted in low HCC rate areas, diabetes also has been
found to be a significant risk factor in areas of high HCC incidence such as Japan. Taken
together, available data suggest that diabetes is a moderately strong risk factor for HCC."
NAFLD and nonalcoholic steatohepatitis contribute to the development of fibrosis and
cirrhosis and therefore, might also contribute to HCC development. The pathogenesis of
NAFLD includes the accumulation of fat in the liver which can lead to reactive oxygen species
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-317 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
in the liver with necrosis factor a (TNFa) elevated in NAFDL and alcoholic liver disease (Seitz
and Stickel, 2006). Abnormal liver enzymes not due to alcohol, viral hepatitis, or iron overload
are present in 2.8 to 5.5% of the United States general population and may be due to NAFLD in
66 to 90% of cases (Adams and Lindor, 2007). Primary NAFLD occurs most commonly and is
associated with insulin-resistant states, such as diabetes and obesity with other conditions
associated with insulin resistance, such as polycystic ovarian syndrome and hypopituitarism also
associated with NAFLD (Adams and Lindor, 2007). The steatotic liver appears to be susceptible
to further hepatotoxic insults, which may lead to hepatocyte injury, inflammation, and fibrosis,
but the mechanisms promoting progressive liver injury are not well defined (Adams and Lindor,
2007). Substrates derived from adipose tissue such as FFA, TNF-a, leptin, and adiponectin have
been implicated with oxidative stress appearing to be important leading to subsequent lipid
peroxidation, cytokine induction, and mitochondrial dysfunction. Liver disease was the third
leading cause of death among NAFLD patients compared to the 13th leading cause among the
general population, suggesting that liver-related mortality is responsible for a proportion of
increased mortality risk among NAFLD patients (Adams and Lindor, 2007).
The relative risk for HCC in type 2 diabetics has been reported to be approximately 4 and
increases to almost 10 for consumption of more than 80 g of alcohol per day (Hassan et al.,
2002). El-Serag and Rudolph (2007) reported that
it has been suggested that many cryptogenic cirrhosis and HCC cases represent
more severe forms of nonalcoholic fatty liver disease (NAFLD), namely
nonalcoholic steato hepatitis (NASH). Studies in the United States evaluating risk
factors for chronic liver disease or HCC have failed to identify HCV, HBV, or
heavy alcohol intake in a large proportion of patients (30-40%). Once cirrhosis
and HCC are established, it is difficult to identify pathologic features of NASH.
Several clinic-based controlled studies have indicated that HCC patients with
cryptogenic cirrhosis tend to have clinical and demographic features suggestive of
NASH (predominance of women, diabetes, and obesity) as compared with age-
and sex-matched HCC patients of well defined vial or alcoholic etiology. The
most compelling evidence for an association between NASH and HCC is indirect
and come from studies examining HCC risk with 2 conditions strongly associated
with NASH: obesity and diabetes. In a large prospective cohort in the US,
followed up for 16 years, liver cancer mortality rates were 5 times greater among
men with the greatest baseline body mass index (range 35-40) compared with
those with a normal body mass index. In the same study, the risk of liver cancer
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-318 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
was not as increase in women, with a relative risk of 1.68. Two other population-
based cohort studies from Sweden and Denmark found excess HCC risk
(increased 2- to 3-fold) in obese men and women compared with those with a
normal body mass index.. .Finally, liver disease occurs more frequently in those
with more severe metabolic disturbances, with insulin resistance itself shown to
increase as the disease progresses. Several developed countries most notably the
United States, are in the midst of a burgeoning obesity epidemic. Although the
evidence linking obesity to HCC is relatively scant, even small increase in risk
related to obesity could translate into a large number of HCC cases.
Thus, even a small increase in risk related to obesity could result in a large number of HCC cases
and the latency of HCC may make detection of increased HCC risk not detectable for several
years.
Other factors are involved as not every cirrhotic liver progresses to HCC. Seitz and
Stickel (2006) suggested that 90 to 100% of those who drink heavily suffer from alcoholic fatty
liver, 10-35% of those evolve to alcoholic steatohepatitis, 8-20% of those evolve to alcoholic
cirrhosis, and 1-2% of those develop HCC. HCV infects approximately 170 million individuals
worldwide with approximately 20% of chronic HCV cases developing liver cirrhosis and 2.5%
developing HCC.
Infection with HBV, a noncytopathic, partially double stranded hepatotropic DNA virus
classified as a member of the hepadnaviridae family, is also associated with liver cancer risk with
several lines of evidence supporting the direct involvement of HBV in the transformation process
(Farazi and Depinho, 2006). El-Serag and Rudolph (2007) suggested that
Epidemiologic research has shown that the great majority of adult-onset HCC
cases are sporadic and that many have at lease 1 established non-genetic risk
factor such as alcohol abuse or chronic HCV or HBV infection. However, most
people with these known environmental risk factors never develop cirrhosis or
HCC, whereas a sizable minority of HCC case develop among individuals without
any known risk factors... Genetic epidemiology studies in HCC, similar to several
other conditions, have fallen short of early expectations that they rapidly and
unequivocally would result in identification of genetic variants conveying
substantial excess risk of disease and thereby establish the groundwork for
effective genetic screening for primary prevention.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-319 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.3.1.4. Issues Associated with Target Cell Identification
Another outstanding and important question in HCC pathogenesis involves the cellular
origin of this cancer. The liver is made up of a number of cell types showing different
phenotypes and levels of differentiation. Which cell types are targets of hepatocarcinogens and
are those responsible for human HCC is a matter of intense debate. Studies over the last decade
provide evidence of several types of cells in the liver that can repopulate the hepatocyte
compartment after a toxic insult. "Indeed, although the existence of a liver stem cell is often
debated, most experts agree that progenitor liver cells are activated, in response to significant
exposure to hepatotoxins. Also, progenitor cells derived from nonhepatic sources, such as bone
marrow and pancreas, have been demonstrated recently to be capable of differentiating into
mature hepatocytes under correct microenvironmental conditions" (Gandillet et al., 2003).
At present, analyses of human HCCs for oval cell markers, comparison of their gene-
expression patterns with rat fetal hepatoblasts and the cellular characteristics of HCC from
various animal models have provided contrasting results about the cellular origin of HCC and
imply dual origins from either oval cells or mature hepatocytes. The failure to identify a clear
cell of origin for HCC might stem from the fact that there are multiple cells of origin, perhaps
reflecting the developmental plasticity of the hepatocyte lineage. The resolution of the HCC cell
of origin issue could affect the development of useful preventative strategies to target nascent
neoplasms, foster an understanding of how HCC-relevant genetic lesions function in that specific
cell-development context and increase our ability to develop more accurate mouse models in
which key genetic events are targeted to the appropriate cellular compartment (Farazi and
Depinho, 2006). Two reviews by Librecht (2006) and Wu and Chen (2006) provide excellent
summaries of the issues involved in identifying the target cell for HCC and the review by
Roskams et al. (2004) provided a current view of the "oval cell" its location and human
equivalent. Recent reports by Best and Coleman (2007) suggest another type of liver cell is also
capable of proliferation and differentiating into small hepatocytes (i.e., small hepatocyte-like
progenitor cell).
The review by Librecht (2006) provides an excellent description of the controversy and
data supporting different views of the cells of origin for HCC.
In recent years, the results of several studies suggest that human liver tumors can
be derived from hepatic progenitor cells rather than from mature cell types. The
available data indeed strongly suggest that most combined hepatocellular-
cholangiocarcinomas arise from hepatic progenitor cells (HPCs) that retained
their potential to differentiate into the hepatocyte and biliary lineages. Hepatic
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-320 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
progenitor cells could also be the basis for some hepatocellular carcinomas and
hepatocellular adenomas, although it is very difficult to determine the origin of an
individual hepatocellular carcinoma. There is currently not enough data to make
statements regarding a hepatic progenitor cell origin of cholangiocarcinoma. The
presence of hepatic progenitor cell markers and the presence and extent of the
cholangiocellular component are factors that are related the prognosis of
hepatocellular carcinomas and combined hepatocellular-cholangiocarcinomas,
respectively... The traditional view that adult human liver tumors arise from
mature cell types has been challenged in recent decades.. .HPCs are small
epithelial cells with an oval nucleus, scant cytoplasm and location in the bile
ductules and canals of Hering. HPCs can differentiate towards the biliary and
hepatocytic lineages. Differentiation towards the biliary lineage occurs via
formation of reactive bile ductules, which are anastamosing ductules lined by
immature biliary cells with a relatively large and oval nucleus surrounded by a
small rim of cytoplasm. Hepatocyte differentiation leads to the formation of
intermediate hepatocyte-like cells, which are defined as polygonal cells with a
size intermediate between than of HPCs and hepatocytes. In most liver diseases,
hepatic progenitor cells are "activated" which means that they proliferate and
differentiate towards the hepatocytic and/or biliary lineages. The extent of
activation is correlated with disease severity.. HPCs and their immediate biliary
and hepatocytic progeny not only have a distinct morphology, but they also
express several markers, with many also present in bile duct epithelial cells.
Immunohistochemistry using antibodies against these markers facilitates the
detection of HPCs. The most commonly used markers are cytokeratin (CK) 19
and CK7.. The proposal that a human hepatocellular carcinoma does not
necessarily arise from mature hepatocyte, but could have HPC origin, has
classically been based on three different observations. Each of them, however,
gives only indirect evidence that can be disputed.. .Firstly, it has been shown that
HPCs are the cells of origin of HCC in some animal models of
hepatocarcinogenesis, which has led to the suggestion that this might also be the
case in humans. However, in other animal models, the HCCs arise from mature
hepatocytes and not from HPCs or reactive bile ductular cells (Bralet et al 2002;
Lin et al 1995- DEN treated rats). Since it is currently insufficiently clear which
of these animal models accurately mimics human hepatocarcinogenesis, one
should be careful about extrapolating data regarding HPC origin of HCC in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-321 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
animal models to the human situation... Secondly, liver diseases that are
characterized by the presence of carcinogens and development of dysplastic
lesions also show HPC activation. Therefore, the suggestion has been made that
HPCs form a "target population" for carcinogens, but this is only a theoretical
possibility not supported by experimental data.. .Thirdly, several studies have
shown that a considerable proportion of HCCs express one or more HPC markers
that are not present in normal mature hepatocytes. Due to the fact that most HPC
markers are also expressed in the biliary lineage, the term "biliary marker" has
been used in some of these studies. The "maturation arrest" hypothesis states that
genetic alterations occurring in a HPC, or its immediate progeny, cause aberrant
proliferation and prevent its normal differentiation. Further accumulation of
genetic alterations eventually leads to malignant transformation of these
incompletely differentiated cells. The resulting HCC expresses HPC markers as
evidence of its origin. However, expression of HPC markers can also be
interpreted in the setting of the "dedifferentiation" hypothesis, which suggests that
the expression of HPC markers is acquired during tumor progression as a
consequence of accumulating mutations. For example, experiments in which
human HCC cells lines were transplanted into nude mice have nicely shown that
the expression of HPC marker, CK19, steadily increased when the tumors became
increasingly aggressive and metastasized to the lung, Thus, the expression of
CK19 in a HCC does not necessarily mean that the tumor has a HPC origin, but it
can also be mutation-induced, acquired expression associated with tumor
progression. Both possibilities are not mutually exclusive. For an individual
HCC that expresses a HPC marker, it remains impossible to determine whether
this marker reflects the cellular origin and/or is caused by tumor progression.
This can only be elucidated by determining whether HCC contains cells that are
ultrastructurally identical to HPCs in nontumor liver.
Similarly, the review by Wu and Chen (2006) also presents a valuable analysis of these
issues and stated:
The question of whether hepatocellular carcinomas arises from the differentiation
block of stem cells or dedifferentiation of mature cells remains controversial.
Cellular events during hepatocarcinogenesis illustrate that HCC may arise for
cells at various stages of differentiation in the hepatic stem cell lineage.. .The role
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-322 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of cancer stem cells has been demonstrated for some cancers, such as cancer of
the hematopoietic system, breast and brain. The clear similarities between normal
stem cell and cancer stem cell genetic programs are the basis of the a proposal
that some cancer stem cells could derived form human adult stem cells. Adult
mesenchymal stem cells (MSC) may be targets for malignant transformation and
undergo spontaneous transformation following long-term in vitro culture,
supporting the hypothesis of cancer stem cell origin. Stem cells are not only units
of biological organization, responsible for the development and the regeneration
of tissue and organ systems, but are also targets of carcinogenesis. However, the
origin of the cancer stem cell remains elusive.. .Three levels of cells that can
respond to liver tissue renewal or damage have been proved (1) mature liver cells,
as "unipotential stem cells," which proliferate under normal liver tissue renewal
and respond rapidly to liver injury, (2) oval cells, as bipotential stem cells, which
are activated to proliferate when the liver damage is extensive and chronic or if
proliferation of hepatocytes is inhibited; and (3) bone marrow stem cells, as
multipotent liver stem cells, which have a very long proliferation potential. There
are two major nonexclusive hypotheses of the cellular origin of cancer; from stem
cells due to maturation arrest or from dedifferentiation of mature cells. Research
on hepatic stem cells in hepatocarcinogenesis has entered a new era of
controversy, excitement and great expectations.. .The two major hypotheses about
the cellular origination of HCC have been discussed for almost 20 years. Debate
has centered on whether or not HCC originates from the differentiation block of
stem cells or dedifferentiation of mature cells. Recent research suggests that HCC
may originate from the transdifferentiation of bone marrow cells. In fact, there
might be more than one type of carcinogen target cell. The argument about the
origination of HCC becomes much clearer when viewed from this viewpoint:
poorly differentiated HCC originate from bone marrow stem cells and oval cells,
while well-differentiated HCC originates form mature hepatocytes.. .The cellular
events during hepatocarcinogenesis illustrate that HCC may arise from cells at
various stages of differentiation in the hepatocyte lineage. There are four levels
of cells in the hepatic stem cell lineage: bone marrow cell, hepato-pancreas stem
cell, oval cell and hepatocyte. HSC and the liver are known to have a close
relationship in early development. Bone marrow stem cells could differentiate
into oval cells, which could differentiate into heptatocytes and duct cells. The
development of pancreatic and liver buds in embryogenesis suggests the existence
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-323 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of a common progenitor cells to both the pancreas and liver. All of the four levels
of cells in the stem cell lineage may be targets of hepatocarcinogenesis.
Along with the cell types described as possible targets and participants in HCC, Best and
Coleman (2007) described yet another type of cell in the liver that can respond to hepatocellular
injury, which they term small hepatocyte-like progenitor cells and conclude that they are not the
progeny of oval cells, but represent a distinct liver progenitor cell population. Another potential
regenerative cell is the small hepatocyte-like progenitor cell (SHPC). SHPCs share some
phenotypes with hepatocytes, fetal hepatoblasts, and oval cells, but are phenotypically distinct.
They express markers such as albumin, transferring, and alpha-fetoprotein (AFP) and possess
bile canaliculi and store glycogen.
A recent review by Roskams et al. (2004) provided a current view of the "oval cell" its
location and human equivalent. They concluded that
while similarities exist between the progenitor cell compartment of human and
rodent livers, the different rodent models are not entirely comparable with the
human situation, and use of the same term has created confusion as to what
characteristics may be expected in the human ductular reaction. For example, a
defining feature of oval cells in many rodent models of injury is production of
alpha-fetoprotein, whereas ductular reactions in humans rarely display such
expression. Therefore we suggest that the "oval cell" and "oval -like cell" no
longer be used in description of human liver.
In the chronic hepatitis and cancer model of Vig et al. (2006) it is not the oval cells or
SHPCs that are proliferating but the mature hepatocytes, thus, supporting theories that it is not
only oval cells that are causing proliferations leading to cancer. Vig et al. (2006) also reported
that studies in mice an humans indicate that oval cells also may give rise to liver tumors and that
oval cells commonly surround and penetrate human liver tumors, including those caused by
hepatitis B. Tarsetti et al. (1993) suggested that although some studies have suggested that oval
cells are directly involved in the formation of HCC, others assert that HCC originates from
preneoplastic foci and nodules derived from hepatocytes and report that HCC evolved in their
model of liver damage from hepatocytes, presumably hepatocellular nodules, and not from oval
cells. They also suggested that proliferation alone may not lead to cancer. Recent studies that
follow the progression of hepatocellular nodules to HCC in humans (see Section E.3.2.4, below)
suggest an evolution from nodule to tumor.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-324 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
E.3 .1.5. Status of Mechanism of Action for Human Hepatocellular Carcinoma (HCC)
The underlying molecular mechanisms leading to hepatocarcinogenesis remain largely
unclear (Yeh et al., 2007). Although HCC is multistep, and its appearance in children suggest a
genetic predisposition exists, the inability to identify most of the predisposing genes and how
their altered expression relates to histological lesions that are the direct precursors to HCC, has
made it difficult to identify the rate limiting steps in hepatocarcinogenesis (Feitelson et al.,
2002). Calvisi et al. (2007) report that although the major etiological agents have been
identified, the molecular pathogenesis of HCC remains unclear and that while deregulation of a
number of oncogenes (e.g., c-Myc, cyclin D1 and fi-catenin and tumor suppressor genes
including P16INK4A, P53, E-cadherin, DLC-1, andpRb) have been observed at different
frequencies in HCC, the specific genes and the molecular pathways that play pivotal roles in
liver tumor development have not been identified. Indeed rather than simple patterns of
mutations, pathways that are common to cancer have been identified through study of tumors
and through transgenic mouse models. Branda and Wands (2006) stated that the molecular
factors and interactions involved in hepatocarcinogenesis are still poorly understood but are
particularly true with respect to genomic mutations, "as it has been difficult to identify common
genetic changes in more than 20% to 30% of tumors." As well as phenotypically heterogeneous,
"it is becoming clear that HCCs are genetically heterogeneous tumors." The descriptions of
heterogeneity of tumors and of pathway disruptions common to cancer are also shown for liver
tumors (see Sections E.3.1.6 and E.3.1.8, below). However, many of these studies focused on
the end process and of examination of the genomic phenotype of the tumor for inferences
regarding clinical course, aggressiveness of tumor, and consistency with other forms of cancer.
As stated above, the events that produce these tumors from patients with conditions that put them
at risk, are not known.
El-Serag and Rudolph (2007) suggested that risk of HCC increases at the cirrhosis stage
when liver cell proliferation is decreased and that acceleration of carcinogenesis at this stage may
result from telomere shortening (resulting in limitations of regenerative reserve and induction of
chromosomal instability), impaired hepatocyte proliferation (resulting in cancer induction by loss
of replicative competition), and altered milieu conditions that promote tumor cell proliferation.
When telomeres reach a critically short length, chromosome uncapping induces
DNA damage signals, cell-cycle arrest, senescence, or apoptosis. Telomeres are
critically short in human HCC and on the single cell level telomere shortening
correlated with increasing aneuploidy in human HCC.. .Chemicals inhibiting
hepatocyte proliferation accelerate carcinogen-induced liver tumor formation in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-325 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
rats as well as the expansion and transformation of transplanted hepatocytes. It is
conceivable that abnormally proliferating hepatocytes would not expand in
healthy regenerating liver but would expand quickly and eventually transform in
the growth restrained cirrhotic liver... .Liver mass is controlled by growth factors
- mass loss through could provide a growth stimulatory macroenvironment. For
the microenvironment, cirrhosis activates stellate cells resulting in increased
production of extracellular matrix proteins, cytokines, growth factors, and
products of oxidative stress.
Like other cancers, genomic instability is a common feature of human HCC with various
mechanisms thought to contribute, including telomere erosion, chromosome segregation defects,
and alteration in DNA damage-response pathways. In addition to genetic events associated with
the development of HCC (p53 inactivation, mutation in P-catenin, overexpression of ErbB
receptor family members, and overexpression of the MET receptor whose ligand is HGF) various
cancer-relevant genes seem to be targeted on the epigenetic level (methylation) in human HCC
(Farazi and Depinho, 2006). Changes in methylation have been detected in the earliest stages of
hepatocarcinogenesis and to a greater extent in tumor progression (Lee et al., 2003). Seitz and
Stickel (2006) report that aberrant DNA hypermethylation (a silencing effect on genes) may be
associated with genetic instability as determined by the loss of heterozygosity and microsatellite
instability in human HCC due to chronic viral hepatitis and that modifications of the degree of
hepatic DNA methylation have also been observed in experimental models of chronic
alcoholism.
Farazi and DePinho (2006) reported that two of the key molecules that involved in DNA
damage response, p53 and BRCA2, seem to have roles in destabilizing the HCC genome (Collin,
2005). The inactivation of p53 through mutation or viral oncoprotein sequestration is a common
event in HCC and p53 knock in mouse models containing dominant point mutations have been
shown to cause genomic instability. However, Farazi and DePinho (2006) noted that despite
documentation of deletions or mutations in these and other DNA damage network genes, their
direct roles in the genomic instability of HCC have yet to be established in many genetic model
systems.
Telomere shortening has been described as a key feature of chronic hyperproliferative
liver disease (Kitada et al., 1995; Miura et al., 1997; Urabe et al., 1996)( Rudolf and DePinho,
2001), specifically occurring in the hepatocyte compartment. These observations have fueled
speculation that telomere shortening associated with chronic liver disease and hepatocyte
turnover contribute to the induction of genomic instability that drives human HCC (Farazi and
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-326 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Depinho, 2006). Defects in chromosome segregation during mitosis result in aneuploidy, a
common cytogenetic feature of cancer cell including HCC (Farazi and Depinho, 2006).
Several studies have attempted to categorize genomic changes in relation to tumor state.
In general, high levels of chromosomal instability seem to correlate with the de-differentiation
and progression of HCC (Wilkens et al., 2004). Several studies have suggested certain
chromosomal changes to be specific to dysplastic lesions, early -stage and late-stage HCCs, and
metastases. It is important to note that the studies that have attempted to compare genomic
profiles and tumor state are few in number, often did not classify HCCs on the basis of etiology,
and used relatively low-resolution genome-scanning platforms (Farazi and Depinho, 2006).
Farazi and DePinho (2006) noted that it should be emphasized that although genome-etiology
correlates reported in some studies are intriguing, several studies have failed to uncover
significant differences in genomic changes between different etiological groups, although the
outcome might related to small sample sizes and the low-resolution genome-scanning platform
used.
E.3.1.6. Pathway and Genetic Disruption Associated with Hepatocellular Carcinoma
(HCC) and Relationship to Other Forms of Neoplasia
In their landmark paper, Hanahan and Weinberg (2000) suggesteded that the vast catalog
of cancer cell genotypes were a manifestation of six essential alterations in cell physiology that
collectively dictate malignant growth; self-sufficiency in growth signals, insensitivity to growth
-inhibitory (antigrowth signals), elevation of programmed cell death (apoptosis), limitless
replication potential, sustained angiogenesis, and tissue invasion and metastasis. They proposed
that these six capabilities are shared in common by most and perhaps all types of human tumors
and, while virtually all cancers must acquire the same six hallmark capabilities, their means of
doing so would vary significantly, both mechanistically and chronologically. It was predicted
that in some tumors, a particular genetic lesions may confer several capabilities simultaneously,
decreasing the number of distinct mutational steps required to complete tumorigenesis. Loss of
the p53 tumor suppressor was cited as an example that could facilitate both angiogenesis and
resistance to apoptosis and to enable the characteristic of genomic instability. The paths that
cells could take on their way to becoming malignant were predicted to be highly variable, and
within a give cancer type, mutation of a particular target genes such as ras or p53 could be found
only in a subset of otherwise histologically identical tumors. Furthermore, mutations in certain
oncogenes and tumor suppressor genes could occur early in some tumor progression pathways
and late in others. Genes known to be functionally altered in "cancer" were identified as
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-327 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
including Fas,Bcl2, Decoy R, Bax, Smads, TFGpR, pl5, pl6, Cycl D, Rb, human papilloma
virus E7, ARF, PTEN, Myc, Fos, Jun, Ras, Abl, NF1, RTK, transforming growth factor alpha
(TGF-a), Integrins, E-cadherin, Src, P-catenin, APC, and WNT.
Branda and Wands (2006) reported that two signal transduction cascades that appear to be
very important are insulin/IFG-l/IRS-l/MAPK and Wnt/Frizzled/p-catenin pathways which are
activated in over 90% of HCC tumors (Branda and Wands, 2006). Feitelson et al. (2002)
reported that
In addition to NF-kB, up-regulated expression of rhoB has been reported in some
HCCs. RhoB is in the ras gene family, is associated with cell transformation, and
may be a common denominator to both viral and non-viral hepatocarcinogenesis.
Activation of ras and NF-kB, combined with down regulation of multiple negative
growth regulatory pathways, then, may contribute importantly to early steps in
hepatocarcinogenesis. Thus viral proteins may alter the patterns of hepatocellular
gene expression by transcriptional trans-regulation.. .Another early event appears
to involve the mutation of P-catenin, which is a component of the Wnt signal
transduction pathway whose target genes include c-myc, c-jun, cyclin Dl,
fibronectin, the connective tissue growth factor WISP, and matrix
metaolloproteinases.
Boyault et al. (2007) reported that
altogether, the principle carcinogenic pathways known to be deregulated in HCC
are inactivation of TP53, Wnt/wingless activation mainly through CTNNB1
mutations activating P-catenin- and AXIN1-inactivating mutations,
retinoblastoma inactivation through RBI and CDKN2A promoter methylation and
rare gene mutations, insulin growth factor activation through IGF2
overexpression, and IGF2R-inactiving mutations.
El-Serag and Rudolph suggested that "in general, the activation of oncogenic pathways in
human HCC appears to be more heterogeneous compared with other cancer types." El-Serag
and Rudolph (2007) reported that the p53 pathway is a major tumor-suppressor pathway that
(1) limits cell survival and proliferation (replicative senescence) in response to telomere
shortening (2) induces cell-cycle arrest in response to oncogene activation (oncogene-induced
senescence), (3) protects genome integrity, and (4) is affected at multiple levels in human HCC.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-328 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
"p53 mutations occur in aflatoxin induced HCC (>50%) and with lower frequency (20-40%) in
HCC not associated with aflatoxin." In addition,
the vast majority of human HCC overexpresses gankyrin, which inhibits both Rb
checkpoint and p53 checkpoint function.. .The pl6/Rb checkpoint is another
major pathway limiting cell proliferation in response to telomere shortening,
DNA damage, and oncogene activation. In human HCC the Rb pathway is
disrupted in more than 80% of cases, with repression of pl6 by promoter
methylation being the most frequent alteration. Moreover, expression of gankyrin
(an inhibitor of p53 and Rb checkpoint function) is increased in the vast majority
of human HCCs, indicating that the Rb checkpoint is dysfunctional in the vast
majority of human HCCs.. The frequent inactivation of p53 in human HCC
indicates that abrogation of p53-dependent apoptosis could promote
hepatocarcinogenesis. The role of impairment of p53-independent apoptosis for
hepatocarcinogenesis remains to be defined.. .Activation of the P-catenin pathway
frequently occurs in mouse and human HCC involving somatic mutations, as well
as transcriptional repression of negative regulators. An activation of the Akt
signaling and impaired expression of phosphatase and tensin homolog (PTEN) (a
negative regulator of Akt) have been reported in 40-60% of Human HCC.
They suggested that although Myc is a potent oncongene inducing hepatocarcinogenesis in
mouse models the data on human HCC are heterogeneous and further studies are required.
E.3.1.7. Epigenetic Alterations in Hepatocellular Carcinoma (HCC)
The molecular pathogenesis of HCC remains largely unknown but it is presumed that the
development and progression of HCC are the consequence of cumulative genetic and epigenetic
events similar to those described in other solid tumors (Calvisi et al., 2006). Calvisi et al. (2007)
provided a good summary of DNA methylation status and cancer as well as its status in regard to
HCC:
Aberrant DNA methylation occurs commonly in human cancers in the forms of
genome-wide hypomethylation and regional hypermethylation. Global DNA
hypomethylation (also known as demethylation) is associated with activation of
protooncogenes, such as c-Jun, c-Myc, and c-HA-Ras, and generation of genomic
instability. Hypermethylation on CpG islands located in the promoter regions of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-329 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
tumor suppressor genes results in transcriptional silencing and genomic
instability. CpG hypermethylation (also known as de novo methylation) acts as
an alternative and/or complementary mechanisms to gene mutations causing gene
inactivation, and it is now recognized as an important mechanism in
carcinogenesis. Although the mechanism(s) responsible for de novo methylation
in cancer are poorly understood, it has been hypothesized that epigenetic silencing
depends on activation of a number of proteins known as DNA methyltransferases
(DNMTs) that posses de novo methylation activity. The importance of DNMTs
in CpG methylation was substantiated by the observation that genetic disruption
of both DNMT1 and DNMT3b genes in HCT116 cell lines nearly eliminated
methyltransferase activity. However, more recent findings indicate that the
HCT116 cells retain a truncated, biologically active form of DNMT1 and
maintain 80% of their genomic methylation. Further reduction of DNMT1 levels
by a siRNA approach resulted in decreased cell viability, increased apoptosis,
enhanced genomic instability, checkpoint defects, and abrogation of replicative
capacity. These data show that DNTM1 is required for cell survival and suggest
that DNTM1 has additional functions that are independent of its methyltransferase
activity. Concomitant overexpression of DNMT1, -3A, and -3b has been found in
various tumors including HCC. However, no changes in the expression of
DNMTs were found in other neoplasms, such as colorectal cancer, suggesting the
existence of alternative mechanisms. In HCC, a novel DNMT3b splice variant,
known as DNMT3b4 is overexpressed. DNMT3b4 lacks DNMT activity and
competes with DNMT2b3 for targeting of pericentromeric satellite regions in
HCC, resulting in DNA hypomethylation of these regions and induction of
chromosomal instability, further linking aberrant methylation and generation of
genomic alterations.
It is now well accepted that methylation changes occur early and ubiquitously in
cancer development. The case has been made that tumor cell heterogeneity is
due, in part, to epigenetic variation in progenitor cells and that epigenetic
plasticity together with genetic lesions drive tumor progression (Feinberg et al.,
2006).
A growing number of genes undergoing aberrant CpG island hypermethylation in
HCC have been discovered, suggesting that de novo methylation is an important
mechanism underlying malignant transformation in the liver. However, most of
the previous studies have focused on a single or a limited number of genes, and
few have attempted to analyze the methylation status of multiple genes in HCC
and associated chronic liver diseases. In addition, the functional consequence(s)
of global DNA hypomethylation and CpG island hypermethylation in human liver
cancer has not been investigated to date. Furthermore, to our knowledge no
comprehensive analysis of CpG island hypermethylation involving activation of
signaling pathways has been performed.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-330 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Calvisi et al. (2007) reported that global gene expression profiles show human HCC to
harbor common molecular features that differ greatly from those of nontumorous surrounding
tissues, and that human HCC can be subdivided into 2 broad but distinct subclasses that are
associated with length of patient survival. They further suggested that aberrant methylation is a
major event in both early and late stages of liver malignant transformation and might constitute a
critical target for cancer risk assessment, treatment, and chemoprevention of HCC. Calvisi et al.
(2007) conducted analysis of methylation status of genes selected based on their capacity to
modulate signaling pathways (Ras, Jak/Stat, Wingless/Wnt, and RELN) and/or biologic features
of the tumors (proliferation, apoptosis, angiogenesis, invasion, DNA repair, immune response,
and detoxification). Normal livers were reported to show the absence of promoter methylation
for all genes examined. At least 1 of the genes involved in inhibition of Ras (ARH1, CLU,
DAB2, hi)A B2IP, HIN-1, HRASL, LOX, NORE1A, PAR4, RASSF1A, RASSF2, RASSF3,
RASSF4, RIG, RRP22, and SPRY2 and -¥), Jak/Stat (ARHl.CIS, SHP1, PIAS-1, PIAS-y, SOCS1,
-2, and -3, SYK, and GRIM-19), and Wnt/p-catenin (APC, E-cadherin, y-catenin, SFRP1, -2, -4,
and -5, DKK-1 and -3, WIF-1 and HDPR1) pathways were affected by de novo methylation in all
HCC. A number of these genes were also reported to be highly methylated in the surrounding
nontumorous liver. In contrast, inactivation of at least 1 of these genes implicated in the RELN
pathway (DAB1, reelin) was detected differentially in HCC of subclasses of tumor that had
difference in tumor aggressiveness and progression. Epigenetic silencing of multiple tumor
suppressor genes maintains activation of the Ras pathway with a major finding in the Calvisi et
al. (2007) study to be the concurrent hypermethylation of multiple inhibitors of the Ras pathway
with Ras was significantly more active in HCC than in surrounding or normal livers. Also
important, was the finding that no significant associations between methylation patterns and
specific etiologic agents (i.e., HVB, HVC, ethanol, etc.) were detected further substantiating the
conclusion that aberrant methylation is a ubiquitous phenomenon in hepatocarcinogenesis.
Current evidence suggests that hypomethylation might promote malignant
transformation via multiple mechanisms, including chromosome instability,
activation of protooncogenes, reactivation of transposable elements, and loss of
imprinting.. .The degree of DNA hypomethylation progressively increased from
nonneoplastic livers to fully malignant HCC, indicating that genomic
hypomethylation is an important prognostic factor in HCC, as reported for brain,
breast, and ovarian cancer.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-331 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Calvisi et al. (2007) also reported that regional CpG hypermethylation was also enhanced during
the course of HCC disease and that the study of tumor suppressor gene promoters showed that
CpG methylation was frequently detected both in surrounding nontumorous livers and HCC.
E.3 .1.8. Heterogeneity of Preneoplastic and Hepatocellular Carcinoma (HCC) Phenotypes
A very important issue for the treatment of HCC in humans is early detection. Research
has focused on identification of lesions that will progress to HCC and to also determine from the
phenotype of the nodule and genetic expression its cell source, likely survival, and associations
with etiologies and MOAs. As with rodent models where preneoplastic foci have been observed
to be associated with progression to adenoma and carcinoma, nodules observed in humans with
high risk for HCC have been observed to progress to HCC. In humans, histomorphology of
HCC is notoriously heterogeneous (Yeh et al., 2007). Although much progress has been made,
there is currently not universally accepted staging system for HCC partly because of the natural
course of early HCC is unknown and the natural progression of intermediated and advanced
HCC are quite heterogeneous (Thorgeirsson, 2006). Nodules are heterogeneous as well with
differences in potential to progress to HCC. Chen et al. (2002) reported that standard clinical
pathological classification of HCC has limited valued in predicting the outcome of treatment as
the phenotypic diversity of cancer is accompanied by a corresponding diversity in gene
expression patterns. There is also histopathological variability in the presentation of HCC in
geographically diverse regions of the world with some slow growing, differentiated HCC
nodules surrounded by a fibrous capsule are common among Japanese but, in contrast, a
"febrile" form of HCC, characterized by leukocytosis, fever, and necrosis within a poorly
differentiated tumor to be common in South African blacks (Feitelson et al., 2002).
A multistep process is suggested histologically, where HCC appears within the context of
chronic hepatitis and/or cirrhosis within regions of the liver cell dysplasia or adenomatous
hyperplasia (Feitelson et al., 2002). Kobayashi et al. (2006) reported that the higher the grade of
the nodule the higher the percentage that will progress to HCC with 18.8% of all nodules and
regenerative lesions going on to become HCC, 53.3% remaining unchanged, and 27.9%
disappearing in the observation period of 0.1 to 8.9 years. Borzio et al. (2003) reported that the
rate of liver malignant transformation was 40% in larger regenerative nodules, low-grade
dysplastic, and high-grade dysplastic nodules with higher grade of dysplasia extranodular
detection of large cell change and hyperchronic pattern associated with progression to HCC.
Yeh et al. (2007) reported that nuclear staining for Ki-67 and Topo II-a (a nuclear protein
targeted by several chemotherapeutic agents) significantly increased in the progression from
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-332 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
cirrhosis, through high grade dysplastic nodules to HCC whereas the scores for TGF-a in these
lesions showed an inverse relationship. "In comparison with 18 HCC arising in noncirrhotic
livers, the expression of TGF-a is significantly stronger in cirrhotic liver than in noncirrhotic
parenchyma and its expression is also stronger in HCC arising in cirrhosis than in HCC arising in
noncirrhotic patients." They concluded that initiation in cirrhotic and noncirrhotic liver may
have different pathways with Transforming growth factor-a (a mitogen activated the EFGR)
playing a relative more important role in HCC from cirrhotic liver. Over expression of TGF-a in
the liver of transgenic mice induced increased proliferation, dysplasia, adenoma and carcinoma.
Yeh et al. (2007) concluded that such high-grade dysplastic nodules are precursor lesions in
hepatocarcinogenesis and that TGF-a may play an important role in the early events of liver
carcinogenesis.
Moinzadeh et al. (2005) reported in a meta-analysis of all available (n = 785) HCCs that
gains and losses of chromosomal material were most prevalent in a number of chromosomes and
that amplifications and deletions occurred on chromosomal arms in which oncogenes (e.g., MYC
and 8q24) and tumor suppressor genes (e.g., RBI on 13ql4) are located as well a modulators of
the WNT-signaling pathway. However, in multifocal HCC, nodules arising de novo within a
single liver have a different spectrum of genetic lesions. "Hence, there are likely to be many
paths to hepatocellular carcinoma, and this is why it has been difficult to assign specific
molecular alterations to changes in hepatocellular phenotype, clinical, or histopathological
changes that accompany tumor development" (Feitelson et al., 2002).
Serum AFP is commonly used as tumor marker for HCC. Several reports have linked
HCC to cytokines in an attempt to find more specific markers of HCC. Jia et al. (2007) reported
that AFP marker allows for identification of a small set of HCC patients with smaller tumors,
and these patients have a relatively long-term survival rate following curative treatment.
Presently the only approach to screen for the presence of HCC in high-risk
populations is the combination of serum AFP and ultrasonagraphy. However,
elevated AFP is only observed in about 60 to 70% of HCC patients and to a lesser
extent (33-65%) in patients with smaller HCCs. Moreover, nonspecific elevation
of serum AFP has been found in 15% to 58% of patients with chronic hepatitis
and 11% to 47% of patients with liver cirrhosis.
Soresi et al. (2006) reported that serum interleukin (IL)-6 levels are low in physiological
conditions, but increase considerably pathological conditions such as trauma, inflammation and
neoplasia. In tumors IL-6 may be involved in promoting the differentiation and growth of target
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-333 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
cells. "Many works have reported high serum IL-6 levels in various lifer diseases such as acute
hepatitis, primary biliary cirrhosis, chronic hepatitis (hepatitis C) and HCV-correlated liver
cirrhosis and in hepatocellular carcinoma." Soresi et al. (2006) reported that patients with HCC
group had higher IL-6 values than those with cirrhosis and that "higher-staged" patients had the
highest IL-6 levels. Hsia et al (2007) also examined IL-6, IL-10 and hepatocyte growth factor
(HGF) as potential markers for HCC.
The expression of IL-6 or IL-10 or higher level of HGF or AFP was observed only
0-3% of normal subjects. Patients with HCC more frequently had higher IL-6 and
IL-10 levels, where as HGF levels in HCC patients were not significantly elevated
compared to patients with chronic hepatitis or non-HCC tumors (but greater than
controls). Among patients with low AFP level, IL-6 or IL-10 expression was
significantly associated with the existence of HCC. Patients with large HCC (>5
cm) more often had increased IL-6, IL-10 or AFP levels. Serum levels of IL-6
and IL-10 are frequently elevated in patients with HCC but not in benign liver
disease or non-HCC tumors.
Nuclear DNA content and ploidy have also been the subjects of several studies through
the years for identification of pathways for prediction of survival or origin of tumors. Nakajima
et al. (2004) report that p53 loss can contribute to the propagation of damaged DNA in daughter
cells through the inability to prevent the transmission of inaccurate genetic material, considered
to be one of the major mechanisms for the emergence of aneuploidy in tumors with inactivated
p53 protein and the increasing ploidy in HCC was associated with disturbance in p53. McEntee
et al. (1992) reported that specimens from 74 patients who underwent curative resection for
primary HCC and analyzed for DNA content, (i.e., tumors were classified as DNA aneuploid if a
separate peak was present from its standard large diploid peak [2C] and tetraploid peak [4C])
33% were DNA diploid, 30% were DNA tetraploid/polyploidy, and 37% were aneuploid of the
primary tumors examined. Nontumor controls were diploid and survival was not different
between patients with diploid versus nondiploid tumors. Zeppa et al. (1998) reported ploidy in
84 hepatocellular carcinomas diagnosed by fine-needle aspiration biopsy to have 68 cases that
were aneuploid and 16 euploid (9 diploid and 7 polyploid), with median survival of 38 months
for patients with diploid HCC and 13 months for aneuploid HCC. Lin et al. (2003) reported in
their study of fine needle aspiration of HCC that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-334 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
the ratio of S and G2/M periods of DNA, which reflect cell hyperproliferation, in
the group with HCC tumors> 3cm in diameter were markedly higher than those of
the group with nodules< 3 cm in diameter and the group with hyperplastic
nodules.. .DNA analysis of aspiration biopsy tissues acquired from intrahepatic
benign hyperplastic nodules showed steady diploid (2c) peak that stayed in G1
period. DNA analysis of aspiration biopsy tissues acquired from HCC nodules
showed S period of hyperproliferation and G2/M period. The DNA analysis of
HCC nodules showed aneuploid peak.
They concluded that in regard to the biological behavior of the cell itself, that the normal tissue,
reactive tissue and benign tumor all have normal diploid DNA but, like most other malignant
tumors, "HCC appears to have polyploid DNA, especially aneuploid DNA."
Attallah et al. (1999) reported small needle liver biopsy data to show HCC to be 21.4%
diploid, 50% aneuploid and 28.6% tetraploid and that higher ploidies (aneuploid and tetraploid)
were observed in human liver cancer than residual tissues, although in some cases there was
increased aneuploidy (cirrhosis, 37%, hepatitis -50%). Of note for the study, is the lack of
appropriate control tissue and uncertainty as to how some of their diploid cells could have been
binucleate tetraploid cells. Anti et al. (1994) reported reduction in binuclearity in the chronic
hepatitis and cirrhosis groups that was significantly correlated with a rise in the
diploid/polyploidy ratio and that precancerous and cancerous nodules within cirrhotic liver show
an increased tendency toward diploidy or the emergence of aneuploid populations. They noted
that a number of investigators have noted significantly increased hepatocyte diploidization
during the early stages of chemically induced carcinogenesis in rat liver, but other experimental
findings indicate that malignant transformation can occur after any type of alteration in ploidy
distribution.
On the other hand, Melchiorri et al. (1994) noted that several studies using flow
cytometric or image cytometric methods reported high DNA ploidy values in 50-77%) of the
examined HCCs and that the presence of aneuploidy was significantly related to a poor patient
prognosis. They reported that the DNA content of mononucleated and binucleated hepatocytes,
obtained by ultrasound-guided biopsies of 10 macroregenerative nodules without histologic signs
of atypia from the lesions with the greater fraction of mononucleated hepatocytes were
diagnosed as HCCs during the clinical follow-up with results also suggesting that diploid and
tetraploid stem cell lines are the main lines of the HCCs as well as a reduction in the percentage
of binucleated hepatocytes in HCC. Gramantieri et al. (1996) reported that the percentage of
binucleated cells was reduced in most of HCC they studied (i.e., the mean percentage of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-335 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
binucleated cells 9% in comparison to 24% found in normal liver) and that most HCC, as many
other solid neoplasms, showed altered nuclear parameters.
Along with reporting pathways that are perturbed in HCC, emerging evidence also shows
that signatures of pathway are predictive of clinical characteristics of HCC. A number of studies
have examined gene expression in tumors to try to determine which pathways may have been
disturbed in an attempt to predict survival and treatment options for the patients and to
investigate possible MO As for the tumor induction and progression. Chen et al. (2002)
described a systematic characterization of gene expression patterns in human liver cancers using
cDNA microarrays to study tumor and nontumor liver tissues in HCC patients, and of note did
quality assurance on their microarray chips (many studies do not report that they have done so),
and examined the effects of hepatitis virus on its subject and identified people with it. Most
importantly, Chen et al. (2002) provided phenotypic anchoring of each tumor with its genetic
profile rather than pooling data.
The hierarchical analysis demonstrated that clinical samples could be divided into two
major clusters, one representing HCC samples and the other with a few exceptions, representing
nontumor liver tissues. Most importantly, expression patterns varied significantly among the
HCC and nontumor liver samples and that samples from HBV-infected, hepatitis C virus
infected, and noninfected individuals were interspersed in the HCC branch. Thus, tumors from
people infected with HVB, HVC and noninfected people with HCC were interspersed in the
HCC pattern and could be discerned based on etiology. One cluster of genes was highly
expressed in HCC samples compared with nontumor liver tissues included a "proliferation
cluster" comprised of genes whose functions are required for cell-cycle progression and whose
expression levels correlate with cellular proliferation rates with most of the genes in this cluster
are specifically expressed in the G2/M phase. Gene profiles for HCC were consistent with fewer
molecular features of differentiated normal hepatocytes.
Chen et al. (2002) noted that both normal and liver tumors are complex tissue compose
of diverse cells and that distinct patterns of gene expression seemed to provide molecular
signatures of several specific cell types including expression of two clusters of genes associated
with T and B lymphocytes, presumably reflecting lymphocytic infiltration into liver tissues, and
genes associated with stellate cell activation. This important finding acknowledges that HCC are
not only heterogeneous in hepatocyte phenotype but are made up of many other nonparenchymal
cell types and that gene expression patterns reflect that heterogeneity. A gene cluster was also
identified at a higher level in HCC that included several genes typically expressed in endothelial
cells, including CD34, which is expressed in endothelial cells in veins and arteries but not in the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-336 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
endothelial cells of the sinusoids in nontumor liver and which may reflect disruption of the
molecular program that normally regulate blood vessel morphogenesis in the liver.
Of great importance was the investigation by Chen et al. (2002) of whether samples from
multiple sites in a single HCC tumor, or multiple separate tumor nodules in one patient, would
share a recognizable gene expression signature. With a few instructive exceptions, all the tumor
samples from each patient clustered were reported to cluster together. To further examine the
relationship among multiple tumor samples from individual patients, they calculated the pairwise
comparison for all pairs of samples and samples some primary tumors multiple times. Tumor
patterns of gene expression were more highly correlated those seen in samples from the same
patient than other patients but every tumor had a distinctive and characteristic gene expression
pattern, recognizable in all samples taken from different areas of the same tumor.
For multiple discrete tumor masses obtained from six patients, three of these patients had
multiple tumors with a shared distinctive gene expression pattern but in three other patients,
expression patterns varied between tumor nodules and the difference providing new insights into
the sources of variation in molecular and biological characteristics of cancers. Thus, in some
patients multiple tumors were from the same clone, as demonstrated by a similar gene expression
profile, but for some patients multiple tumors were arising from differing clones within the same
liver. In regard to whether the distinctive expression patterns characteristic of each tumor reflect
the individuality of the tumor or are determined by the patient in whom the tumor arose, analysis
of the expression patterns observed in the two tumor nodules from one patient showed that the
two tumors were not more similar than those of an arbitrary pair of tumors from different
patients. These results show the heterogeneity of HCC and that "one gene pattern" will not be
characteristic of the disease.
However, HCC did have a pattern that differed from other cancers. Chen et al. (2002)
analyzed the expression patterns of 10 randomly selected HCC samples and 10 liver metastases
of other cancers and reported that the HCC samples and the metastatic cancers clustered into two
distinct groups, based on difference in their patterns of gene expression. Although some of the
HCC samples were poorly differentiated and expressed the genes of the liver-specific cluster at
very low levels compared to with either normal liver or well-differentiated HCC, the genes of the
liver-specific cluster were reported to be consistently expressed at higher levels in HCC than in
tumors of nonliver origin. Metastatic cancers originating from the same tissue typically clustered
together, expressing genes characteristic of the cell types of origin. Thus, liver cancer was
distinguishable from other cancer even though very variable in expression and differentiation
state.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-337 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In an attempt to create molecular prognostic indices that can be used for identification of
distinct subclasses of HCC that could predict outcome, Lee et al. (2004a) reported two subclasses
of HCC patients characterized by significant differences in the length of survival. They also
identified expression profiles of a limited number of genes that accurately predicted the length of
survival. Total RNAs from the 19 normal livers, including "normal liver in HCC patients," were
pooled and used as a reference for all microarray experiments and thus variations between
patients, and especially differences due to conditions predisposing HCC, were not determined.
DNA microarray data using hierarchical clustering was reported to yield two major clusters, one
representing HCC tumors, and the other representing nontumor tissues with a few exceptions that
were not characterized by the authors. Lee et al. (2004a) reported that, along with 2 distinctive
subtypes of gene expression patterns in HCC, there was heterogeneity among HCC gene
expression profiles and that one group had an overall survival time of 30.8 months and the other
83.7 months. Only about half the patients in each group were reported to have cirrhosis.
Expression of typical cell proliferation markers such as PCNA and cell cycle regulators such as
CDK4, CCNB1, CCNA2, and CKS2 was greater in one class than the other of HCC.
The report by Boyault et al. (2007) attempted to compare etiology and genetic
characterization of the tumors they produce and confirmed the heterogeneity of HCC, some
without attendant genomic instability. Boyault et al. (2007) reported that genetic alterations are
indeed closely associated with clinical characteristics of HCC that define 2 mechanisms of
hepatocarcinogenesis.
The first type of HCC was associated with not only a high level of chromosome
instability and frequent TP53 and AXIN1 mutations but also was closely linked to
HBV infections and a poor prognosis. Conversely, the second subgroup of HCC
tumors was chromosome-stable, having a high incidence of activating P-catenin
alteration and was not associated with viral infection.
Boyault et al. (2007) reported that in a series of 123 tumors, mutations in the CTNNB1
(encoding P-catenin), TP53, ACINI, TCF1, PIK3CA and KRAS genes in 34, 31, 13, 5, 2, and
1 tumors were identified, respectively. No mutations were found in NRAS, HRAS, and EGFR.
Hypermethylation of the CDKN2A and CDH1 promoter was identified in 35 and 16% of the
tumors, respectively. Boyault et al. (2007) grouped tumors by genomic expression as well as
other factors. HCC groups associated with high rate of chromosomal instability were reported to
be enriched with over expression of cell-cycle/proliferation/DNA metabolism genes. They
concluded that "the primary clinical determinant of class membership is HBV infection and the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-338 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
other main determinants are genetic and epigenetic alterations, including chromosome instability,
CTNNB1 and TP53 mutations, and parental imprinting. Tumors related to HCV and alcohol
abuse were interspersed across subgroups G3-G6." Boyault et al. (2007) suggested that their
results indicated that HBV infection early in life leads to a specific type of HCC that has
immature features with abnormal parental gene imprinting selections, possibly through the
persistence of fetal hepatocytes or alternatively through partial dedifferentiation of adult
hepatocytes. "These G1 tumors are related to high-risk populations found in epidemiological
studies."
E.3.2. Animal Models of Liver Cancer
There are obvious differences between rodents and primate and human liver, and there is
a difference in background rates of susceptibility to hepatocarcinogenesis. With strains of mice
there are large differences in responses to hepatotoxins (e.g., acetaminophen) and to
hepatocarcinogens as well as background rates of hepatocarcinogenecity. Boyault et al. (2007)
reported that modulators of murine hepatocarcinogenesis, such as diet, hormones, oncogenes,
methylation, imprinting, and cell proliferation/apoptosis are among multiple mechanistically
associated factors that impact this target organ response in control as well as in treated mice, and
suggested that there is no one simple paradigm to explain the differential strain sensitivity to
hepatocarcinogenesis. Because of the variety of studies with differing protocols used to generate
susceptibility data, direct comparisons among strains and stocks is problematic but in regard to
susceptibility to carcinogenicity the C3H/HeJ and C57BL/6J mouse have been reported to have
up to a 40-fold difference in liver tumor multiplicity Boyault et al. (2007).
However, as noted above, TCE causes liver tumors in C6C3F1 and Swiss mice with
studies of trichloroethylene metabolites dichloroacetic acid, trichloroacetic acid, and CH
suggesting that both dichloroacetic acid and trichloroacetic acid are involved in
trichloroethylene-induced liver tumorigenesis. Many effects reported in mice after
dichloroacetic acid exposure are consistent with conditions that increase the risk of liver cancer
in humans and can involve GST Xi, histone methylation, and overexpression of insulin-like
growth factor-II (IGF-II) (Caldwell and Keshava, 2006). The heterogeneity of liver phenotype
observed in mouse models is also consistent with human HCC. These data lend support to the
qualitative relevance of the mouse model for TCE-induced cancer risk.
Bannasch et al. (2003) made important observations that have implications regarding the
differences in susceptibility between rodent and human liver cancer. They stated that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-339 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Although the classification of such nodular liver lesions in rodents as hyperplastic
or neoplastic has remained controversial, persistent nodules of this type are
considered neoplasms, designated as adenomas. In human pathology, the
situation appears to be paradoxical because adenomas are only diagnosed in the
noncirrhotic liver, yet a confusing variety terms avoiding the clearcut
classification as an adenoma has been created for nodular lesions in liver
cirrhoses, not withstanding that the vast majority hepatocellular carcinomas
develop in cirrhotic livers. Even if a portion of these nodular lesions would be
regarded as adenomas, being integrated into an adenoma-carcinoma sequence as
observed in many animal experiments, clinical and epidemiological records of
liver neoplasms, including both benign and malignant forms, would increase
considerably. This would not only bring hepatic neoplasia further into focus of
human neoplasia in general, but also shed new light on the classification of some
chemicals producing high incidence of liver neoplasms in rodents, but appearing
harmless to humans according to epidemiological evaluations solely based on the
incidence of hepatocellular carcinoma in exposed populations.
Thus, in humans only HCCs are recorded but in animals adenomas are counted as neoplasms,
indicating that the scope of the problem of liver cancer in humans may be underestimated.
Tumor phenotype differences have been reported for several decades through the work of
Bannasch et al. The predominant cell line of foci of altered hepatocytes (FAH) have excess
glycogen storage early in development that appears to be similar to that shown by DCA
treatment. Bannasch et al. (2003) reported that "the predominant glycogenotic-basophilic cell
line FAH reveals that there is an overexpression of the insulin receptor, the IGF-1 receptor, the
insulin receptor substrates-1/2 and other components of the insulin-stimulated signal transduction
pathway." Bannasch stated that foci of this type have increased expression of GST-71 and insulin
has also been shown to induce the expression of GST-pi but that hyperinsulin-induced foci do
not show increased GST-71. Cellular dedifferentiation during progression from glycogenotic to
basophilic cell populations is associated with downregulation in insulin signaling. The
amphophilic-basophilic cell lineage of peroxisome proliferators and hepadnaviridae were
reported to have foci that mimic effects of thyroid hormone with mitochondrial proliferation and
activation of mitochondrial enzymes. Bannasch et al. (2003) stated that
the unequivocal separation of 2 types of compounds, usually classified as
initiators and promoters, remains a problem at the level of the foci because at least
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-340 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
the majority of chemical hepatocarcinogens seem to have both initiating and
promoting activity, which may differ in quantitative rather than qualitative terms
from one compound to another.. .Whereas genetic mutations have been
predominantly postulated to initiate hepatocarcinogenesis for many years, more
recently epigenetic changes have been increasingly discussed as a plausible cause
of the evolution of preneoplastic foci characterized by metabolic changes
including the expression of GSTpi.
Su and Bannasch (2003) reported that glycogen-storing foci represents early lesion with
the potential to progress to more advance glycogen-poor basophilic lesions through mixed cell
foci and resulting hyperproliferative lesions and are associated with HCC in man. Small-cell
change (SCC) of liver parenchyma (originally called liver cell dysplasia of small cell size) is
reported to share cytological and histological similarities to early well defined HCC. Close
association between SCC and more advanced (basophilic) foci indicates that foci often progress
to HCC through SCC in humans. SCC were reported to be present in all basophilic foci.
Previous studies were cited that showed that the biochemical phenotype of human FAH, mainly
including glycogen storing clear cell foci and clear cell-predominated mixed cell foci, were
observed in more than 50% of cirrhotic livers with or without HCC. FAH of clear and mixed
cell types were observed in almost all livers bearing HCC, and in chronic liver diseases without
HCC but at a lower frequency. Su and Bannasch (2003) reported that
the finding of mixed cell foci (MCF) mainly in livers with high-risk or
cryptogenenic cirrhosis indicates that these are more advanced precursor lesions
in man, in line with earlier observations in experimental animals. Considering
their preferential emergence in cirrhotic livers of the high-risk group, their
unequivocally elevated proliferative activity, and the resulting large size with
frequent nodular transformation, we suggest that mixed cell populations are
endowed with a high potential to progress to HCC in humans, as previously
shown in rats.
In human HCC, irregular areas of liver parenchyma with marked cytoplasmic
amphophilia, phenotypically similar to the amphophilic preneoplastic foci in rodent liver
exposed to different hepatocarcinogenic chemicals (e.g., DHEA a peroxisome proliferator) or the
hepadnaviruses, were reported to present in 45% of the specimens from cirrhotic livers
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-341 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
examined. "However, more data are needed to elucidate the nature of the oncocytic and
amphophilic lesions regarding their role in HCC development."
With respect to the ability respond to a mitogenic stimulus, differences between primate
and rodent liver response to a powerful stimulus, such as partial hepatectomy, have been noted
that indicate that primate and human liver respond differently (and much more slowly) to such a
stimulus. Gaglio et al. (2002) reported after 60% partial hepatectomy in Rhesus macaques
(Macaca mulatto), the surface area of the liver remnant was restored to its original preoperative
value over a 30 day period. The maximal liver regeneration occurred between days 14 and 21,
with thickening of liver cell plates, binucleation of hepatocytes, Ki-67 and PCNA expression
(occurring in hepatocytes throughout the lobule at a maximum labeling index of 30%), and
mitoses parallel increased most prominently between posthepatectomy days 14 and 30.
However, cytokines associated with inducing proliferation were elevated much earlier.
TGF-a, IL-6, HGF, IL-6 and TNF-a mRNA persisted until Day 14, with peak elevations of IL-6,
TNF-a, occurring 24 hours later surgery, and IL-6 reduced to control levels by Day 14. Gaglio
et al. (2002) suggested that their results clearly indicate that the pattern and timing of liver
regeneration observed in this nonhuman primate model are significantly different when
comparing different species (e.g., peak expression of Ki-67 in a 60% partial hepatectomy model
in rats occurs within hours following partial hepatectomy) and that the difference in timing and
pattern of maximal hepatocellular regeneration cannot be explained simply by differences in size
of animals (e.g., 60% partial hepatectomy in dogs produced liver regeneration peaks at 72 hours
with weights approximating the weights of the Rhesus macaques). They noted that previous
studies in humans, who underwent 40-80%) partial hepatectomy, reveal a similar delay in peak
liver regeneration based on changes in serum levels of ornithine decarboxylase and thymidine
kinase, further highlighting significant interspecies differences in liver regeneration.
For C57BL/6 X 129 mice Fujita et al. (2001) reported that after partial hepatectomy, the
liver had recovered more than 90% of its weight within 1 week. This difference in response to a
mitogenic stimulus has impacts on the interpretations of comparisons between rodent and
primate liver responses to chemical exposures which give a transient increases in DNA synthesis
or cell proliferation such as PPARa agonists. Also, as stated above, the primate and human liver,
while having a significant polyploidy compartment, do not have the extent of polyploidization
and the early onset of that has been observed in the rodent. However, as noted by Lapis et al.
(1995), exposure to DEN has proven to be a highly potent hepatocarcinogen in nonhuman
primates, inducing malignant tumors in 100% of animals with an average latent period of 16
months when administered at 40 mg/kg intraperitoneally every 2 weeks.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-342 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
In regard to species extrapolation of epigenomic changes between humans and rodents,
Weidman et al. {, 2007, 732081 cautioned that
Although we do predict some overlap between mouse and human candidate
imprinted genes identified through our machine-learning approach, it is likely that
the most significant criterion in species-specific identification will differ. This
difference underscored the importance for increased caution when assessing
human risk from environmental agents that alter the epigenome using rodent
models; the molecular pathways targeted may be independent.
Despite species differences, the genome of the mouse has been sequenced and many
transgenic mouse models are being used to study the consequences of gene expression
modulation and pathway perturbation to study human diseases and treatments. However, the use
of transgenic models must be used with caution in trying to determine to determine MO As and
the background effects of the transgene (including background levels of toxicity) and specificity
of effects must be taken into account for interpretation of MO A data, especially in cases where
the knockout in the mouse causes significant liver necrosis or steatosis {Caldwell, 2006,
701418;Keshava, 2006, 700361;Caldwell, 2008, 630407}. For the determination of effects of
pathway perturbation and similarity to human HCC phenotype, mouse transgenic models have
been particularly useful with tumors produced in such models shown to correlate with tumor
aggressiveness and survival to human counterparts.
E.3 .2.1. Similarities with Human and Animal Transgenic Models
Mice transgenic for transforming growth factor-a (a member of the EGF family and a
ligand for the ErfB receptors) develop HCCs (Farazi and Depinho, 2006). Compound TGFa and
MYC transgenic mice show increase hepatocarcinogenesis that is associated with the disruption
of TGF-pi signaling and chromosomal losses, some of which are syntenic to those in human
HCCs that include the retinoblastoma (RB) tumor suppressor locus (Sargent et al., 1999).
Lee et al. (2004b) investigated whether comparison of global expression patterns of
orthologous genes in human and mouse HCCs would identify similar and dissimilar tumor
phenotypes, and thus, allow the identification of the best-fit mouse models for human HCC. The
molecular classification of HCC on the basis of prognosis in Lee et al. (2004a) was further
compared with gene-expression profiles of HCCs from seven different mouse models (Lee et al.,
2004b). Lee et al. (2004b) characterized the gene expression patters of 68 HCC from seven
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-343 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
different mouse models; two chemically induced (Ciprofibrate and diethylnitrosamine), four
transgenic (targeted overexpression of Myc, E2F1, Myc andE2Fl, and Myc and Tgfa in the
liver). HCCs from some of these mice (MYC, E2F1 and MYC-E2F1 transgenics) showed
similar gene-expression patterns to the ones of HCCs from patients with better survival. Murine
HCCs derived for MYC-TGF-a transgenic model or diethylnitrosamine-treated mice showed
similar gene-expression patterns to HCCs from patients with poor survival. The authors reported
that Myc Tgfa transgenic mice typically have a poor prognosis, including earlier and higher
incident rates of HCC development, higher mortality, higher genomic instability and higher
expression of poor prognostic markers (e.g., AFP) and that Myc and Myc/E2fl transgenic mice
have relatively higher frequency of mutation in P-catenin (Catnb) and nuclear accumulation of P-
catenin that are indicative of lower genomic instability and better prognosis in human HCC.
Lee et al. (2004b) indentified three distinctive HCC clusters, indicating that gene
expression pattern of mouse HCC are clearly heterogeneous and reported that Ciprofibrate-
induced HCCs and HCCs from Acox -/- mice were closely clustered and well separated from
other mouse models. However, there are several issues regarding this study that give limitations
to some of its conclusions regarding the Acox -/- mouse and Ciprofibrate treatment. The Acox -
/- mouse is characterized by profound hepatonecrosis, which confounds conclusions regarding
gene expression related to PPARa agonism made by the authors. There was very limited
reporting of the animal models (DEN and Clofibrate) protocols used. Only three tumors were
examined for Clofibrate treatment and it is unknown if the tumors were from the same animals.
Similarly only three tumors were examined from DEN treatment, which has been shown to
produce heterogeneous tumors and to produce necrosis in some paradigms of exposure.
Myc/E2F1 and E2F1 mice were split in both clusters that were compared with human HCCs.
The authors used previously published data from Meyer et al. (2003) for tumors from Acoxl"1"
null mice, DENA-treated mice and Ciprofibrate-treated mice.
Meyer et al. (2003) examined three tumors from 2 C57BL/6j mice fed Ciprofibrate for
19 months and three tumors from 2 C57BL/6j mice injected with DEN at 2-3 months but the age
at which tumors appeared was not given by the authors. Pooled mRNA from animals of varying
age (5-15 months old) was used for controls. mRNAs that differed by 2-fold in tumors were
reported to have: 60 genes up-regulated and 105 genes down-regulated in Acoxl"1" null mice
tumors; 136 genes up-regulated and 156 genes down-regulated in Ciprofibrate-induced tumors;
and 61 genes up-regulated and 105 genes down-regulated in DEN-induced tumors. The authors
stated that "Each tumor class revealed a somewhat different unique expression pattern." There
were "genes that were general liver tumor markers in all three types of tumors" with 38 genes
commonly deregulated in all three tumor types. Of note, the cell cycle genes (CDK4,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-344 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
CDC25Am CDC7 and MAPK3) cited by Lee et al. (2004b) as being more highly expressed in
DEN-induced tumors were not reported to be changed in DEN tumors in Meyer et al. (2003) or
to be altered in the Acoxl"1" null mice or mice treated with Ciprofibrate. Finally, the distinction
between groups may be dominated by gene expression changes in a large number of genes that
are related to PPAR activation but not related to hepatocarcinogenesis.
Calvisi et al. (2004a) used transgenic mice to study pathway alterations and tumor
phenotype and to further examine the premise that genomic alterations (genetic and epigenetic)
characteristic of HCC can describe tumors into 2 broad categories, the first category
characterized by activation of the Wnt/Wingless pathway via disruption of P-catenin function
and chromosomal stability and the second by chromosomal instability. Increased coexpression
of c-myc with TGF-a or E2F-1 transgenic mice was reported to result in a dramatic synergistic
effect on liver tumor development when compared with respective monotransgenic lines,
including shorter latency period, and more aggressive phenotype. P-catenin activation is
relatively common in HCCs developed in c-myc and c-myc/TGF-pi transgenic mice and rare in
the c-myc/TGF-a transgenic line which also has genomic instability.
Calvisi et al. (2004a) also reported that P-catenin staining correlated with histopathologic
type of liver tumors. Eosinophilic tumors with abnormal nuclear staining of P-catenin were
predominant in neoplastic lesions characteristic of c-myc and c-myc/E2Fl lesions. Poorly
differentiated HCCs with basophilic or clear-cell phenotypes developed more frequently in c-
myc/TGF-a and TGF-a mice and often showed a reduction or loss of P-catenin
immunoreactivity. P-catenin mutation was associated with a more benign phenotype. These
observations regarding tincture and aggressiveness are consistent with those of Bannasch (1996)
and Carter et al (2003). Calvisi et al. (2004a) noted that the relationship between P-catenin
activation, tumor grade, and clinical outcome in human HCC remains controversial.
There are studies that show a significant correlation between P-catenin
nuclear accumulation, a high grade of HCC tumor differentiation, and a better
prognosis, whereas others find that nuclear accumulation of P-catenin may be
associated with poor survival or that it does not affect clinical outcome.
Calvisi et al. (2004b) reported that for E-cadherin a variety of morphologenetic events,
including cell migration, separation, and formation of boundaries between cell layers and
differentiation of each cell layer into functionally distinct structures. Loss of expression of E-
cadherin was reported to result in dedifferentiation, invasiveness, lymph node or distant
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-345 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
metastasis in a variety of human neoplasms including HCC and that the role of E-cadherin might
be more complex that previously believed.
In order to elucidate the role of E-cadherin in the sequential steps of liver
carcinogenesis, we have analyzed the expression patterns of E-cadherin in a
collection of preneoplastic and neoplastic liver lesions from c-Myc, E2F1,
c-Myc/TGF-a and c-Myc/E2Fl transgenic mice. In particular, we have
investigated the relevance of genetic, epigenetic, and transcriptional mechanisms
on E-cadherin protein expression levels. Our data indicate that loss of E-cadherin
contributes to HCC progression in c-Myc transgenic mice by promoting cell
proliferation and angiogenesis, presumably through the upregulation of HIF-la
and VEGF proteins.
The c-Myc line, was most like wild-type and lost E-cadherin in the tumors. c-Myc/TGF-a
dysplastic lesion were reported to show overexpression of E-cadherin mainly in pericentral areas
with E2F1 clear cell carcinoma showed intense staining of E-cadherin. Reduction or loss of E-
cadherin expression is primarily determined by loss of heterozygosity at the E-cadherin locus or
by its promoter hypermethylation in human HCC. Calvisi et al. (2004b) determined the status of
the E-cadherin locus and promoter methylation in wild-type livers and tumors from transgenic
mice by microsatellite analysis and methylation specific PCR, respectively.
Wild-type livers and HCCs, regardless of their origins, showed the absence of
LOH at the E-cadherin locus. E-cadherin promoter was not hypermethylated in
wild-type, c-Myc/TGF-a and E2F1 livers. No E-cadherin promoter
hypermethylation was detected in c-Myc and c-Myc/E2Fl HCCs with normal
levels of E-cadherin protein. In striking contrast, seven of 20 (35%) of c-Myc and
two of four (50%) c-Myc/E2Fl HCCs with downregulation of E-cadherin
displayed E-cadherin promoter hypermethylation. These results suggest that
promoter hypermethylation might be responsible for E-cadherin downregulation
in a subset of c-Myc and c-Myc/E2Fl HCCs.. .The molecular mechanisms
underlying down-regulation of E-cadherin in c-Myc tumors remain poorly
understood at present. No LOH at the E-cadherin locus was detected in the c-
Myc HCCs whereas only a subset of c-Myc tumors displayed hypermethylation of
the E-cadherin promoter. Furthermore, no association was detected between
E-cadherin downregulation and protein levels of transcriptional repressors, Snail,
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-346 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Slug or the tumor suppressor WT1, in disagreement with the finding that
overexpression of Snail suppresses E-cadherin in human HCC.. .E-cadherin might
play different and apparently opposite roles, which depend on specific tumor
requirements in both human and murine liver carcinogenesis.
Importantly, the results of Calvisi et al. (2004b) showed that hypermethylation of
promoters can be associated with down regulation of a gene in mouse liver tumors similar to
human HCC and that tumors can have the same behavior with methylation change as with loss of
hetererozygosity.
This report also gave evidence of the usefulness of the mouse model to study human liver
cancer as it shows the similarity of dysfunctional regulation in mouse and human cancer and the
heterogeneity within and between mouse lines tumors with differing dysfunctions in gene
expression. These findings parallel human cancer where there is heterogeneity in tumors from
one person and every tumor has its own signature. Finally, this report correlates differing
pathway perturbations with mouse liver phenotypes similar to those reported in experimental
carcinogenesis models and for TCE and its metabolites.
Farazi and DePinho (2006) suggested that
as comparative array CGH analysis of various murine cancers has shown that such
aberrations often target syntenic loci in the analogous human cancer type, we
further suggest that comparative genomic analysis of available mouse model of
mouse HCC might be particularly helpful in filtering through the complex human
cancer genome. Ultimately, mouse models that share features with human HCCs
could serve as valuable tools for gene identification and drug development.
However, one needs to keep in mind key differences between mice and humans.
For example, as noted in certain human HCC cases, telomere shortening might
drive the genomic instability that enables the accumulation of cancer-relevant
changes for hepatocarcinogenesis. As mice have long telomeres, this aspect of
hepatocarcinogenesis might be fundamentally different between the species and
provide additional opportunities for model refinement and testing of this
mechanism through use of a telomere deficient mouse model. These and other
cross-species difference, and limitations in the use of human cell-culture systems,
must be considered in any interpretation of data from various model systems
(Farazi and Depinho, 2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-347 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Thus, these mouse models of liver cancer inductions are qualitatively able to mimic
human liver cancer and support the usefulness of mouse models of cancer.
E.3 .3. Hypothesized Key Events in HCC Using Animal Models
E.3 .3 .1. Changes in Ploidy
As stated above in Section E.l.l, increased polyploidization has been associated with
numerous types of liver injury and appears to result from exposure to TCE and its metabolites as
well as changes in the number of binucleate cells. Hortelano et al. (1995) reported that cytokines
and NO can affect ploidy and further suggest a role of these changes for carcinogenesis in
general. Vickers and Lucier (1996) noted that while both DEN and 17 a-ethinylestradiol have
been reported to enhance the proportion of diploid hepatocytes, initiators like N-
nitrosomorpholine are reported to increase the proportion of hypertrophied and polyploidy
hepatocytes. The relationship of such changes to cancer induction has been studied in transgenic
mouse models and in models involved with mitogens of differing natures.
Melchiorri et al. (1993) reported the response pattern of the liver to acute treatment with
primary mitogens in regard to ploidy changes occurring in rat liver following two different types
of cell proliferation: compensatory regeneration induced by surgical partial hepatectomy (PH)
and direct hyperplasia induced by the mitogens lead nitrate and Nafenopin (a PPARa agonist) in
8 week old male Wistar rats. Feulgen stain was used and DNA content quantified by image
cytometry in mononucleate and binucleate cells. Mitotic index was determined in the same
samples. The term "diploid" was used to identify cells with a single, diploid nucleus and
tetraploid for cells containing 2 diploid nuclei or one tetraploid nucleus referred (bi- and
mononucleate, respectively). Octoploid cells were identified as either binucleate or
mononucleate.
During liver regeneration following surgical PH an increase in the mitotic index
with a peak at 24 hours was observed. The most striking effect associated with
the regenerative response was the almost complete disappearance of binucleate
cells, tetraploid (2 X 2c) as well as octoploid (4 X 2c) with only < 10% of the
control values being present 3 days after PH... Concomitantly, an increase in
mononucleate tetraploid (4c) as well as mononucleate octoploid (8c) cells was
observed, resulting at 3 days after PH in a population made up of almost entirely
(98%) by mononucleated cells.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-348 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Lead nitrate treatment was reported to induce rapid increase in the formation of
binucleate cells occurring 3 days after treatment, their number accounting for 40% of the total
cell population versus 22% binucleate cells in control rats and 2% in PH animals killed at the
same time point. The increased binuclearity was reported to be observed only in the 4 X 2c cells
(25 vs. 6% of the controls) and in 8 X 2c cells (3.7 vs. 0.1% of controls). The increase in 4 X 2c
and 8 X 2c cells was reported to be accompanied by a concomitant reduction in 2 X 2c cells with
the change induced in cellular ploidy by lead nitrate resulting in 37% of cells being either 8c or
16c. However, at the same time point, cells having a ploidy higher than 4c were reported to
account for only 11% in PH rats and 9% in control animals. Changes in the ploidy pattern were
reported to be preceded by an increased mitotic activity, which was maximal 48 hours after
treatment with lead nitrate. The increase in mitotic index in lead nitrate-treated rats was
associated with a striking increase in the labeling index of hepatocytes (60.1 vs. 3% of control
rats) and to an almost doubling of hepatic DNA content in 3 days after lead nitrate.
Melchiorri et al. (1993) concluded that the entire cell cycle appeared to be induced by
lead nitrate but that the finding of a high increase of binucleate cells suggested that lead nitrate-
induced liver growth, unlike liver regeneration induced by partial hepatectomy, was
characterized by an uncoupling between cell cycle and cytokinesis. This raised questions
whether lead nitrate-induced liver growth resulted in a true increase in cell number or is only the
expression of an increased hepatocyte ploidy. They reported that part of the increase in DNA
content observed 3 days after lead nitrate was indeed expression of polyploidizing process due to
acytokinetic mitoses but that a consistent increase in cells number (+26%) was also induced by
lead nitrate treatment.
After Nafenopin treatment, Melchiorri et al. (1993) reported that the increase in DNA
content was increased 22% over controls and was much lower than induced by lead nitrate and
that Nafenopin did not induce significant changes in binucleate cell number. However, a shift
towards a higher ploidy class (8c) was reported to be observed following Nafenopin and the 21%
increase in DNA content seen after Nafenopin treatment was almost entirely due to increase in
the ploidy state with only 7% increase in cell number.
Melchiorri et al. (1993) examined whether hepatocytes characterized by high ploidy
content (highly differentiated cells) would be preferentially eliminated by apoptosis. An increase
in apoptotic bodies was reported to be associated with the regression phase after lead nitrate
treatment (when liver mass is reduced) but despite the elimination of excess DNA, the changes in
ploidy distribution induced by lead nitrate were found to persist suggested that polyploidy cells
were not preferentially eliminated by apoptosis during the regression phase of the liver.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-349 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Melchiorri et al. (1993) noted that other studies in rat exposed to the mitogens cyproterone
acetate (CPA) and the peroxisome proliferator MCP also reported a very strong decline in
binucleate cells with a concomitant increase in mononucleate tetraploid cells in the liver similar
to the pattern described after partial hepatectomy.
Lalwani et al. (1997) reported the results of 1,000 ppm WY-14,643 exposure in male
Wistar rats after 1, 2, and 4 weeks and suggested that an early wave of nuclear division occurred
at the early stages of exposure without cumulative effects on cell proliferation. Consistent with
hepatomegaly, WY-14,643-treated rat were reported to exhibit multifocal hepatocellular
hypertrophy and karyomegaly by routine microscopic analysis. For binucleate hepatocytes, there
were no reported differences between WY-14,643-treated and control groups for days 4 and 11
but an increase in the number at Day 25 in WY-14,643-treated animals compared to controls.
Increases in the diameter of nuclei were shown by WY-14,643 treatment from Day 11 and 25
with increasing numbers of cells displaying larger nuclear diameters. The mitotic index was
reported not to be significantly changed in WY-14,643 treated rats compared to controls. Mitotic
figures did not appear to survive the treatment necessary for flow cytometric analyses. PCNA
was increased on Day 4 in WY-14,643- treated animals compared to controls whereas no
differences were found on days 11 and 25.
However, immunohistochemistry was reported to show remarkable increases in BrdU-
labeled nuclei in liver sections after 4 days of labeling with the populations of BrdU-labeled cell
declining over the course of treatment. The labeling index was high and approximately 80% of
the BrdU-labeled cells were in periportal areas. PCNA-expressing cells were increased in the
periportal area of the liver. Intense nuclear staining of PCNA was evident as an indicator of
DNA replication in S phase. Microscopic examination showed BrdU labeling only in periportal
hepatocytes, whereas no significant labeling was observed in nonparenchymal cells, indicating
that the replicative activity was confined to the liver cells.
Lalwani et al. (1997) suggested that their results showed that events related to cell
proliferation occur in the initial phase of WY-14,643 treatment in rats but not followed by
changes in the rate of DNA synthesis as the treatment progressed. They note that Marsman et al.
"3
(1988) observed constant increases in DNA synthesis by [ H]-thymidine authoradiography with
up to 1 year of continuous administration of WY-14,643, whereas the rate of DNA synthesis or
the BrdU labeling index in their study declined after the first 4 weeks of treatment. They suggest
that the increased percentage of cells appearing in G2-M phase and the analysis of liver nuclear
profiles suggest that the progression of these additional cells (i.e., cells that are stimulated to
enter the cell cycle by the test agent) through the cell cycle is arrested in the late stages of the cell
cycle. They state
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-350 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Unlike BrdU labeling, which demonstrated DNA synthesis activity over the 4-day
labeling period, the PCNA labeling index represents levels of the protein product
at an interval post treatment. PCNA expression in cells exposed to chemicals or
to WY may not provide true representation of S phase or proliferative activity
because PCNA-expressing nuclei were also found in G0=G1 and G2-M phases.
Lalwani et al. (1997) concluded that cell proliferation alone does not appear to constitute
a determining process leading to tumors in most tissues and sustained cell replication may not be
a primary feature of peroxisome proliferator-induced hepatocarcinogenesis.
Miller et al. (1996) noted that studies with MCP in Alpk:AP rats indicate that DNA
synthesis occurs primarily in one hepatocyte subpopulation as defined by ploidy status, the
binucleated tetraploid (2 X 2N) hepatocytes, and that this preferential hepatocyte DNA synthesis
is manifested by dramatic alterations in hepatocyte ploidy subclasses (i.e., significant increases
in mononucleate tetraploid (4N) hepatocytes concomitant with decreases in 2 X 2N hepatocytes).
They reported results in male Fischer 344 rats that were 13 weeks old (an age in which
polyploidization had reached a plateau) exposed to 1,000 ppm WY-14,643 and MCP (gavage via
corn oil at 8 mg/mL or 25 mg/kg MCP once daily) for 2, 5, and 10 days (n = 4). WY-14,643 and
MCP were reported to induce significant increases in the octoploid hepatocyte class that
coincided with decreases in the tetraploid hepatocyte class. However, MCP did not induce this
shift until Day 5 of exposure. These results showed an approximate doubling of mononuclear
octoploid (8N) hepatoctyes but still a very low number of the total hepatocyte population that did
not reach greater than 7% and was still only approximately twice that of control values. Thus,
this finding does not indicate a very large target population. There was no real effect on 4N
hepatocytes due to these treatments and the percent of hepatocytes that were 4N stayed -70%
and were thus, the majority cell type in the liver. Miller et al. (1996) noted the importance of
maturation and/or strain for these analyses there are maturation-dependent differences in the
distribution and mitogenic sensitivity of hepatoctyes in the various subclasses.
Hasmall and Roberts (2000) noted that despite their differing abilities to induced liver
cancer, both DCB (a nonhepatocarcinogen in Fischer 344 rats) and DEHP, at the doses and
routes used in the NTP bioassays, induced similar profiles of S-phase LI. A large and rapid peak
during the first 7 days (1,115 and 1,151% of control for DEHP and DCB, respectively) was
followed by a return to control levels. They suggested that the size of the S-phase response does
not necessarily determine hepatocarcinogenic risk and that the subpopulation in which S-phase is
induced may be a better correlate with subsequent hepatocarcinogenecity.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 51 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
They compared the effects on polyploidy/nuclearity and on the distribution of S-phase
labeled cells with ETU, the peroxisome proliferator MCP, and phenobarbitone. Male F334 rats
7-9 weeks old were exposed to MCP (0.1% in diet), ETU 83 ppm diet, phenobarbitone (500
mg/mL drinking water) for 7 days. The number of rats for 7 day study was not given by the
authors. Hasmall and Roberts (2000) reported that treatment of rats with MCP, ETU or
phenobarbitone for 7 days had no significant effect on the ploidy profile as compared with corn
oil controls (data not shown) but that MCP and phenobarbitone did induce significant changes in
nuclearity. MCP reduced the 2 X 2N population and increased the 8N population.
Phenobarbitone similarly increased the proportion of cells in the 4N population. ETU had no
effect on the nuclearity profile as compared with control. However, what the authors describe
for their results in polidy and nuclearity are different than those presented in their figures. There
were significant differences between controls that the authors did not characterize and there
appeared to be a greater difference between controls than some of the treatments.
Gupta (2000) reported that in transgenic mice with overexpression of TGF-a, liver-cell
turnover increases, along with the onset of hepatic polyploidy, whereas hepatocellular carcinoma
originating in these animals contain more diploid cells. Coexpression of c-Myc and TGF-a
transgenes in mouse hepatocytes was associated with greater degrees of polyploidy as well as
increased development of hepatocellular carcinoma. Gupta (2000) noted that in the presence of
ongoing liver injury and continuous depletion of parenchymal cells, hepatic progenitor cells
(including oval cells) are eventually activated but what roles polyploid cells play in this process
requires further study. In the working model by Gupta (2000), sustained disease by chronic
hepatitis, metabolic disease, toxins, etc., may lead to hepatocyte polyploidy and loss, and the
emergence of rapidly cycling progenitor or escape cell clones with the onset of liver cancer.
Conner et al. (2003) described the development of transgenic mouse models in which
E2F1 and/or c-Myc was overexpressed in mouse liver. The E2F1 and c-Myc transcription
factors are both involved in regulating key cellular activities including growth and death and,
when overexpressed, are capable of driving quiescent cells into S-phase in the absence of other
mitogenic stimuli and are potent inducers of apoptosis operating at least through one common
pathway involving p53. Deregulation of their expression is also frequently found in cancer cells
(Conner et al., 2003). Conner et al. (2003) reported that although both c-Myc and E2F1 mono-
transgenic mice were prone to liver cancer, E2F1 mice developed HCC more rapidly and with a
higher frequency and that the combined expression of these two transcription factors
dramatically accelerated HCC growth compared to either E2F1 or c-Myc mono-transgenic mice.
All three transgenic lines were reported to show a low but persistent elevation of hepatocyte
proliferation before an onset of tumor growth. Ploidy was shown to be affected differently by
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 52 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
c-Myc and E2F1, and suggested distinct differences by which these two transcription factors
control liver proliferation/maturation. Both transgenic alterations induced liver cancer but had
differing effects on polyploidization suggestive that liver cancer can arise from either type of
mature hepatocyte.
c-Myc single-transgenic mice showed a continuous high cell proliferation that preceded
the appearance of preneoplastic lesions, which was also true, although to a lesser extent, in the
E2F1 mice. At 15 weeks of age, all of the transgenic mouse lines were reported to have a high
incidence (>60%) of hepatic dysplasia with mitotic indices equivalent in c-Myc/E2Fl, and c-
Myc livers, but 2-fold higher than the mitotic index in E2F1 and very low in wild-type mice.
Thus, the combination of the two transgenes did not have an additive effect on proliferation. An
analysis of the DNA content in hepatocyte nuclei isolated from 4- to 15-week old mice was
reported to show that in young wild-type livers, the majority of nuclei had a diploid DNA
content with a smaller proportion of tetraploid nuclei. As the mice aged, the number of
tetraploid and octoploid nuclei increased consistent with the previous findings of others.
However, c-Myc mice were reported to demonstrate a premature polyploidization with
the number of 2N nuclei in c-Myc livers almost 2-fold less, while the proportion of 4N nuclei
increased more than 2.5-fold at 4 weeks of age. The most prominent ploidy alteration was an
increase in the fraction of hepatocytes with octaploid nuclei (~200-fold higher). The percentage
of polyploidy cells was reported to continue to rise in 15 week old c-Myc livers. The majority of
hepatocytes had nuclei with 4N and 8N DNA content, with an attendant increase in binucleated
hepatocytes and increase in average cell size.
In striking contrast, E2F1 hepatocytes were reported not to undergo normal
polyploidization with aging. The majority of E2F1 nuclei were reported to remain in the diploid
state and to be almost identical in E2F1 mice at 4 and 15 weeks of age. The percentage of
binucleated hepatocytes was also reduced. In c-Myc/E2Fl mice, the age-related changes in
ploidy distribution were reported to resemble those found in both c-Myc and in E2F1 single
transgenic mice.
At a young age, c-Myc/E2Fl mice, similar to E2F1 mice, were reported to retain
significantly more diploid nuclei than c-Myc mice. However, as mice aged, the majority of c-
Myc/E2F1 hepatocytes, similar to c-Myc cells but in contrast to findings in E2F1 cells, became
polyploid. Consistent with a more progressive polyploidization, the DNA content was
significantly higher in both c-Myc/E2Fl and c-Myc livers. Conner et al. (2003) reported that
other known modulators of ploidy in the liver are the tumor suppressor p53, pRb, and the cell
cycle inhibitor p21 as well as, genes involved in the control of the cell cycle progression such as
cyclin A, cyclin B, cyclin D3, and cyclin E.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-353 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Along with increased liver cancer, Conner et al. (2003) noted that the C-Myc mice also
experienced a persistent liver injury as evidenced by significant elevation of circulating levels of
aspartate aminotransferase, alanine aminotransferase, and alkaline phosphatase along with the
appearance of a frequent oval/ductular proliferation. However, oval cell proliferation may be a
marker of hepatocyte damage but not be the cells responsible for tumor induction (Tarsetti et al.,
1993). Conner et al. (2000) reported that if E2F1 is overexpressed in the liver, there is both
oncogenic and tumor-suppressive properties. In regard to liver morphological changes, E2F1
transgenic mice were reported to uniformly develop pericentral dysplasia and foci adjacent to
portal tracts followed by the abrupt appearance of adenomas and subsequent malignant
conversion with all of the animals having foci by 2-4 months and by 8-10 months most having
adenomas with dysplastic changes remaining confined to the pericentral regions of the liver
lobule.
In regard to phenotype, the majority of the foci were composed of small round cells, with
clear-cell phenotype but eosinophilic, mixed, and basophilic foci were also seen. In adenomas
with malignant transformation to HCC, there appeared to be high mitotic indices, blood vessel
invasion, and central collection of deeply basophilic cells with large nuclei giving a "nodule- in-
nodule" appearance. Macrovesicular hepatic steatosis was first noted in some E2F1 transgenic
livers at 6-8 months and by 10-12 months 60% of animals had developed prominent fatty
change. Hepatic steatosis has been noted in several transgenic mouse models of liver
carcinogenesis (Conner et al., 2000). These results raise interesting points of regional difference
in tumor formation which can be lost in analyses using whole liver and that the phenotype of foci
and tumors are similar to those seen from chemical carcinogenesis. The occurrence of
hepatotoxicity in these transgenic mice is also of note.
E.3.3.2. Hepatocellular Proliferation and Increased DNA Synthesis
Caldwell et al. (2008a) presented a discussion of the role of proliferation in cancer
induction. They stated that
in the case of CC14 exposure, hepatocyte proliferation may be related to its ability
to induce liver cancer at necrogenic exposure levels, but the nature of this
proliferation is fundamentally different from peroxisome proliferators or other
primary mitogens that cause hepatocyte proliferation without causing cell death
(Columbano and Ledda-Columbano, 2003; Coni et al., 1993; Ledda-Columbano
et al., 1993; Ledda-Columbano et al., 1998; Ledda-Columbano et al., 2003;
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 54 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Menegazzi et al., 1997). After initiation with a mutagenic agent, the transient
proliferation induced by primary mitogens has not been shown to lead to cancer-
induction, while partial hepatectomy or necrogenic treatments of CC14 result in
the development of tumors (Gelderblom et al., 2001; Ledda-Columbano et al.,
1993).
Roskams et al. (2003) noted that partial hepatectomy does not cause hepatocellular carcinoma in
normal mice without initiation. Melchiorri et al. (1993) reported that a series of studies has shown that
acute proliferative stimuli provided by primary mitogens, unlike those of the regenerative type such as
those elicited by surgical or chemical partial hepatectomy, do not support the initiation phase and do not
effectively promote the growth of initiated cells (Columbano et al., 1990; Columbano et al., 1987; Ledda-
Columbano et al., 1989). They noted that, the finding that most of these chemicals, with the exception of
WY, induce only a very transient increase in cell proliferation raises the question whether such a transient
induction of liver cell proliferation might be related to liver cancer appearing 1-2 years later. They noted
that mitogen-induced liver growth differs from compensatory regeneration in several aspects (1) it does
not require an increased expression of hepatocyte growth factor mRNA in the liver (2) it is not necessarily
associated with an immediate early genes such as c-fos and c-jun; (3) it results in an excess of tissue and
hepatic DNA content that is rapidly eliminated by apoptotic cell death following withdrawals of the
stimulus.
Other studies have questioned the importance of a brief wave of DNA synthesis in induction of
liver cancer. Chen et al. (1995) noted that Jirtle et al. (1991) and Schulte-Hermann et al. (1986) reported
that during a 2-week period of treatment with lead, DNA synthesis was increased most in centrolobular
hepatocytes and that the predominantly centrilobular distribution of the labeled nuclei may have been due
largely to the brief wave of mitogenic response, because from the fifth day onward DNA synthesis
activity returned to control level even though lead nitrate treatment continued. They concluded that
sustained cell proliferation may be more important than a brief wave of increased DNA synthesis. Chen
et al. (1995) also noted that a number of different agents acting via differing MOAs will induce periportal
proliferation.
Vickers and Lucier (1996) reported that mitogenic response induced by acute 17
a-ethinylestradiol administration is randomly distributed throughout the hepatic lobule, while continuous
administration increases the proportion of diploid cells. Richardson et al. (1986) reported that the lobular
distribution of the correlation of hepatocyte initiation and akylation reported in their model of
carcinogenicity did "not support that early proliferation is associated with cancer as at 7 days there is a
transient increase in the lobes least likely to get a tumor and no difference between the lobes at 14 and 28
days DEN although there is a difference in tumor formation between the lobes." Thus, cells undergoing
DNA synthesis may not be in the same zone of the liver where other hypothesized "key events" take
place.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-355 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Tanaka et al. (1992) noted that the distribution of hepatocyte proliferation in the periportal area
was in contrast to the distribution of peroxisome proliferation in the centrilobular area of Clofibrate
treated rats. Melnick et al. (1996) noted that replicative DNA synthesis commonly has been evaluated by
measurement of the fraction of cells incorporating BrdU or tritiated thymidine into DNA during S-phase
of the cell cycle (S-phase labeling index), but that the S-phase labeling index would not be identical to the
cell division rate when replication of DNA does not progress to formation of two viable daughter cells.
"The general view at an international symposium on cell proliferations and chemical carcinogenesis was
that although cell replication is involved inextricably in the development of cancers, chemically enhanced
cell division does not reliably predict carcinogenicity" (Melnick et al., 1993). They noted that the finding
that enzyme-altered hepatic foci were not induced in rats fed WY-14,643 for 3 weeks followed by partial
hepatectomy indicates that early high levels of replicative DNA synthesis and peroxisome proliferation
are not sufficient activities for initiation of hepatocarcinogenesis.
Baker et al. (2004) reported that, similar to the pattern of transient increases in DNA synthesis
reported for TCE metabolites, Clofibrate exposure induced the upregulation of a variety of cell
proliferation-associated genes (e.g., G2/M specific cyclin Bl, cyclin-dependent kinase 1, DNA
topoisomerase II alpha, c-myc protooncogene, pololike serien-threonine protein kinase, and cell divisions
control protein 20) began on or before Day 1 and peaked at some point between days 3 and 7. By Day 7,
cell proliferation genes were down regulated. The chronology of this gene expression agrees with the
histologic diagnosis of mitotic figures in the tissue, where an increase in mitotic figures was detected in
the Day 1 and most notably Day 3 high and low-dose groups. However, by Day 7, the incidence of
mitotic figures had decreased. The clustering of genes associated with the G2/M transition point suggests
that in the rats, the polyploid cells arrested at G2/M are those that are proceeding through the cell cycle.
A dose-response for increased DNA-synthesis also seems to be lacking for the model
PPARa agonist, WY-14,643 suggesting that the transient increases in DNA synthesis reported by
Eacho et al.(1991) for this compound at lower levels that then increase later at necrogenic
exposure levels, are not related to its carcinogenic potential. Wada et al. (1992) reported that in
male Fischer 344 rats exposed to a range of WY-14,643 concentrations (5-1,000 ppm) that liver
weight gain occurred at the lowest dose that gave a sustained response for many weeks but gave
increased cell labeling only in the first week. Peroxisomes proliferation, as measure by electron
microscopy, increases started at 50 ppm exposures. By enzymatic means, peroxisomal activities
were elevated at the 5 ppm dose. Of note is the reported difference in distribution in
hepatocellular proliferation, which was not where the hypertrophy or where the lipofuscin
increases were observed. The authors noted that these data suggest that 50 and 1,000 ppm WY-
14,643 should give the same carcinogenicity if peroxisome proliferation or sustained
proliferation are the "key events."
The study of (Marsman et al., 1992) is very important in that it not only shows that
clofibric acid (another PPAR a agonist) does not have sustained proliferation, but it also shows
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-356 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
that it and WY-14,643 at 50 ppm did not induce apoptosis in rats. It is probable that use of WY-
14,643 at high concentrations may induce apoptosis in a manner not applicable to other
peroxisome proliferators or to treatment with WY-14,643 at 50 ppm. This study also confirmed
that exposure to WY-14,643 at 50 ppm and WY-14,643 at 1,000 ppm induces similar effects in
regards to hepatocyte proliferation and peroxisomal proliferation.
The study by Eacho et al. (1991) also gave a reference point for the degree of hepatocytes
undergoing transient DNA synthesis from WY-14,643 and Clofibrate and how much smaller it is
for TCE and its metabolites, which generally involve less than 1% of hepatocytes.
The labeling index of BrdU was 7.2% on day 3 and 15.5% on day 6 after clofibric
acid but by day 10 and 30 labeling index was the same as controls at —1-2%... .For
WY the labeling index was 34.1% at day 3 and 18.6% at day 6. At day 10 the
labeling index was 3.3% and at day 30 was 6%, representing 6.6- and 15-fold of
respective controls. Control levels were -0.5 to 1%.... The labeling index was
increased to 32% by 0.3% LY171883 and to 52% by 0.05% Nafenopin. The
0.005%) and 0.1% dietary doses of WY increased the 7 day labeling index to a
comparable level (55% - 58%).
Yeldandi et al. (1989) reported that until foci appear, cell proliferation has ceased to
increase over controls after the first week for ciprofibrate-induced hepatocarcinogenesis. The
results also showed the importance of using age matched controls and not pooled controls for
comparative purposes of proliferation as well as how low proliferative rates are in control
animals.
The results of Barrass et al. (1993) are important in suggesting that age of animals is
important when doing quantitation of labeling indexes. Studies such as that conducted by
Pogribny et al. (2007) that only give the replication rate as a ratio to control will make the
proliferation levels look progressive when in fact they are more stable with time as it is just the
controls that change with age as a comparison point.
E.3 .3 .3. Nonparenchymal Cell Involvement in Disease States Including Cancer
The recognition that not only parenchymal cells but also nonparenchymal cells play a
role in HCC has resulted in studies of their role in initiation as well as progression of neoplasia.
The role of the endothelial cell in controlling angiogenesis, a prerequisite for neoplastic
progression, and the role of the Kupffer cell and its regulation of the cytokine milieu that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-357 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
controls many hepatocyte functions and responses have been reported. However, as pointed out
by Pikarsky et al. (2004) and by the review by Nickoloff et al. (2005), the roles of inflammatory
cytokines in cancer are context and timing specific and not simple. For TCE, nonparenchymal
cell proliferation has been observed after inhalation (Kjellstrand et al., 1983a) and gavage (Goel
et al., 1992) exposures of ~4 weeks duration.
E.3 .3 .3 .1. Epithelial cell control of liver size and cancer—angiogenesis.
The epithelium is key in controlling restoration after partial hepatectomy and not
surprisingly HCC growth. Greene et al. (2003) hypothesized that the control of physiologic
organ mass was similar to the control of tumor mass in the liver and that specifically, the
proliferation of hepatocytes after partial hepatectomy, like the proliferations of neoplastic cells in
tumors, requires the synthesis of new blood vessels to support the rapidly increasing mass. They
reported that a peak in hepatocyte production of vascular endothelial growth factor (VEGF), an
endothelial mitogen, corresponds to an increase of VEGF receptor expression on endothelial
cells after partial hepatectomy and the rate of endothelial proliferation. Fibroblast growth factor
and transforming growth factor-alpha (TGfox), which stimulate endothelial cells, are secreted by
hepatoctyes 24 hours after partial hepatectomy. However, endothelial cells were reported to
secrete hepatocyte growth factor, a potent hepatocyte mitogen, that is also proangiogenic. The
secretion of transforming growth factor -beta by (TGfox) endothelial cells 72 hours after partial
hepatectomy was reported to inhibit hepatocyte proliferation. Thus, Greene et al. (2003)
suggested that endothelial cells and hepatocytes of the regenerating liver influence each other,
and both populations are required for the regulation of the regenerative process.
E.3 .3 .3 .2. Kupffer cell control of proliferation and cell signals, role in early and late effects
Vickers and Lucier (1996) have reported that Kupffer cells are increased in number in
prenoplastic foci but are decreased in hepatocellular carcinoma, and that other studies have
demonstrated that both sinusoidal endothelial cells and Kupffer cells within hepatocellular
carcinoma cells in humans stain positive for mitotic activity although the number of
nonparenchymal cells compared to parenchymal cells may be reduced. Lapis et al. (1995)
reported that Kupffer cells contain lysozyme in their cytoplasmic granules, vacuoles and
phagosomes, some cells show a positive reaction in the rough endoplasmic reticulum,
perinuclear cisternae and the Golgi zone, and that in human monocytes the lysozyme is
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-358 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
colocalized with the CD68 antigen and myeloperoxidase. They also reported that, in rodent
hepatocarcinogenesis, increased numbers of Kupffer cells were observed in preneoplastic foci,
whereas abnormally low numbers were present following progression to hepatocellular
carcinoma. They also noted that "the Kupffer cell count in human HCC has also been shown to
be very low and varies with different histological form." They reported that for monkey HCCs,
that the proportion of endothelial elements remained constant (the parenchymal/endothelial cell
ratio), however, there was a striking reduction in the areas occupied by Kupffer cells. While
healthy control livers contained the highest number of Kupffer cells, in the tumor-bearing cases
the nonneoplastic, noncirrhotic liver adjacent to the HCC nodules had a significantly lower
number of Kupffer cells and the number decreased further in the nonneoplastic portions of
cirrhotic livers. Within HCC nodules the Kupffer cell count was greatly reduced with no
significant changes were observed between the cirrhotic areas and the carcinomas, however, the
tumors contained fewer lysozyme and CD68 positive cells. Lapis et al. (1995) noted that
since other cell types within the liver sinusoids (monocytes and polypmorphs) and
portal macrophage were also positive, it was important to identify the star-like
morphology of the Kupffer cells. The results of the two independent observers
assessment of the morphology and enumeration of Kupffer cells were quite
consistent and differed by only 3%." "The loss of Kupffer cells in the HCC may
possibly result from capillarization of the sinusoids, which has been observed
during the process of liver cirrhosis and carcinogenesis. Capillarization entails the
sinusoidal lining endothelial cells losing their fenestrations.
E.3 .3 .3 .3. Nf-kB and TNF-a - context, timing and source of cell signaling molecules
A large body of literature has been devoted to the study of nuclear factor k B for its role
not only in inflammation and a large number of other processes, but also for its role in
carcinogenesis. However, the effects of these cytokines are very much dependent on their
cellular context and the timing of their modulation. As described by Adli and Baldwin (2006),
The classic form of NF-kB is composed of a heterodimer of the p50 and p65
subunits, which is preferentially localized in the cytoplasm as an inactive complex
with inhibitor proteins of the IkB family. Following exposure to a variety of
stimuli, including inflammatory cytokines and LPS, IkBs are phosphorylated by
the IKKa/p complexes then accumulate in the nucleus, where they
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-359 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
transcriptionally regulate the expression of genes involved in immune and
inflammatory responses.
The five members of the mammalian NF-kB family, p65 (RelA), RelB, c-Rel, P50/pl05
(NF-KB1) and p52/pl00 (NF-kB2), exist in unstimulated cells as homo- or heterodimers bound
to IkB family proteins. Transcriptional specificity is partially regulated by the ability of specific
NF-kB dimmers to preferentially associate with certain members of the IkB family. Individual
NF-kB responses can be characterized as consisting of waves of activation and inactivation of
the various NF-kB members (Hayden and Ghosh, 2004). While the function of NF-kB in many
contexts have been established, it is also clear that there is great diversity in the effects and
consequences of NF-kB activation with NF-kB subunits not necessarily regulating the same
genes in an identical manner and in all of the different circumstances in which they are induced.
The context within which NF-kB is activated, be it the cell type or the other stimuli to which the
cell is exposed, is therefore, a critical determinant of the NF-kB behavior (Perkins and Gilmore,
2006).
Balkwill et al. (2005) reported that
the NF-kB pathway has dual actions in tumor promotion: first by preventing cell
death of cells with malignant potential, and second by stimulating production of
proinflammatory cytokines in cells of infiltrating myeloid and lymphoid cells.
The proinflammatory cytokines signal to initiated and/or otherwise damaged
epithelial cells to promote neoplastic cell proliferation and enhance cell survival.
However, the tumor promoting role of NF-kB may not always predominate. In
some cases, especially early cancers, activation of this pathway may be tumor
suppressive (2004). Inhibiting NF-kB in keratinocytes promotes squamous cell
carcinogenesis by reducing growth arrest and terminal differentiation of initiated
keratinocytes (Seitz et al., 1998).
Other inflammatory mediators have also been associated with oncogenesis. Balkwill et
al. (2005) reported that TNFa is frequently detected in human cancers (produced by epithelial
tumor cells, as in for instance, ovarian and renal cancer) or stromal cells (as in breast cancer).
They also report that the loss of hormonal regulation of IL-6 is implicated in the pathogenesis of
several chronic diseases, including B cell malignancies, renal cell carcinoma, and prostate,
breast, lung, colon, and ovarian cancers. Over 100 agents, such as antioxidants, proteosome
inhibitors, NSAIDs, and immunosuppressive agents are NF-kB inhibitors with none being
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 60 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
entirely specific (Balkwill et al., 2005). Thus, alterations in these cytokines, and the cells that
produce them, are implicated as features of "cancer" rather than specific to HCC.
Balkwill et al. (2005) reported that
Two mouse models of inflammation-associated cancer now implicate the gene
transcription factor NF-kB and the inflammatory mediator known as tumor-
necrosis factor a (TNF- a) in cancer progression. Using a mouse model of
inflammatory hepatitis that predisposes mice to liver cancers, Pikarsky et al.
present evidence that the survival of hepatocytes - liver cells - and their
progression to malignancy are regulated by NF-kB. NF-kB is an important
transcription factor that controls cell survival by regulating programmed cell
death, proliferation, and growth arrest. Pikarsky et al. find that the activation state
of NF-kB, and its localization in the cell, can be controlled by TNF-a produced by
neighboring inflammatory cells (collectively known as stromal cells).
Pikarsky et al. (2004) reported that that the inflammatory process triggers hepatocyte NF-
kB through upregulation of TNF-a in adjacent endothelial and inflammatory cells. Switching off
NF-kB in mice from birth to seven months of age, using hepatocyte-specific inducible IicB-super
repressor transgene, had no effect on the course of hepatitis, nor did it affect early phases of
hepatocyte transformation. By contrast, suppressing NF-kB inhibition through anti-TNF-a
treatment or induction of the IicB-super repressor in later stages of tumor development resulted in
apoptosis of transformed hepatocytes and failure to progress to hepatocellular carcinoma. The
Mdr2 knockout hepatocytes in Pikarsky's model of hepatocarcinogenicity were distinguishable
from wild-type cells by several abnormal features; high proliferation rate, accelerated
hyperploidy and dysplasia. Pikarsky et al. (2004) reported that NF-kB knockout and double
mutant mice displayed comparable degrees of proliferation, hyperploidy and dysplasia implying
that NF-kB is not required for early neoplastic events. Thus, activation of NF-kB was not
important in the early stages of tumor development, but was crucial for malignant conversion.
It was noted that
Greten et al. reporting in Cell, come to a similar conclusion by studying a mouse colitis-
associated cancer model. Their work does not directly implicate TNF-a, but instead
found enhanced production of several pro-inflammatory mediators (cytokines) including
TNF-a„ in the tumor microenvironment during the development of cancer. An important
feature of both studies is that NF-kB activation was selectively ablated in different cell
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-361 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
compartments in developing tumor masses, and at different stages of cancer
development.
Balkwill et al. (2005) also noted that TNF-a and NF-kB have many different effects, depending
on the context in which they are called into play and the cell type and environment.
In contrast, El-Serag and Rudolph (2007) noted that "the influence of inflammatory
signaling on hepatocarcinogenesis can be context dependent; deletion of Nf-KB-dependent
inflammatory responses enhanced HCC formation in carcinogen treated mice (Sakurai et al.,
2006)." Similarly, deletion of Nf-KB essential modulator/I kappa P kinase (NEMO/IKK), an
activator of Nf-KB, induced steatohepatitis and HCC in mice (Luedde et al., 2007).
Maeda et al. (2005) reported that hepatocyte specific deletion of IKKp (which prevents
NF-kB activation) increased DEN-induced hepatocarcinogenesis and that a deletion of IKKp in
both hepatocytes and hematopoietic-derived cells, however, had the opposite effect, decreasing
compensatory proliferation and carcinogenesis. They suggested that these results, differ from
previous suggestion that the tumor-promoting function of NF-kB is excreted in hepatocytes
(Pikarsky et al., 2004), and suggest that chemicals or viruses that interfere with NF-kB activation
in hepatocytes may promote HCC development.
Alterations in NF-kB levels have been suggested as a key event for the
hepatocarcinogenicity by PPARa agonists. The event associated with PPAR effects has been
the extent of NF-kB activation as determined through DNA binding. As reported by Tharappel
(2001), NF-kB activity is assayed with electrophoretic modibility shift assay with nuclear
extracts prepared from frozen liver tissue as a measure of DNA binding of NF-kB. Increase
transcription of downstream targets of NF-kB activity have also been measured. It has been
suggested that PPARa may act as a protective mechanism against liver toxicity. Ito et al. (2007)
cite repression of NF-kB by PPARa to be the rationale for their hypothesis that PPARa-null
mice may be more vulnerable to tumorigenesis induced by exposure to environmental
carcinogens. However, as shown in Section E.3.4.1.2, although DEHP was reported to also
induce glomerularnephritis more often in PPARa-null mice, as suggested (Kamijima et al., 2007)
to be due of the absence of PPARa- dependent anti-inflammatory effect of antagonizing the
oxidative stress and NF-kB pathway, there was no greater or lesser susceptibility to DEHP-
induced liver carcinogenicity in the PPARa null mice.
Because PPARa is known to exert anti-inflammatory effects by inducing expression of
LcBa, which antagonizes NFkB signaling, the expression of iKBa has been measured in some
studies (Kamijima et al., 2007) as well as expression of TNR1 mRNA to evaluate the sensitivity
to the inflammatory response. Ito et al. (2007) reported that in wild-type mice there did not
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 62 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
appear to be a difference between controls and DEHP treatment for p65 immunoblot results.
DEHP treatment was also reported to not induce p65 or p52 mRNA either or influence the
expression levels of TNFa, IkBa, IkBp and IL-6 mRNA in wild-type mice.
Tharappel et al. (2001) treated rats with WY-14,643, gemfibrozil or Dibutyl phthalate
and reported elevated NF-kB DNA binding in rats with WY-14,642 to have sustained response
but not others. WY-14,643 increased DNA binding activity of NF-kB at 6, 34 or 90 days.
Gemfibrozil and DEHP increased NF-kB activity to a lesser extent and not at all times in rats.
For gemfibrozil, there was only a 2-fold increase in binding at 6 days with no increase at 34 days
and increase only in low dose at 90 days. In rats treated with Dibutyl phthalate, there no change
at 6 days, at 34 days there was an increase at high and low dose, at 90 days only low dose
animals showed a change. In pooled tissue from WY-14,643- treated animals, the complex that
bound the radiolabeled NF-kB fragment did contain both p50 and p65. Both WY-14,643 and
gemfibrozil were reported to produce tumors in rats with Dibutyl pthalate untested in rats for
carcinogenicity. Thus, early changes in NF-kB were not supported as a key event and WY-
14,643 to have a pattern that differed from the other PPARa agonists examined.
In regard to the links between inflammation and cancer, Nickoloff et al. (2005) in their
review of the issue, cautioned that such a link is not simple. They noted that
dissecting the mediators of inflammation in cutaneous carcinogenic pathways has
revealed key roles for prostaglandins, cyclooxygenase-2, tumor necrosis factor-a,
AP-1, NF-kB, signal transducer and activator of transcription (STAT)3, and
others. Several clinical conditions associated with inflammation appear to
predispose patients to increased susceptibility for skin cancer including discoid
lupus erythematosus, dystrophic epidermolysis bullosa, and chronic wound sites.
Despite this vast collection of data and clinical observations, however, there are
several dermatological setting associated with inflammation that do not
predispose to conversion to lesions into malaignancies such as psoriasis, atopic
dermatitis, and Darier's disease.
Nickoloff et al. (2005) suggested that such a
link may not be as simple as currently portrayed because certain types of
inflammatory processes in skin (and possibly other tissues as well) may also serve
a tumor suppressor function. Over the past few months, several publications in
leading biomedical journals grappled with an important issue in oncology, namely
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 63 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
defining potential links between chronic tissue damage, inflammation, and the
development of cancer. Balkwill and Coussens (2004) reviewed the role of the
NF-kB signal transduction pathway that can regulate inflammation and also
promote malignancy. Their review summarized the latest findings revealed in a
letter to Nature by Pikarsky et al. (2004). Using Mdr2 knockout mice in which
hepatitis is followed by hepatocellular carcinoma, Pikarsky et al. implicated
TNFa upregulation in tumor promotion of HCC, and suggest that TNFa and NF-
kB are potential targets for cancer prevention in the context of chronic
inflammation. A similar conclusion was reached with respect to NF-kB by an
independent group of investigators using a model of experimental dextran sulfate-
induced colitis, in which inactivation of the IkB kinase resulted in reduced
colorectal tumors (Greten et al., 2004). Although there are many other clinical
condition supporting the concept of inflammation is a critical component of tumor
progression (e.g., reflux esophagitis/esophageal cancer; inflammatory bowel
disease/colorectal cancer), there is at least one notable example that does not fit
this paradigm. As described below, psoriasis is a chronic cutaneous inflammatory
disease, which is seldom if ever accompanied by cancer suggesting the
relationship between tissue repair, inflammation, and development may not be as
simple as portrayed by the aforementioned reviews and experimental results.
Besides psoriasis, other noteworthy observations pointing to more complexity
include the observation that in the Mdr2 knockout mice, we rarely detect bile duct
tumors despite extensive inflammation, NF-kB activation, and abundant
proliferation of bile ducts in portal spaces (Pikarsky et al., 2004). Moreover, in a
skin-cancer mouse model, NF-kB was shown to inhibit tumor formation (Dajee et
al., 2003). Thus, the composition of inflammatory mediators, or the properties of
the responding epithelial cells (e.g., signaling machinery, metabolic status), may
dictate either tumor promotion or tumor suppression. Chronic inflammation and
tissue repair can trigger pro-oncogenic events, but also that tumor suppressor
pathways may be upregulated at various sites of injury and chronic cytokine
networking.
One cannot easily dismiss the many dilemmas raised by the psoriatic
plaque that confound a simple link between the tissue repair, inflammation, and
carcinogenesis. Since it is easily visible to the naked eye, and patients may suffer
from such lesions for decades, it is difficult to argue that various skin cancers
such as squamous cell carcinoma, basal cell carcinoma, or melanoma actually do
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 64 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
develop within plaques by are being overlooked by patients and dermatologists.
Remarkably, psoriatic plaques are intentionally exposed to mutagenic agents
including excessive sunlight, topical administration of crude coal tar, or parenteral
DNA cross-linking agent -psoralen followed by ultraviolet light. Moreover these
treatments are known to induce skin cancer in nonlesional skin. Thus since
psoriatic skin is characterized by altered differentiation, angiogenesis, increased
telomerase activity, proliferative changes, and apoptosis resistance, one would
expect that each and every psoriatic plaque would be converted to cancer, or at
least serve as fertile soil for the presence of non-epithelial skin cancers over
time... .In conclusion, it would seem prudent to remember the paradigm proposed
by Weiss (1971) in which he suggested that premalignant cells do not comprise an
isolated island, but are a focus of intense tissue interactions. The myriad
inflammatory effects of the tumor microenvironment are important for
understanding tumor development, as well as tumor suppression and senescence,
and for the design for efficacious prevention strategies against inflammation-
associate cancer (Nickoloff et al., 2005).
E.3.3.4. Gender Influences on Susceptibility
As discussed previously, male humans and rodents are generally more likely to get HCC.
The increased risk of liver tumors from estrogen supplements in women has been documented.
In mice male TCE exposure has been shown to have greater variability in response and greater
effects on body weight in males (Kjellstrand et al., 1983a; Kjellstrand et al., 1983b) but to also
induce dose-related increases in liver weight and carcinogenic response in female mice as well as
males (see Section E.2.3.3.2). Recent studies have attempted to link differences in inflammatory
cytokines and gender differences in susceptibility.
Lawrence et al. (2007) suggested that
studies of Naugler et al. (2007) and Rakoff-Nahoum and Medzhitov (2007),
advance our understanding of the mechanisms of cancer-related inflammation.
They describe an important role for an intracellular signaling protein called
MyD88 in the development of experimental liver and colon cancers in mice.
MyD88 function has been well characterized in the innate immune response
(Akira and Takeda, 2004), relaying signals elicited by pathogen-associated
molecules and by the inflammatory cytokine interleukin-1 (IL-1).... The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-365 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
conclusion from Naugler et al. (2007) and Rakoff-Nahoun and Medzhitov is that
MyD88 may function upstream of NF-kB in cells involved in inflammation-
associated cancer. Immune cells infiltrate the microenvironment of a tumor.
Naugler et al. (2007) and Rakoff-Nahoun and Medzhitov (2007) suggest that the
development of liver and intestinal cancers in mice may depend on a signaling
pathway in infiltrating immune cells that involved the protein MyD88, the
transcription factor NF-kB, and the pro-inflammatory cytokine 11-6. TLR binds a
ligand which acts on MyD88 which acts on NF-kB which leads to secretion of
inflammatory cytokine IL-6 which leads to promotion of tumor cell survival and
proliferation.
Naugler et al. (2007) suggested gender disparity in MyD88-dependent IL-6 production was
linked to differences in cancer susceptibility using the DEN model (a mutagen with concurrent
regenerative proliferation at a single high dose) with a single injection of DEN. Partial hepatectomy was
reported to induce no gender-related difference in IL-6 increase. After DEN treatment the male mouse
had 275 ng/mL as the peak IL-6 levels 12 hours after DEN and for female mice the peak was reported to
be 100 ng/mL 12 hours after DEN administration. This is only about a 2.5-fold difference between
genders. 11-6 mRNA induction was reported for mice 4 hours after DEN while at 4 hours, at a time when
there was no difference in serum IL-6 between male and female mice. It was not established that the 4-
hour results in mRNA translated to the differences in serum at 12 hour between the sexes. The magnitude
of mRNA differences does not necessarily hold the same relationship as the magnitude in serum protein.
In fact, there was not a linear correlation between mRNA induction and IL-6 serum levels.
A number of issues complicate the interpretation of the results of the study. The study examined
an acute response for the chronic endpoint of cancer and may not explain the differences in gender
susceptibility for agents that do not cause necrosis. The DEN was administered in 15-day old mice
(which had not reached sexual maturity) for tumor information at a much lower dose than used in short-
term studies of inflammation and liver injury in which mature mice were used. If large elevations of IL-6
are the reason for liver cancer, why does not a partial hepatectomy induce liver cancer in itself?
The percentage of proliferation at 36 and 48 hours after partial hepatectomy was the same
between the sexes. If a 2.5-fold difference in IL-6 confers gender susceptibility, it should do so after
partial hepatectomy and lead to cancer. For female mice, partial hepatectomy showed alterations in a
number of parameters. However, partial hepatectomy does not cause cancer alone. The 5-fold increase 4
hours after DEN induction of IL-6 mRNA in male mice is in sharp contrast to the 27-fold induction of IL-
6 1 hour after partial hepatectomy (in which at 4 hours the IL-6 had diminished to 6-fold). There
appeared to be variability between experiments. For example, the difference in males between
experiments appears to be the same magnitude as the difference between male and female in one
experiment and the baseline of IL-6 mRNA induction appeared to be highly variable between experiments
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 66 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
as well as absolute units of ALT in serum 24 and 48 hours after DEN treatment that tended to be greater
that the effects of treatments. The experiments used very few animals (n = 3) for most treatment groups.
Of note is that the MyD88 -/- male mice still had a background level of necrosis similar to that of WT
mice at 48 hours after DEN treatment, a time, long after the peak of IL-6 mRNA induction and IL-6
serum levels were reported to have peaked.
One of the key issues regarding this study is whether difference in IL-6 reported here lead to an
increase proliferation and does that difference within 48 hours of a necrotizing dose of a carcinogen
change the susceptibility to cancer? This report shows that male and female mice have a difference in
necrosis after CCL4 and a difference in proliferation. Are early differences in IL-6 at 4 hours related to
the same kind of stimulus that leads to necrosis and concurrent proliferation? The amount of proliferation
(as measured by DNA synthesis) between male and female mice 48 hours after DEN was very small and
the study was conducted in a very few mice (n = 3). At 36 hours the degree of proliferation was almost
the same between the genders and about 0.6% of cells. The baseline of proliferation also differed
between genders but the variation and small number of animals made it insignificant statistically. At 48
hours the differences in proliferation between male and female mouse were more pronounced but still
quite low (2% for males and -1% for females). Is the change in proliferation just a change in damage by
the agent? Given the large variation in serum ALT and by inference necrosis, is there an equal amount of
variability in proliferation? This study gives only limited information for DEN treatment.
The difference in incidence of HCC was reported to be greater than that of "proliferation"
between genders and of other parameters although differences in tumor multiplicity or size between the
genders are never given in the paper. Most importantly, comparisons between the short-term changes in
cytokines and indices of acute damage are for adult animals that are sexually mature and at doses that are
4 times (100 vs. 25 mg/kg) that of the sexually immature animals who are going through a period of rapid
hepatocyte proliferation (15 day old animals).
It is therefore, difficult to extrapolate between the two paradigms to distinguish the effects of
hormones and gender on the response. Finally, the work of Rakoff-Nahoum and Medzhitov (2007)
showed that it is the effect of tumor progression and not initiation that is affected by MyD88 (a signaling
adaptor to Toll-like receptors). Thus, examination of parameters at the initiation phase at necrotic doses
for liver tumors may not be relevant.
E.3.3.5. Epigenomic Modification
There are several examples of chemical exposure to differing carcinogens that have lead
to progressive loss of DNA methylation (i.e., DNA hypomethylation) including TCE and its
metabolites. The evidence for TCE and its metabolites is specifically discussed in
Section E.3.4.2.2, below. Other examples of carcinogens exposures or conditions that have been
noted to change DNA methylation are early stages of tumor development include ethionine
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 67 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
feeding, phenobarbitol, arsenic, dibromoacetic acid, and stress. However, it has not yet been
established whether epigenetic changes induced by carcinogens and found in tumors play a
causative role in carcinogenesis or are merely a consequence of the transformed state (Tryndyak
et al., 2006).
Pogribny et al. (2007) reported the effects of WY-14,643 on global mouse DNA
hypomethylation exposed at 1,000 ppm for 1 week, 5 weeks, or 5 months. What is of particular
note in this study is that at this exposure level, one commonly used for MOA studies using
WY-14,643 to characterize the effects of PPARa agonists as a class, there was significant
hepatonecrosis and mortality reported by Woods et al. (2007a).
Both wild-type and PPARa -/- null mice were examined. In wild-type mice DNA
syntheses was elevated 3-, 13-, and 22-fold of time-matched controls after 1 week, 5 weeks, and
5 months of WY 14,543 treatment. Changes in ploidy were not examined. After 5 weeks of
exposure, the ratio of unmethylated CpG cites in whole liver DNA was the same for WY-14,643
treatment and control but by 5 months there was an increase in hypomethylation in WY-14,643
treated wild-type mice. The authors did not report whether foci were present or not which could
have affected this result. The similarity in hypomethylation at 5 days and 5 weeks, a time point
that also had a small probability of foci development, is suggestive of foci affecting the result at
5 months.
For PPAR -/- mice there was increased hypomethylation reported at 1 week and 5 weeks
after WY-14,643 treatment that was not statistically significant with so few animals studied. At
5 months the null mice had decreased hypomethylation compared to 1 and 5 weeks. The authors
noted that, methylation of c-Myc genes was reported to not be affected by long-term dietary
treatment with WY-14,643 even though WY-14,643-related hypomethylation of c-Myc gene
early after a single dose of WY-14,643 has been observed (Ge et al., 2001a). The authors
concluded "thus, alterations in the genome methylation patterns with continuous exposure to
nongenotoxic liver carcinogens, such as WY, may not be confined to specific cell proliferation-
related genes."
Pogribny et al. (2007) reported Histone H3 and H4 trimethylation status in wild-type and
PPAR null mice to show a rapid and sustained loss of histone H3K9 and histone H4K20
trimethylation in wild-type mice fed WY-14,643 from 1 week to 5 months. There was no
progressive loss in histone hypomethylation, with the same amount of demethylation occurring
at 5 days, 5 weeks, and 5 months in wild-type mice fed WY-14,643. The change from control
was -60% reduction. The control values with time were not reported and all controls were
pooled to give one value (n = 15). For PPAR -/-1 mice there was a slight decrease with WY-
14,643 treatment (-15%) reported. In wild-type mice, WY-14,643 treatment was reported to
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-368 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
have no effect on the major histone methyltransferase, Suv39hl, while expression of another
(PRDM/Rizl) increased significantly as early as on week of treatment and remained elevated for
up to five months. The effect on expression of Suv420h2 (responsible for histone H4K20
trimethylation) was more gradual and the amounts of this protein in livers of mice fed Wy-
14m643 were reported to be lower than in control.
The authors did not examine these parameters in the null mice so the relationship of
these effects to receptor activation cannot be determined. Pogribny et al. (2007) reported
hypomethylation of retroelements (LTR IAP, LINE1 and LINE2 retrotransposons) following
long-term exposure to WY-14,643, which the authors concluded, can have effects on the stability
of the genome. Again, these results are for whole liver that may contain foci.
Nevertheless, these findings raise questions about other target organs and a more general
mechanism for WY-14,643 effects than a receptor mediated one. The lack of effects on c-Myc
and the irrelevance of the transient proliferation through it reported here gives more evidence of
the irrelevance of a MOA dependent on transient proliferation. The authors noted that studies
show that a sustained loss of DNA methylation in liver is an early and indispensable event in
hepatocarcinogenesis induced by long-term exposure of both genotoxic and nongenotoxic
carcinogens in rodents. Thus, this statement argues against making such a distinction in MOA
for "genotoxic" and "nongenotoxic" carcinogens. Finally, the use of a dose which Woods et al.
(2007a) demonstrate to have significant hepatonecrosis and mortality, limits the interpretation of
these results and their relevance to models of carcinogenesis without concurrent necrosis.
Strain sensitivity to hepatocarcinogenicity has been investigated in terms of short-term
changes in methylation. Bombail et al. (2004) reported that a tumor-inducing dose of
phenobarbital reduced the overall level of liver DNA methylation in a tumor-sensitive (B6C3F1)
mouse strain but that the same dose of phenobarbital did not alter global methylation level in a
more tumor-resistant strain (C57BL/6), although the compound increased hepatocyte
proliferation as measured by increased DNA synthesis in both strains (Counts et al., 1996).
Bombail et al. reported that "In a similar study, Watson and Goodman (2002) used a PCR-based
technique to measure DNA methylation changes specifically in GC-rich regions of the mouse
genome." Watson and Goodman (2002) found that, that in these areas of the genome, exposure
to phenobarbital caused an increase in methylation in dosed animals compared with control
animals. Again, the change was more pronounced in tumor-prone C3H/He and B6C3F1 strains
than in the less sensitive C57BL/6 strain. They also reported increased DNA synthesis in
C57BL/6 mice but decreased global methylation in the B6C3F1 strain after PB administration
1-2 weeks. The lifetime spontaneous tumor rates were reported to be less than 5% in C57BL/6
mice but up to 80% in C3H/He mice.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 69 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Counts et al. (1996) reported cell proliferation and global hepatic methylation status in
relatively liver tumor susceptible B6C3F1 with relatively resistant C57BL6 mice following
exposure to PB and/or chlorine/methionine deficient (CMD) diet. Cell proliferation (i.e, DNA
synthesis) was reported to be higher in C57BL/6 mice while transient hypomethylation occurred
to a greater extent in B6C3F1 mice after phenobarbital treatment. Dual administration of CMD
and PB led to enhanced cell proliferation and greater global hypomethylation with similar trends
in terms of strain sensitivities in comparison to with either treatment alone (i.e., greater increase
in cell proliferation in C57BL/6 and greater levels of hypomethylation in B6C3F1). Thus, the
authors concluded that B6C3F1 mice have relatively low capacity to maintain the nascent
methylation status of their hepatic DNA.
However, on the whole, the control values for methylation for the C57BL/6 mice appear
to be slightly higher than the B6C3F1 mice. Claims that the liver tumor sensitive B6C3F1 had
more global hypomethylation after a promoting stimulus, which could be related to tumor
sensitivity, are tempered by the fact that resistant strain had a higher control baseline of
methylation. The baseline level of LI or hepatocyte proliferation also appears to be slightly
higher in the C57BL/6 mouse. In addition, the largest strain difference in hypomethylation after
a CMD diet was at Week 12 (135% of control for the B6C3F1 strain and 151% of control for the
C57BL/6 strain) and this pattern was opposite that for the 1 week time point. Thus, the
suggestion by Counts et al. (1996), that the inability to maintain methylation status by the
B6C3F1 strain, is also not supported by the longer duration data for CMD diet.
E.3.4. Specific Hypothesis for Mode of Action (MOA) of Trichloroethylene (TCE)
Hepatocarcinogenicity in Rodents
E.3 .4.1. PPARa Agonism as the Mode of Action (MOA) for Liver Tumor Induction—The
State of the Hypothesis
PPARa receptor activation has been suggested to be the MOA for TCA liver tumor
induction and for TCE liver tumor induction to occur primarily as a result of the presence of its
metabolite TCA (NRC, 2006). However, as discussed previously (see Section E.2.1.10), TCE-
induced increases in liver weight have been reported in male and female mice that do not have a
functional PPARa receptor (Nakajima et al., 2000; Ramdhan et al., 2010). The dose-response
for TCE-induced liver weight increases differs from that of TCA (see Section E.2.4.2). The
phenotype of the tumors induced by TCE have been described to differ from those by TCA and
to be more like those occurring spontaneously in mice, those induced by DC A, or those resulting
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 70 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
from a combination of exposures to both DCA and TCA (see Section E.2.4.4). As to whether
TCA-induced tumors are induced through activation of the PPARa receptor, the tumor
phenotype of TCA-induced mouse liver tumors has been reported to have a pattern of H-ras
mutation frequency that is opposite that reported for other peroxisome proliferators (see Section
E.2.4.4.; Bull et al., 2002; Fox et al., 1990; Hegi et al., 1993; Stanley et al., 1994). While TCE,
DCA, and TCA are weak peroxisome proliferators, liver weight induction from exposure to these
agents has not correlated with increases in peroxisomal enzyme activity (e.g., PCO activity) or
changes in peroxisomal number or volume. However, liver weight induction from subchronic
exposures appears to be a more accurate predictor of carcinogenic response for DCA, TCA, and
TCE in mice (see Section E.2.4.4). The database for cancer induction in rats is much more
limited than that of mice for determination of a carcinogenic response to these chemicals in the
liver and the nature of such a response.
The MOA for peroxisome proliferators has been the subject of research and debate for
several decades. It has evolved from an "oxidative damage" due to increased peroxisomal
activity to a MOA framework example developed by Klaunig et al. (2003) that described causal
inferences for hepatocarcinogenesis after a chemical exposure was shown to activate of the
PPAR-a receptor with concurrent perturbation of cell proliferation and apoptosis, and selective
clonal expansion. Of note, although inhibition of apoptosis was proposed as part of the sequellae
of PPARa activation, as noted in Section E.2.4.1, no changes in apoptosis in mice exposed to
TCE have been reported with the exception of mild enhanced apoptosis at 1,000 mg/kg/d dose.
More importantly, for mice the rate of apoptosis decreases as mice age and appear to be lower
than that of rats. While DCA exposure has been noted to reduce apoptosis, the significance of
DCA-induced reduction in apoptosis from a level that is already inherently low in the mouse, is
difficult to apply as the MOA for DCA-induced liver cancer.
Klaunig et al. based causal inferences on the attenuation of these events in PPAR-a-null
mice in response to the prototypical agonist WY-14,643 with a number of intermediary events
considered to be associative (e.g., expression of peroxisomal and nonperoxisome genes,
peroxisome proliferation, inhibition of gap junction intracellular communication, hepatocyte
oxidative stress as well as Kupffer cell-mediated events). The data set for DEHP was
prominently featured as an example of "PPAR-a induced hepatocarcinogenesis." For DEHP
PPAR-a activation was described as the initial key event with evidence lacking for a direct effect
but supported primarily supported by evidence from PPAR-a-knockout mice treated with
WY-14,643. Klaunig et al. concluded that".. .all the effects observed are due only to the
activation of this receptor and the downstream events resulting from this activation and that no
other modes of action are operant"
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-371 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Although that PPARa receptor activation is the sole MOA for DEHP has been cited by
several reports (including IARC, 2000), several articles have questioned the adequacy of this
proposed MOA (Caldwell and Keshava, 2006; Caldwell et al., 2008a; Guyton et al., 2009;
Keshava and Caldwell, 2006; Keshava et al., 2007; Melnick et al., 2007) FIFRA SAP, 2004;.
New information is now available that also questions several of the assumptions inherent in the
proposed MOA by Klaunig et al. and the dismissal of PPARa agonists as posing a health risk to
humans. These issues were recently examined in Guyton et al. (2009) and are discussed below.
Furthermore, IARC has recently concluded that additional mechanistic information has become
available, including studies with DEHP in PPAR-a-null mice, studies with several transgenic
mouse strains,carrying human PPARa or with hepatocyte-specific constitutively activated
PPARa and a study in humans exposed to DEHP from the environment that has changed its
conclusions regarding the relevance of rodent tumor data to human risk (Grosse et al.. 2011).
Data from these new studies suggest that many molecular signals and pathways in several cell
types in the liver, rather than a single molecular event, contribute to cancer development in
rodents with IARC concluding that the human relevance of the molecular events leading to
DEHP-induced cancer in several target tissues (eg, liver and testis) in rats or mice could not be
ruled out, resulting in the evaluation of DEHP as a Group-2B agent, rather than Group 3.
Specific questions have been raised about the use of WY-14,643 as a prototype for
PPARa (especially at necrogenic doses) and use of the PPARa -/- null mouse in abbreviated
bioassays to determine carcinogenic hazard.
E.3.4.1.1. Heterogeneity of PPARa agonist effects and inadequacy of WY-14,643 paradigm
as prototype for class. Inferences regarding the carcinogenic risk posed to humans by
PPARa agonists have been based on limited epidemiology studies in humans that were not
designed to
detect such effects. However, as noted by Nissen et al. (2007) the PPARa receptor is pleiotropic,
highly conserved, has "cross talk" with a number of other nuclear receptors, and plays a role in
several disease states. "The fibrate class of drugs, which are PPARa agonists intended to treat
dyslipidemia and hypercholesterolemia, have recently been associated with a number of serious
side effects." While these reports of clinical side effects are for acute or subchronic conditions
and do not (and would not be expected to) be able to detect liver cancer from fibrate treatment,
they clearly demonstrate that compounds activating the PPAR receptors may produce a spectrum
of effects in humans and the difficulty in studying and predicting the effects from PPAR
agonism. Graham et al. (2004) recently reported significantly increased incidence of
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 72 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
hospitalized rhabdomyolysis in patients treated with fibrates both alone and in combination with
statins. Even though pharmaceutical companies have spent a great deal of effort to develop
agonists which are selective for desired effects, the pleiotropic nature of the receptor continues to
be an obstacle.
Also, fibrates, WY-14,643 and other PPARa agonists are pan agonists for other PPARs.
Shearer and Hoekstra (2003) noted that fibrates, including Fenofibrate, Clofibrate, Bezafibrate,
Ciprofibrate, Gemfibrozil, and Beclofibrate are all drugs that were discovered prior to the
cloning of PPARa and without knowledge of their mechanism of action but with optimization of
lipid lowering activity carried out by administration of candidates to rodents. They report that
many PPARa ligands, including most of the common fibrate ligands, show only modest
selectivity over the other subtypes with, for example, fenofibric acid and WY-14,643 showing
<10-fold selectivity for activation of human PPARa compared to PPARy and/or PPARS. In
human receptor transactivation assays they report:
Human receptor transactivation assays of median effective concentration (EC50):
WY-14,643 = 5.0 |im for PPARa, 60 |im for PPAR y, 35 |im for PPARS.
Clofibrate = 55 [j,m for PPARa, -500 [j,m for PPAR y, inactive at 100 [j,m for PPARS
Fenofibrate = 30 jam for PPARa, 300 [j,m for PPAR y, inactive at 100 [j,m for PPARS
Bezafibrate = 50 [j,m for PPARa, 60 [j,m for PPAR y, 20 |iin for PPARS.
Murine receptor transactivation assay of EC50:
WY = 0.63 jam for PPARa, 32 [j,m for PPAR y, inactive at 100 [j,m for PPARS
Clofibrate = 50 [j,m for PPARa, -500 [j,m for PPAR y, inactive at 100 [j,m for PPARS
Fenofibrate =18 |im for PPARa, 250 [j,m for PPAR y, inactive at 100 [j,m for
PPARS
Bezafibrate = 90 [j,m for PPARa, 55 [j,m for PPAR y, 110 jam for PPARS.
Thus, these data show the relative effective concentrations and "potency for PPAR
activity" of various agonists in humans and rodents, rodent and human responses may vary
depending on agonist, agonists vary in what they activate between the differing receptors, and
that there is a great deal of transactivation of these drugs.
For fibrates specifically, a study by Nissen et al. (2007) reported that in current practice,
2 fibrates, Gemfibrozil and Fenobibrate, are still widely used to treat a constellation of lipid
abnormalities known as atherogenic dyslipidemia and note that currently available fibrates are
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 73 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
weak ligands for the PPARa receptor and may interact with other PPAR systems. They noted
that the pharmaceutical industry has sought to develop new, more potent and selective agents
within this class but, most importantly, that none of the novel PPARa agonists has achieved
regulatory approval and that according to a former safety officer in the U.S. Food and Drug
Administration (El-Hage, 2007) that more than 50 PPAR modulating agents have been
discontinued due to various types of toxicity (e.g., elevations in serum creatinine, rhabdomylosis,
"multi-species, multi-site increases in tumor with no safety margin for clinical exposures," and
adverse cardiovascular outcomes) but without scientific publications describing the reasons for
termination of the development programs. Nissen et al. reported differences in effect between a
more highly selective and potent PPARa agonist and the less potent and specific one in humans.
They noted
a recent large study of Fenofibrate in patients with diabetes showed no significant
reduction in morbidity but a trend toward increased all-cause mortality (Keech et
al., 2006; Keech et al., 2005). Whether this potential increase in mortality is
derived from compound specific toxicity of Fenofibrate or is an adverse effect of
PPARa activation remains uncertain."
In addition to the lack of publication of effects from PPAR agonists in human
trials in which toxicity can be examined as noted by Nissen et al., the Keech study
is illustrative of the problem in trying to ascertain liver effects from fibrate
treatment in humans as the focus of the outcomes was coronary events in a study
of 5 years duration in a older diabetic population. As stated above, the challenges
the pharmaceutical industry and the risk assessor face in determining the effects
of PPAR agonists is "that these compounds and drugs modulate the activity of a
large number of genes, some of which produce unknown effects."
Nissen et al. further noted that
Accordingly, the beneficial effects of PPAR activation appear to be associated
with a variety of untoward effects which may include, oncogenesis, renal
dysfunction, rhabdomylosis, and cardiovascular toxicity. Recently, the FDA
began requiring 2-year preclinical oncogenicity studies for all PPAR-modulating
agents prior to exposure of patients for durations of longer than 6 months
(El-Hage, 2007).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 74 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Guyton et al. (2009) further explored the status of the PPARa epidemiological database and
describe its inability to discern a cancer hazard from the available data. Thus, while existing
evidence for liver cancer in humans is null rather than negative, there remains a concern for
oncogenicity and many obstacles for determining such effects through human study. The
heterogeneity in response to PPARa agonists and the heterogeneity of effects they cause
(Keshava and Caldwell, 2006) are evident from these reports.
Many studies have used the effects of WY-14,643 at a very high dose and extrapolated
those findings to PPARa agonists as a class. However, this diverse group of chemicals have
varying potencies and effects for the "key events" described by Klaunig et al. (Keshava and
Caldwell, 2006; 2003). The standard paradigm used with WY-14,643 to induced liver tumors in
all mice exposed to 1 year (an abbreviated bioassay), uses a large dose that has also has been
reported to produced liver necrosis, which can have an effect of cell proliferation and gene
expression patterns, and to also induce premature mortality (Woods et al., 2007a).
As stated above, WY-14,643 also has a short peak of DNA synthesis that peaks after a
few days of exposure, recedes, and then unlike most PPARa agonists studied (e.g., Clofibrate,
clofibric acid, Nafenopin, Ciprofibrate, DEHP, DCA, TCA and LY-171883) has a sustained
proliferation at the doses studied (Barrass et al., 1993; Carter et al., 1995; David et al., 1999;
Eacho et al., 1991; Lake et al., 1993; Marsman et al., 1988; Marsman et al., 1992; Sanchez and
Bull, 1990; Tanaka et al., 1992; Yeldandi et al., 1989). Clofibrate has been shown to have a
decrease in proliferation gene expression shortly after its peak (see Section E.3.2.2).
As shown above for WY-14,643, hepatocellular increases in DNA synthesis did not
appear to have a dose-response (see Section E.3.4.2), only WY-14,643 had a sustained elevation
of Nf-KB (gem and dibutyl phthalate did not) (see Section E.3.4.3.3). The effects on DNA
methylation occurred at 5 months and not earlier time points (when Foci were probably present)
and effects of histone trimethylation were observed to be the same from 1 weeks to 5 months
(see Section E.3.4.5). Such effects on the epigenome suggest other effects of WY-14,643, other
than receptor activation, are not specific to just WY-14,643 and are found in a number of
conditions leading to cancer and in tumor progression (see Sections E.3.2.1 and E.3.2.7.).
In their study of PPARa-independent short-term production of reactive oxygen species
from induced by large concentrations of WY-14,643 and DEHP in the diet, Woods et al. (2007a)
examined short-term exposures to (0.6% w/w DEHP or 0.05% or 500 pm WY-14,643 for 3 days,
1 weeks or 3 weeks) and reported that WY-14,643 induced a dramatic increase in bile flow that
was not observed from DEHP exposure. By 1 week of exposure there was a 5% increase in bile
flow for DEHP treatment but a 240% increase in bile flow for WY-14,643 treatment. By
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-375 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
3 weeks the difference in bile volume between treated and control was 12% for DEHP and
1,100% for WY-14,643 treated animals.
In this study oxygen radical formation, as measured by spin trapping in the bile, was
reported to be decreased after 3 days of treatment after DEHP and WY-14,643 treatment.
However, the large changes in bile flow by WY-14,643 treatment limit the interpretation of these
data along with a small number of animals examined in this study (e.g., 6 control and DEHP
animals and 3 animals exposed to WY-14,643 at 3 days), a 30% variation in percent liver/body
weight ratios between control groups, and the insensitivity of the technique. In an earlier study
oxidative stress appears to be correlated with neither cell proliferation nor carcinogenic potency
(Woods et al., 2006). Woods et al. (2006) reported WY-14,643Y or DEHP to induce an increase
in free radicals at 2 hrs, a decrease at 3 days then an increase at 3 weeks for both. However,
radical formation did not correlate with the proliferative response, as DEHP fails to produce a
sustained induction of proliferative response in rodent liver but WY-14,643 does, and both WY-
14,643 and DEHP gave a similar pattern of radical formation that did not vary much from
controls which is in contrast to their carcinogenic potency.
Although assumed to be a reflection of cell proliferation in many studies of WY-14,643
and by Klaunig et al. (2003), DNA synthesis recorded using the standard exposure paradigm for
WY-14,643, can also be a reflection of hepatocyte, nonparenchymal cell or inflammatory cell
mitogenesis (in the case of necrosis induced inflammation), from changes in hepatocyte ploidy,
or a combination of all. Other peroxisome proliferators have been shown to have a decrease in
proliferation gene expression shortly after their peaks (e.g., Clofibrate, see Section E.3.2.2) and
both Methylclofenapate and Nafenopin have been shown to increase cell ploidy with Nafenopin
having the majority of its DNA synthesis a reflection of increased ploidy with only a small
percentage as increases in cell number (see Section E.3.4.1). Several authors have also noted
increases in ploidy for WY-14,643 (see Section E.3.4.1).
The Tg.AC genetically modified mouse was used to study 14 chemicals administered by
the topical and oral (gavage and/or diet) routes by Eastin et al. (2001). Clofibrate was considered
clearly positive in the topical studies but not WY-14,643 regardless of route of administration.
Based on the observed responses, it was concluded by the workgroup (Assay Working Groups)
that the Tg.AC model was not overly sensitive and possesses utility as an adjunct to the battery
of toxicity studies used to establish human carcinogenic risk. The difference in result between
Clofibrate and WY-14,643 is indicative of a different MOA for the two compounds.
Similarly, at large exposure concentrations, (2004) investigated the response of male and
female lacZ-plasmid transgenic mice treated at 4 months of age with 6 doses of 2,333 mg/kg
DEHP, 200 mg/kg WY-14,643 or 90 mg/kg Clofibrate over a two week period. Mutation
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 76 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
frequencies were assayed at 21 days following the last exposure. DEHP and WY-14,643 were
shown to significantly elevate the mutant frequency in both male and female liver DNA while
Clofibrate, at the dose level studied, was apparently nonmutagenic in male and female liver (i.e.,
six-dose exposure to DEHP or WY-14,643 over a two week period significantly increased the
mutant frequency in liver of both female and male mice by approximately 40%). The author
noted that
the laxZ plasmid-based transgenic mouse mutation assay is somewhat unique
among other commercially available models (e.g. mutamouse and big blue), by
virtue of its ability to accurately quantify both point mutations and large deletions
including those which originate in the lacZ plasmid catamer and extend into the 3'
flanking genomic region. It should be noted that to date there is no single, agreed
upon protocol for conducting mutagenicity assays with transgenic rodents
although several aspects have been upon by the Transgenic Mutation Assays
workgroup of the International Workshop on Genotoxicity Procedures.
For several chemicals both rats and mice demonstrate evidence of receptor activation
through peroxisome proliferation and peroxisome-related gene expression but only one develops
cancer. The herbicide, 2,4-dichlorophenoxyacetic acid (2,4-D), is a striking example of the
problems that would be associated with only using evidence of PPARa receptor activation to
make conclusions about MO A of liver tumors. 2,4-D is structurally similar to the PPARa
agonist Clofibrate and has been shown at similar concentrations to increase peroxisome number
and size, increase hepatic carnitine acetyltransferase activity and catalase, and decrease serum
triglycerides and cholesterol in rats (Vainio et al., 1983). Peroxisome number was also increased
in Chinese hamsters to a similar level as with Clofibrate at the same exposure concentration after
9 days of exposure to 2,4-D (Vainio et al., 1982). In mice, Lundgren et al. (1987) reported that
2,4-D exposure statistically increased the liver-somatic index over controls after a few days
exposure and increased mitochondrial protein, microsomal protein, carnitine acetyltransferase,
PCO activity, cytochrome oxidase, cytosolic epoxide hydrolase, microsomal epoxide hydrolase,
microsomal P450 content, and hepatic cytosolic epoxide hydrolase in mouse liver. Thus, 2,4-D
activates the PPARa receptor, with associated changes in peroxisome-related gene expression, in
multiple species and at similar doses to Clofibrate. However, Charles et al. (1996) and Charles
and Leeming (1998) reported that in several 2-year studies that there were no 2,4-D-induced
increases in liver tumors in F344 rats, CD-I rats, B6C3F1 mice and CD-I mice.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 77 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Another example, is provided by Gemfibrozil, known as (5-2[2,5-dimethylphenoxy]
2-2-dimethylpentanoic acid) and [2,2-dimethyl-5-(2,5-xylyoxy) valeric acid], a therapeutic agent
that activates the PPARa receptor and is a peroxisome proliferator, but is carcinogenic only in
male rats but not female rats, nor in either gender of mouse (Contrera et al., 1997). Gemfibrozil
causes tumors in pancreas, liver, adrenal, and testes of male rats and causes increases in absolute
and relative liver weights in both rats and mice (Fitzgerald et al., 1981). Gemfibrozil, is a highly
effective lipid and cholesterol lowering drugs in humans and in mice (Olivier et al., 1988).
However, although Gemfibrozil activates the PPARa receptor and induces peroxisome
proliferation in mice, it does not induce liver tumors in that species.
In the long-term study of Bezafibrate, Hays et al. (2005) noted that the role of this
receptor in hepatocarcinogenesis has only been examined using one relatively specific PPARa
agonist (WY-14,643) and report that Bezafibrate can induce the expression of a number of
PPARa target genes (acyl CoA oxidase and CYP4a) and increased liver weight in PPARa
knockout mice that is not dependent on activation of PPARP or PPARy. As noted by Boerrigter
(2004)
In contrast to DEHP and WY-14,643, Clofibrate produced hepatocellular
carcinomas in rats only while no increase in the incidence of tumors was reported
in mice (Gold and E, 1997). However, Clofibrate induces peroxisome
proliferation in both rats and mice (Lundgren and DePierre, 1989) but only
produced hepatocellular carcinomas in rats (Gold and E, 1997).
Melnick et al. (1996) noted that similar levels of peroxisomal induction were observed in
rats exposed to DEHP and di(2-ethylhexyl) adipate (DEHA) at doses comparable to those used in
the bioassays of these chemicals. However, DEHP but not DEHA gave a positive liver tumor
response in 2-year studies in rats. In an evaluation of the carcinogenicity of tetrachloroethylene,
an expert panel of the International Agency for Research on Cancer concluded that the weak
induction of peroxisome proliferation by this chemical in mice was not sufficient to explain the
high incidence of liver tumors observed in an inhalation bioassay.
In adult animals, apoptosis acts as a safeguard to prevent cells with damaged DNA from
progressing to tumor, but like cell proliferation, alterations in apoptosis are common to many
MO As. In addition, only short-term data are available on changes in apoptosis due to PPARa
agonists, and long-term changes have not been investigated (Rusyn et al., 2006). For example,
although a decrease in apoptosis has also suggested to be an important additional molecular
event that may affect the number of cells in rodent liver following exposure to the peroxisome
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-378 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
proliferator DEHP, apoptosis rates have not investigated past 4 days of exposure and thus, the
time-course of this event is uncertain. The antiapoptotic effects of PPAR agonists appear to be
also dependent on nonparenchymal cells (i.e., Kupffer cells) which do not express PPARa and
could be a transient event (Rusyn et al., 2006). Morimura et al. (2006) reported evidence for
exposure to WY-14,643 that does not support a role for PPARa-mediated apoptosis in tumor
formation (see Section E.3.5.1.3, below) as well as appearing to be specific to WY-14,643 (see
Section E.3.4.3.3).
The lack of a causal relationship of transient DNA synthesis increases and
hepatocarcinogenesis has been raised by many (Caldwell et al., 2008a) and is discussed in
Section E.3.4.2 as well as the changes in ploidy (see Section E.3.4.1). In regard to gene
expression profiles, many studies have focused on gene profiles during the early transient
proliferative phase or have identified genes primarily associated with peroxisome proliferation as
"characteristic" or relevant to those associated with tumor induction. Several have focused on
the number of genes whose expression "goes up" or "goes down" from a small number of
animals. Caldwell and Keshava (2006) presented information on WY-14,643, dibutyl phthalate,
Gemfibrozil and DEHP, and noted inconsistent results between PPARa agonists, paradoxes
between mRNA and protein expression, strain, gender, and species differences in response to the
same chemical, and time-dependent differences in response for several enzymes and glutathione.
E.3.4.1.2. New information on causality and sufficiency for PPARa receptor activation. In
its review of the U.S. EPA's draft risk assessment of perfluorooctanoic acid (PFOA), the
Science Advisory Panel (FIFRA SAP, 2004) expressed concerns about whether PPARa
agonism
constitutes the sole MO A for PFOA effects in the liver and the relevance to exposed fetuses,
infants, and children. In part based on uncertainties regarding the Klaunig et al. (2003) proposed
MO A, they concluded that the tumors induced by PFOA were relevant to human risk assessment.
The hypothesis that activation of the PPARa receptor is the sole mode of action
hepatocarcinogenesis induced by DEHP and many other chemicals is further called into question
by recent studies. In the case of DEHP, Klaunig et al. (2003) assumed that WY-14,643 and
DEHP would operate through the same key events and that long-term bioassays of DEHP in
PPARa -/- knockout mice would be negative and hence demonstrate the need for receptor
activation for hepatocarcinogenesis from DEHP.
The fallacy of these assumptions is illustrated by the recent report of the first 2-year
bioassay of DEHP in PPARa -/- knockout mice (Sv/129 background strain) that reported DEHP-
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 79 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
induced hepatocarcinogenesis (Ito et al., 2007). Further discussion was provided by Guyton et
al. (Guyton et al., 2009). Similar to other studies, the PPAR -/- mice had slightly increased liver
weights in comparison to controls and treated wild-type mice (-12% increase over controls). In
fact statistical analysis of the incidence data show that adenomas were significantly increased in
PPARa -/- mice compared with wild-type mice exposed to 500 ppm DEHP and that a significant
dose-response trend for adenomas and adenomas plus carcinomas was observed in PPARa -/-
mice (Figure E-5). Overall, the cancer incidences were consistent with a previous study of
DEHP (David et al., 1999) in B6C3F1 mice at the same doses for nearly the same exposure
duration. A strength of this study is that it was conducted at much lower more environmentally
relevant doses that did not significantly increase liver enzymes as indications of toxicity.
As noted by Kamijo et al. (2007), DEHP was reported also to induce glomerularnephritis
more often in PPARa-null mice because of the absence of PPARa-dependent anti-inflammatory
effect of antagonizing the oxidative stress and NF-kB pathway (Kamijo et al., 2007). Thus, these
data support that hypothesis that there is no difference in liver tumor incidences between PPARa
-/- mice and wild-type mice in a standard nonabbreviated exposure bioassay that does not exceed
the maximal tolerated doses and that DEHP can induce hepatotoxicity as well as other effects
independent of action of the PPARa receptor.
The study of Yang et al. (2007) informs as to the sufficiency of PPARa receptor
activation and subsequent molecular event for hepatocarcinogenesis in mice. The study used a
VP16PPARa transgene under control of the liver-enriched activator protein (LAP) promoter to
activate constitutively the PPARa receptor in mouse hepatocytes. LAP-VP16PPARa transgenic
mice showed a number of effects associated with PPARa receptor activation including decreased
serum triglycerides and free fatty acids, peroxisome proliferation, enhanced hepatocyte DNA
synthesis and induction of cell-cycle genes and those described as "PPARa targets" to
comparable levels reported for WY-14,643 exposure. Hepatocyte proliferation, as determined by
the labeling index of hepatocyte nuclei, was increased after 2 weeks of WY-14,643 treatment
over controls (20.5 vs. 1.6% in control livers) with the LAP-VP16PPARa mice giving a similar
results (20.8 vs. 1.0% in control livers).
The authors noted that transgenic mice did not appear to have positive labeling of
nonparenchymal cell nuclei that were present in the WY-14,643 treated animals. The
transferase-mediated dUTP nick end-labeling assay results were reported to show that there was
no difference in apoptosis in wild-type mice treated with WY-14,643, the transgenic mice, or
controls. In a small number of animals, microsomal genes (CYP4A), peroxisomal (Acox,
BIEN—the bifunctional enzyme) and mitochondrial fatty oxidation genes (LCAD—long chain
acyl CoA dehydrogenase and VLCAD—very long chain acyl CoA dehydrogenase) were
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 80 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
expressed in the transgenic mice with WY-14,643 also increasing expression of these genes in
wild-type mice but with less lipoprotein lipase (LPL) than the transgenic mice. Hepatic CoA
oxidation, were increased to a similar level in wild-type mice treated with WY-14,643 and the
transgenic mice (n = 3-4) and were statistically different than controls. LAP- VP16PPARa
transgenic mice (8 weeks of age) exhibited hepatomegaly (-50 increase percent body/liver
weight over controls), and an accumulation of lipid due to triglycerides but not cholesterol.
However, compared to wild-type mice exposed to WY-14,643 for two weeks, the extent
of hepatomegaly was reduced (i.e., percent liver/body weight increase of ~2.5-fold with
WY-14,643 treatment), no hepatocellular hypertrophy or eosinophilic cytoplasms and no
evidence of nonparenchymal cell proliferation were observed in the LAP-VP16PPARa
transgenic mice.
Adenomas
35.0%
30.0%
25.0%
20.0%
15.0%
10.0%
5.0%
0.0%
1	



I* |m
il =
1 00
|>|>m DEHP
500
~ ltd-Wild ~ tto-knockout ~ David-79wk ~ David-total
+ - p<=0.05 by 1 -tail Fisher exact test as compared to
control; * - p <=0.05 by 1 - and 2-tail Fisher exact test as
compared to control in same study
Carcinomas
25.0%
20.0%
£ 15.0%
« 1 0.0%
5.0%
0.0%

1 oo
l>l>m DEHP
500
~	it o-Wild	~ It o-knock out
~	David-79wk ~ David-total
No statistically-significant differences across all
studies and doses.
At ~1 year of age, Yang et al. (2007) reported there to be no evidence of preneoplastic
lesions or hepatocellular neoplasia in LAP- VP16PPARa transgenic mice, in contrast to results
after 11 months of exposure to WY-14,643 in wild-type mice. Microscopic examination of liver
sections were consistent with the gross findings, as hepatocellular carcinomas and hepatic lesions
were observed in the long-term WY-14,643 treated wild-type mice, but not in >20.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-381 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Figure E-5. Comparison of Ito et al. and David et al. data for DEHP tumor
induction from (Guyton et al., 2009).
LAP-VP16PPARa mice at the age of over 1 year in the absence of dox. There was no
quantitative information on tumors given nor of foci development in the WY-14,643 mice. As
noted by Yang et al. (2007), PPARa activation only in mouse hepatocytes is sufficient to induce
peroxisome proliferation and increased DNA synthesis but not to induce liver tumors.
Thus, "hepatocyte proliferation" indentified by Klaunig et al. (2003) as a "causal event"
in their PPARa MOA is not sufficient to induce hepatocarcinogenesis. These data not only call
into question the adequacy of the MOA hypothesis proposed by Klaunig et al. (2003), but
suggest multiple mechanisms and also multiple cell types may be involved in
hepatocarcinogenicity caused by chemicals that are also PPARa agonists.
E.3.4.1.3. Use of the PPAR -/- knockout and humanized mouse. Great importance has been
attached to the results reported for PPARa -/- mice and their humanized counterpart with
respect to inferences regarding the MOA or peroxisome proliferators and whether short-
term chemical
exposures or abbreviated bioassays conducted with these mice can show that a PPARa
MOA is involved. Consequently, the use of these models warrants scrutiny.
Compared to untreated wild-type mice, liver weights in knockout mice or humanized
mice have been reported to be elevated (Laughter et al., 2004; Morimura et al., 2006; Voss et al.,
2006) and within 10% of each other (Peters et al., 1997). In order to be able to assign affects to a
test chemical tested in knockout mice, a better characterization is needed of the baseline
differences between PPARa -/- knockout and wild-type mice. This is particularly important for
examining weak agonists because the changes they induce may be small and need to be
confidently distinguished from differences due to the loss of the receptor alone. As shown by the
Ito et al. (2007)study and as noted by Maronpot et al. (2004), there is a need for lifetime studies
to characterize background or spontaneous tumor patterns and life spans (including those of the
background strain). While the original work by Lee et al. (1995) describes "the mice
homozygous for the mutation were viable, healthy, and fertile and appeared normal," the authors
did not describe the survival curves for this model nor their background tumor rate. In fact,
further work has shown that they carry a background of chronic conditions, including: (1)
chronic diseases such as obesity and steatosis (Akiyama et al., 2001; Costet et al., 1998); (2)
altered hepatic of hepatocellular structure and function, such as vacuolated hepatocytes
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 82 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
(Anderson et al., 2004; Voss et al., 2006), also seen in "humanized" mice (Cheung et al., 2004);
and (3) altered lipid metabolism, including reduced glycogen stores, blunted hepatic and cardiac
fatty acid oxidation enzyme system response to fasting, elevated plasma free fatty acids, fatty
liver (steatosis), impaired gluconeogenesis, and significant hepatic insulin resistance (Lewitt et
al., 2001). Howroyd et al. (2004) reported decreased longevity and enhancement of age-
dependent lesions in PPARa -/- mice.
These baseline differences from wild-type mice may render them more susceptible to
toxic responses or shorten their lifespans with chemical exposure. For example, after
administration of 250 microliters CCU/kg, all male and 40% of female PPARa knockout mice
were dead or moribund after 2 days of treatment, whereas 25% of male wild-type mice and none
of the female wild-type mice exhibited outward signs of toxicity (Anderson et al., 2004). Hays
et al. (2005) reported that 100% of PPARa knockout have cholestasis after 1 year of Bezafibrate
treatment with higher bile acid concentration than wild-type mice. As described in Section
E.2.1.15, Ramdhan et al (2010) have provided data that not only indicated greater susceptibility
of TCE liver toxicity in PPARa-null mice and humanized null mice, but that there is a
background dysregulation of the number of gene and protein expressions and triglyceride
accumulation in the liver in these strains.
Lewitt et al. (2001) noted that male knockout mice have more marked accumulation of
hepatic fat, hypercholesterolemia and to be particularly sensitive to fasting with some dying if
fasted for more than 24 hours. Sexual dimorphism but especially increased susceptibility of the
male mouse has been reported for knockout mice with pure Sv/129 backgrounds (Anderson et
al., 2004; Lewitt et al., 2001) as well as those with a suggested C57BL/6N background (Costet et
al., 1998; Djouadi et al., 1998). Akiyama et al. (2001) showed an apparent greater sexual
dimorphism in mice with a pure Sv/129 background than C57BL/6N in regard to weight gain
from 2 to 9 months but not in changes in body weight or liver weight between wild-type and
knockout animals. Adipose tissue, serum triglycerides and cholesterol were altered in the
knockout animals. Given that the experiment was only carried out for 9 months, changes in body
fat, liver weight and lipid levels may be greater as the animals get older and steatosis is more
prevalent.
The dramatic effect on survival as well as gender difference by the increased expression
of lipoprotein lipase in the PPARa knockout mouse with further genetic modification is
demonstrated by Nohammer et al. (2003) who reported 50% mortality in 6 months and 100%
mortality within 11 months of age while females survived. These differences could affect the
results of tumor induction for PPARa agonists with less potency than WY-14,643 that do not
produce tumors so rapidly.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 83 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In addition, these studies suggest the need for careful consideration of the effects of use
of different background strains for the knockout and the need for careful characterization of the
background responses of the mouse model and the effects of the use of different background
strains for the knockout. Morimura et al. (2006) reported that, using the B6 background strain,
there were only foci at time periods but knockouts with the SV129 background had multiple
tumors after WY-14,643 treatment.
PPARa knockout mice have also been used to examine the dependence of PPARa on
changes in cell signaling, protein production, or liver weight. However, to be useful, the changes
incurred just by loss of the PPARa should also be well described. Reported differenced between
PPARa-knockout and wild-type mice can impact the sensitivity and specificity of these markers
of for the hypothesized MO A.
In regards to altered cell signaling, Wheeler et al. (2003) note that in normal cells p21waf
and p27kipl inhibit the Cdk/cyclin complexes responsible for cell cycle progression through Gl/S
transition. While these cellular signaling molecules are down-regulated in response to partial
hepatectomy in normal mice, they remain elevated in PPARa knockout mice along with
decreased DNA synthesis.
Fumonisins are hepatocarcinogens that have been associated changes in apoptosis and
tissue generation, and increased acyl-CoA oxidase and CYP4A (markers of PPARa activation)
(Martinez-Larranaga et al., 1996). Voss et al. (2006) report that the average number of hepatic
apoptotic foci per mouse induced by Fumonisins were 3-fold higher and liver mitotic figures
counts were 2-fold lower in PPARa knockout in comparison to wild-type mice, thus, illustrating
a difference in proliferative response in the mice. PPARa-null mice have been reported to have
increased apoptosis and decreased mitosis with fumonisin treatment.
Voss et al. (2006) also report several differences in gene expression in wild-type and
PPARa knockout mice that ranged from 0.3 to 483% of the activity of wild-type mice. The
complex expression patterns of gene expression and determination of their mechanistic
implications in regard to hepatotoxicity and carcinogenicity are difficult. Certainly the large
number of genes whose expression is affected by WY-14,643 (1,012 genes as cited by Voss et
al., 2006) illustrates such complexity. Voss et al. (2006) concluded that studies should consider
dose- and time course-related effect as well as species and strain-related differences in the
expression of gene products.
The "humanized" PPARa mouse has a human copy of PPARa inserted into a PPARa
knockout mouse. It is inserted in a tetracycline response system so that in the absence of DOX
only human PPARa is transcribed in humanized mouse liver and not in other tissues. A rigorous
examination of newly emerging studies regarding the "humanized" mouse is warranted. The
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 84 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
humanized PPARa mouse has been studied in the reports of Cheung et al. (2004), Morimura et
al. (2006), and Ramdhan et al. (2010) (see Section E.2.1.15). Many of the issues described
above for PPARa -/- mice are of concern for the humanized knockout mouse. In addition, the
placement of the humanized PPAR gene is a potential confounding factor, as discussed by
Morimura et al. (2006):
It also cannot be ruled out that the hPPARa mice are resistant to the hepatotoxic
effects of peroxisome proliferators due to the site of expression of the human
receptor. The cDNA was placed under control of the tetracycline regulatory
system and the liver-specific Cebp/B promoter that is preferentially expressed in
hepatocytes.
In the Cheung (2004) report, the humanized mouse was fed WY-14,643 for 2 or 8 weeks
(age not given for the mice). WY-14,643 and Fenobrate were reported to decrease serum total
triglyceride levels in wild and humanized mice to about the level seen in PPARa -/- mice (which
were already suppressed without treatment). Hepatomegaly and increase in hepatocyte size were
observed in the PPARa -humanized mice fed WY-14,643 for 2 weeks but less than that of wild
mice. By contrast, Morimura et al. (2006) stated that the humanized mice did not exhibit
hepatomegaly after treatment with WY-14,643.
Cheung et al. (2004) present figures that showed increased vacuolization of hepatocytes
in a control humanized mouse in comparison to wild-type mice. Vacuolization increased with
WY-14,643 treatment in the humanized mouse. Therefore, there was a background level of liver
dysfunction in these mice even with humanized PPARa. Vacuolization is consistent with fatty
liver observed in the nonhumanized PPARa -/- mouse. As reported by Ramdhan et al. (2010)
untreated humanized mice had increased triglyceride levels in their livers in comparison to
untreated wild type mice.
The authors reported that the humanized mouse did not have increased #s of
peroxisomes after WY treatment. However, they present a figure for genes encoding
peroxisomal, mitochondrial, and microsomal fatty acid oxidation enzymes that shows they were
still markedly increased in PPARa -humanized mice following 8 weeks of exposure to
WY-14,643. Therefore, there is a paradox in these reported results.
Morimura et al. (2006) provided a useful example to illustrate the many issues associated
with interpreting studies with genetically-altered animals. While this study is suggestive of a
difference in susceptibility to tumor induction between wild-type and PPARa humanized mice, a
conclusion that human PPARa is refractory to liver tumor induction is not sufficiently supported.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-385 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
This study had uneven durations of exposure and follow-up and reported substantial
toxicity or mortality that limit the interpretation of the observed tumor rates. For example, the 6
week-old male "humanized" mice had a 44-week experimental period but for wild-type mice that
period was 38 weeks. In addition, for humanized mice, 10 mice were treated with 0.1% WY-
14,643 with 20 controls, but for wild-type mice, 9 mice were given 0.1% WY with 10 controls.
Furthermore, wild-type, WY-14,643-treated animals had suppressed growth and only a 50%
survival to 38 weeks, so an effective LD50 has been used for this length of exposure.
Specifically, of the 10 wild-type WY-14,643 treated mice, 3 died of toxicity and 2 were killed
due to morbidity and their tissues examined. Humanized mice had similar growth for animals
treated with WY-14,643 or controls with only one mouse killed because of morbidity.
Therefore, the reported results, including tumor numbers, are for a mixture of different exposure
durations and ages of animals. In addition the results of the study were reported for only on
exposure level.
Furthermore, it is interesting that while control humanized mice had no adenomas,
WY-14,643 treated humanized mice had one. Morimura et al. (2006) noted that this adenoma
had a morphology "similar to spontaneous mouse liver tumor with basophilic and clear
hepatocytes," whereas the tumors in wild-type mice treated with WY-14,643 were more
diffusely basophilic. If the humanized animals were allowed to live their natural lifespan, this
raises the possibility that WY-14,643 may induce tumors that are similar to other carcinogens
rather than those that have been described as "characteristic" of peroxisome proliferators (see
Section E.3.5.1.5) when human PPARa is present. Therefore, the humanized PPARa rather than
mouse PPARa may have an association with a tumor phenotype characteristic of other MO As
but this study need to be carried out for a longer period of exposure and with more animals to
make that determination.
The baseline tumor response of PPARa humanized mice needs to be characterized as
well as tumors exposure to WY-14,643 or other carcinogens acting through differing MO As.
The numbers of foci were not reported, but "altered foci" were detected in one humanized mouse
with WY-14,643 treatment and one without treatment. The phenotypes of the foci were not
given by the authors.
As discussed above, changes in liver weights have been associated with susceptibility to
liver tumor induction and the issues regarding baseline differences in PPARa -/- mice are equally
relevant for PPARa humanized mice. Morimura et al. (2006) reported that absolute liver weight
for control humanized mice at 44 weeks was 1.57 g (n = 10). The absolute liver weight for wild
control mice was 1.1 g (n = 9) at 38 weeks. The final body weights differed by 14% but liver
weights differed by 30%. Therefore, even though comparing different aged mice, the control
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 86 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
humanized mice had greater liver size than the wild-type control mice on an absolute and relative
basis. This is consistent with humanized knockout mice having greater sized livers and a
baseline of hepatomegaly. Ramdhan et al. (2010) reported significantly elevated liver/body
weight ratios in untreated humanized mice.
With treatment, Morimura et al. (2006) reported that PPARa humanized mice treated
with WY-14,643 had greater absolute and relative liver weights than controls but less elevations
than wild-type treated animals. However, because half of the wild-type animals died, it is
difficult to discern if liver weights were reported for moribund animals sacrificed as well as
animals that survived to 38 weeks for wild-type mice treated with WY-14,643. However, it
appears that moribund animals were included that were sacrificed early for treated groups and
that values from the animal killed at 27 weeks were added in with those surviving till 45 weeks
in the PPARa humanized mice treated with WY-14,643 group.
With respect to the gene expression results reported by Morimura et al. (2006), it is
important to note that they are for liver homogenates with a significant portion of the nuclei from
nonparenchymal cell of the liver (e.g., Kupffer and stellate cells). Thus, the results represent
changes resulting from a mixture of cell types and from differing zones of the liver lobule, with
potentially different gene changes merged together. Livers without macroscopic nodules were
used for western blot and but could have contained small foci in the homogenate as well. The
gene expression results were also reported for an exposure level of WY-14,643 that is an LD50 in
wild-type mice and could reflect toxicity responses rather than carcinogenic ones. The samples
were also obtained at the end of the experiment (with a mix of durations of exposure) and may
not reflect key events in the causation of the cancer but events that are downstream.
These limitations notwithstanding, it is interesting that expression of p53 gene was
reported by Morimura et al. (2006) to be increased in PPARa humanized mice treated with
WY-14,643 compared to all other groups. Furthermore, of the cell cycle genes that were tested,
(i.e., CD-I, Cyclin-dependent Kinases 1 and -/, and c-myc) there was a slightly greater level of
c-myc and CD-I in control PPARa humanized mice than control wild-type mice as a baseline.
This could indicate that there was already increased cell cycling going on in the control PPARa
humanized mouse and could be related to the increased liver size. Treatment with WY-14,643
induced an increase in cycling genes in wild-type mice in relation to its control, but whether that
induction was greater than control levels for PPARa humanized mice for c-myc and CDk4 was
not reported by the authors.
Apoptosis genes were reported to have little difference between control PPARa
humanized and wild-type mice but to have a greater response induced by WY-14,643 in
humanized mice forp53 andp21. There was no consistent or large change in apoptosis genes in
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 87 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
response to exposure to WY-14,643 in wild-type mice. The increased response of apoptosis
genes in PPARa humanized mice without corresponding tumor formation does not support that
response as a key event in the MOA (neither does the lack of response from WY-14,643 in wild-
type mice). For genes associated with PPARa peroxisomal (Acox), microsomal (CYP4a)
mitochondrial fatty oxidation (Mead) and especially malic enzyme, there was a greater response
in wild-type than PPARa humanized mouse after treatment with WY-14,643. However, this is
somewhat in contrast to Cheung et al. (2004), who reported increased in some genes encoding
peroxisomal, mitochondrial, and microsomal fatty oxidation enzymes in the PPARa humanized
mouse after treatment with WY-14,643.
The results reported by Yang et al. (2007) use another type of "humanized" mouse to
study PPARa effects. Yang et al. (2007) used a PPARa humanized transgenic mouse on a PPAR
-/- background that has the complete human PPARa (hPPARa) gene on a PAC genomic clone,
introduced onto the mouse PPARa-null background and express hPPARa not only in the liver
but also in other tissues. Mice were administered WY-14,643 or Fenofibrate [0.1% or 0.2%
(w/w)]. The authors showed a figure representing expression of the hPPARa for two mice with
the tissue used for the genotyping exhibiting great variation in expression between the two
cloned mice as indicated by intensity of staining. The authors stated that in agreement with
mRNA expression, hPPARa protein was highly expressed in the liver of hPPARaPAC mice to an
extent similar to the mPPARa in wild-type mice. They reported that following two weeks of
Fenofibrate treatment, a robust induction of mRNA expression of genes encoding enzymes
responsible for peroxisomal (Acox), mitochondrial (MCAD and LCAD), microsomal (CYP4A)
and cytosolic (ACOT) fatty acid metabolism were found in liver, kidney and heart of both wild-
type and hPPARaPAC mice indicating that hPPARa functions in the same manner as mPPARa to
regulate fatty acid metabolism and associated genes.
However, the authors did no measures in Fenofibrate treated animals, only WY-14,643,
raising the issue of whether there was a difference in the relative mRNA expression of genes for
ACOX etc. and lipids between the two peroxisomal proliferator treatments. The expression of
enzymes associated with PPARa induction was presented only for mice treated with Fenofibrate.
However, the lipids results were presented only for mice treated with WY-14,643. Therefore, it
cannot be established that these two agonists give the same response for both parameters. Also
for the enzymes, the relative expressions compared to wild-type controls, the absolute
expression, and variation between animals is not reported.
It appears that the peroxisomal enzyme induction by Fenofibrate is the same in the wild-
type and transgenic mice. However, in Figure 4 of the paper the mice treated with WY-14,643
instead of Fenofibrate were presented for the peroxisomal membrane protein 70 (PMP70) in total
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-388 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
liver protein gel. There appears to be more PMP70 in the transgenic mice than wild-type mice as
a baseline. The PMP70 appeared to be similar after WY-14,643 treatment. However, only one
gel was given and no other quantitation was given by the authors.
The authors stated that "in addition WY-14,643 and Fenofibrate treatment produced
similar effect to the liver specific humanized PPARa mouse line (Cheung et al., 2004)."
However, the results were not the same between Fenofibrate and WY-14,643 and the mouse line
used by Cheung et al. had background differences in response and pathology. In one figure in
the paper there appears to be a difference in background level of serum total triglyceride between
the wild-type and hPPARaPAC mice that the authors did not note. The power of using such few
mice does not help discern any significant differences in background level of triglycerides.
The authors note that WY-14,643 treatment also resulted in decreased serum triglycerides
levels in hPPARaPAC mice consistent with the induction of expression of genes encoding fatty
acid metabolism and that the hypolipidemic effects of fibrates are generally explained by
increased expression of LPL and decreased expression of apolipoprotein C- III (Apo C-III)
(Auwerx et al., 1996). However, the alteration of these genes by WY-14,643 treatment was only
observed in wild-type mice and not in hPPARaPAC mice suggesting that the hypolipidemic effect
observed in hPPARaPAC mice are not through LPL and APO C-III. The authors do not note that
there could be a difference in the regulation of these pathways by the transgene rather than how
the normal gene is regulated and the pathways it affects. The rationale for examining this
question with WY-14,643 treatment rather than with Fenofibrate treatment is not addressed by
the authors, especially since the other "markers" of peroxisomal gene induction appear to be
affected by Fenofibrate in the wild-type and hPPARaPAC mice.
Hepatomegaly was reported to be observed in the hPPARaPAC mice following two weeks
of WY-14,643 treatment as revealed by the increase liver to body weight ratio compared to
untreated hPPARaPAC mice but to be markedly lower when compared to wild-type mice under
the same treatment.
Histologically, the livers of the wild-type mice treated with WY-14,643 were
hypertrophic with clear eosinophilic regions. These phenotypic effects were observed in both
wild-type and hPPARaPAC mice. The percent liver/body weight was reported to increase from
-4% in wild-type mice to -9% after WY-14,643 treatment and from -4% in hPPARaPAC to little
less that 6% after treatment with WY-14,643.
In wild-type mice treated with WY-14,643 the labeling index was 21.8% compared with
1.1% in untreated wild-type controls. In hPPARaPAC mice, WY-14,643 treatment was reported
to give an average labeling index of 1.0% compared with 0.8% in the untreated control
hPPARaPAC mice. Treatment with WY-14,643 treatment was reported to result in a marked
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 89 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
induction in the expression of CDK4 and cyclin D1 in the livers of wild-type mice but to be
unaffected hPPARaPAC mice treated with WY-14,643. These data were reported to be in
agreement with the liver-specific PPARa-humanized mice that showed not increase in
incorporation of BrdU into hepatocytes upon treatment with WY-14,643 (Cheung et al., 2004)
and further confirmed that activation of hPPARa dose not induce hepatocyte proliferation.
However, the authors present a figure as an example with one liver each with no
quantitation given by the authors for BrdU incorporation. It is not clear whether the pictures
were taken from the same area of the liver or how representative they are. The numbers of mice
were never reported for the labeling index. The data presented do suggest that there was
hypertrophy and hepatomegaly in the humanized mice and but not proliferation in this particular
WY,-14,643 model. Of interest would be investigation of proliferation by other peroxisome
proliferators besides WY-14,643 at this necrogenic dose as it is WY-14,643 that is the anomaly
to continue to induce proliferation or DNA synthesis at 2 weeks. The photomicrographs
presented by the authors are so small and at such low magnification that little detail can be
discerned from them. There are no portal triads or central veins to orient the reader as to what
region of the liver has been affected and where if any there would be hepatocellular
vacuolization.
To determine whether peroxisome proliferation occurred in the hPPARaPAC mice upon
administration of PPs, Yang et al. (2007) examined by Western Blot analysis the protein levels
of the major PMP70 a marker of peroxisome proliferation). After two weeks treatment of
1,000 ppm WY-14,643, induction of PMP70 was reported to be observed in the wild-type mice
as well as in hPPARaPAC mice. The authors suggested that this result indicates that peroxisomal
proliferator treatment induced peroxisomal proliferation in hPPARaPAC mice. The results of this
study indicate that hepatomegaly and peroxisome proliferation occur in this humanized mouse
model when treated with large concentrations of WY-14,643. Thus, these results are inconsistent
with claims that peroxisome proliferators cannot cause hepatomegaly or peroxisome proliferation
in humans or that humans are refractory to these effects. Like the lipid effects, they suggest a
broader spectrum of effects may occur in humans and decreases the specificity of these effects as
species specific. However, due to the model compound being WY-14,643 at a necrogenic dose
of 1,000 ppm, the effect may not be seen in humans using the lower potency peroxisome
proliferators. It would have been useful for this study to include an examination of these effects
with Fenofibrate rather than WY-14,643 and then attempting to extrapolate such effects to other
peroxisome proliferators. The authors often attributed the effects of peroxisome proliferators to
those reactions induced by WY-14,643 and did not acknowledge that the changes induced by
WY-14,643 may be different. This is especially true in regards to hepatocellular DNA synthesis
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 90 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
in which other peroxisome proliferators can cause liver tumors without the sustained
proliferation that WY-14,643 induces, especially at a necrogenic dose.
Yang et al. (2007) reported the results of induction of various genes by WY-14,643 in
wild-type and hPPARaPAC mice by microarray analysis followed by confirmation and
quantitation by qPCR and report that more genes were induced by WY-14,643 in wild-type mice
than in hPPARaPAC mice. They reported that
importantly, the oncogene c-myc was not induced in hPPARaPAC mice.
Moreover, genes encoding cell surface proteins such as Anxa2, CD39, CD63,
Ly6D, and CD24a, and several other genes such as Cidea, Cidec, Dhrs8 and
Hsdllb were also not induced in hPPARaPAC mice. Interestingly, Sult2al was
only induced in hPPARaPAC mice and not in WT mice; this gene is also induced
in human hepatocytes by PP (Fang et al., 2005). The regulation of several of
these genes has previously been demonstrated through a PPARa-dependent
mechanism. Additional studies will be necessary to fully explore the molecular
regulatory mechanism and the functional implication associated with these
differently regulated genes.
The authors did not indicate the context of how the mice were treated, whether these are
pooled results, and when the samples were taken. It is assumed to be whole liver. As stated in
Section E.3.2.2 above, there are several limitations for interpretations of the results such as those
presented by Yang et al. (2007) which include the lack of phenotypic anchoring for the results.
The authors have shown changes from whole liver and have listed changes in genes between
wild-type and humanized mice on a PPAR -/- background that in itself with bring about changes
in gene expression. The authors acknowledge difficulties in determining what their reported
gene changes mean.
Yang et al. (2007) reported that "activation of PPARa alters hepatic miRNA expression
(Shah et al., 2007)." They report that let-7C, a miRNA critical in cell growth and shown to
target c-myc, was inhibited by WY-14,643 treatment in wild-type mice and that the expression
levels of both pri-let-7C and mature let-7C were significantly higher in hPPARaPAC mice
compared to wild-type mice. Treatment with WY-14,643 was reported to decrease the
expression of Pri-let-7C and mature let-7C in wild-type mice but in hPPARaPAC mice. The
authors noted that
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-391 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
in addition, the induction of c-myc by WY-14,643 treatment in wild type mice did
not occur in WY-14,643 treated hPPARaPAC mice. This is in agreement with the
previous observation in liver-specific humanized PPARa (Shah et al., 2007) and
further indicates the activation of human PPARa does not cause a change in
hepatic miRNA and c-myc gene expression.
A qPCR analysis of pri-let-7C following 2 weeks WY-14,632 treatment was reported for
wild-type and hPPARaPAC mice (n = 3-4). There appeared to be -20 times more let-7C
expression in hPPARaPAC mice than control wild mice as a baseline. The gel given by the
authors showed a very small difference in wild-type mice in let-7C northern blot analysis
between a control wild-type and WY-14,643-treated wild-type mouse. There appeared to be no
difference in the hPPARaPAC mice between control and WY-14,643 treatment and a larger
stained area than the control wild-type mice. The relative c-Muc expression between the
hPPARaPAC mice and wild-type control mice did not correlate with changes in let-7C expression.
Thus, the amount of decrease by treatment with WY-14,632 in wild-type mice appeared
to be extremely small compared to the much greater baseline expression in the hPPARaPAC mice.
The change brought by WY-14,632 treatment in wild-type mice was a small change compared to
the 20-fold greater baseline expression in the hPPARaPAC mice. The authors stated that the
expression of the c-Myc regulator was higher in the hPPARaPAC mice indicating over regulation
of cell division and an inability for hepatocytes to proliferate. However, their results showed that
there was a greater difference in regulatory baseline function of the PPAR using this paradigm
and this construct. Are these differences due to human PPAR or to the way PPAR was put back
into PPAR -/- mouse and expected to function? If the experiment included mouse PPAR put
back in this way on a null background, what would such an experiment show? Are these results
representative of the PPAR or how it is now controlled and expressed? In addition, what would
the study of other peroxisome proliferators besides WY-14,643 show in regard to changes in
miRNA. Are these results reflective of a just the transient effect that is prolonged in a special
case?
As discussed in Section E.3.2.2 there are issues with microarray data in addition to the
newly emerging field of miRNA arrays, which include phenotypic anchoring and whether they
are from whole liver or pooled samples. The results given in this report are for relative Let-7C
expression given and not absolute values. The changes in baseline Let-7C expression between
the wild-type and the hPPARaPAC mice did not correlate with the magnitude of difference in
northern blot analysis and did not correlate at all with c-myc expression reported in this study.
Thus, a direct correlation between the effect of Let-7C expression and function and effects from
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 92 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
WY-14,643 was not supported. The relative expression was reported but the variation of
baseline expression of the "PPAR controlled genes" was not. Given that one of the first figures
reported a large difference between animals in expression of the human PPAR gene in the
transgenic animals, how did this difference affect the results given here as relative changes
downstream?
Yang et al. (2007) concluded that the hPPARaPAC mice represent the most relevant model
for humans since, the tissue distribution of PPARa is similar to that observed in wild-type mice
and the hPPARa in hPPARaPAC mice is under regulation of its native promoter. Indeed up-
regulation of hepatic mPPARa in wild-type mice by fasting was mirrored by the hPPARa in
hPPARaPAC mice. However, there was no demonstration that the artificial chromosome that is
replicating along with other DNA is controlled sterically by the same control since it is not on
the mouse genome in the same place as the native PPAR. There is also not a demonstration of
how stable the baseline of PPAR DNA expression is in this mouse model—does it vary as much
or more than native PPAR between mice? The authors stated that
induction of PPARa target genes for fatty acid metabolism and a decrease in
serum triglycerides by PP in hPPARaPAC mice indicates that hPPARa is
functional in the mouse environment with respects to regulation of fatty acid
metabolism. This is in agreement with the liver-specific PPARa humanized mice
that also exhibit these responses (Cheung et al., 2004). Indeed the DNA binding
domain of hPPARa is 100% homologous with that of the mouse suggesting that
both bind to the same PPRE binding site in the promoter region of target genes.
Transfection of hPPAR into murine hepatocytes increased PPs induced
peroxisome proliferation related effects (Macdonald et al., 1999). These results
suggest that hPPARa and mPPARa do not differ in induction of target genes with
known PPRE.
However, replacement with human PPAR in the Cheung et al. model is not sufficient to
prevent the same types of toxicity as seen with PPAR knockouts on the hepatocytes such as
steatosis.
Yang et al. (2007) note that
the increased LPL and decreased expression of apo C-III are proposed to explain
the hypolipidemic effects of PPS (Auwerx et al., 1996). However, hPPARaPAC
mice treated with PP exhibit lowered serum triglycerides without alteration of the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 93 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
expression of LPL and apo C-III. This indicates the hypolipidemic effects in
rodents are mediated via other molecular regulatory mechanisms. It is also
suggested that the activation of PPARa by PPs stimulates hepatic fatty acid
oxidation and thereby diminishing their incorporation into triglycerides and
secretion of VLDL (Frayland et al., 1997). Consistent with this idea, a robust
induction of the genes encoding enzymes for fatty acid oxidation by PP in
hPPARaPAC mice were observed. Thus, the exact mechanism by which PPs exert
their hypolipidemic effects needs reexamination.
However, the use of two different peroxisome proliferators (i.e., WY-14,643 and
Fenofibrate) for two types of effects (peroxisomal and lipid) may be the cause of some paradoxes
here in terms of MO A for lipid effects. The baseline differences in the hPPARaPAC mice for
serum total triglycerides was not explored by these authors and the small number of animals used
make conclusions difficult about the magnitude of difference. The differences in baseline
expression for LPL are not discernable in the graphic representation of the results.
Yang et al. (2007) noted that
on the other hand, the difference in the affinity of ligands for the human
and mouse PPARa receptor was proposed to account for the species difference.
The ligand binding domain of hPPARa is 94% homologous with that of the
mouse. In vitro transactivation assays have previously shown that WY has a
higher affinity for rodent PPARa than human PPARa, while Fenofibrate has
similar affinity for rodent and human PPARa (Shearer and Hoekstra, 2003; Sher
et al., 1993). In the present study WY and Fenofibrate exhibit the same capacity
to induce known PPARa target genes in the liver, kidney and heart in both wild-
type and hPPARaPAC mice.
The statement by the authors that Fenofibrate and WY-14,643 had the same affinity "as
shown by this study" is not correct. The two treatments were not studied for the same enzymes
or genes in the data reported in the study. Both WY-14,643 and Fenofibrate can induce PPARa
targets but it was not shown to the same extent. Yang et al. (2007) stated that
This is in agreement with the liver-specific PPARa humanized mice that
also exhibit a similar capacity to induce PPARa target genes in liver by WY and
Fenofibrate (Cheung et al., 2004). Thus, the ligand affinity difference between
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 94 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
mouse and human PPARa may not be critical under the conditions of these
studies.
Alternatively, these results could reflect that these studies were conducted with two
different agonists with different affinities and responses due to receptor activation.
Finally, a useful comparison to make are the differences between wild-type mice,
PPARa -/- mice that serve as the background for the transgenic human mouse models, and both
transgenic models. The small and variable number of animals examined in these studies is
readily apparent. The results of the Cheung et al. (2004) humanized mouse model and those
reported for Yang et al. (2007) show differences in the study designs including PPARa agonists
studied for particular effects and results reported for similar treatments (see Table E-18).
As shown above, the effect on the PPARa -/- by the knockout included decreased
triglyceride levels and slightly increased liver weight. Although treatment with WY-14,643 and
Fenofibrate were reported to decrease triglyceride levels in wild-type mice, paradoxically so did
knocking out the receptor. Exposures to WY-14,643 appeared to induce a slight increase and
Fenofibrate a slight decrease in triglyceride levels in PPARa -/- mice but the variability of
response and small number of animals in the experiments limited the ability to discern a
quantitative difference in the treatments.
In the study by Cheung et al. (2004) it appears that the insertion of humanized PPARa
restored the baseline and treatment responses for triglyceride levels. Of note, use of the same
humanized mode in Ramdhan et al. (2010) showed accumulation of triglycerides in the liver of
untreated mice. Overall, the results reported by Yang et al. (2007) appeared to show a lower
level of triglycerides in control wild-type mice that was similar in magnitude to the treatment
effect reported by Fenofibrate by Cheung et al. (2004). However, there also appeared to be
restoration of this effect in the humanized mouse model of Yang et al. (2007).
In regard to DNA synthesis, both Cheung et al. (2004) and Yang et al. (2007) only gave
results for WY-14,643 and for different durations of exposure so they were not comparable. It
appeared that -60% of hepatocytes were labeled by 8 weeks of WY-14,643 treatment (Cheung
et al., 2004) compared to -20% after 2 weeks of exposure. Again this highlights the difference
between using WY-14,643 as a model for the PPARa as a class at times when almost all other
PPARa agonists have ceased to increase DNA synthesis or have reductions in this parameter.
The background changes due to the PPARa -/- knockout were not reported so that the effects of
the knockout could not be ascertained. It appeared that insertion of humanized PPARa did not
result in restoration of WY-14,643 -induced DNA synthesis. The correlation with this
parameter and any focal areas of necrosis were not discussed by the authors of the study.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-395 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
In regard to hepatomegaly, Fenofibrate and WY-14,643 appeared to both give an
increase in liver weight in the humanized mouse model of Cheung et al. (2004) with little effect
in the knockout mouse. For Fenofibrate there was little difference in liver weight gain in the
wild-type mouse and that of the humanized mouse model of Cheung et al. (Cheung et al., 2004).
However, Fenofibrate was not tested in the humanized mouse model of Yang et al. (2007). In
that model, only WY-14,643 was used but there was still an increase in liver weight. Thus, in
terms of effects on liver weight gain and triglyceride levels both models gave comparable results
and appeared to indicate that insertion humanized PPARa would restore some of the effects of
the knockout. However, the results from both experiments highlight the need for adequate
numbers of animals and other PPARa agonists to be tested besides WY-14,463 at such a high
dose and certainly for longer periods of time to ascertain whether such manipulations will
affects carcinogenicity.
The study by Ramdhan et al. (2010) is more definitive in regard to characterization of
the effects of the knockout and insertion of human PPARa (see Section E.2.1.15). From this
study dysregulation by the knockout and by reinsertion of human PPARa at levels of greater
than 10-fold protein expression can be observed and include alterations in a number of gene and
protein expression levels and an underlying background level of hepatic steatosis and
triglyceride accumulation.
E.3.4.1.4. NF-kB activation. NF-kB activation has also been proposed as a key event in the
induction of liver cancer through PPARa activation. As discussed in Sections E.3.2.6 and
E.3.4.3.3, activation of the NF-kB pathway is implicated in carcinogenesis, nonspecific for a
particular MOA for liver cancer, and is context dependent on its effects. Its specific
actions depend on the cell type and type of agent or signal that activates translocation of the
complex. NF-kB is not only involved in biological processes other than tumor induction, but also
exhibits some apparently contradictory behaviors (Perkins and Gilmore, 2006). Although many
studies point to a tumor-promoting function of NF-kB subunits, evidence also exists for tumor
suppressor functions. NF-kB actions are associated with TNF and JNK among many other cell
signaling systems and molecules and it has functions that alter proliferation and apoptosis. NF-
kB activation reported in some studies may be associated with early Kupffer cell responses and
be associative but not key events in the carcinogenic process. However, most assays look at total
NF-kB expression in the whole liver and at the early periods of proliferation and apoptosis. The
origin of the NF-kB is crucial as to its effect in the liver. For instance, hepatocyte specific
deletion of IKKp increased DEN-induced hepatocarcinogenesis but a deletion of IKKp in both
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 96 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
hepatocytes and Kupffer cells however, were reported to have the opposite effect (Maeda et al.,
2005).
E.3 .4.1.5. Phenotype as an indicator of a PPARa mode of action (MOA). As discussed
previously (see Sections E.3.1.5, and E.3.1.8) FAH precede both hepatocellular adenomas
and carcinomas in rodents and, in humans with chronic liver diseases that predispose them
to
hepatocellular carcinomas. Striking similarities in specific changes of the cellular phenotype of
preneoplastic FAH are emerging in experimental and human hepatocarcinogenesis, irrespective
of whether this was elicited by chemicals, hormones, radiation, viruses, or, in animal models, by
transgenic oncogenes or Helicobacter hepaticus. Several authors have noted that the detection
of phenotypically similar FAH in various animal models and in humans prone to developing or
bearing hepatocellular carcinomas favors the extrapolation from data obtained in animals to
humans (Bannasch et al., 2003; Bannasch et al., 2001; Su and Bannasch, 2003). In regard to
phenotype by tincture Caldwell and Keshava (2006) stated:
In addition, the term "basophilic" in describing preneoplastic foci or tumors can
be misleading. The different types of FAH have been related to three main
preneoplastic hepatocellular lineages: 1) the glycogenotic-basophilic cell lineage,
2) its xenomorphic-tigroid cell variant, and 3) the amphophilic-basophilic cell
lineage. Specific changes of the cellular phenotype of the first two lineages of
FAHs are similar in experimental and human hepatocarcinogenesis, irrespective
of whether they were elicited by DNA-reactive chemicals, hormones, radiation,
viruses, transgenic oncogenes and local hyperinsulinism as described by the first
two FAHs and this similarity favors extrapolation from data obtained in animals
to humans (Bannasch et al., 2003; Bannasch et al., 2001; Su and Bannasch,
2003). In contrast, the amphophilic cell lineage of hepatocarcinogenesis has
been observed mainly after exposure of rodents to peroxisome proliferators or to
hepadnaviridae (Bannasch et al., 2001).
Bannasch (1996) describes "amphophilic" FAH and tumors induced by
peroxisome proliferators to maintain the phenotype as the foci progress to
tumors. They are glycogen poor from the start with increased numbers of
mitochondria, peroxisomes and ribosomes. The author further states that the
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-3 97 DRAFT—DO NOT CITE OR QUOTE
-------
1	"homogenous basophilic" descriptions by others of foci induced by WY are
2	really amphophilic. Agents other than peroxisome proliferators can induce
3	"acidophilic" or "eosinophilic" (due to increased smooth endoplasmic reticulum)
4	or glycognotic foci which tend to progress to basophilic stages (due to increased
5	ribosomes).
6
7	Tumors and foci induced by peroxisome proliferators have been suggested to
8	have a phenotype of increased mitochondrial proliferation and mitochondrial
9	enzymes (thyromimetic rather than insulinomimetic) (2006).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	E-398 DRAFT—DO NOT CITE OR QUOTE
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
>S
X |
to'
*
^S
«
s
TO
Co
0
S
1
a.
to
o
Table E-18. Comparison between results for Yang et al. (2007) and Cheung et al. (2004)a


PPAR -/- knockout
Humanized mice
Humanized PAC
Effect
Wild type mice
mice
(liver only)
mice
Triglyc
Cheung
Cheung
Cheung

erides
(n = 6-9)
(n = 6-9)
(n = 6-9)


Control 145 mg/mL
Control 100 mg/mL
Control 175 mg/mL


0.1% WY-14,643 60
0.1% WY-14,643 115
0.1%WY-14,643 60


mg/mL
mg/mL
mg/mL


(2 wks)
(2 wks)
(2 wks)


0.2% Fenofibrate 85
0.2% Fenofibrate 85
0.2% Fenofibrate 85


mg/mL
mg/mL
mg/mL


(2 wks)
(2 wks)
(2 wks)
Yang




(// = 4-6)

Yang


Control 120 mg/mL

(n = 4-6)


0.1%WY-14,643 75

Control 95 mg/mL


mg/mL

0.1 % WY-14,643 55 mg/mL


(2 wks)

(2wks)



BrdU
Cheung
Not done
Cheung

incorporation
(n = 5)

(w = 5)


Control 1.6%

Control 1.6%


0.1% WY-14,643 57.9%

0.1% WY-14,643 2.8%


(8 wks)

(8




wks)


Yang


Yang

(n = 4-6)


(n = 4-6)

Control 1.1%


Control 0.8%

0.1% WY-14,643 21.8%


0.1% WY-14,643 1.0%

(2 wks)


(2 wks)
-------

§•
to
s
a
to
s
>3
o
o
s
Co
I?
^ 5
Oq ^
53
^S
§*
Co
S5
S*
<§>
>
to
*
^s
«
s
TO
Co
0
s
1
a.
VO
to
o
-------
S"4
>}'
§•
to
£
a
to
s
3"
a
a,
>r
>s
Table E 18. Comparison between results for Yang et al. (2007) and Cheung et al. (2004) (continued)


PPAR -/- knockout
Humanized mice
Humanized PAC
Effect
Wild type mice
mice
(liver only)
mice
Hepato
Cheung
Cheung
Cheung

megalyb
(n = 5-9)
(n = 5-9)
(n = 5-9)

(%
Control 4%
Control 5%
Control 4.5%

liver body
0.1% WY-14,643
0.1% WY-14,643 5%
0.1% WY-14,643 7%

weight ratio)
11%
(2 wks)
(2 wks)


(2 wks)
0.2% Fenofibrate 5.5%
0.2% Fenofibrate 7%


0.2% Fenofibrate 8.5%
(2 wks)
(2 wks)


(2 wks)


Yang

Yang


I
II

(n = 4-6)


Control 4%

Control 4%


0.1% WY 6%

0.1% WY-14,643 9%


(2 wks)

(2 wks)



>3
o
o
s
Co
I?
^ 5
Oq ^
53
to
*
«
s
TO
Co
0
S
1
a.
to
o
VO
aThe ages of the humanized knockout mice are not given for Cheung et al. (2004) but are 8-10 weeks for Yang et al. (2007).
Percentages are approximate values extrapolated from figures for hepatomegaly.
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Tumors from peroxisome proliferators in Kraupp-Grasl et al. (1990) and
Grasl-Kraupp et al. (1993) for rat liver tumors were characterized as weakly basophilic with
some eosinophilia and as similar to the description given by Bannasch et al. as amphophilic.
However, a number of recent studies indicate that other "classic" peroxisome proliferators may
have a different phenotype than has been attributed to the class through studies of WY-14,643.
A recent study of DEHP, another peroxisome proliferator assumed to induce liver tumors
through activation of the PPARa receptor, reported the majority of liver FAH to be of the first
two types after a lifetime of exposure to DEHP with a dose-related tendency for increased
numbers of amphophilic FAHs in rats (Voss et al., 2005). As stated previously, the MOA of
DEHP-induced liver tumors in mice also appears not to be dependent on PPARa activation.
Michel et al. (2007) reported the phenotype of tumors and foci in rats treated with
clofibric acid at a very large dose (5,000 ppm for 20 months) and noted that in controls the first
type of foci to appear was tigroid on Day 264 and their incidence increased with time
representing the most abundant type in this group. They reported no adenomas or carcinomas
after up to 607 days after giving saline injection in the control animals.
DEN treatment was examined up to 377 days only with tigroid, eosinophilic and clear
cell foci observed at that time. Clofibric acid was examined up to 607 days with tigroid and
clear cell foci reported to be the first to appear on Day 264 no other foci class. By Day 377,
there were tigroid, eosinophilic and clear cell foci but no basophilic foci reported with clofibric
acid treatment and, although only a few animals were examined, 2/5 had adenomas but not
carcinomas. By Day 524 all types of foci were seen (including basophilic for the first time) and
there were adenomas and carcinomas in 2/5 animals. By 607 days a similar pattern was
observed without adenomas but 3/6 animals showing carcinomas.
Although the number of animals examined was very small, these results indicate that
clofibric acid was not inducing primarily "basophilic foci" as reported for peroxisome
proliferators but the first foci are tigroid and clear cell foci. Basophilic foci did not appear until
Day 524 similar to control values for foci development and distribution. However, unlike
controls, clofibric acid induced eosinophilic and clear cell foci earlier. This is inconsistent with
the phenotype ascribed to peroxisome proliferators as exemplified by WY-14,643.
In regard to GST-71 and y-transpeptidase (GGT), Rao et al. (1986) fed 2 male F344 rats a
diet of 0.1% WY-14,643 for 19 months or 3 F344 rats 0.025% Ciprofibrate for 15-19 months
and reported "altered areas,"(AA) "neoplastic nodules" (NN), and hepatocellular carcinomas
(HCC). For WY-14,643 treatment 107 AA, 75 NN, and 5 HCC, and for Ciprofibrate treatment
107 AA, 27 NN, and 16 HCC were identified. In the WY-14,643-treated rats, HCC, and NN
were both GGT and GST-71 negative (96-100%) with 87% of AA was negative for both. In
This document is a draft for review purposes only and does not constitute Agency policy.
E-402 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Ciprofibrate-treated rats NN and HCC were negative for both markers (95%) but only 46% of
AA were negative for both markers. Thus, a different pattern for tumor phenotype was reported
for WY-14,643 and another peroxisome proliferator, Ciprofibrate, in this study as well.
In addition, GGT phenotype is reported not to be specific to weakly basophilic foci.
GGT staining was reported to be negative in eosinophilic tumors after initiation and promotion.
Kraupp-Grasl et al. (1990) noted differences among PPARa agonists in their ability to promote
tumors and suggested they not necessarily be considered a uniform group. Caldwell and
Keshava (2006) suggested that the reports of a simple designation of "basophilic" is not enough
to associate a foci as caused by peroxisome proliferators (Bannasch, 1996; Grasl-Kraupp et al.,
1993; Kraupp-Grasl et al., 1990). Increased basophilia of tumors and increased numbers of
carcinomas is consistent with the progressive basophilia described by Bannasch (1996), as many
adenomas progress to carcinomas.
It should be noted that the amphophilic foci and tumors described by Bannasch et al.
were primarily studied in rats. Morimura et al. (2006) noted that WY-14,643 induced diffusely
basophilic tumors in mice and therefore, identified the WY-14,643 tumors in a way consistent
with the descriptions of amphophilic tumors by Bannasch et al. The tumor induced by
WY-14,643 in their humanized mouse was reported to be similar to those arising spontaneously
in the mouse. However, the mouse response could differ from the rat, especially for PPARa
agonists other than WY-14,643.
H-ras activation and mutation studies have attempted to assign a pattern to peroxisome
proliferator-induced tumors as noted in Section E.2.3.3.2, above. However, also as noted in
Section E.2.3.3.2, the genetic background of the mice used, the dose of carcinogen and the stage
of progression of "lesions" (i.e., foci vs. adenomas vs. carcinomas) may affect the number of
activated H-ras containing tumors that develop. Fox et al. (1990) noted that tumors induced by
Ciprofibrate (0.0125% diet, 2 years) had a much lower frequency of H-ras gene activation than
those that arose spontaneously (2-year bioassays of control animals) or induced with the
"genotoxic" carcinogen benzidine-2 HC1 (120 ppm, drinking H20, 1 year) and that the
Ciprofibrate-induced tumors were reported to be more eosinophilic as were the surrounding
normal hepatocytes than spontaneously occurring tumors. Anna et al. (1994) also stated that
mice treated with Ciprofibrate had a markedly lower frequency of tumors with activated H-ras
but that the spectrum of mutations in tumors was similar those in "spontaneous tumors."
Hegi et al. (1993) tested Ciprofibrate-induced tumors from Fox et al. (1990) in the NIH3T3
cotransfection-nude mouse tumorigenicity assay and concluded that ras protooncogene
activation, were not frequent events in Ciprofibrate-induced tumors and that spontaneous tumors
were not promoted with it.
This document is a draft for review purposes only and does not constitute Agency policy.
E-403 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
Stanley et al. (1994) studied the effect of MCP, a peroxisome proliferator, in B6C3F1
(relatively sensitive) and C57BL/10J (relatively resistant) mice for H-ras codon 61-point
mutations in MCP-induced liver tumors (hepatocellular adenomas and carcinomas). In the
B6C3F1 mice, -24% of MCP-induced tumors had codon 61 mutations and for C57BL/10J mice
-13%. The findings of an increased frequency of H-ras mutation in carcinomas compared to
adenomas in both strains of mice is suggestive that these mutations were related to stage of
progression. Thus, in mice, the phenotype of tumors did not appear to be readily distinguishable
from spontaneous tumors based on tincture for peroxisome proliferators other than WY-14,643,
but did have more of a signature in terms of H-ras mutation and activation.
The expression of c-Jun has been used to discern TCE tumors from those of its
metabolites. However, as pointed out by Caldwell and Keshava (2006), although Bull et al.
(2004) have suggested that the negative expression of c-jun in TCA-induced tumors may be
consistent with a characteristic phenotype shown in general by peroxisome proliferators as a
class, there is no supporting evidence of this. While increased mitochondrial proliferation and
mitochondrial enzymes (thyromimetic rather than insulinomimetic) properties have been
ascribed to peroxisome proliferator-induced tumors, the studies cited in Bull et al. (2004) have
not examined TCA-induced tumors for these properties.
E.3.4.1.6. Human relevance. In its framework for making conclusions about human
relevance, the U.S. EPA Cancer Guidelines (U.S. EPA, 2005c) asks that critical similarities
and differences between test animals and humans be identified. Humans possess PPARa at
sufficient
levels to mediate the human hypolipidemic response to peroxisome-proliferating fibrate drugs.
Fenofibrate and Ciprofibrate induce treatment related increases in liver weight, hypertrophy,
numbers of peroxisomes, numbers of mitochondria, and smooth endoplasmic reticulum in
cynomologous monkeys at 15 days of exposure (Hoivik et al., 2004). Given the species
difference in the ability to respond to a mitogenic stimulus such as partial hepatectomy (see
Section E.3.3), lack of hepatocellular DNA synthesis at this time point is not unexpected and, as
Rusyn (2006) noted, examination at differing time point may produce differing results. It is
therefore, generally acknowledged that "a point in the rat and mouse key events cascade where
the pathway is biologically precluded in humans in principle cannot be identified." (Klaunig et
al., 2003)(NAS, 2006). Thus, from a qualitative standpoint, the effects described above are
plausible in humans.
This document is a draft for review purposes only and does not constitute Agency policy.
E-404 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
As for quantitative differences, there are two key issues. First, as stated in the Cancer
Guidelines, when considering human relevance, "Any information suggesting quantitative
differences between animals and humans is flagged for consideration in the dose-response
assessment." Therefore, while Klaunig et al. (2003) and NAS (2006) go on to suggest that
"this mode of action is not likely to occur in humans based on differences in several key steps
when taking into consideration kinetic and dynamic factors," under the Cancer Guidelines,
such "kinetic and dynamic factors" need to be made explicit in the dose-response assessment,
and should not be part of the qualitative characterization of hazard. Second, the discussion
above points to the lack of evidence supporting associations between the postulated events and
carcinogenic potency. Thus, because interspecies differences in carcinogenicity do not appear
to be associated with interspecies differences in postulated events, they do not provide reliable
metrics with which to make inferences about relative human sensitivity.
E.3.4.2. Other Trichloroethylene (TCE) Metabolite Effects That May Contribute to its
Hepatocarcinogenicity
While the focus of most studies of TCA has been its effects on peroxisomal proliferation,
DCA has been investigated for a variety of effects that are also observed either in early stages of
oncogenesis (glycogen deposition) or conditions that predispose patients to liver cancer. Some
studies have examined microarray profiles in attempt to study the MOA or TCE (see
Section E.3.2.2 for caveats regarding such approaches). Caldwell and Keshava have provided a
review of these studies, which is provided below.
E.3.4.2.1. DCA effects and glycogen accumulation correlations with cancer. As noted
previously, DCA administration has been reported to increase the observable amount of
glycogen in mouse liver via light microscopy and, although to not be primarily responsible
for DCA-induced liver mass increases, to be increase whole liver glycogen as much by 50%
(Kato-Weinstein et al., 2001). Given that TCE and DCA tumor phenotypes indicate a role for
DCA in TCE hepatocarcinogenicity (see Section E.2.3.3.2, above), Caldwell and Keshava (2006)
described the correlations with effects induced by DCA that have been associated with
hepatocarcinogenicity.
A number of studies suggest DCA-induced liver cancer may be linked to its
effects on the cytosolic enzyme glutathione (GST)-S-transferase-zeta. GST-zeta
This document is a draft for review purposes only and does not constitute Agency policy.
E-405 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
is also known as maleylacetoacetate isomerase and is part of the tyrosine
catabolism pathway whose disruption in type 1 hereditary tyrosinemia has been
linked to increased liver cancer risk in humans. GST-zeta metabolizes
maleylacetoacetate (MAA) to fumarylacetoacetate (FAA) which displays
apoptogenic, mutagenic, aneugenic, and mitogenic activities (Bergeron et al.,
2003; Jorquera and Tanguay, 2001; Kim et al., 2000). Increased cancer risk has
been suggested to result from FAA and MAA accumulation (Tanguay et al.,
1996). Cornett et al. (1999) reported DC A exposure in rats increased
accumulation of maleylacetone (a spontaneous decarboxylation product of
MAA), suggesting MAA accumulation. Ammini et al. (2003) report depletion of
the GST-zeta to be exclusively a post-transcriptional event with genetic ablation
of GST-zeta causing FAA and MAA accumulation in mice. Schultz et al. (2002)
report that elimination of DCA is controlled by liver metabolism via GST-zeta in
mice, and that DCA also inhibits the enzyme (and thus its own elimination) with
young mice being the most sensitive to this inhibition. On the other hand, older
mice (60 weeks) had a decreased capacity to excrete and metabolize DCA in
comparison with younger ones. The authors suggest that exogenous factors that
deplete or reduce GST-zeta will decrease DCA elimination and may increase its
carcinogenic potency. They also suggest that, due to suicide inactivation of
GST-zeta, an assumption of linear kinetics can lead to an underestimation of the
internal dose of DCA at high exposure rates. In humans, GST-zeta has been
reported to be inhibited by DCA and to be polymorphic (Blackburn et al., 2001;
Blackburn et al., 2000; Tzeng et al., 2000). Board et al. (2001) report one variant
to have significantly higher activity with DCA as a substrate than other GST zeta
isoforms, which could affect DCA susceptibility.
Individuals with glycogen storage disease or with poorly controlled diabetes have
excessive storage of glycogen in their livers (glycogenosis) and increased risk of
liver cancer (Adami et al., 1996; La Vecchia et al., 1994; Rake et al., 2002;
Wideroff et al., 1997). In an animal model where hepatocytes are exposed to a
local hyperinsulinemia from transplanted islets of Langerhans and the remaining
tissue is hypoinsulinemic, insulin induces alterations that resemble preneoplastic
foci of altered hepatocytes (FAH) and develop into hepatocellular tumors in later
stages of carcinogenesis (Evert et al., 2003). A number of studies have reported
suppression of apoptosis, decreases in insulin, and glycogenosis in mice liver by
This document is a draft for review purposes only and does not constitute Agency policy.
E-406 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DCA at levels that also induce liver tumors (Bull, 2004b; Bull et al., 2004;
Lingohr et al., 2001). In isolated murine hepatocytes, Lingohr et al. (2002)
reported DCA-induced glycogenosis was dose related, occurred at very low
doses (10 |iM), occurred without the presence of insulin, was not affected by
insulin addition, was dependent on phosphatidylinositol 3-kinase (P13K)
activity, and was not a result of decreased glycogen breakdown. The authors
noted that PI3K is also known to regulate cell proliferation and apoptosis in
hepatocytes, and that understanding these mechanisms may be important to
understanding DCA-induced carcinogenesis. They also report insulin receptor
(IR) protein levels decreased to 30% of controls in mice liver after up to 52
weeks of DCA treatment. Activation of the IR is also the principal pathway by
which insulin stimulates glycogen synthetase (the rate limiting enzyme of
glycogen biosynthesis). However, in DCA-induced liver tumors IR protein was
elevated as well as mitogen-activated protein kinase (a downstream target protein
of the IR) phosphorylation. DCA-induced tumors were glycogen poor (Lingohr
et al., 2001). The authors suggest that normal hepatocytes down-regulate
insulin-signaling proteins in response to the accumulation of liver glycogen
caused by DCA and that the initiated cell population, which does not accumulate
glycogen and is promoted by DCA treatment, responds differently from normal
hepatocytes to the insulin-like effects of DCA.
Gene expression studies of DCA show a number of genes identified with cell
growth, tissue remodeling, apoptosis, cancer progression, and xenobiotic
metabolism to be altered in mice liver at high doses (2 g/L DCA) in drinking
water (Thai et al., 2003; Thai et al., 2001). After 4 weeks, RNA expression was
altered in 4 known genes (alpha-1 protease inhibitor, cytochrome B5, stearoyl-
CoA desaturase and caboxylesterase) in two mice (Thai et al., 2001). Except for
Co-A desaturase, a similar pattern of gene change was reported in DCA-induced
tumors (10 tumors from 10 different mice) after 93 weeks. Using cDNA
microarray in the same mice, Thai et al. (2003) identified 24 genes with altered
expression, of which 15 were confirmed by Northern blot analysis after 4 weeks
of exposure. Of the 15 genes, 14 revealed expression suppressed two- to fivefold
and included: MHR 23 A, cytochrome P450 (CYP), 2C29, CYP 3A11, serum
paraoxonase/arylesterase 1, liver carboxylesterase, alpha-1 antitrypsin, ERp72,
GST-pi 1, angiogenin, vitronectin precursor, cathepsin D, plasminogen precursor
This document is a draft for review purposes only and does not constitute Agency policy.
E-407 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
(contains angiostatin), prothrombin precursor and integrin alpha 3 precursor. An
additional gene, CYP 2A4/5, had a twofold elevation in expression. After 93
weeks of treatment with 3.5 g/L DC A, Northern blot analyses of total RNA
isolated from DCA-induced hepatocellular carcinomas showed similar alteration
of expression (11 of 15). It was noted that peroxisome proliferator-activated
receptor (PPAR)a and IR gene expression were not changed by DCA treatment.
Genes involved in glycogen or lipid metabolism were not tested.
Although it has not been possible to determine directly whether DCA is produced
from TCE at carcinogenic levels, there is indirect evidence that DCA is formed
from TCE in vivo and contributes to liver tumor development. Pretreatment with
either DCA or TCE inhibits GST-zeta while TCA pretreatment does not (Bull et
al., 2004; Schultz et al., 2002). TCE treatment decreased Vmax for DCA
metabolism to 49% of control levels with a 1 g/kg TCE dose resembling effects
those of 0.05 g/L DCA (Schultz et al., 2002).
E.3.4.2.2. Genetic profiling data for Trichloroethylene (TCE): gene expression and
methylation status studies. Caldwell and Keshava (2006) and Keshava and Caldwell (2006)
reported on both genetic expression studies and studies of changes in methylation status
induced by TCE and its metabolites (see Sections E.2.3.2 and E.2.3.3, above) as well as
differences and
difficulties in the patterns of gene expression between differing PPARa agonists. In
Section E.4.2.2 (below), the effects of coexposures of DCA, TCA and Chloroform on
methylation status are discussed. In particular are concerns for the interpretation of studies that
employ pooling of data as well as interpretation of "snapshots in time of multiple gene
changes."
For the Laughter et al. (2004) study in particular, it is not clear whether transcription
arrays were performed on pooled data (no data on variability between individual animals was
provided and the methodology section of the report is not transparently written in this regard).
The issue of phenotypic anchoring also arises as data on percent liver/body weight indicates
significant variability within TCE treatment groups, especially in PPARa-null mice. For studies
of gene expression using microarrays Bartosiewicz et al. (2001) used a screening analysis of
148 genes for xenobiotic-metabolizing enzymes, DNA repair enzymes, heat shock proteins,
cytokines, and housekeeping gene expression patterns in the liver in response TCE. The TCE-
This document is a draft for review purposes only and does not constitute Agency policy.
E-408 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
induced gene induction was reported to be highly selective; only Hsp 25 and 86 and Cyp2a were
up-regulated at the highest dose tested. Collier et al. (2003) reported differentially expressed
mRNA transcripts in embryonic hearts from S-D rats exposed to TCE with sequences down-
regulated with TCE exposure appearing to be those associated with cellular housekeeping, cell
adhesion, and developmental processes. TCE was reported to induce up-regulated expression of
numerous stress-response and homeostatic genes.
Laughter et al. (2004) reported transcription profiles using macroarrays containing
approximately 1,200 genes in response to TCE exposure. Forty-three genes were reported to be
significantly altered in the TCE-treated wild-type mice and 67 genes significantly altered in the
TCE-treated PPARa knockout mice. Out of the 43 genes expressed in wild-type mice upon
TCE exposure, 40 genes were reported by the authors to be dependent on PPARa and included
genes for CYP4al2, epidermal growth factor receptor, and additional genes involved in cell
growth. However, the interpretation of this information is difficult because in general, PPARa
knockout mice have been reported to be more sensitive to a number of hepatotoxins partly
because of defects in the ability to effectively repair tissue damage in the liver (Mehendale,
2000; Shankar et al., 2003) and because a comparison of gene expression profiles between
controls (wild-type and PPARa knockout) were not reported.
As stated previously, knockout mice in this study also responded to TCE exposure with
increased liver weight, had increased background liver weights, and also had higher baseline
levels of hepatocyte proliferation than wild-type mice. Nakajima et al. (2000) reported that the
number of peroxisomes in hepatocytes increased by 2-fold in wild-type mice but not in PPARa
knockout mice. However, TCE induced increased liver weight in both male and female wild-
type and knockout mice, suggesting hepatic effects independent of PPARa activation. Ramdhan
et al (2010) also reported increased liver weight after TCE exposure in male wild type, PPARa-
null, and PPARa humanized mice to a similar extent.
In regards to toxicity, after three weeks of TCE treatment (0 tol,500 mg/kg via gavage),
Laughter et al. (2004) reported toxicity at thel,500 mg/kg level in the knockout mice that was
not observed in the wild-type mice — all knockout mice were moribund and had to be removed
from the study. Differences in experimental protocol made comparisons between TCE effects
and those of its metabolites difficult in this study (see Section E.2.1.13, above). After short-
term inhalation exposure, Ramdhan et al. (2010) reported increased TCE induction of toxicity in
PPARa-null and humanized mice in terms of hepatic steatosis and minimal levels of necrosis.
As reported by Voss et al. (2006), dose-, time course-, species-, and strain-related
differences should be considered in interpreting gene array data. The comparison of differing
PPARa agonists presented in Keshava and Caldwell (2006) illustrate the pleiotropic and varying
This document is a draft for review purposes only and does not constitute Agency policy.
E-409 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
liver responses of the PPARa receptor to various agonists, but did imply that these responses
were responsible for carcinogenesis.
As discussed above in Section E.3.3.5 and in Caldwell and Keshava (2006),
Aberrant DNA methylation has emerged in recent years as a common hallmark of
all types of cancers, with hypermethylation of the promoter region of specific
tumor suppressor genes and DNA repair genes leading to their silencing (an effect
similar to their mutation) and genomic hypomethylation (Ballestar and Esteller,
2002; Berger and Daxenbichler, 2002; Herman et al., 1998; Pereira et al., 2004b;
Rhee et al., 2002). Whether DNA methylation is a consequence or cause of cancer
is a long-standing issue (Ballestar and Esteller, 2002). Fraga et al. (2005; 2004)
reported global loss of monoacetylation and trim ethyl ati on of hi stone H4 as a
common hallmark of human tumor cells; they suggested, however, that
genomewide loss of 5-methylcytosine (associated with the acquisition of a
transformed phenotype) exists not as a static predefined value throughout the
process of carcinogenesis but rather as a dynamic parameter (i.e., decreases are
seen early and become more marked in later stages).
Although little is known about how it occurs, a hypothesis has also been proposed that
that the toxicity of TCE and its metabolites may arise from its effects on DNA methylation status.
In regard to methylation studies, many are coexposure studies as they have been conducted in
initiated animals, and as stated above, some are very limited in regard to the reporting and
conduct of the study.
Caldwell and Keshava (2006) reviewed the body of work regarding TCE, DCA, and TCA
for this issue. Methionine status has been noted to affect the emergence of liver tumors. As
noted by Counts et al. (1996) a choline/methionine deficient diet for 12 months did not increase
liver tumor formation in C3H/HeN mice but is tumorigenic to B6C3F1 mice. Tao et al. (2000a)
and Pereira et al. (2004b) have studied the effects of excess methionine in the diet to see if it has
the opposite effects as a deficiency (i.e., and reduction in a carcinogenic response rather than
enhancement). As noted above for Tao et al. (2000a), the administration of excess methionine in
the diet is not without effect. The data of Tao et al. (2000a) suggested that percent liver/body
weight ratios are affected by short-term methionine exposure (300 mg/kg) in female B6C3F1
mice.
Pereira et al. (2004b) reported that very high level of methionine supplementation to an
AIN-760A diet, affected the number of foci and adenomas after 44 weeks of coexposure to
This document is a draft for review purposes only and does not constitute Agency policy.
E-410 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
3.2.g/L DCA. While the highest concentration of methionine (8.0 g/kg) was reported to decrease
both the number of DCA-induce foci and adenomas, the lower level of methionine coexposure
(4.0 g/kg) increased the incidence of foci. Coexposure of methionine (4.0 or 8.0 g/kg) with 3.2
g/L DCA was reported to decrease by -25% DCA-induced glycogen accumulation, increase
mortality, but not to have much of an effect on peroxisome enzyme activity (which was not
elevated by more than 33% over control for DCA exposure alone).
Methionine treatment alone at the 8 g/kg level was reported to increase liver weight,
decrease lauroyl-CoA activity and to increase DNA methylation. The authors suggested that
their data indicate that methionine treatment slowed the progression of foci to tumors. Given that
increasing hypomethylation is associated with tumor progression, decreased hypomethylation
from large doses of methionine are consistent with a slowing of progression. Whether, these
results would be similar for lower concentrations of DCA and lower concentrations of
methionine that were administered to mice for longer durations of exposure, cannot be
ascertained from these data. It is possible that in a longer-term study, the number of tumors
would be similar. Whether, methionine treatment coexposure had an effect on the phenotype of
foci and tumors was not presented by the authors in this study. Such data would have been
valuable to discern if methionine coexposure at the 4.0 mg/kg level that resulted in an increase in
DCA-induce foci, resulted in foci of a differing phenotype or a more heterogeneous composition
than DCA treatment alone. Finally, a decrease in tumor progression by methionine
supplementation is not shown to be a specific event for the MOA for DCA-induced liver
carcinogenicity.
Tao et al. (2000a) reported that 7 days of gavage dosing of TCE (1,000 mg/kg in corn oil),
TCA (500 mg/kg, neutralized aqueous solution), and DCA (500 mg/kg, neutralized aqueous
solution) in 8-week old female B6C3F1 mice resulted in not only increased liver weight but also
increased hypomethylation of the promoter regions of c-Jun and c-Myc genes in whole liver
DNA (data shown for 1-2 mice per treatment). Treatment with methionine was reported to
abrogate this response only at a 300 mg/kg i.p. dose with 0-100 mg/kg doses of methionine
having no effect. Ge et al. (2001b) reported DCA- and TCA-induced DNA hypomethylation and
cell proliferation in the liver of female mice at 500 mg/kg and decreased methylation of the
c-Myc promoter region in liver, kidney and urinary bladder. However, increased "cell
proliferation" preceded hypomethylation. Ge et al. (2002) also reported hypomethylation of the
c-myc gene in the liver after exposure to the peroxisome proliferators 2,4-dichlorophenoxyacetic
acid (2,4-D)(l,680 ppm), dibutyl phthalate (20,000 ppm), Gemfibrozil (8,000 ppm), and
WY-14,643 (50-500 ppm, with no effect at 5 or 10 ppm) after six days in the diet. Caldwell and
Keshava (2006) concluded that hypomethylation did not appear to be a chemical-specific effect
This document is a draft for review purposes only and does not constitute Agency policy.
E-411 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
at these concentrations. As noted above in Section E.3.3.5, chemical exposure to a number of
differing carcinogens have been reported to lead to progressive loss of DNA methylation..
Caldwell and Keshava (2006) also noted similar changes in methylation after initiation
and treatment with DC A or TCA.
After initiation by N-methyl-N-nitrosourea (25 mg/kg) and exposure to 20 mmL/L
DCA or TCA (46 weeks), Tao et al. (2004a) report similar hypomethylation of
total mouse liver DNA by DCA and TCA with tumor DNA showing greater
hypomethylation. A similar effect was noted for region-2 (DMR-2) of the
insulin-like growth factor-II (IGF-II) gene. The authors suggest that
hypomethylation of total liver DNA and the IGF-II gene found in non-tumorous
liver tissue would appear to be the result of a more prolonged activity and not cell
proliferation, while hypomethylation of tumors could be an intrinsic property of
the tumors. Over expression of IGF-II gene in liver tumors and preneoplastic foci
has been shown in both animal models of hepatocarcinogenesis and humans, and
may enhance tumor growth, acting via the over-expressed IGF-I receptor (Scharf
et al., 2001; Werner and Le Roith, 2000). IGF-I is the major mediator of the
effects of the growth hormone; it thus has a strong influence on cell proliferation
and differentiation and is a potent inhibitor of apoptosis (Fiirstenberger and Senn,
2002). Normally, expression of IGF-II in liver is greater during the fetal period
than the adult, but is over-expressed in human hepatocarcinomas due to activation
of fetal promoters (Scharf et al., 2001) and loss of imprinting (Khandwala et al.,
2000). Takeda et al. (1996) report IGF-II expression in the liver is monoallelic
(maternally imprinted) in the fetal period is relaxed during the postnatal period,
(resulting in biallelic expression), and is imbalanced in human hepatocarcinomas
(leading to restoration of monoallelic IG-II expression).
However, Bull (2004b) and Bull et al. (2004) have recently suggested that
hypomethylation and peroxisome proliferation occur at higher exposure levels than those that
induce liver tumors for TCE and its metabolites. They reported that a direct comparison in the
no-effect level or low-effect level for induction of liver tumors in the mouse and several other
endpoints shows that, for TCA, liver tumors occur at lower concentrations than peroxisome
proliferation in vivo but that PPARa activation occurs at a lower dose than either tumor formation
or peroxisome proliferation. A similar comparison for DCA shows that liver tumor formation
occurs at a much lower exposure level than peroxisome proliferation, PPARa activation, or
This document is a draft for review purposes only and does not constitute Agency policy.
E-412 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
hypomethylation. In addition, they reported that these chemicals are effective as carcinogens at
doses that do not produce cytotoxicity.
E.3 .4.2.3. Oxidative Stress. Several studies have attempted to study the possible effects of
"oxidative stress" and DNA damage resulting from TCE exposures. The effects of
induction of metabolism by TCE, as well as through coexposure to ethanol, have been
hypothesized in itself
to increase levels of "oxidative stress" as a common effect for both exposures (see
Section E.4.2.4, below). Oxidative stress has been hypothesized to be the MOA for peroxisome
proliferators as well, but has been found to neither be correlated with cell proliferation nor
carcinogenic potency of peroxisome proliferators (see Section E.3.4.1.1). As a MOA, it is not
defined or specific as the term "oxidative stress" is implicated as part of the pathophysiologic
events in a multitude of disease processes and is part of the normal physiologic function of the
cell and cell signaling.
In regard to measures of oxidative stress, Rusyn (2006) noted that although an
overwhelming number of studies draw a conclusion between chemical exposure, DNA damage,
and cancer based on detection of 8-OHdG, a highly mutagenic lesion, in DNA isolated from
organs of in vivo treated animals, a concern exists as to whether increases in 8-OHdG represent
damage to genomic DNA, a confounding contamination with mitochondrial DNA, or an
experimental artifact. As described in Section E.2.2.8, the study by Channel et al. (1998)
demonstrated that corn oil as vehicle had significant effects on measures of "oxidative stress"
such as TBARS. Also as noted previously (see Sections E.2.1.1 and E.2.2.11), studies of TCE
which employ the i.p. route of administration can be affected by inflammatory reactions resulting
from that routes of administration and subsequent toxicity that can involve oxygen radical
formation from inflammatory cells.
The issues with interpretation of the Channel et al. (1998) study of TCE administered via
corn oil gavage to mice have already been discussed in Section E.2.1.7, above. The TBARS
results indicated suppression of TBARS with increasing time of exposure to corn oil alone with
data presented in such a way for 8-OHdG and total free radical changes that the pattern of corn
oil administration was obscured. It was not apparent from that study that TCE exposure induced
oxidative damage in the liver.
Toraason et al. (1999) measured 8-OHdG and a "free radical-catalyzed isomer of
arachidonic acid and marker of oxidative damage to cell membranes, 8-Epi-prostaglandin F2a
(8epiPGF)," excretion in the urine and TBARS (as an assessment of malondialdehyde and marker
This document is a draft for review purposes only and does not constitute Agency policy.
E-413 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of lipid peroxidation) in the liver and kidney of male Fischer rats (150-200 g) exposed to single
0, 100, 500, or 1,000 mg/kg TCE i.p. injections in Alkamuls vehicle (n = 6/group). Two
sequential urine samples were collected 12 hours after injection and animals were sacrificed at
24 hours with DNA collected from liver tissues and TBARS measured in liver homogenates. The
mean body weights of the rats were reported to vary by 13% but the liver weights varied by 44%
after the single treatments of TCE. In contrast to the large volume of the literature that reports
TCE-induced increases in liver weight, the 500 and 1,000 mg/kg exposed rats were reported to
have reduced liver weight by 44% in comparison to the control values.
Using this paradigm, 500 mg/kg TCE was reported to induce stage II anesthesia and a
1,000 mg/kg TCE to induce Level III or IV (absence of reflex response) anesthesia and burgundy
colored urine with 2/6 rats at 24 hours comatose and hypothermic. The animals were sacrificed
before they could die and the authors suggested that they would not have survived another 24
hours. Thus, using this paradigm there was significant toxicity and additional issues related to
route of exposure. Urine volume declined significantly during the first 12 hours of treatment and
while water consumption was not measured, it was suggested by the authors to be decreased due
to the moribundity of the rats. Given that this study examined urinary markers of "oxidative
stress" the effects on urine volume and water consumption, as well as the profound toxicity
induced by this exposure paradigm, limit the interpretation of the study.
The authors noted that because both using volume and creatinine excretion were affected
by experimental treatment, urinary excretion of 8-OHdG changed significantly based on the
mode of data expression. Excretion of 8epiPGF was reported to be no different from controls
12-24 hours and decreased 24 hours after TCE exposure at the two highest levels. Excretion of
8-OHdG was reported to not be affected by any exposure level of TCE and, if expressed on the
basis of 24-hours, decreased. TBARS concentration per gram of liver was reported to be
increased at the 500 and 1,000 mg/kg TCE exposure levels (-2-3-fold). The effects of
decreased liver size in the treated animals for this measure in comparison to control animals, was
not discussed by the authors. For 8-OHdG measures in the liver and lymphocytes, the authors
reported that "cost prohibited analysis of all of the tissues samples" so that a subset of animals
was examined exhibiting the highest TBARS levels. The number of animals used for this
determination was not given nor the data except for 500 mg/kg TCE exposure level. TCE was
reported to increase 8-OHdG/dG in liver DNA relative to controls to about the same extent in
lymphocytes from blood and liver (~2-fold) with the results for liver reported to be significant.
The issues of bias in selection of the data for this analysis, as well as the issues already stated for
this paradigm limit interpretation of these data while the authors suggest that evidence of
oxidative damage was equivocal.
This document is a draft for review purposes only and does not constitute Agency policy.
E-414 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
DCA and TCA have also been investigated using similar measures. Larson and Bull
(1992b) exposed male B6C3F1 mice [26 ± 3 g (SD)] to a single dose of 0, 100, 300, 1,000, or
2,000 mg/kg/d TCA or 0, 100, 300, or 1,000 mg/kg/d DCA in distilled water by oral gavage
(n = 4). Fischer 344 rats (237 ± 4 g) received a single oral dose of 0, 100, or 1,000 mg/kg DCA
or TCA (n = 4 or 5) TBARS was measured from liver homogenates and assumed to be
malondialdehyde. The authors stated that a preliminary experiment had shown that maximal
TBARS was increased 6 hours after a dose of DCA and 9 hours after a dose of TCA in mice (data
shown) and that by 24 hours TBARS concentrations had declined to control values (data not
shown). However, time-course information in rats was not presented and the same times used for
both species, (i.e., 6- and 9-hours time periods after administration of DCA and TCA) for
examination of TBARS activity. A dose of 100 mg/kg DCA (rats or mice) or TCA (mice) did
not elevate TBARS concentrations over that of control liver with this concentration of TCA not
examined in rats.
For TCA, there was a slight dose-related increase in TBARS over control values starting
at 300 mg/kg in mice (i.e., 1.68-, 2.02-, and 2.70-fold of control for 300, 1,000, and 2,000 mg/kg
TCA). For DCA there were similar increases over control for both the 300 and 1,000 mg/kg dose
levels in mice (i.e., 3.22- and 3.45-fold of control, respectively).
For rats the 1,000 and 2,000 mg/kg levels of TCA were reported to show a statistically
significant increase in TBARS over control (i.e., 1.67- and 2.50-fold, respectively) with the 300
and 1,000 mg/kg level of DCA showing similar increases but with only the 300 mg/kg-induced
change statistically significant different than control values (i.e., 3.0- and 2.0-fold of control,
respectively). Of note, is the report that the induction of TBARS in mice is transient and had
subsided within 24 hours of a single dose of DCA or TCA, that the response in mice appeared to
be slightly greater with DCA than TCA at similar doses, and that for DCA, there was similar
TBARS induction between rats and mice at similar dose levels.
A study by Austin et al. (1996) appears to a follow-up publication of the preliminary
experiment cited in Larson and Bull (1992b). Male B6C3F1 mice (8 weeks old) were treated
with single doses of DCA or TCA in buffered solution (300 mg/kg) with liver examined for 8-
OHdG. The authors stated that in order to conserve animals, controls were not employed at each
time point. For DCA the time course of 8-OHdG was studied at 0, 4, 6, and 8 hours after
administration and for TCA at 0, 6, 8, and 10 hours after of a 300 mg/kg dose (n = 6). There was
a statistically significant increase over controls in 8-OHdG for the 4- and 6-hour time points for
DCA (-1.4- and 1.5-fold of control, respectively) but not at 8 hours in mice. For TCA, there was
a statistically significant increase in 8-OHdG at 8 and 10 hours for TCA (-1.4- and 1.3-fold of
control, respectively).
This document is a draft for review purposes only and does not constitute Agency policy.
E-415 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The results for PCO and liver weight for Parrish et al. (1996) are discussed in
Section E.2.3.2.2 above for maleB6C3Fl mice exposed to TCA orDCA (0, 0.01, 0.5, and
2.0 g/L) for 3 or 10 weeks (n = 6). The study focused on an examination of the relationship with
measures of peroxisome proliferation and oxidative stress. The dose-related increase in PCO
activity at 21 days (-1.5-, 2.2-, and ~4.1-fold of control, for 0.1, 0.5, and 2.g/L TCA) was
reported not to be increased similarly for DC A. Only the 2.0 g/L dose of DC A was reported to
induce a statistically significant increase at 21-days of exposure of PCO activity over control
(~1.8-fold of control). After 71 days of treatment, TCA induced dose-related increases in PCO
activities that were approximately twice the magnitude as that reported at 21 days (i.e., ~9-fold
greater at 2.0 g/L level). Treatments with DCA at the 0.1 and 0.5 g/L exposure levels produced
statistically significant increase in PCO activity of-1.5- and 2.5-fold of control, respectively.
The administration of 1.25 g/L clofibric acid in drinking water, used as a positive control, gave
~6-7-fold of control PCO activity at 21 and 71 days exposure.
Parrish et al. (1996) reported that laurate hydroxylase activity was reported to be elevated
significantly only by TCA at 21 days and to approximately the same extent (-1.4 to 1.6-fold of
control) increased at all doses tested. At 71 days both the 0.5 and 2.0 g/L TCA exposures
induced a statistically significant increase in laurate hydroxylase activity (i.e., 1.6- and 2.5-fold of
control, respectively) with no change reported after DCA exposure. The actual data rather than
percent of control values were reported for laurate hydroxylase activity with the control values
varying 1.7-fold between 21 and 71 days experiments. Levels of 8-OHdG in isolated liver nuclei
were reported to not be altered from 0.1, 0.5, or 2.0 g/L TCA or DCA after 21 days of exposure
and this negative result was reported to remain even when treatments were extended to 71 days of
treatment.
The authors noted that the level of 8-OHdG increased in control mice with age (i.e., -2-
fold increase between 71-day and 21-day control mice). Clofibric acid was also reported not to
induce a statistically significant increase of 8-OHdG at 21 days, but to produce an increase (-1.4-
fold of control) at 71 days. Thus, the increases in PCO activity noted for DCA and TCA were
not associated with 8-OHdG levels (which were unchanged) and, also, not with changes laurate
hydrolase activity observed after either DCA or TCA exposure. Of note is the variability in both
baseline levels of PCO and laurate hydrolase activity. Also of note, is that the authors report
taking steps to minimize artifactual responses for their 8-OHdG determinations. The authors
concluded that their data does not support an increase in steady state oxidative damage to be
associated with TCA initiation of cancer and that extension of treatment to time periods sufficient
to insure peroxisome proliferation failed to elevate 8-OHdG in hepatic DNA. The increased 8-
OHdG at 10 weeks after Clofibrate administration but lack of 8-OHdG elevation at similar levels
This document is a draft for review purposes only and does not constitute Agency policy.
E-416 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
of PCO induction by were also noted by the authors to suggest that peroxisome proliferative
properties of TCA were not linked to oxidative stress or carcinogenic response.
As noted above for the study of Leakey et al. (2003a) (see Section E.2.3.4), hepatic
malondialdehyde concentration in ad libitum fed and dietary controlled mice did not change
with CH exposure at 15 months but the dietary controlled groups were all approximately half
that of the ad libitum-fed mice. Thus, while overall increased tumors observed in the ad libitum
diet correlated with increased malondialdehyde concentration, there was no association between
CH dose and malondialdehyde induction for either diet.
E.4. EFFECTS OF COEXPOSURES ON MODE OF ACTION (MOA)—INTERNAL
AND EXTERNAL EXPOSURES TO MIXTURES INCLUDING ALCOHOL
Caldwell et al. (2008b) published a review of the issues and studies involved with the
effects of coexposures to TCE metabolites that could be considered internal (i.e., an internal
coexposure for the liver) and coexposures to metabolites and other commonly occurring
chemicals that are present in the environment. As they stated:
Human exposure to a pollutant rarely occurs in isolation. EPA's Cumulative
Exposure project and subsequent National Air Toxics Assessment have
demonstrated that environmental exposure to a number of pollutants, classified
as potential human carcinogens, is widespread [U.S. EPA, 2006;(Woodruff et al.,
1998). Interactions between carcinogens in chemical mixtures found in the
environment have been a concern for several decades. Furthermore, how these
interactions affect the mode of action (MOA) by which these chemicals operate
and how such effects may modulate carcinogenic risk is of concern as well.
Thus, an understanding of the MOA(s) of a pollutant can help elucidate its
potential carcinogenic risk to humans, and can also help identify susceptible
subpopulations through their intrinsic factors (e.g., age, gender, and genetic
polymorphisms of key metabolic and clearance pathways) and extrinsic factors
(e.g. co-exposures to environmental contaminants, ethanol consumption, and
pharmaceutical use). Trichloroethylene (TCE) can be a useful example for
detailing the difficulties and opportunities for investigating such issues because,
for TCE, there is both internal exposure to a "chemical mixture" of multiple
carcinogenic metabolites (Chiu et al., 2006a; Chiu et al., 2006b) and co-
exposures from environmental contamination of TCE metabolites, and from
This document is a draft for review purposes only and does not constitute Agency policy.
E-417 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
pollutants that share common metabolites, metabolic pathways, MO As, and
targets of toxicity with TCE.
Typically, ground water or contaminated waste sites can have a large number of
pollutants that vary in regard to information available to support the
characterization of their potential hazard, and that have differing MO As and
targets. For example, Veeramachaneni et al. (2001) reported reproductive effects
in male rabbits, resulting from exposure to drinking water containing
concentrations of chemicals typical of ground water near hazardous waste sites.
The drinking water exposure mixture contained arsenic, chromium, lead,
benzene, chloroform, phenol, and TCE. Even at 45 weeks after the last
exposure, mating desire/ability, sperm quality, and Leydig cell function were
subnormal. However, while the exposure levels are relevant to human
environmental exposures, design of this study precludes a conclusion as to which
individual toxicant, or combination of the seven toxicants, caused the effects.
Thus, this study exemplifies he problems associated with studying a multi-
mixture milieu. Studies of the interactions of TCE metabolites or common co-
exposures that report the interactions of 2 or 3 chemicals at one time are easier to
interpret.
Since EPA published its 2001 draft assessment, several approaches have been
reported that include examination of tumor phenotype, gene expression, and
development of physiologically-based pharmacokinetic (PBPK) models to assess
possible effects of co-exposure. They attempt to predict whether such co-
exposures would produce additivity of response or if co-exposure would change
the nature of responses induced by TCE or its metabolites. In addition, new
studies on co-exposure to DBA may help identify a co-exposure of concern.
These studies may give potential insights into possible MO As and modulators of
TCE toxicity. More recent information on the toxicity of individual metabolites
of TCE (Caldwell and Keshava, 2006) may be helpful in trying to identify which
are responsible for TCE toxicity, but may also identify the effects of
environmental co-exposures.
Recently, EPA sought advice from the National Academy of Sciences (NAS)
(Chiu et al., 2006a) with the NAS charge questions including the following. (1)
This document is a draft for review purposes only and does not constitute Agency policy.
E-418 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
What TCE metabolites, or combinations of metabolites, may be plausibly
involved in the toxicity of TCE? (2) What chemical co-exposures may plausibly
modulate TCE toxicity? (3) What can be concluded about the potential for
common drinking water contaminants such as other solvents and/or haloacetates
to modulate TCE toxicity? (4) What can be concluded about the potential for
ethanol consumption to modulate TCE toxicity? Thus, the understanding of the
effects of co-exposure, in the context of MO A, is an important element in
understanding the risk of a potential human carcinogen.
U.S. EPA's draft TCE risk assessment (U.S. EPA, 2001) identified several
factors involving co-exposure to TCE metabolites, environmental contaminants,
and ethanol that could lead to differential sensitivity to TCE toxicity. Research
needs identified there, as well as in previous reviews (Bull, 2000; Pastino et al.,
2000), included further elucidation of the interaction of TCA and DCA in TCE-
induced liver tumors and a better understanding of the functional relationships
among risk factors. The complexity of TCE's potential interactions with
chemical co-exposures from either common environmental co-contaminants or
common behaviors such as alcohol consumption mirrors the complexity of the
metabolism and the actions of TCE metabolites. Thus, TCE presents a good case
study for further exploration of the effects of co-exposure on MOA.
The following sections first reiterate the findings of Bull et al. (2002) in regard to simple
coexposures of DCA and TCA which can be experienced as an internal coexposure after TCE
exposure. A number of studies have examined the effects of TCE or its metabolites after
previous exposure to presumably genotoxic carcinogen to not only determine the effect of the
coexposure on liver carcinogenicity but also to use such paradigms to distinguish between the
effects of TCA and DCA. Finally, not only is TCE a common coexposure with its own
metabolites, but is also a common coexposure with other solvents, and the brominated analogues
of TCA and DCA. The available literature is examined for potential similarities in target and
effects that may cause additional concern. The effects of ethanol on TCE toxicity is examined
as well as the potential pharmacokinetic modulation of risk using recently published reports of
PBPK models that may be useful in predicting coexposure effects.
This document is a draft for review purposes only and does not constitute Agency policy.
E-419 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
E.4.1. Internal Coexposures to Trichloroethylene (TCE) Metabolites: Modulation of
Toxicity and Implications for TCE Mode of Action (MOA)
Exposure to TCE will produce oxidative metabolites in the liver as an internal
coexposure. As stated above, the phenotypic analysis of TCE-induced tumors have similarities
to combinations of DCA and TCA and in some reports to resemble more closely DCA-induced
tumors in the mouse. Results from Bull et al. (2002) are presented in Section E.2.2.22 for the
treatment of mice to differing concentrations of DCA and TCA in combination and the
resemblance of tumor phenotype to that of TCE. In regard to cancer dose-response, the most
consistent treatment-related increase in response occurred with combinations of exposure to
DCA and TCA that appeared to increase lesion multiplicity when compared to effects from
individual chemicals separately. Bull et al. (2002) presented results for "selected" lesions
examined for pathology characterization that suggest coexposure of 0.5 g/L DCA with either 0.5
or 2 g/L TCA had a greater than additive effect on the total number of hyperplastic nodules. In
addition coexposure to 0.1 g/L DCA and 2 g/L TCA was reported to have a greater than additive
effect on the total number of adenomas, but not carcinomas, induced. The random selection of
lesions for the determination of potential treatment-related effects on incidence and multiplicity,
rather than characterization of all lesions, increases the uncertainty in this finding.
E.4.2. Initiation Studies as Coexposures
There is a body of literature that has focused on the effects of TCE and its metabolites
after rats or mice have been exposed to "mutagenic" agents to "initiate" hepatocarcinogenesis.
Given that most of these "initiating agents" have many effects that are not only mutagenic but
also epigenetic, that the dose and exposure paradigm modify these effects, that "initiators" can
increased tumor responses alone, and the tumors that arise from these protocols are reflective of
simultaneous actions of both "initiator" and "promoter," paradigms that first expose rats or mice
to a "mutagen" and then to other carcinogenic agents can be described as a coexposure
protocols.
As stated previously, DEN and A-nitrosomorpholine have been reported to increase
differing populations of mature hepatocytes with DEN not only being a mutagen but also able to
induce concurrent hepatocyte regeneration at a high dose. Thus, the effects of the TCE or its
metabolites are hard to discern from the effects of the "initiating" agent in terms of MOA.
As demonstrated in the studies of Pereira et al. (1997) below, the gender also
determines the nature of the tumor response using these protocols. In addition, when the
endpoint for examination is tumor phenotype the consequences of tumor progression are hard to
This document is a draft for review purposes only and does not constitute Agency policy.
E-420 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
discern from the MOA of the agents using paradigms of differing concentrations, different
durations of exposure, lesions counted as "tumors" to include different stages of tumor
progression (foci to carcinoma), and highly variable and low numbers of animals examined.
However, differences in phenotype of tumors resulting from such coexposures, like the
coexposure studies cited above for just TCE metabolites, can help determine that exposure to
TCE metabolites results in differing actions as demonstrated by differing effects in the presence
of cocarcinogens. As stated above, Kraupp-Grasl et al. (1990) use the same approach and note
differences among PPARa agonists in their ability to promote tumors suggest they should not
necessarily be considered a uniform group.
E.4.2.1. Herren-Freund et al. (1987)
The results of TCE exposure alone were reported previously (E.2.2.17) for this study.
This study's focus was on the effect of TCE, TCA, DCA and Phenobarbital on
hepatocarcinogenicity in male B6C3F1 mice after "initiation" at 15 days with 2.5 or 10 j_ig/g
body weight of ethylnitrosourea (ENU) and then subsequent exposure to TCE and other
chemicals in drinking water begging at 4 weeks of age (an age when the liver is already
undergoing rapid growth). DCA and TCA were given in buffered solutions and sodium chloride
given in the water of control animals. The experiment was reported to be terminated at 61
weeks because the "mice started to exhibit evidence of tumors." Concentrations of TCE were 0,
3 and 40 mg/L, of DCA and TCA 0, 2 and 5 g/L, and of Phenobarbital 0 and 500 mg/L. The
number of animals examined in each group ranged from 16 to 32. ENU alone in this paradigm
was reported to induce statistically significant increases in adenomas and hepatocellular
carcinomas (39% incidence of adenomas and 39% incidence of carcinomas vs. 9 and 0% for
controls) at the 10 jag/g dose (n = 23), but not at 2.5 j_ig/g dose (n = 22).
The effects of high doses of DCA and TCA alone have already been discussed for other
studies, as well as the lack of statistical power using a paradigm with so few and variable
numbers of animals, the limitations of an abbreviated duration of exposure which does not allow
for full expression of a carcinogenic response, and problems of volatilization of TCE in drinking
water. DCA and TCA treatments at these levels (5 g/L) were reported to increase adenomas and
carcinomas irrespective of ENU pretreatment and to approximately the same extent with and
without ENU. TCE at the highest dose was reported to increase the number of animals with
adenomas (37 vs. 9% in control) and carcinomas (37 vs. 0% in controls) but only the # of
adenomas/animal was statistically significant as the number of animals examined was only 19 in
This document is a draft for review purposes only and does not constitute Agency policy.
E-421 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
the TCE group. Phenobarbital was reported to have no effect on ENU tumor induction using
this paradigm.
E.4.2.2. Parnell et al. (1986)
This study used a rat liver foci bioassay (y-glutamyltranspeptidase, i.e., GGT) for hepatic
foci after at 3 and 6 month using protocols that included partial hepatectomy, DEN (10 mg/kg)
or TCA (1,500 ppm in drinking water) treatment, and then promotion with 5,000 ppm TCA (i.e.,
5	g/L) for 10, 20, or 30 days and phenobarbital (500 ppm) in male S-D rats (5-6 weeks old at
partial hepatectomy). The number of animals per group ranged from 4-6. PCO activities were
given for various protocols involving partial hepatectomy, DEN, TCA and Phenobarbital
treatments but there was no controls values given that did not have a least one of these
treatments.
Overall, it appeared there was a slight decrease of PCO activity in rats treated with
partial hepatectomy/DEN/Phenobarbital treatments and a slight increase over other treatments
for rats treated with partial hepatectomy/DEN/5,000 ppm TCA or just TCA from 2 weeks to
6	months of sampling. In regard to GGT-positive foci, the partial
hepatectomy/DEN/Phenobarbital group (n = 6) was reported to have more positive foci at 3 or
6 months than rats "initiated" with TCA and PB after partial hepatectomy or partial
2	2
hepatectomy/Phenobarbital treatment alone (2.05 foci/cm vs. -.05-0.10 foci/cm for all other
groups). The number of GGT positive foci in rats without any treatment were not studied or
presented by the authors. For "promotion" protocols the number of GGT positive foci induced
by the partial hepatectomy/DEN/Phenobarbital protocol at 3 and 6 months, appeared to be
reduced when Phenobarbital exposure was replaced by TCA coexposure but there was no dose-
response between the 50, 500 and 5,000 ppm. However, TCA treatment along with partial
hepatectomy and DEN treatment did increase the levels of foci (means of 0.71-0.39 foci/cm at
3 months and 1.83-2.45 foci/cm at 6 months) over treatment of just partial hepatectomy and
DEN (0.05 ± 0.20 foci/cm2 at 3 months and 0.30 ± 0.39 foci/cm2 at 6 months).
For the TCA animals treated only with 5,000 ppm TCA, the number of GGT positive
2	2
foci at 3 months was 0.23 ±0.16 foci/cm and at 6 months 0.03 ± 0.32 foci/cm with no values
for untreated animals presented. For the positive control (partial
hepatectomy/DEN/Phenobarbital) the number of GGT positive foci increased from 3 to 6
months (1.65 ± 0.23 foci/cm2 and at 6 months 7.61 ± 0.72 foci/cm2). The authors concluded that
This document is a draft for review purposes only and does not constitute Agency policy.
E-422 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
although TCA is reported to cause hepatic peroxisomal stimulation in rats and
mice, the results of this study indicate that it is unlikely TCA's effects are related
to the promoting ability seen here. The minimal stimulation of, 10 to 20% over
controls of peroxisomal associated, PCO activity in TCA exposed rats was seen
only at the 5000 ppm level and only within the promotion protocol. This finding
is in contrast to the promoting activity seen at all three concentrations of TCA.
E.4.2.3. Pereira and Phelps (1996)
The results for mice that were not "initiated" by exposure to MNU, but exposed to DCA
or TCA, are discussed in Section E.2.3.2.6. However, differences in responses after initiation
are useful for showing differences between single and coexposures as well as differences
between DCA and TCA effects. On Day 15 of age, female B6C3F1 mice received an i.p.
injection of MNU (25 mg/kg) and at 7 weeks of age received DCA (2.0, 6.67, or 20 mmol/L),
TCA (2.0, 6.67 mmol, or 20 mmol/L), orNaCl continuously for 31 or 51 weeks of exposure.
The number of animals studied ranged from 6 to 10 in 31-week groups and 6 to 39 in the
52-week groups. There was a "recovery group" in which mice received either 20 mmol/L
DCA (2.58 g/L DCA) (n = 12) or TCA (3.27 g/L TCA) (n = 11) for 31 weeks and then
switched to saline for 21 weeks until sacrifice at 52 weeks. Strengths of the study included the
reporting of hepatocellular lesions as either foci, adenomas, or carcinomas and the presentation
of incidence and multiplicity of each separately reported for the treatment paradigms.
Limitations included the low and variable number of animals in the treatment groups.
MNU was reported to not "significantly" induce foci or altered hepatocytes, adenomas,
or carcinomas at 31 (n= 10) or 51 weeks (n = 39). However, MNU did increase the incidence
and number/mouse of foci, adenomas and carcinomas at the 52 week sacrifice time in
comparison to saline controls, albeit at lower levels than observed in DCA or TCA
cotreatments groups (e.g., 10 vs. 0% foci, 17.5 vs. 2.5% adenomas, and 10 vs. 0% incidence of
carcinomas at 52 weeks for MNU-treated mice vs. saline control). Coexposure of DCA
(20.0 mmol/L) for 52 weeks in MNU-treated mice increased the number of foci and
hepatocellular adenomas with the authors reporting "the yield of total lesions/mouse increased
as a second order function of the concentration of DCA (correlation coefficients > 0.998)."
TCA coexposure in MNU-treated mice was reported not to result in a significant difference in
yield of foci or altered hepatocytes with either continuous 52 week or 31-week exposure, but
exposures to 20.0 or 6.67 mmol/L TCA did result in increased yield of liver tumors with both
exposure protocols (see below).
This document is a draft for review purposes only and does not constitute Agency policy.
E-423 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
For TCA treatment in MNU treated mice, the incidences of foci were similar (12.5 vs.
18.2%) but the number of foci/mouse was ~3-fold greater in the cessation protocol than with
continuous exposure. The incidence of adenomas was reported to be the same (-66%) as well
as the number of adenomas/animal between continuous and cessation exposures. For
carcinomas, there was a greater incidence for mice with continuous TCA exposure (83 vs.
36%) as well as a greater number of carcinomas/mouse (~4-fold) than for those initiated mice
with cessation of TCA exposure. As noted above, the number of animals treated with TCA
was low and variable (e.g., 23 mice studied at 52 weeks 20.0 mmol/L TCA, and 6 mice at
52 weeks 6.67 mmol/L TCA), limiting the ability to discern a statistically significant effect in
regard to dose-response. The concentration-response relationship for tumors/mouse after 31
and 51 weeks was reported to be best represented by linear progression.
A comparison of results for animals treated with MNU and 20.0 mmol/L DCA or TCA
for 31 weeks and sacrificed at 31 weeks and those which were treated with MNU and DCA or
TCA for 31 weeks and then sacrificed at 52 weeks is limited by the number of animals exposed
(n = 10 for 31 week sacrifice DCA or TCA, n= 12 for DCA recovery group, and n= 11 for
TCA recovery group). No carcinoma data were reported for animals exposed at 31 weeks and
sacrificed at 31 weeks making comparisons with recovery groups impossible for this parameter
and thus, determinations about progression from adenomas to carcinomas. For the MNU and
DCA-treated animals, the incidence or number of animals reported to have foci at 31 weeks
was reported to be 80% but 38.5% for in the recovery group. For adenomas, the incidence was
reported to be 50% for DCA-treated animals at 31 weeks and 46.2% for the recovery group.
For MNU and TCA-treated animals, the incidence of foci at 31 weeks was reported to 20 and
18.2%) for the recovery group. For adenomas, the incidence was reported to be 60% for the
TCA-treated animals at 31 weeks and 63.6% for the recovery group. Thus, this limited data set
shows a decrease in incidence of foci for the MNU and DCA-treated recovery group but no
change in incidence of foci for TCA or for adenomas for DCA- or TCA-treatment between
those sacrificed at 31 weeks and those sacrificed 21 weeks later.
In regard to multiplicity, the number of foci/mouse was reported to be 2.80 ± 0.20 for
the 31-week DCA group and 0.46 ± 0.18 for the recovery group (mean ± SEM). The number
of adenomas/mouse was reported to be 1.80 ± 0.83 for the 31-week group and 0.69 ± 0.26 for
the recovery group. Thus, both the number of foci and adenomas per mouse was reported to be
decreased after the recovery period for MNU and DCA treated mice. Given that the number of
animals with foci was decreased by half, the concurrent decrease in foci/mouse is not
surprising. For TCA treatments, the numbers of foci/mouse were reported to be 0.20 ±0.13 for
the 31-week group and 0.45 ± 0.31 for the recovery group. The number of adenomas/mouse
This document is a draft for review purposes only and does not constitute Agency policy.
E-424 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
for TCA-treatment groups was reported to be 1.30 ± 0.45 for the 31-week group and 0.91 ±
0.28 for the recovery group. For the MNU and TCA-treated mice, the numbers of foci/mouse
were reported to be increased and the number of adenomas/mouse reported to be slightly lower.
Because carcinoma data are not presented for the 31 week group, it is impossible to determine
whether the TCA adenomas regressed to foci or the TCA adenomas progressed to carcinomas
and more foci apparent with increased time.
For the comparison of the numbers of foci, adenomas, or carcinomas per mouse that
were reported for the mice exposed at 31 weeks and sacrificed and those exposed for 52 weeks,
issues arise as to the impact of such few animals studied at 31 weeks, and the differing
incidences of lesions reported for these mice on tumor multiplicity estimates. The number of
animals studied who treated with MNU and 20.0 mmol/L DCA or TCA for 31 weeks and then
sacrificed was n= 10, while the number of animals exposed to 20.0 mmol/L DCA or TCA for
52 weeks was 24 for the DCA group and 23 for the TCA group. The number of animals treated
at lower concentrations of DCA or TCA were even lower at the 31-week sacrifice (e.g., n = 6
for MNU and 6.67 mmol/L DCA at 31 weeks) and also for the 52-week durations of exposure
(e.g., n = 6 for MNU and 6.6.7 mmol/L TCA).
At 31 weeks, 80% of the animals were reported to have foci and 50% to have foci after
52 weeks of exposure to 20.0 mmol/L DCA and MNU treatment. Thus, similar to the
"recovery" experiment, the number of animals with foci decreased even with continuous
exposure between 31 and 52 weeks. For adenomas, 20.0 mmol DCA exposure for 31 weeks
was reported to induce adenomas in 50% of mice and after 52 weeks of exposure to induce
adenomas in 73% of mice. For TCA, the number of animals with foci was reported to be 20%
at 31 weeks and 12% at 52 weeks after exposure to 20.0 mmol/L TCA after MNU treatment
and similar to the incidence of foci reported for the TCA-recovery group. For 20.0 mmol TCA,
adenomas reported in 60% of mice after 31 weeks and in 67% of mice after 52 weeks of
exposure and also similar to the incidence of adenomas reported for the TCA-recovery group.
In regard to multiplicity, the number of foci/mouse was decreased from 2.80 ± 0.20 to
1.46 ± 0.48 between 31 weeks and 52 weeks of 20.0 mmol DCA in MNU exposed mice. The
number of adenomas/mouse was reported to be increased from 1.80 ± 0.83 to 3.62 ± 0.70
between 31 weeks and 52 weeks of 20.0 mmol DCA and MNU exposed mice. For
20.0 mmol/L TCA, the number of foci/mouse was 0.20 ± 0.13 and 0.13 ± 0.7 for 31- and
52-week exposures. The number of adenomas/mouse was reported to be 1.30 ± 0.45 and
1.29 ± 0.24 for 31- and 52-week exposures. Thus, by only looking at foci and adenoma
multiplicity data, there would not appear to be a change between 31 and 52-weeks.
This document is a draft for review purposes only and does not constitute Agency policy.
E-425 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
However, during progression a shift may occur such that foci become adenomas with
time and adenomas become carcinomas with time. For carcinomas there was no data reported
for 31-week exposure in MNU and DCA- or TCA-treated mice. However, at 52 weeks 20.0
mmol DCA was reported to induce carcinomas in 19.2% of mice and 20.0 mmol TCA to
induce carcinomas in 83% of mice. The corresponding numbers of carcinomas/mouse was
0.23 ±0.10 for 20.0 mmol/L DCA treatment and 2.79 ± 0.48 for 20.0 mmol/L TCA treatment
at 52 weeks in MNU treated mice. Thus, although fewer than 20% of MNU-treated mice were
reported to have foci at 20.0 mmol TCA, by 52 weeks almost all had carcinomas with -67%
also having adenomas. For DCA, many more mice had foci at 31 weeks (80%) than for TCA
and by 52 weeks -70% had adenoma with only -20% reported to have carcinomas. The
incidence data are suggestive that as these high doses of DCA and TCA, TCA was more
efficient inducing progression of a carcinogenic response than DCA in MNU-treated mice.
The authors interpreted the decrease in foci and adenomas between animals treated with
MNU and 20.0 mmol/L DCA for 31 weeks and sacrificed and those sacrificed 21 weeks later
to indicate that these lesions were dependent on continued exposure. However, the total
number of lesions cannot be ascertained because carcinoma data were not reported for 31-week
exposures. Carcinomas were reported in the recovery group at 52 weeks
(0.15 ± 0.10 carcinomas/mouse in 15.4% of animals). Of note is that not only did the number
of foci/mouse and incidence decrease between the 31-week group and the recovery group, but
also between 31- and 52-weeks of continuous exposure for the MNU and 20.0 mmol/L DCA
treated groups. Although derived from very few animals, the 6.67 mmol/L DCA group
reported no change for foci/mouse but a decrease in the incidence of foci between 31- and
52-weeks of exposure in MNU treated mice (i.e., 0.67 ± 0.18 foci/mouse in 50% of the animals
at 31 weeks and 0.50 ± 0.34 foci/mouse in 20% of mice treated for 52 weeks). The numbers of
foci/mouse for both MNU-treated and untreated control mice were reported to be decreased
between 31 and 51 weeks as well.
As noted in Section E.3.1.8. the number of "nodules" in humans, which may be
analogous to foci and adenomas, can spontaneously regress with time rather than becoming
hepatocellular carcinomas. Also as tumors get larger with progression, the number of
tumors/mouse can decrease due to coalescence of tumors and difficulty distinguishing between
them. While data are suggestive of a decrease in the number of adenomas/mouse after
cessation of DCA exposure, the incidence data are similar between the 31-week exposure and
recovery groups.
Of note is that the number of carcinomas/mouse and the incidence of carcinomas was
reported to be similar between the MNU-treated mice exposed continuously to 20.0 mmol/L
This document is a draft for review purposes only and does not constitute Agency policy.
E-426 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
DCA for 52 weeks and those which were treated for 31 weeks and then sacrificed at 52 weeks.
Also of note is that, although incidences and multiplicities of foci and adenomas was reported
to be relatively low in the 2.0 mmol/L DCA exposure groups, at 52-weeks 40% of the mice
tested had carcinomas with 0.70 ± 0.40 carcinomas/mouse. This was a greater percentage of
animals with carcinomas and multiplicity than that reported for the highest dose of DCA. This
result suggests that the effects in regard to tumor progression, and specifically for carcinoma
induction, differ between the lowest and highest doses used in this experiment. However, the
low numbers of animals examined for the lower doses, 31-weeks exposures, and in the
recovery group decrease the confidence in the results of this study in regard to the effects of
cessation of exposure on tumor progression.
In regard to tumor phenotype, in MNU-treated female mice that were not also exposed
to either DCA or TCA, all four foci and 86.7% of 15 adenomas were reported to be basophilic
and 13.3%) eosinophilic at the end of the 52 week-study. However, when MNU-treated female
mice were also exposed to DCA the number eosinophilic foci and tumors increased with
increasing dose after 52 weeks of continuous exposure. At the 20.0 mmol/L level all 38 foci
examined were eosinophilic and 99% of the tumors (almost all adenomas) were eosinophilic.
At the 2.0 mmol/L DCA exposure there were no foci examined but about 5 of 9 tumors
examined (-2:1 carcinoma:adenoma ratio) were basophilic and the other 4 were eosinophilic.
For TCA coexposure in MNU-treated mice, the 20 mmol/L TCA treatment was reported
to give results of 1 of the 3 foci examined to be basophilic and 2 that were eosinophilic. For
the 98 tumors examined (-2:1 carcinoma/adenoma ratio) 71.4% were reported to be basophilic
and 28.6% were eosinophilic. At the 2.0 mmol/L TCA exposure level, the 2 foci examined
were reported to be basophilic while the 6 tumors (all adenomas) were reported to be 50%
eosinophilic and 50% basophilic. Thus, after 52 weeks female mice treated with MNU and a
high dose of DCA had eosinophilic foci and adenomas and those treated with the high dose of
TCA had a mixture of basophilic and eosinophilic foci and tumors with a 3:1 ratio of tumors
(mostly carcinomas) being basophilic. At the lower doses of either DCA or TCA the tumors
tended to be mostly carcinomas for DCA and adenomas for TCA but both were -50%
basophilic and 50% eosinophilic. The tumors observed from MNU treatment alone were all
adenomas and mostly 87% basophilic. Thus, not only did treatment concentrations of DCA
and TCA give a different result for tumor multiplicity and incidence, but also for tumor
phenotype in MNU treated female mice. Eosinophilic foci and tumors were reported to be
consistently GST-71 positive while basophilic lesions "did not contain GST-71, except for a few
scattered cells or very small area comprising less than 5% of the tumor."
This document is a draft for review purposes only and does not constitute Agency policy.
E-427 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
Thus, exposure to either DCA or TCA increased incidence and number of animals with
lesions (foci, adenomas, or carcinomas) in MNU- versus nontreated mice (see
Section E.2.3.2.6, above). These results suggest that the pattern of foci, adenoma and
carcinoma incidence, multiplicity, and progression appeared to differ between TCA and DCA
in MNU-treated female mice. However, the low and variable number of animals used in this
study, make quantitative inferences between DCA and TCA exposures in "initiated" animals,
problematic.
E.4.2.4. Tao et al (2000a)
The source of liver tumors for this analysis was reported to be the study of Pereira and
Phelps (1996). Samples of liver "tumors" and "noninvolved" liver was homogenized for
protein expression for c-Jun and c-Myc and therefore, contained homogeneous cell types for
study. The term "liver tumors" was not defined so it cannot be ascertained as to whether the
lesions studied were altered foci, hepatocellular adenomas, or carcinomas. Liver tissues were
reported to be frozen prior to study which raises issues of m-RNA quality. Although this study
reports that there were no MNU-induced "tumors" the original paper of Pereira and Phelps
(1996) reports that there were four foci and 15 adenomas in MNU-only treated mice. The
authors reported no difference in c-Jun and c-Myc m-RNA from DCA or TCA-induced tumors
from mice "initiated" with MNU. DNA methyltransferase was reported to be decreased in
noninvolved liver in MNU-only treated mice in comparison to that from TCA- and DCA-
treated mice. For a comparison between noninvolved liver and tumors, tumors were reported
to have a greater level than did noninvolved liver.
E.4.2.5. Latendresse and Pereira (1997)
This study used the tumors from Pereira and Phelps (1996), except for the MNU-treated
only groups and those groups treated with either DCA or TCA but not MNU initiation, to further
study various biomarkers. The omissions were cited as to be due to insufficient tissue. For
immunohistochemical evaluation of the molecular biomarkers other than GST-71, liver
specimens from 7 MNU/20.0 mmol DCA- (i.e., 2.58 g/L DCA) treated and 6 MNU/20.0 mmol
TCA - (i.e., 3.27 g/L TCA) treated female mice randomly selected. For GST-71, the number of
animals from which lesion specimens were derived, was 24 MNU/DCA-treated and
23 MNU/TCA-treated mice.
This document is a draft for review purposes only and does not constitute Agency policy.
E-428 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
The DCA treated mice were reported to have 1-9 lesions/mouse and TCA treated mice
1-3 lesions/mouse. The number of lesions examined for each biomarker varied greatly. For
TCA-induced foci, no foci were examined for any biomarker except 3 lesions for GST-71, while
for DCA 12-15 foci were examined for each biomarker and 38 lesions examined for GST-71.
Similarly for TCA-induced adenomas, there were 8-10 lesions examined for all biomarkers with
32 lesions examined GST-71, while for DCA 12 lesions for all biomarkers with 94 lesions
examined for GST-71. Finally, for TCA-induced carcinomas there were 3-4 lesions examined
per group with 64 lesions examined for GST-71, while for DCA-induced carcinomas there were
no lesions examined for any biomarker except 3 examined for GST-71. The biomarkers used
were: GST-71, TGF-a, TGF-P, c-Jun, c-Fos, c-Myc, cytochrome oxidase CYP2E1, and
cytochrome oxidase CYP4A1.
MNU/DCA treatment was reported to produce "predominantly eosinophilic lesions" with
in general, the hepatocytes of DCA-promoted foci and tumors were less
pleomorphic and uniformly larger and had more distinctive cell borders than the
hepatocytes in lesions caused by TCA. Parenchymal hepatocytes of DCA-
promoted mice were uniformly hypertrophied, with prominent cell borders, and
the cytoplasm was markedly vacuolated, which was morphologically consistent
with the previous description of glycogen deposition in these lesions. In contrast,
TCA-promoted proliferative lesions tended to be basophilic, as previously
reported, and were composed of hepatocytes with less distinct cell borders, slight
cytoplasmic vacuolization, and greater variability in nuclear size and cellular size.
The hepatocytes of altered foci and hepatocellular adenomas from MNU-treated female
mice also treated with DCA were reported to stain positively for TGF-a, c-Jun, c-Myc,
CYP2E1, CYP4A1, and GST-71. The authors do not present the data for foci and adenomas
separately but as an aggregate and as the number of lesions with <50% cells stained or the
number of lesions with >50% cells stained either "minimally to mildly" or "moderately to
densely" stained.
Because no carcinomas for DCA were examined and especially because no foci for TCA
analyses were included in the aggregates, it is difficult to compare the profile between TCA and
DCA exposure in initiated animals and to separate these results from the effects of differences in
tumor progression. Thus, any differences seen in these biomarkers due to progression from foci
to adenoma in DCA-induced lesions or from progression of adenoma to carcinoma in TCA-
induce lesions, was lost. If the results for adenomas had been reported separately, there would
This document is a draft for review purposes only and does not constitute Agency policy.
E-429 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
have been a common stage of progression from which to compare the DCA and TCA effects on
initiated female mice liver tumors. For DCA-induced "lesions" (-50% foci and -50%
adenomas), most lesions had >50% cells staining with moderate to dense levels for TGF-a, and
CYP2E1, CYP4A1, and GST-71 and most lesions had <50% cells staining for even minimally to
mild staining for TGF-P and c-Fos. For c-.Jun and c-Myc the aggregate DCA-induced "lesions"
were heterogeneous in the amount of cells and the intensity of cell staining for these biomarkers
in MNU-treated female mice.
For the TCA "lesions" (—60% adenomas and -30% carcinomas) the authors note that
in general, the hepatocytes of tumors promoted by TCA demonstrated variable
immunostaining. With the exception of c-Jun, greater than 50% of the
hepatocytes in TCA lesions were essentially negative or stained only minimally to
mildly for the protein biomarkers studies. In some instances, particularly in TCA-
promoted tumors, there was regional staining variability within the lesions,
including immunoreactivity for c-Jun and c-Myc proteins, consistent with clonal
expansion or tumor progression.
As stated above, the term "lesion" refers to foci and adenomas for DCA but for
adenomas and carcinomas for TCA making inferences as to differences in the actions of the two
compounds through the comparisons of biomarkers confounded by the effects of tumor
progression. The largest differences in patterns between TCA induced "lesions" and those by
DCA appeared to be TGF-a (with no lesions having >50% cells stained mildly or
moderately/densely for TCA-induced lesions), CYP2E1 (with few lesions having >50%) stained
moderately/densely for TCA-induced lesions), CYP4A1 (with no lesions having >50%) stained
mildly or moderately/densely for TCA-induced lesions), and GST-71 (with all lesions having
<50%> cells stained even mildly for TCA-induced lesions). However, as shown by these data,
while the "lesions" induced by TCA and DCA had some commonalities within each treatment,
there was heterogeneity of lesions produced by both treatments in female mice already exposed
to MNU. Overall, the tumor biomarker pattern suggests differences in the effects of DCA and
TCA through differences in tumor phenotype they induce as coexposures with MNU treated
female mice.
The authors noted that nonlesion parenchymal hepatocytes in DCA-treated initiated mice
stained mostly negative for CYP2E1 and CYP4A1, while in TCA-treated mice staining patterns
in parenchymal nonlesions hepatocytes were centrilobular for CYP2E1 and panlobular for
CYP4A1 (a pattern for CYP4A1 that is opposite of that found in the TCA-induced lesions).
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 0 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E.4.2.6. Pereira et al. (1997)
This study used a similar paradigm as that of Pereira and Phelps (1996) to study
coexposures of TCA and DCA to female B6C3F1 mice already exposed to MNU. At 15 days
the mice received 25 mg/kg MNU and starting at 6 weeks of age neutralized solutions of either
0, 7.8, 15.6, 25.0 mmol/L DCA (n = 30 for control and 25 mmol/L DCA and n = 20 for 7.8 and
15.6 mmol/L DCA), 6.0 or 25.0 mmol/L TCA (n = 30 for 25.0 mmol/L TCA and n = 20 for
6.0 TCA), or combinations of DCA and TCA that included 25.0 mmol/L TCA + 15.6 mmol/L
DCA (n = 20), 7.8 mmol/L DCA + 6.0 mmol/L TCA (n = 25), 15.6 mmol/L DCA + 6.0 mmol/L
TCA (45), 25.0 mmol/L DCA + 6.mmol/L TCA (n = 25). The corresponding concentrations of
DCA and TCA in g/L is 25 mmol = 3.23 g/L, 15.6 mmol = 2.01 g/L and 7.8 mmol = 1.01 g/L
DCA and 25 mmol = 4.09 g/L and 6.0 mmol = 0.98 g/L TCA. Accordingly, the number of
animals at the beginning of the study varied between 20 and 45. At terminal sacrifice (after
44 weeks of exposure) the numbers of animals examined were less with the lowest number
examined to be 17 mice in the 7.8 mmol/L DCA group and the largest to be 42 in the
15.6 mmol/L DCA + 6.0 mmol/L TCA exposed group.
The authors reported that only a total of eight hepatocellular carcinomas were found in
the study (i.e., 25.0 mmol/L DCA induced 3 carcinomas, 7.8 mmol DCA + 6.0 mmol TCA
induced one carcinoma, and 25.0 mmol/L TCA induced 4 carcinomas). Thus, they presented
data for foci/mouse, and adenomas/mouse and their sum of both as "total lesions." The
incidences of lesions (i.e., how many mice in the groups had lesions) were not reported. The
shortened duration of exposure (i.e., 44 weeks), the omission of carcinomas from total "lesion"
counts (precluding consideration of progression of adenomas to carcinomas), the lack of
reporting of tumor incidences between groups, and the variable and low numbers of animals
examined in each group make quantitative inferences regarding additivity of these treatments
difficult. MNU treated mice did have a neoplastic response, albeit low using this paradigm.
For mice that were only exposed to MNU (n = 30 at terminal sacrifice) the mean number
of foci, adenomas and "lesions" per mouse were 0.21, 0.07 and 0.28, respectively. No data were
given for mice without MNU treatment but few lesions would be expected in controls. Pereira
and Phelps (1996) reported that saline-only treatment in 40 female mice for 51 weeks resulted in
0% foci, 0.03 adenomas/mouse in 2.5% of mice, and 0% carcinomas. In general, it appeared
that the numbers of foci, adenomas and the combination of both reported as "lesions" per mouse
that would have been predicted by the addition of multiplicities given for DCA, TCA, and MNU
treatments alone, were similar to those observed as coexposure treatments. The largest numbers
This document is a draft for review purposes only and does not constitute Agency policy.
E-431 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
of foci and adenomas/mouse were reported for the 25.0 mmol/L DCA and 6.0 mmol/L TCA
treatments in MNU treated mice (mean of 6.57 "lesions'Vmouse) with the lowest number
reported for 7.8 mmol/L DCA and 6 mmol/L TCA (mean of 1.16 "lesions'Vmouse).
The authors reported that the foci of altered hepatocytes were predominantly eosinophilic
in DCA-treated female mice initiated with MNU, while those observed after MNU and TCA
treatment were basophilic. MNU treatment alone induced 4 basophilic and 2 eosinophilic foci,
and 2 basophilic adenomas. MNU and DCA treatment was reported to produce only
eosinophilic foci and adenomas at the 25.0 mmol/L DCA exposure level. At the 7.8 mmol/L
DCA level of treatment in MNU-treated mice, 2 foci were basophilic, 4 were eosinophilic and
the 1 adenoma observed was reported to be eosinophilic. Thus, the concentration of exposure
appeared to alter the tincture of the foci observed after MNU and DCA exposure using this
paradigm. In this study, MNU and TCA treatment was reported to induce foci and adenomas
that were all basophilic at both 25.0 mmol/L TCA and 6.0 mmol/L TCA exposures. After
7.8 mmol/L DCA + 6.0 mmol/L TCA exposure, 2/23 foci were basophilic and 21/23 foci were
reported to be eosinophilic while all 4 adenomas reported for this group were eosinophilic.
Irrespective of treatment, eosinophilic foci for were reported to be GST-71 positive and
basophilic foci to be GST-71 negative. An exception was the 4 carcinomas in the group treated
with 25 mmol/L TCA which were reported to be predominantly basophilic but contained small
areas of GST-71 positive hepatocytes.
It should be noted that the increased dose (up to 3.23 g/L DCA and 4/09 g/L TCA) raises
issues of toxicity and effects on water consumption as other studies have noted toxicity at highly
doses of DCA and TCA. The use of an abbreviated duration of exposure in the study raises
issues of sensitivity of the bioassay at the lower doses used in the experiment. In particular, was
enough time provided to observe the full development of a tumor response? Finally, a question
arises as what can be concluded from the low numbers of foci examined in the study and the
affect of such low numbers on the ability to discern differences in these foci by treatment. As
with Pereira and Phelps, there appeared to be a difference the nature of the response induced by
coexposure of MNU to relatively high versus low DCA concentrations. Of note is that while
this experiment reported no hepatocellular carcinomas at the lowest dose of DCA at 44 weeks
(7.8 mmol DCA), Pereira and Phelps (1996) reported that in 9 mice treated with MNU and
2.0 mmol DCA for 52 weeks, there were no foci but 20% of mice had adenomas
(0.20 adenomas/mouse) and 40% of mice had carcinomas (0.70 carcinomas/mouse).
These results suggest that DCA coexposure affects TCA-induced lesions. The authors
concluded that mixtures of DCA and TCA appear to be at least additive and likely synergistic
and similar to the pathogenesis of DCA.
This document is a draft for review purposes only and does not constitute Agency policy.
E-432 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
E.4.2.7. Tao et al. (1998)
The focus of this study was an examination of tumors resulting from MNU and DCA or
TCA exposure in mice with the source of tumors was reported to be the study of Pereira et al.
(1997). Thus, similar concerns discussed above for that study paradigm are applicable to the
results of this study. The authors stated that there were also two recovery groups in which
exposure was terminated 1 week prior to euthanization at Week 44. The Pereira et al. (1997)
study does not report a cessation group in the study. In this study the number of animals treated
in the cessation group, the incidences of tumors in the mice, and the number of tumors examined
were not reported. Another group of female B6C3F1 mice (7-8 weeks old) were reported to not
be administered MNU but given 25 mmol/L DCA (3.23 g/L DCA), 25 mmol TCA (4.09 g/L
TCA), or control drinking water for 11 days (n = 7).
Hepatocellular adenomas in DCA-treated mice, adenomas and carcinomas in TCA-
treated mice were reported to be analyzed for percent-5-methylcytosine in the DNA of tumor
tissues. The levels of 5-methylcytosine in liver DNA of mice administered DCA or TCA for
11 days were reported to be reduced in comparison to control tissues (reduced to -36% of
control for DCA and -41% of control for TCA with the control value reported to be —3.5% of
DNA methylated). The number of animals examined was reported to be 7-10 animals per
group.
For control liver from mice that had received MNU but not DCA or TCA, and
noninvolved liver after 44 weeks of exposure to either, the levels of 5-methylcytosine were
similar and not different from the —3.5% of DNA methylated in untreated mice in the 11-days
experiment. Thus, initial decreases in methylated DNA shown by exposure to DCA or TCA
alone for 11 days, were not observed in "noninvolved" liver of animals exposed to either DCA
or TCA and MNU.
In regard to tumor tissues, the level of 5-methylcytosine in DNA of hepatocellular
adenomas receiving DCA and MNU was reported to be decreased by 36% in comparison to
noninvolved liver from the same animals. When exposure to DCA was terminated for 1 week
prior to sacrifice the level of 5-methylcytosine in the adenomas was reported to be higher and no
longer differed statistically from the noninvolved liver from the same animal or liver from
control animals only administered MNU. The number of samples was reported to be
9-16 samples without identification as to how many samples were used for each tumor analysis
or how many animals provided the samples (i.e., were most of the adenomas from on animal?)
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
For TCA the 5-methylcytosine level was reported to be reduced by 40% in hepatocellular
adenomas and 51% reduction in hepatocellular carcinomas in comparison to noninvolved liver
from the same animals. These levels were also reported to be less than that the control animals
administered only MNU.
Termination of exposure to TCA 1 week prior to sacrifice was reported to not produce a
statistically significant change in the level of 5-methylcytosine in either adenomas or
carcinomas. The levels of 5-methylcytosine were reported to be lower in carcinomas than
adenomas (-20% reduction) and to be lower in the "recovery" carcinomas than continuous
carcinomas (—25%) but were not reported as statistically significant. The results are reported to
have been derived from 8-16 "samples each." Again information on the number of animals
with tumors, whether the tumors were from primarily from one animal, and which DNA results
are from 8 versus 16 samples, was not provided by the authors.
Given that Pereira et al. (1997), the source for material of this study, reported that
treatment of MNU and 25.0 mmol/L TCA treatment for 44 weeks induced only 4 carcinomas, a
question arises as to how many carcinomas were used for the 44-week 5-methylcytosine results
in this study for carcinomas (i.e., how can 8-16 samples arise from 4 carcinomas?). In addition,
a question arises as to whether there was a difference in tumor-response in those animals with
and without one week of cessation of exposure which cannot be discerned from this report. The
use of highly variable number of samples between analysis groups and lack of information as to
how many tumors were analyzed adds uncertainty to the validity of these findings. There did
not appear to be a difference in methylation activity from short-term exposure to either DCA or
TCA alone in whole liver DNA extracts. However, the authors conclude that the difference in
methylation status between tumors resulting from MNU and DCA or TCA exposures supports
differences in the action between DCA and TCA.
E.4.2.8. Stauber et al. (1998)
In this study, 5-8 week old male B6C3F1 mice were used for isolation of primary
hepatocytes which were subsequently isolated and cultured in DCA or TCA. In a separate
experiment 0.5 g/L DCA was given to mice as pretreatment for 2 weeks prior to isolation. The
authors note that and indication of an "initiated cell" is anchorage-independent growth. DCA
and TCA solutions were neutralized before use. The primary hepatocytes from 3 mice per
concentration were cultured for 10 days with DCA or TCA colonies (8 cells or more)
determined in quadruplicate. The levels of DCA used were 0, 0.2, 0.5 and 2.0 mM DCA or
TCA. At concentrations of 0.5 mM or more DCA and TCA both induced an increase in the
This document is a draft for review purposes only and does not constitute Agency policy.
E-434 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
number of colonies that was statistically significant and increased with dose with DCA giving a
slightly greater effect. The authors noted that concentrations greater than 2.0 mM were
cytotoxic but did not show data on toxicity for this study.
Of great interest is the time-course experiment fro7m this study in which the number of
colonies from DCA treatment in vitro peaked by 10 days and did not change through days
15-25 at the highest dose. For the lower concentrations of DCA, increased time in culture
induced similar peak levels of colony formation by days 20-25 as that reached by 10 days at the
higher dose. Therefore, the number of colonies formed was independent of dose if the cells
were treated long enough in vitro. The number of colonies that formed in control hepatocyte
cultures also increased with time but at a lower rate than those treated with DCA (2.0 mM DCA
gave ~2-fold of control by 25 days of exposure to hepatocytes in culture). However, the level
reached by cells untreated in tissue culture alone by 20 days was similar to the level induced by
0.5 mM DCA by 10 days of exposure. This finding raises the issue of what these "colonies"
represent as tissue culture conditions alone transform these cells to what the authors suggest is
an "initiated" state. TCA exposure was not tested with time to see if it had a similar effect with
time as did DCA.
At 10 days, colonies were tested for c-Jun expression with the authors noting that
"colonies promoted by DCA were primarily c-Jun positive in contrast to TCA promoted
colonies that were predominantly c-Jun negative." For colonies that arose spontaneously from
tissue culture conditions, 10/13 (76.9%) were reported to be c-Jun +, those treated with DCA
28/34 (82.3%) were c-Jun +, and those treated with TCA 5/22 (22.7%) were c-Jun +. These
data show heterogeneity in cell in colonies although more were c-Jun + with DCA than TCA.
The number of colonies reported in the c-Jun labeling results represent sums between
experiments and thus, present total numbers of the control and the of colonies derived from
doses of DCA and TCA at 0.2 to 2.0 mM at 10 days. Thus, changes in colony c-Jun+ labeling
due to increasing dose cannot be determined.
The authors reported that with time (24, 48, 72, and 96 hours) of culture conditioning the
number of c-Jun+ colonies was increased in untreated controls. DCA treatment was reported to
delay the increase in c-Jun+ expression induced by tissue culture conditions alone in untreated
controls. TCA treatment was reported to not affect the increasing c-Jun+ expression that
increased with time in tissue culture. In this instance, tissue culture environment alone was
shown to transform cells and can be viewed as a "coexposure." DCA pretreatment in vivo was
reported to increase the number of colonies after plating which reached a plateau at 0.10 mM
and gave changes as at low a concentration of 0.02mM DCA administered in vitro. The
background level of colony formation varied between controls (i.e., 2-fold different in
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
pretreatment experiments and nonpretreatment experiments). Therefore, although the number of
colonies was greater for pretreatment with DCA, the magnitude of difference over the control
level was the same after DCA treatment in vitro with and without pretreatment.
The authors presented a comparison of "tumors" from Stauber and Bull (1997) and state
that DCA tumors were analyzed after 38 weeks of treatment but that TCA tumors were analyzed
after 52 weeks. They note that 97.5% of DCA-induced "tumors" were c-Jun + while none of the
TCA-induced "tumors" were c-Jun +. The concentrations used to give tumors in vivo for
comparison with in vitro results were not reported. What was considered to be "tumors" from
the earlier report for this analysis was also not noted. Stauber and Bull (1997) reported results
for combination of foci and tumors raising issues as to what was examined in this report. The
authors stated that because of such short time, no control tumors results were given. The short
and variable time of duration of exposure increases the possibility of differences between the in
vivo data resulting from differences in tumor progression as well as a decreased ability by the
shortened time of observation for full expression of the tumor response.
E.4.3. Coexposures of Haloacetates and Other Solvents
As noted by Caldwell et al. (2008b) drinking water exposure data suggest coexposure of
TCE and its haloacetic acid metabolites, TCA and DCA, is not an uncommon event as DCA and
TCA are the two most abundant haloacetates in most water supplies (Boorman et al., 1999;
Weisel et al., 1999). Dibromoacetic acid (DBA) concentrations have also been reported to range
up to approximately 20 [j,g/L in finished water and distribution systems (Weinberg et al., 2002).
Caldwell et al. (2008b) have also noted that coexposure in different media also occurs with
solvents like perchloroethylene (PERC) that may share some MO As, targets of toxicity, and
common metabolites that can therefore, potentially affect TCE health risk (Wu and Schaum,
2000). Some of the information contain in the following sections have been excerpted from the
discussions by Caldwell et al. (2008b) regarding the implications for the risk of TCE exposure
as modulated by coexposures to haloacetates and other solvents that have been studied and
reported in the literature.
E.4.3.1. Carbon tetrachloride, Dichloroacetic Acid (DCA), Trichloroacetic Acid (TCA):
Implications for Mode of Action (MOA) from Coexposures
Studies of specific combinations of TCE and chemicals colocated in contaminated areas
have been reported by Caldwell et al. (2008b). For carbon tetrachloride
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Pretreatment with TCE in drinking water at levels as low as 15 mM for three days
has been reported to increase susceptibility to liver damage to subsequent
exposure to a single IP injection of 1 mM/kg carbon tetrachloride (CCI4) in
Fischer 344 rats (Steup et al., 1991). Potential mechanistic explanations for this
observation included altered metabolism, decreased hepatic repair capability,
decreased detoxification ability, or combination of one or more of the above
activities. Simultaneous administration of an oral dose of TCE (0.5ml/kg) has
also been reported to increase the liver injury induced by an oral dose of 0.05
ml/kg CCI4 (Steup et al., 1993). The authors suggested that TCE appeared to
impair the regenerative activity in the liver, thus leading to increased damage
when CCI4 is given in combination with TCE.
As discussed above in Section E.4.2, initiation studies are in themselves a coexposure.
The study of Bull et al. (2004b) is included here as it not only used a coexposure of vinyl
carbamate with TCE metabolites, but also used carbon tetrachloride as a coexposure as well.
The rationale for this approach was that coexposure of TCE (and therefore, to its metabolites)
and CCI4 are likely to occur as they are commonly found together at contaminated sites.
Bull et al. (2004b) hypothesized that modification of tumor growth rates is an indication
of promotion rather than effects on tumor number, and that by studying tumor growth rates they
could classify carcinogens by their MO As. B6C3F1 male mice were initiated with vinyl
carbamate (3 mg/kg) at 2 weeks of age and then treated with DCA, TCA, CCI4, (0.1, 0.5, or 2.0
g/L for DCA and TCA; 50, 100 or 500 mg/kg CCL4 in 5% Alkamuls via gavage) in pair-wise
combinations of the three for 18 to 36 weeks. The exposure level of CCL4 to 5, 20 and 50
mg/kg was reported to be reduced at Week 24 due to toxicity for CCI4. The number of mice in
each group was reported to be 10 with the study divided into 5 segments. There were evidently
differences between treatment segments as the authors state that "because of some significant
quantitative differences in results that were obtained with replicate experiments treated in
different time frames, the simultaneous controls have been used in the analysis and presentation
of these data."
As with Bull et al. (2002), the interpretation of the results of the study is limited by a low
number of animals per group, short duration time of exposure and limited examination and
reporting of results. For example, a sample of 100 out of the 8,000 lesions identified in the
study was examined to verify that the general descriptor of neoplastic and nonneoplastic lesion
was correctly labeled with "tumors" describing a combination of hyperplastic nodules,
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
adenomas, and carcinomas. No incidence data were reported by the authors, but general lesion
growth information included mean lesion volume and multiplicity of lesions (numbers of
lesions/mouse). Using these reported indices, there appeared to be differences in treatment-
related effects.
As discussed in Caldwell et al. (2008b):
Each treatment was examined alone and then in differing combinations with each
other. Mice initiated with vinyl-carbamate, but without further exposure to the
other toxicants, were reported to have a few lesions that were of small size during
the examination period (20-36 weeks). At 30 weeks of CC14 exposure, there was
a dose-related response reported for multiplicity but mean lesion size was smaller
at the highest dose in initiated animals. At 36 weeks, DCA exposure was reported
to increase multiplicity at the two highest exposure levels and increased lesion
size at all levels compared to initiated-only animals. However, at a similar level
of induction, multiplicity and mean size of those lesions resulting from DCA
treatment were reported to be much smaller in comparison with CC14 treatment
(i.e., a 20-fold difference for lesion volume). At 36 weeks, treatments with the
same concentration of TCA or DCA induced similar multiplicity, but the mean
lesion volume was reported to be approximately 4-fold greater in tumors induced
by DCA as compared to TCA, and in animals treated with DCA multiplicity had
reached a plateau by 24 weeks rather than 36 for those treated with TCA.
Thus, using multiplicity of lesions and lesion volume as indicators of differences in
MO A, exposure to CCI4, DCA, and TCA appeared to produce distinct differences in results in
animals previously treated with vinyl carbamate.
As discussed in Caldwell et al. (2008b):
Simultaneous coexposure of differing combinations of CCI4, DCA, and TCA were
reported to give more complex results between 24 and 36 weeks of observation
but to show that coexposure had effects on lesion multiplicity and volume in
initiated animals. At 36 weeks, TCA coexposure appeared to reduce the lesion
volume of either DCA- or CCU-induced lesions after vinyl carbamate treatment.
Similarly, DCA coexposure was reported to reduce the lesion volume of either
TCA- or CCU-induced lesions when each was given alone after vinyl carbamate
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
treatment. With regard to multiplicity, TCA coexposure was reported to reduce
DCA-induced multiplicity only at the lowest dose of TCA while coexposure with
DCA increased multiplicity of CCU-induced lesions at all exposure levels. At 24
weeks, there appeared to be little effect on mean lesion volume by any of the
coexposures but DCA coexposure decreased multiplicity of TCA-induced lesions
(up to 3-fold) while TCA treatment slightly increased the number of CCU-induced
multiplicity (1.6-fold). This study confirms that short duration of exposure to all
three of these chemicals can cause lesions in already exposed to vinyl carbamate,
and suggests that combinations of these agents differentially influence lesion
number and growth rates. The authors have interpreted their results to indicate
differences in MOA between such treatments. However, the limitations of the
study limit conclusions regarding how such coexposure may be able to affect
toxicity and tumor induction and what the MOA is for each of these agents. This
is especially true at lower and more environmentally relevant concentrations
given for longer durations to uninitiated animals.
E.4.3.2. Chloroform, Dichloroacetic Acid (DCA), and Trichloroacetic Acid (TCA)
Coexposures: Changes in Methylation Status
In Section E.3.4.2.2, information on the effects of TCE and its metabolites was presented
in regard to effects on methylation status. After 7 days of gavage dosing, TCE, TCA and DCA
were reported to increased hypomethylation of the promoter regions of c-Jun and c-Myc genes
in mouse whole liver DNA, however, Caldwell and Keshava (2006) concluded that
hypomethylation did not appear to be a chemical-specific effect at the concentration used. Bull
et al. (2004b) suggested that hypomethylation occurs at higher exposure levels than those that
induce liver tumors for TCE and its metabolites. Along with studies of methylation changes
induced by a exposure to a single agent a Pereira et al. (2001) have attempted to examine the
effects on methylation changes from coexposures. This study was also reviewed by Caldwell et
al. (2008b).
Pereira et al. (2001) hypothesized that changes in the methylation status of DNA can be a
key event for MOA for DCA- and TCA-induced liver carcinogenicity through changes in gene
regulation, and that chloroform (CHCI3) coexposure may result in modification of DNA
methylation. As discussed in Caldwell et al. (2008b),
This document is a draft for review purposes only and does not constitute Agency policy.
E-43 9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
After 17 days of exposure of exposure to CHCI3 (0, 400, 800, 1,600 mg/L, n = 6
mice per treatment group) female B6C3F1 mice were coexposed to DCA or TCA
(500 mg/kg) during the last 5 days of exposure to chloroform. As noted by
Caldwell et al. (2008a), Pereira et al. (2001) reported the effects of
hypomethylation of the promoter region of c-Myc gene and on expression of its
mRNA in the whole livers of female B6C3F1 mice and thus, these results
represent composite changes in DNA methylation status from a variety of cell
types and for hepatocytes lumped from differing parts of the liver lobule. When
given alone, DCA, TCA, and to a lesser extent, the highest concentration of
CHCI3 (1,600 mg/L), was reported to decrease methylation of the c-myc promoter
region. Coadministration of CHCI3 (at 800 and 1,600 mg/L) was reported to
decrease DCA-induced hypomethylation while exposure to CHCI3 had no effect
on TCA-induced hypomethylation. Treatment with DCA, TCA, and, to a lesser
extent CHCI3 was reported to increase levels of c-myc mRNA. While expression
of c-myc mRNA was increased by DCA or TCA treatment, increasing
coexposures to CHCI3 were reported to attenuate the actions of DCA but not
TCA. Thus, differences in methylation status and expression of the c-myc gene
induced by DCA or TCA exposure was reported to be differentially modulated by
coexposure to CHCI3. The authors suggest these differences support differing
actions by DCA and TCA. However, whether these changes represent key events
in the induction of liver cancer is a matter of debate, especially as a "snapshot in
time" approach for such a nonspecific endpoint.
In a coexposure study in which an "initiating agent" was used as a coexposure along with
other coexposure, Pereira et al. (2001) treated male and female 15-day old B6C3F1 mice with
MNU (a cause of liver and kidney tumors) and then, starting at 5 weeks of age, treated them
with DCA (3.2 g/L) or TCA (4.0 g/L) along with coexposure to CHCI3 (0, 800, or 1,600 mg/L)
for 36 weeks. Mice were reported to be examined for evidence of promotion of liver and kidney
tumors. The numbers of animals in the exposure groups were highly variable, ranging from 25
(female initiated mice exposed to DCA) to 6 (female initiated mice exposed to DCA and
1,600 mg/L CHCI3), thus, limiting the power of the study to ascertain treatment-related changes.
However, unlike Bull et al., (2004b) all liver tissues were examined with incidences of foci,
adenomas, carcinomas, and both adenoma and carcinoma reported separately for treatment
groups. Multiplicity for a combination of adenomas and carcinomas were reported as well as
the tincture of foci and tumors.
This document is a draft for review purposes only and does not constitute Agency policy.
E-440 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Although as noted by Caldwell et al. (2008b):
[T]he statistical power of the study to detect change was very low, an examination
of the pattern of tumors induced by coexposure to MNU and TCE metabolites in
female mice suggested that: (1) DC A exposure increased the incidence of
adenomas but not carcinomas; (2) TCA increased incidence of carcinomas with
little change in adenoma incidence; (3) coexposure to 800 and 1,600 mg/L of
CHCI3 decreased adenoma incidence by DCA treatment but not TCA; and (4)
CHCI3 coexposure decreased multiplicity of TCA-induced tumors and foci, but
not for DCA. Caldwell et al. (2008a) also note that this study suggests a gender-
related effect on tumor induction from this study with; (1) adenoma incidences
similar in male and female mice treated with DCA, but carcinoma incidence
greater in males; (2) adenoma and carcinoma incidences greater in males than
females treated with TCA; (3) tumor multiplicity similar in both genders for DCA
treatments, but lower in females mice for TCA; and (4) less of an inhibitory effect
by CHCI3 on adenoma incidence from DCA exposure in male mice.
Pereira et al. (2001) also described the tinctural characteristics of the specific lesions
induced by their coexposure treatments. Both foci and tumors induced by DCA exposure in
"initiated" mice were reported to be over 95% eosinophilic in females, while in males, 89% of
the foci were eosinophilic and 91% of tumors were basophilic. Thus, not only was there a
gender-related difference in the incidences of tumors and foci but also foci and tumor
phenotype. CHCI3 coexposure was reported to change the DCA-induced foci from primarily
eosinophilic to basophilic (i.e., 11 vs. 75% basophilic) in male mice coexposed to MNU. In
male and female mice, TCA-induced tumors and foci were basophilic with no effect of CHCI3
on phenotype in MNU treated mice.
E.4.3.3. Coexposures to Brominated Haloacetates: Implications for Common Modes of
Action (MOAs) and Background Additivity to Toxicity
As noted by Caldwell et al. (2008b), along with chlorinated haloacetates and other
solvents, "coexposures with TCE and brominated haloacetates may occur through drinking
water. These compounds may affect TCE toxicity in a similar fashion to their chlorinated
counterparts. As bromide concentrations increase, brominated haloacetates increase in the water
supply."
This document is a draft for review purposes only and does not constitute Agency policy.
E-441 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Kato-Weinstein et al. (2001) administered dibromoacetate (DBA), bromochloroacetate
(BCA), bromodichloroacetate (BDCA), TCA, and DCA in drinking water at concentrations of
0.2-3 g/L for 12 weeks to B6C3F1 male mice. The focus of the study was to determine the
similarity in action between the brominated and chlorinated haloacetates. Each of the
haloacetates, given individually, were reported to increase liver/body weight ratios in a dose-
dependent manner.
The dihaloactates, DCA, BCA and DBA, caused liver glycogen accumulation both by
chemical measurements in liver homogenates and in ethanol-fixed liver sections (to preserved
glycogen) stained with PAS. For DCA, a maximal level of glycogen increase was observed at 4
weeks of exposure at a 2 g/L exposure concentration. They report a 1.60-fold of control percent
liver/body weight and 1.50-fold of control glycogen content after 8 weeks of exposure to 2 g/L
DCA in male B6C3F1 mice. The baseline level of glycogen content (-60 mg/g) and the
increase in glycogen after DCA exposure was consistent with the results reported by Pereira et
al. (2004b). The percent liver/body weight data for control mice was for animals sacrifice at 20
weeks of age. The 4-12 week exposure to DCA were staggered so that all animals would be 20
weeks of age at sacrifice. Thus, the animals were at differing ages at the beginning of DCA
treatments between the groups.
However, as with Pereira et al. (2004b) the -10% increase in liver mass that the
glycogen increases represent are lower than the total increase in liver mass reported for DCA
exposure. The authors noted possible contamination of BCA with small percentages of DCA
and DBA in their studies (i.e., 84% BCA, 6% DCA and 8% DBA). The trihaloacetates (TCA
and low concentrations of BDCA) were reported to produce slight decreases in liver glycogen
content, especially in the central lobular region in cells that tended to accumulate glycogen in
control animals. These effects on liver glycogen were reported at the lowest dose examined
(i.e., 0.3 g/L). At the highest concentration, BDCA was reported to induce a pattern of glycogen
distribution similar to that of DCA in mice.
All dihaloacetates were reported to reduce serum insulin levels at high concentrations.
Conversely, trihaloacetates were reported to have no significant effects on serum insulin levels.
For the study of peroxisome proliferation and DNA synthesis, mice were treated to BCA, DBA,
and BDCA for 2, 4, or 26 weeks. The effects on DNA synthesis were small for all brominated
haloacetates with only DBA reported to show a significant increase in DNA synthesis at 2 and 4
weeks but not at 26 weeks (increase in DNA synthesis was 3-fold of the highest control level).
Of note is the highly variable level of DNA synthesis reported for control values that varied to a
much higher degree (~3-6-fold variation within control groups at the same time points) than did
treatment-related changes. DBA was the only brominated haloacetate that was reported to
This document is a draft for review purposes only and does not constitute Agency policy.
E-442 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
consistently increased PCO activity as a percentage of control values (actual values and
variability between controls were not reported) with a 2-3-fold increase in PCO activity at 0.3
to 3.0 g/L DBA. DBA-induced PCO activity increases were reported to be limited to 2-4 weeks
of treatment in contrast to TCA, which the authors reported to increase PCO activity
consistently over time.
Tao et al. (2004a) reported DNA methylation, glycogen accumulation and peroxisome
proliferation after exposure of female B6C3F1 mice and male Fischer 344 rats exposed to 1 or
2 g/L DBA in drinking water for 2 to 28 days. DBA was reported to induce dose-dependent
DNA hypomethylation in whole mouse and rat liver after 7 days of exposure with suppression
sustained for the 28-day exposure period. The expression of mRNA for c-Myc in mice and rats
and mRNA expression of the IGF-II gene in female mice were reported to be increased during
the same period. Both rats and mice were reported to exhibit increased glycogen with mice
having increased levels at 2 day and rats at 4 days. DBA was reported to cause an increase in
lauroyl-CoA oxidase activity (a marker of peroxisome proliferation) in both mice (after 7 days)
and rats (after 4 days) that was sustained for 28 days.
Methylation changes reported here for DBA exposure in rats and mice are consistent
with those reported for TCA and DCA by Pereira et al. (2001) in mice. The pattern of glycogen
accumulation was also similar to that reported for DCA by Kato-Weinstein et al. (2001) and
suggests that the brominated analogues of TCE metabolites exhibited similar actions as their
chlorinated counterparts. In regard to peroxisomal enzyme activities Kato-Weinstein et al.
(2001) reported PCO activity to be limited to 2-4 weeks with Tao et al. (2004a) reporting
lauroyl-CoA oxidase activity to be sustained for the lengths of the study (28-days) for DBA.
As noted by Caldwell et al. (2008b), "given the similarity of DCA and DBA effects, it is
plausible that DBA exposure also induces liver cancer. Melnick (2007) reported the results of
DBA exposure to F344/N rats and B6C3F1 mice exposed to DBA for 3 months or 2 years in
drinking water (0, 0.05, 0.5, or 1.0 g/L DBA for 2 years). Neoplasms at multiple sites were
reported in both species exposed to DBA for 2 years with no effects on survival and little effect
on mean body weight in either species. Similar to TCE, DCA and TCA, the liver was reported
to be a target of DBA exposure. After 3-months of exposure, there were dose-related increases
in hepatocellular vacuolization and liver weight reported in rats and mice described as
'glycogen-like.'" The authors report that the major neoplastic effect of DBA in rats was
induction of malignant mesotheliomas in males and increased incidence of mononuclear cell
leukemia in males and females. For mice, the major neoplastic effect of DBA exposure was
reported to be the increased incidence of hepatocellular adenomas and carcinomas at all
exposure levels.
This document is a draft for review purposes only and does not constitute Agency policy.
E-443 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
In addition to these liver tumors, hepatoblastomas were also reported to be increased in
all exposure groups of male mice and exceeded historical control rates. The incidence of
alveolar/bronchiolar adenoma and carcinoma was reported to be increased in the 0.5 g/L group
of male mice along with marginal increases in alveolar hyperplasia in DBA-treated groups. The
authors reported that the increases in hepatocellular neoplasms were not associated with
hepatocellular necrosis or regenerative hyperplasia and concluded that an early increase in
hepatocyte proliferation were not likely involved in the MOA for DBA because no increases in
hepatocyte DNA labeling index were observed in mice exposed for 26 days and the slight
increase that occurred in male F344 rats was not accompanied by an increase in liver tumor
response.
As noted by Caldwell et al. (2008b),
[T]he results of Kato-Weinstein et al. (2001), Tao et al. (2004a), and Melnick et
al. (2007) are generally consistent for DBA and show a number of activities that
may be common to TCE metabolites (i.e., brominated and chlorinated haloacetate
analogues generally have similar effects on liver glycogen accumulation, serum
insulin levels, peroxisome proliferation, hepatocyte DNA synthesis, DNA
methylation status, and hepatocarcinogenicity). It is therefore, plausible that these
effects may be additive in situations of coexposure. However, as noted by
(Melnick et al., 2007), methylation status, events associated with PPARa
agonism, hepatocellular necrosis, and regenerative hyperplasia are not established
as key events in the MOA of these agents, and the MO As for DCA- and DBA-
induced liver tumors are unknown.
E.4.3 .4. Coexposures to Ethanol: Common Targets and Modes of Action (MOAs)
As noted in the U.S. EPA's draft TCE assessment (U.S. EPA, 2001), alcohol
consumption is a common coexposure that has been noted to affect TCE toxicity with TCE
exposure cited as potentially increasing the toxicity of methanol and ethanol, not only by
altering their metabolism to aldehydes, but also by altering their detoxification (e.g., similar to
the "alcohol flush" reported for those who have an inactive aldehyde dehydrogenase allele). As
noted by Caldwell et al. (2008b) "chemical co-exposures from both the environment and
behaviors such as alcohol consumption may have effects that overlap with TCE in terms of
active agents, pharmacokinetics, pharmacodynamics, and/or target tissue toxicity."
Caldwell et al. (2008b) also noted:
This document is a draft for review purposes only and does not constitute Agency policy.
E-444 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
In their review of solvent risk (including TCE), Brautbar and Williams (2002)
suggest that laboratory testing that is commonly used by clinicians to detect liver
toxicity may not be sensitive enough to detect early liver hepatotoxicity from
industrial solvents and that the final clinical assessment of hepatotoxicity and
industrial solvents must take into account synergism with medications, drugs of
use and abuse, alcohol, age-dependent toxicity, and nutrition. Although many of
these factors may be important, the focus in this section is on the effects of
ethanol. Contemporary literature reports effects similar to those of TCE's and
previous reports indicate ethanol consumption impacts TCE toxicity in humans,
affects the pharmacokinetics and toxicity of TCE in rats, and is also a risk factor
for cancer.
The association between malignant tumors of the upper gastrointestinal tract and
liver and ethanol consumption is based largely on epidemiological evidence, and
thought to be causally related (Badger et al., 2003; Bradford et al., 2005).
Studies of the mechanisms of ethanol carcinogenicity have suggested the
importance of its metabolism, including induction of CYP2E1 associated
increases in production of reactive oxygen species and enhanced activation of a
variety of pro-carcinogens, alteration of retinol and retinoic acid metabolism, and
the actions of the metabolite acetaldehyde (Badger et al., 2003). While ethanol is
primarily metabolized by alcohol dehydrogenase, it undergoes simultaneous
oxidation to acetate by hepatic P450s, primarily CYP2E1. Both chronic ethanol
consumption as well as TCE treatment induces CYP2E1 (U.S. EPA, 2001).
Oneta et al. (2002) report that even at moderate chronic ethanol consumption,
hepatic CYP2E1 is induced in humans, which they suggest, may be of
importance in the pathogenesis of alcoholic liver disease; of ethanol, drug, and
vitamin A interactions; and in alcohol-associated carcinogenesis. Induction of
CYP2E1 can cause oxidative stress to the liver from nicotinamide dinucleotide
phosphate (NADPH)-dependent reduction of dioxygen to reactive products even
in the absence of substrate, and subsequent apoptosis (Badger et al., 2003).
Bradford et al. (2005) suggest that CYP2E1, and notNADPH oxidase, is
required for ethanol-induced oxidative DNA damage to rodent liver but that
NADPH oxidase-derived oxidants are critical for the development of ethanol-
induced liver injury.
This document is a draft for review purposes only and does not constitute Agency policy.
E-445 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
There is increasing evidence that acetaldehyde, which is toxic, mutagenic, and
carcinogenic, rather than alcohol is responsible for its carcinogenicity (Badger et
al., 2003). Mitochondrial aldehyde dehydrogenase (ALDH2) disposes of
acetaldehyde generated by the oxidation of ethanol, and ALDH2 inactivity
through mutation or polymorphism has been linked to esophageal cancer in
humans (everyday drinkers and alcoholics) (Badger et al., 2003). For instance,
increased esophageal cancer risk was reported for patients with the ALDH3*1
polymorphism as well as increased acetaldehyde in their saliva. TCE exposure
has also been reported to induce a similar alcohol flush in humans which may be
linked to its ability to decrease ALDH activities at relatively low concentrations
and thus conferring a similar status to individuals with inactive ALDH2 allele
(Wang et al., 1999). Whether the MOA for the buildup of acetaldehyde after
ethanol and TCE co-exposure is: (1) the induction of CYP2E1 by TCE resulting
in increased metabolism to acetaldehyde; (2) inhibition of ALDH and thus
reduced clearance of acetaldehyde, or (3) a number of other actions are
unknown. Crabb et al. (2001) reported 20-30% reductions in ALDH2 protein
level by PPARa agonists (Clofibrate treatment in rats and WY treatment in both
wild and PPARa-null mice). This could be another pathway for TCE-induced
effects on ethanol metabolism. It is an intriguing possibility that the reported
association between the increased risk of human esophageal cancer and TCE
exposure (Scott and Chiu, 2006) could be related to TCE effects on
mitochondrial ALDH, given a similar association of this endpoint with ethanol
consumption or ALDH polymorphism.
Finally, ethanol ingestion may have significant effects on TCE
pharmacokinetics. Baraona et al. (2002a; 2002b) reported that chronic, but not
acute, ethanol administration increased the hepatotoxicity of peroxynitrite, a
potent oxidant and nitrating agent, by enhancing concomitant production of nitric
oxide and superoxide. They also reported that nitric oxide mediated the
stimulatory effects of ethanol on blood flow. Ethanol markedly enhanced portal
blood flow (2-fold increase), with no changes in the hepatic, splenic, or
pancreatic arterial blood flows in rats.
This document is a draft for review purposes only and does not constitute Agency policy.
E-446 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
E.4.3.5. Coexposure Effects on Pharmacokinetics: Predictions Using Physiologically Based
Pharmacokinetic (PBPK) Models
Along with experimental evidence that has focused on chronic and acute experiments
using rodents, the potential pharmacokinetic modulation of risk has also been recently published
reports using PBPK models that may be useful in predicting coexposure effects. Caldwell et al.
(2008b) also examined and discussed these approaches and noted:
An important issue for prediction of the effects and relationship on MO As by
co-exposure is the degree to which modulation of TCE toxicity by other agents
can be quantified. Pharmacokinetics or the absorption, distribution, metabolism,
and elimination of an agent, can be affected by internal and external co-exposure.
Such information can help to identify the chemical species that may be causally
associated with observed toxic responses, the MO A, and ultimately identify
potentially sensitive subpopulations for an effect such as carcinogenicity.
Physiologically based pharmacokinetic (PBPK) models are often used to
estimate and predict the toxicologically relevant dose of foreign compounds in
the body and have been developed to predict effects on pharmacokinetics that are
additive or less or greater than additive. One of the first such models was
developed for TCE (Andersen et al., 1987b). Given that TCE, PERC, and
methyl chloroform (MC) are often found together in contaminated groundwater,
Dobrev et al. (2001) attempted to investigate the pharmacokinetic interactions
among the three solvents to calculate defined "interaction thresholds" for effects
on metabolism and expected toxicity. Their null hypothesis was defined as
competitive metabolic inhibition being the predominant result for TCE given in
combination with other solvents. Gas uptake inhalation studies were used to test
different inhibition mechanisms. A PBPK model was developed using the gas
uptake data to test multiple mechanisms of inhibitory interactions (i.e.,
competitive, noncompetitive, or uncompetitive) with the authors reporting
competitive inhibition of TCE metabolism by MC and PERC in simulations of
pharmacokinetics in the rat. Occupational exposures to chemical mixtures of the
three solvents within their Threshold Limit Value (TLV)/TWA limits were
predicted to result in a significant increase (22%) in TCE blood levels compared
with single exposures.
This document is a draft for review purposes only and does not constitute Agency policy.
E-447 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Dobrev et al. (2002) extended this work to humans by developing an interactive
human PBPK model to explore the general pharmacokinetic profile of two
common biomarkers of exposure, peak TCE blood levels, and total amount of
TCE metabolites generated in rats and humans. Increases in the TCE blood
levels were predicted to lead to higher availability of the parent compound for
GSH conjugation, a metabolic pathway that may be associated with kidney
toxicity/carcinogenicity. A fractional change in TCE blood concentration of
15% for combined TLV exposures of the three chemicals (25/50/350 ppm of
PERC/TCE/MC) resulted in a predicted 27% increase of the S-(l, 2-
dichlorovinyl)-L-cysteine (DCVC) metabolites, indicating a nonlinear risk
increase due to combined exposures to TCE. Binary combinations of the
solvents produced GST-mediated metabolite levels almost twice as high as the
expected rates of increase in peak blood levels of the parent compound. The
authors suggested that using parent compound peak blood levels (a less sensitive
biomarker) would result in two to three times higher (i.e., less conservative)
estimates of potentially safe exposure levels. In regard to the detection of
metabolic inhibition by PERC and MC, the simulations showed TCE blood
concentrations to be the more sensitive dose metric in rats, but the total of TCE
metabolites to be the more sensitive dose measure in humans. Finally,
interaction thresholds were predicted to be occurring at lower levels in humans
than rats.
Thrall and Poet (2000) investigated the pharmacokinetic impact of low-dose
co-exposures to toluene and TCE in male F344 rats in vivo using a real-time
breath analysis system coupled with PBPK modeling. The authors report that,
using the binary mixture to compare the measured exhaled breath levels from
high- and low-dose exposures with the predicted levels under various metabolic
interaction simulations (competitive, noncompetitive, or uncompetitive
inhibition), the optimized competitive metabolic interaction description yielded
an interaction parameter Ki value closest to the Michaelis-Menten affinity
parameter (KM) of the inhibitor solvent. This result suggested that competitive
inhibition is the most plausible type of metabolic interaction between these two
solvents.
This document is a draft for review purposes only and does not constitute Agency policy.
E-448 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Isaacs et al. (2004) have reported gas uptake co-exposure data for CHCI3 and
TCE. The question as to whether it is possible to use inhalation data in
combination with PBPK modeling to distinguish between different metabolic
interactions was addressed using sensitivity analysis theory. Recommendations
were made for design of optimal experiments aimed at determining the type of
inhibition mechanisms resulting from a binary co-exposure protocol. This paper
also examined the dual nature of inhibition of each chemical in the pair to each
other, which is that TCE and CHCI3 were predicted to interact in a competitive
manner. Even though as stated by Dobrev et al. (2001), other solvents inhibit
TCE metabolism, it is also possible to quantify the synergistic interaction that
TCE has on other solvents, using techniques such as gas uptake inhalation
exposures.
Haddad et al. (2000) has developed a theoretical approach to predict the
maximum impact that a mixture consisting of co-exposure to dichloromethane,
benzene, TCE, toluene, PERC, ethylbenzene, m-, p-, and o-xylene, and styrene
would have on venous blood concentration due to metabolic interactions in
Sprague-Dawley rats. Two sets of experimental co-exposures were conducted.
The first study evaluated the change in venous blood concentration after a 4 hour
constant inhalation exposure to the 10 chemical mixtures. This experiment was
designed to examine metabolic inhibition for this complex mixture. The second
study was designed to study the impact of possible enzyme induction by using
the same inhalation co-exposure after a 3 day pretreatment with the same 10
chemical mixture. The resulting venous concentration measurements for TCE
from the first study were consistent with metabolic inhibition theory. The 10-
chemical mixture was the most complex co-exposure used in this study. The
authors stated that as mixture complexity increased, the resulting parent
compound concentration time courses changed less, an observation which is
consistent with metabolic inhibition. For the pretreatment study, the authors
found a systematic decrease in venous concentration (due to higher metabolic
clearance) for all chemicals except PERC. Overall, these studies suggest a
complex metabolic interaction between TCE and other solvents.
A PBPK model for TCE including all its metabolites and their interactions can
be considered a mixtures model where all metabolites have a common starting
This document is a draft for review purposes only and does not constitute Agency policy.
E-449 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
point in the liver. An integrated approach taking into account TCE metabolites
and their metabolic inhibition and interactions among each other is suggested in
Chiu et al. (2006b).
E.5. POTENTIALLY SUSCEPTIBLE LIFE STAGES AND CONDITIONS THAT MAY
ALTER RISK OF LIVER TOXICITY AND CANCER
As described in Sections E.1.2, E.3.2.2, E.3.2.6, E.4.2.1, E.4.2.2, E.4.2.3, and E.4.2.4,
there are a number of conditions that are associated with increased risk of liver cancer and
toxicity that include age, use of a number of prescription medications including fibrates and
statins, disease state (e.g., diabetes, NALD, viral infections) and exposure to external
environmental contaminants that have an affect on TCE toxicity and targets. Obviously
epigenetic and genetic factors play a role in determining the risk to the individual. In terms of
liver cancer, there is general consensus that despite the associations that have been made with
etiological factors and the risk of liver cancer, the mechanism is still unknown. The MOA of
TCE toxicity is also unknown but exposure to TCE and its metabolites have shown in rodent
models to induce liver cancer and in a fashion that is not consistent with only a hypothesized
MOA of PPARa receptor activation that is in need of revision. However, multiple TCE
metabolites have been shown to also induce liver cancer with varying effects on the liver that
have also been associated with early stages of neoplasia (glycogen storage) or other actions
associated with risk of hepatocarcinogenicity. The growing epidemic of obesity has been
suggested to increase the risk of liver cancer and may reasonably increase the target population
for TCE effects on the liver.
Lifestyle factors such as ethanol ingestion have not only been shown to increase liver
cancer risk in those who already have fatty liver, but also to increase the toxicity of TCE.
However, as noted by Caldwell et al. (2008b), while there is evidence to suggest that TCE
exposure may increase the risk of liver toxicity and cancer, there are not data to support a
quantitative estimate of how coexposures may modulate that risk.
These findings can also serve to alert the risk manager to the possibility that
multiple internal and external exposures to TCE that may act via differing MOAs
for the production of liver effects. This information suggests a possible lack of
"zero" background exposures and can help identify potential susceptible
populations.
This document is a draft for review purposes only and does not constitute Agency policy.
E-450 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Background levels of haloacetates in drinking water may add to the cumulative
exposure an individual receives via the metabolism of TCE. The brominated
haloacetates apparently share some common effects and pathways with their
chlorinated counterparts. Thus, concurrent exposure of TCE, its metabolites, and
other haloacetates may pose an additive response as well as an additive dose.
However, personal exposures are difficult to ascertain and the effects of such co-
exposures on toxicity are hard to quantify. EPA's guidance on cumulative risk
assessments directs "each office to take into account cumulative risk issues in
scoping and planning major risk assessments and to consider a broader scope that
integrates multiple sources, effects, pathways, stressors, and populations for
cumulative risk analyses in all cases for which relevant data are available" [U.S.
EPA, 1997], Widespread exposure to possible background levels of TCE
metabolites or co-contaminants and other extrinsic factors have the potential to
affect TCE toxicity. However, the available data for co-exposures on TCE
toxicity appears inadequate for quantifying these effects, particularly at
environmental levels of contamination and exposure. Thus, the risk manager and
assessor are going to be limited by not having information regarding either (1)
the type of exposure data necessary to assess the magnitude of co-exposures that
may affect toxicity, or (2) the potential quantitative impacts of these co-
exposures that would enable specific adjustments to risk. Nonetheless, the risk
manager should be aware that qualitatively a case can be made that extrinsic
factors may affect TCE toxicity.
E.6. UNCERTAINTY AND VARIABILITY
Along with general conclusions about the coherence of data that enable conclusions
about effects on the liver shown through experimental studies of TCE, there have also been
extensive discussions throughout this report regarding the specific limitations of experimental
studies whose design was limited by small and varying groups of animals and variability in
control responses as well as reporting deficiencies. Section E.3.2.5 has brought forward the
uncertainty in the MO A for liver cancer in general. The consistency of different animal models
with human HCC is described in Section E.3.3, with Section E.3.2.2 providing a discussion of
the promise and limitations of emerging technologies to study the MO As of liver can in general
and for TCE specifically. Issues regarding the target cell for HCC and the complexities of
studying the MOA for a heterogeneous disease are described in Sections E.3.2.4 and E.3.2.8,
This document is a draft for review purposes only and does not constitute Agency policy.
E-451 DRAFT—DO NOT CITE OR QUOTE
-------
1	respectively. Finally, the uncertainty regarding key events in how activation of the PPARa
2	receptor my lead to hepatocarcinogenesis and the problems with extrapolation of results using
3	the common paradigm to study them (exposure to high levels of WY-14,643 in abbreviated
4	bioassays in knockout mice) are outlined in Section E.3.5.1. As such uncertainties are identified
5	future research can focus on resolving them.
6
1
This document is a draft for review purposes only and does not constitute Agency policy.
E-452 DRAFT—DO NOT CITE OR QUOTE
-------
APPENDIX F
TCE Noncancer Dose-Response Analyses
This document is a draft for review purposes only and does not constitute Agency policy.
F-i DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS—Appendix F: TCE Noncancer Dose-Response Analyses
LIST OF TABLES	F-iv
LIST OF FIGURES	F-v
APPENDIX F: TCE NONCANCER DOSE-RESPONSE ANALYSESError! Bookmark not defined.
F.l. DATA SOURCES	F-l
F.2. DOSIMETRY	F-l
F.2.1. Estimates of Trichl or ethylene (TCE) in Air From Urinary Metabolite
Data Using Ikeda et al. (1972)	F-l
F.2.1.1. Results for Chia et al. (1996)	F-l
F.2.1.2. Results for Mhiri et al. (2004)	F-5
F.2.2. Dose Adjustments to Applied Doses for Intermittent Exposure	F-5
F.2.3. Estimation of the Applied Doses for the Oral Exposure (Feed) Study
of George et al. (1986)	F-6
F.2.4. Physiologically Based Pharmacokinetic (PBPK) Model-Based Internal
Dose Metrics	F-7
F.3. DOSE-RESPONSE MODELING PROCEDURES	F-7
F.3.1. Models for Dichotomous Response Data	F-8
F.3.1.1. Quantal Models	F-8
F.3.1.2. Nested Dichotomous Models	F-8
F.3.2. Models for Continuous Response Data	F-8
F.3.3. Model Selection	F-9
F.3.4. Additional Adjustments for Selected Data Sets	F-9
F.4. DOSE-RESPONSE MODELING RESULTS	F-10
F.4.1. Quantal Dichotomous and Continuous Modeling Results	F-10
F.4.2. Nested Dichotomous Modeling Results	F-l 1
F.4.2.1. Johnson et al. (2003) Fetal Cardiac Defects	F-l 1
F.4.2.2. Narotsky et al. (1995)	1-15
F.4.3. Model Selections and Results	F-23
F.5. DERIVATION OF POINTS OF DEPARTURE	1-30
F.5.1. Applied Dose Points of Departure	F-30
F.5.2. Physiologically Based Pharmacokinetic (PBPK) Model-Based Human
Points of Departure	F-30
F.6. SUMMARY OF POINTS OF DEPARTURE (PODs) FOR STUDIES AND
EFFECTS SUPPORTING THE INHALATION REFERENCE
CONCENTRATION (RfC) AND ORAL REFERENCE DOSE (RfD)	F-31
This document is a draft for review purposes only and does not constitute Agency policy.
F-ii DRAFT—DO NOT CITE OR QUOTE
-------
F.6.1. National Toxicology Program (NTP, 1988)—Benchmark Dose (BMD)
Modeling of Toxic Nephropathy in Rats	F-32
F.6.1.1. Dosimetry and Benchmark Dose (BMD) Modeling	F-32
F.6.1.2. Derivation of HEC99 and HED99	F-32
F.6.2. Woolhiser et al. (2006)—Benchmark Dose (BMD) Modeling of
Increased Kidney Weight in Rats	F-34
F.6.2.1. Dosimetry and Benchmark Dose (BMD) Modeling	F-34
F.6.2.2. Derivation of HEC99 and HED99	F-37
F.6.3. Keil et al. (2009)—Lowest-Observed-Adverse-Effect Level (LOAEL)
for Decreased Thymus Weight in Mice	F-37
F.6.3.1. Dosimetry	F-37
F.6.3.2. Derivation of HEC99 and HED99	F-38
F.6.4. Johnson et al. (2003)—Benchmark Dose (BMD) Modeling of Fetal
Heart Malformations in Rats	F-38
F.6.4.1. Dosimetry and Benchmark Dose (BMD) Modeling	F-39
F.6.4.2. Derivation of HEC99 and HED99	F-40
F.6.5. Peden- Adams et al. (2006)—Lowest-Ob served-Adverse-Effect Level
(LOAEL) for Decreased PFC Response and Increased Delayed-Type
Hypersensitivity in Mice	F-40
This document is a draft for review purposes only and does not constitute Agency policy.
F-iii DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF TABLES
Table F-l. Dose-response data from Chia et al. (1996)	F-l
Table F-2. Data on TCE in air (ppm) and urinary metabolite concentrations in workers reported
by Ikeda et al. (1972)	F-2
Table F-3. Estimated urinary metabolite and TCE air concentrations in dose groups from Chia et
al. (1996)	1-5
Table F-4. Data on fetuses and litters with abnormal hearts from Johnson et al. (2003)	F-12
Table F-5. Comparison of observed and predicted numbers of fetuses with abnormal hearts from
Johnson et al. (2003), with and without the high-dose group, using a nested model	F-12
Table F-6. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, on the basis of applied dose (mg/kg/d in drinking water)...F-
13
Table F-7. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, using the TotOxMetabBW34 dose metric	F-16
Table F-8. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, using the AUCCBld dose metric	F-17
Table F-9. Analysis of LSCs with respect to dose from Narotsky et al. (1995)	F-18
Table F-10. Results of nested log-logistic and Rai-VanRyzin model for fetal eye defects from
Narotsky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water)	F-19
Table F-l 1. Comparison of results of nested log-logistic (without LSC or IC) and quantal log-
logistic model for fetal eye defects from Narotsky et al. (1995)	F-21
Table F-12. Results of nested log-logistic and Rai-VanRyzin model for prenatal loss from
Narotsky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water)	F-23
Table F-13. Model selections and results for noncancer dose-response analyses	F-26
This document is a draft for review purposes only and does not constitute Agency policy.
F-iv DRAFT—DO NOT CITE OR QUOTE
-------
LIST OF FIGURES
Figure F-l. Regression of TCE in air (ppm) and TCA in urine (mg/g creatinine) based on data
from Ikeda et al. (1972)	F-4
Figure F-2. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
applied dose, without LSC, with IC, and without the high-dose group, using a BMR of 0.05 extra
risk (top panel) or 0.01 extra risk (bottom panel)	F-14
Figure F-3. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
TotOxMetabBW34 dose metric, without LSC, with IC, and without the high-dose group, using a
BMR of 0.01 extra risk	F-16
Figure F-4. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
AUCCBld dose metric, without LSC, with IC, and without the high-dose group, using a BMR of
0.01 extra risk	F-17
Figure F-5. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested log-
logistic model, with applied dose, with both LSC and IC, using a BMR of 0.05 extra risk	F-20
Figure F-6. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested log-
logistic model, with applied dose, without either LSC or IC, using a BMR of 0.05 extra risk. F-21
Figure F-7. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested Rai-
VanRyzin model, with applied dose, without either LSC or IC, using a BMR of 0.05 extra risk.F-
22
Figure F-8. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested log-
logistic model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
panel) or 0.01 extra risk (bottom panel)	F-24
Figure F-9. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested Rai-
VanRyzin model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
panel) or 0.01 extra risk (bottom panel)	F-25
Figure F-10. BMD modeling of NTP (1988) toxic nephropathy in female Marshall rats	F-34
Figure F-l 1. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from NTP
(1988) toxic nephropathy in rats	F-34
Figure F-12. BMD modeling of Woolhiser et al. (2006) for increased kidney weight in female S-
D rats	F-3 6
Figure F-13. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from
Woolhiser et al. (2006) for increased kidney weight in rats	F-38
Figure F-14. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from
Keil et al. (2009) for decreased thymus weight in mice	F-39
This document is a draft for review purposes only and does not constitute Agency policy.
F-v DRAFT—DO NOT CITE OR QUOTE
-------
Figure F-15. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from Johnson
et al. (2003) for increased fetal cardiac malformations in female S-D rats using the total oxidative
metabolism dose metric	F-40
This document is a draft for review purposes only and does not constitute Agency policy.
F-vi DRAFT—DO NOT CITE OR QUOTE
-------
1
2
F.l. DATA SOURCES
3	Data sources are cited in the body of this report in the section describing dose-response
4	analyses (see Chapter 5).
5
F.2. DOSIMETRY
6	This section describes some of the more detailed dosimetry calculations and adjustments
7	used in Section 5.1.
8
F.2.1. Estimates of Trichlorethylene (TCE) in Air From Urinary Metabolite Data Using
Ikeda et al. (1972)
F.2.1.1. Results for Chia et al. (1996)
9	Chia et al. (1996) demonstrated a dose-related effect on hyperzoospermia in male
10	workers exposed to trichloroethylene (TCE), lumping subjects into four groups based on range of
11	trichloroacetic acid (TCA) in urine (see Table F-l).
12
13	Table F-l. Dose-response data from Chia et al. (1996)
14
TCA, mg per g creatinine
No. of subjects
No. with hyperzoospermia
0.8 to <25
37
6
50 to <75
18
8
75 to <100
8
4
>100 to 136.4
5
3
15
16	Minimum and maximum TCA levels are reported in the text of Chia et al. (1996), the other data, in their
17	Table 5.
18
19
20	Data from Ikeda et al. (1972) were used to estimate the TCE exposure concentrations
21	corresponding to the urinary TCA levels reported by Chia et al. (1996). Ikeda et al. (1972)
This document is a draft for review purposes only and does not constitute Agency policy.
F-l DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
studied 10 workshops, in each of which TCE vapor concentration was "relatively constant."
They measured atmospheric concentrations of TCE and concentrations in workers' urine of total
trichloro compounds (TTC), TCA, and creatinine, and demonstrated a linear relation between
TTC/creatinine (mg/g) in urine and TCE in the work atmosphere. Their data are tabulated as
geometric means (the last column was calculated by us, as described in Table F-2).
Table F-2. Data on TCE in air (ppm) and urinary metabolite concentrations
in workers reported by Ikeda et al. (1972)
n
TCE
(ppm)
TTC
(mg/L)
TCA
(mg/L)
TTC (mg/g
creatinine)
TCA (mg/g
creatinine)
9
3
39.4
12.7
40.8
13.15127
5
5
45.6
20.2
42.4
18.78246
6
10
60.5
17.6
47.3
13.76
4
25
164.3
77.2
122.9
57.74729
4
40
324.9
90.6
221.2
61.68273
5
45
399
138.4
337.7
117.137
5
50
418.9
146.6
275.8
96.52012
5
60
468
155.4
359
119.2064
4
120
915.3
230.1
518.9
130.4478
4
175
1210.9
235.8
1040.1
202.5399
These data were used to construct the last column "TCA.cr.mg.g" (mg TCA/g creatinine),
as follows: TCA (mg/g creatinine) = TCA (mg/L) x TTC (mg/g creatinine)/TTC (mg/L). The
regression relation between TCE (ppm) and TCA (mg/g creatinine) was evaluated using these
data. Ikeda et al. (1972) reported that the measured values are lognormally distributed and
exhibit heterogeneity of variance, and that the reported data (above) are geometric means. Thus,
the regression relation between loglO(TCA [mg/g creatinine]) and loglO(TCE [ppm]) was used,
assuming constant variances and using number of subjects as weights. Figure F-l shows the
results.
Next, a Berkson setting for linear calibration was assumed, in which one wants to predict
X(TCE, ppm) from means for Y (TCA, mg/g creatinine), with substantial error in Y (Snedcor and
Cochran, 1980). Thus, the inverse prediction for the data of Chia et al. (1996) was used to infer
their mean TCE exposures. The relation based on data from Ikeda et al. (1972) is
This document is a draft for review purposes only and does not constitute Agency policy.
F-2 DRAFT—DO NOT CITE OR QUOTE
-------
1	loglO(TCA, mg/g creatinine) = 0.7098 + 0.7218*logl0(TCE, ppm)	(Eq. F-l)
2
3	and the inverse prediction is
4
This document is a draft for review purposes only and does not constitute Agency policy.
F-3 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
log10(TCA, mg/g.creatinine in urine) = 0.7098 + 0.7218 * log10(TCE, ppm)
|t|)
(Intercept) 0.7098	0.1132	6.2688	0.0002
logl0(TCE.ppm) 0.7218	0.0771	9.3578	0.0000
Residual standard error: 0.3206 on 8 degrees of
freedom
Multiple R-Squared: 0.9163
F-statistic: 87.57 on 1 and 8 degrees of freedom,
the p-value is 0.0000139
Figure F-l. Regression of TCE in air (ppm) and TCA in urine (mg/g
creatinine) based on data from Ikeda et al. (1972).
loglO(TCE) = [loglO(TCA) - 0.7098]/0.7218	(Eq. F-2)
TCE, ppm = 10A( [loglO(TCA) - 0.7098]/0.7218)
Because of the lognormality of data reported by Ikeda et al. (1972), the means of the
logarithms of the ranges for TCA (mg/g creatinine) in Chia et al. (1996), which are estimates of
the median for the group, were used. The results are shown in Table F-3.
This document is a draft for review purposes only and does not constitute Agency policy.
F-4 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Table F-3. Estimated urinary metabolite and TCE air concentrations in dose
groups from Chia et al. (1996)
TCA, mg per g
Creatinine
Estim. TCA
median"
Logl0(TCA
median)
Estim. ppm
TCEb
0.8 to <25
4.47
0.650515
0.827685
50 to <75
61.2
1.787016
31.074370
75 to <100
86.6
1.937531
50.226119
>100 to 136.4
117
2.067407
76.008668
11 10A(mean[logl0(TCA limits in first column)]).
b 10A([logl0(TCA median)] - 0.7098)/0.7218.
Dose-response relations for the data of Chia et al. (1996) were modeled using both the
estimated medians for TCA (mg/g creatinine) in urine and estimated TCE (ppm in air) as doses.
The TCE-TCA-TTC relations are linear up to about 75 ppm TCE (Figure 1 of Ikeda et al.
(1972)), and certainly in the range of the benchmark dose (BMD). As noted below (see
Section F.2.2), the occupational exposure levels are further adjusted to equivalent continuous
exposure for deriving the point of departure (POD).
F.2.1.2. Results for Mhiri et al. (2004)
The lowest-observed-adverse-effect level (LOAEL) group for abnormal trigeminal nerve
somatosensory evoked potential reported in Mhiri et al. (2004) had a urinary TCA concentration
of 32.6 mg TCA/mg creatinine. Using Eq. F-2, above gives an occupational exposure level =
10A([logl0(32.6) - 0.7098J/0.7218) = 12.97404 ppm. As noted below (see Section F.2.2), the
occupational exposure levels are further adjusted to equivalent continuous exposure for deriving
the POD.
F.2.2. Dose Adjustments to Applied Doses for Intermittent Exposure
The nominal applied dose was adjusted for exposure discontinuity (e.g., exposure for
5 days per week and 6 hours per day reduced the dose by the factor [5/7]*[6/24]). The
physiologically based pharmacokinetic (PBPK) dose metrics took into account the daily and
weekly discontinuity to produce an equivalent average dose for continuous exposure. No dose
This document is a draft for review purposes only and does not constitute Agency policy.
F-5 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
adjustments were made for duration of exposure or a less-than-lifetime study, as is typically done
for cancer risk estimates, though in deriving the candidate reference values, an uncertainty factor
for subchronic-to-chronic exposure was applied where appropriate.
For human occupational studies, inhalation exposures (air concentrations) were adjusted
by the number of work (vs. nonwork) days and the amount of air intake during working hours as
3	3
a fraction of the entire day (10 m during work/20 m for entire day). For the TCE ppm in air
converted from urinary metabolite data using Ikeda et al. (1972), the work week was 6 days, so
the adjustment for number of work days is 6/7.
F.2.3. Estimation of the Applied Doses for the Oral Exposure (Feed) Study of George et al.
Female F334 rats were exposed for 19 weeks in their feed. Average body weights (Wt)
are reported (Table A2, p. 53) for time periods having durations (dt) of 1-4 weeks. Proportions
of the 19 weeks of feeding were calculated for each time period as
Average daily feed consumed (Ft) is reported (Table A3,) for the same time periods as body
weight. Concentration (%w/w) of TCE in feed (Table 1, p.31) is reported for weeks 1, 6, 12, and
18.13 Two determinations are reported, which we averaged. The grouping of TCE feed
concentrations into time periods (Table 1) differs from that used for body weight and feed
consumption (Tables A2, A3). This was reconciled by linear interpolation of feed concentrations
to produce concentrations (denoted Ct) for the time periods presented in Tables A2 and A3. We
then calculated mg TCE consumed per kg-day, for each time period, as the product of:
Ct/100	feed concentration, %w/w, divided by 100 to give a fraction
Ft	feed consumed (grams)
1000	1000 (conversion of grams to mg)
1/Wt	l/[ body weight, kg ]
13 "Study Week 1" is repeated in the table, which is a typo for week 6, confirmed positively by the text on pages 19-
20 "Analysis of Task 2 feed formulations at six week intervals ... Similarly, during week 6 of Task 2, the 0.15%,
0.30%, and 0.60% TCE formulations assayed at 27%, 71% and 82% of the theoretical concentration, respectively
(Table 1)".
This document is a draft for review purposes only and does not constitute Agency policy.
(1986)
t
F-6 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
And found the time-weighted average of these for each dose group:
x ((Ct x Ft x 1000)/Wt)}
The results were:
Nominal %w/w concentration in feed	Calculated mg/kg/day
0	0
0.15	72
0.30	186
0.60	389
F.2.4. Physiologically Based Pharmacokinetic (PBPK) Model-Based Internal Dose Metrics
PBPK modeling was used to estimate levels of dose metrics corresponding to different
exposure scenarios in rodents and humans (see Section 3.5). The selection of dose metrics for
specific organs and endpoints is discussed under Section 5.1.
The PBPK model requires an average body weight. For most of the studies, averages
specific to each species, strain, and sex were used. Where these were not reported in the text of
an article, data were obtained by digitizing the body weight graphics (Maltoni et al., 1986) or by
finding the median of weekly averages from graphs (NCI, 1976; NTP, 1988, 1990). Where
necessary, default adult body weights specific to the strain were used (U.S. EPA, 1994b).
F.3. DOSE-RESPONSE MODELING PROCEDURES
Where adequate dose-response data were available, models were fitted with the
BenchMark Dose Software (BMDS) (http://www.epa.gov/ncea/bmds) using the applicable
applied doses or PBPK model-based dose metrics for each combination of study, species, strain,
sex, endpoints, and benchmark response (BMR) under consideration.
This document is a draft for review purposes only and does not constitute Agency policy.
F-7 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
F.3.1. Models for Dichotomous Response Data
F.3.1.1. Quantal Models
For dichotomous responses, the log-logistic, multistage, and Weibull models were fitted.
These models adequately describe the dose-response relationship for the great majority of data
sets, specifically in past TCE studies (Falk Filipsson and Victorin, 2003). If the slope parameter
of the log-logistic model was less than 1, indicating a supralinear dose-response shape, the model
with the slope constrained to 1 was also fitted for comparison. For the multistage model, an
order one less than the number of dose groups was used, in addition to the 2nd-order multistage
model if it differed from the preceding model, and the first-order ('linear') multistage model
(which is identical to a Weibull model with power parameter equal to 1). The Weibull model
with the power parameter unconstrained was also fitted t.
F.3.1.2. Nested Dichotomous Models
In addition, nested dichotomous models were used for developmental effects in rodent
studies to account for possible litter effects, such maternal covariates or intralitter correlation.
The available nested models in BMDS are the nested log-logistic model, the Rai-VanRyzin
models, and the NCTR model. Candidates for litter-specific covariates (LSC) were identified
from the studies and considered legitimate for analysis if they were not significantly dose-related
(determined via regression, analysis of variance). The need for a LSC was indicated by a
difference of at least 3 in the Akaike Information Criteria (AIC) for models with and without a
LSC. The need to estimate intralitter correlations (IC) was determined by presence of a high
correlation coefficient for at least one dose group and by AIC. The fits for nested models were
also compared with the results from quantal models.
F.3.2. Models for Continuous Response Data
For continuous responses, the distinct models available in BMDS were fitted: power
model (power parameter unconstrained and constrained to >1), polynomial model, and Hill
model. Both constant variance and modeled variance models were fit; but constant variance
models were used for model parsimony unless the />value for the test of homogenous variance
was <0.10, in which case the modeled variance models were considered. For the polynomial
model, model order was selected as follows. A model of order 1 was fitted first. The next higher
This document is a draft for review purposes only and does not constitute Agency policy.
F-8 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
order model (up to order n~ 1) was accepted if AIC decreased more than 3 units and the />value
for the mean did not decrease.
F.3.3. Model Selection
After fitting these models to the data sets, the recommendations for model selection set
out in U.S. Environmental Protection Agency (U.S. EPA)'s Benchmark Dose Technical
Guidance Document (Inter-Agency Review Draft, (U.S. EPA, 2000a) were applied. First,
models were generally rejected if the p-w alue for goodness of fit was <0.10. In a few cases in
which none of the models fit the data with p > 0.10, linear models were selected on the basis of
an adequate visual fit overall. Second, models were rejected if they did not appear to adequately
fit the low-dose region of the dose-response relationship, based on an examination of graphical
displays of the data and scaled residuals. If the benchmark dose lower bound (BMDL) estimates
from the remaining models were "sufficiently close" (a criterion of within 2-fold for "sufficiently
close" was used), then the model with the lowest AIC was selected. The AIC is a measure of
information loss from a dose-response model that can be used to compare a set of models.
Among a specified set of models, the model with the lowest AIC is considered the "best." If two
or more models share the lowest AIC, the BMD Technical Guidance Document (U.S. EPA,
2000a) suggests that an average of the BMDLs could be used, but averaging was not used in this
assessment (for the one occasion in which models shared the lowest AIC, a selection was made
based on visual fit). If the BMDL estimates from the remaining models are not sufficiently
close, some model dependence is assumed. With no clear biological or statistical basis to choose
among them, the lowest BMDL was chosen as a reasonable conservative estimate, as suggested
in the Benchmark Dose Technical Guidance Document, unless the lowest BMDL appeared to be
an outlier, in which case further judgments were made.
F.3.4. Additional Adjustments for Selected Data Sets
In a few cases, the dose-response data necessitated further adjustments in order to
improve model fits.
The behavioral/neurological endpoint "number of rears" from Moser et al. (1995)
consisted of counts, measured at five doses and four measurement times (with eight observations
each). The high dose for this endpoint was dropped because the mean was zero, and no
monotone model could fit that well. Analysis of means and standard deviations for these counts
This document is a draft for review purposes only and does not constitute Agency policy.
F-9 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
suggested a Box-Cox power transform (Box et al., 1978) of V2 (i.e., square root) to stabilize
variances (i.e., the slope of the regression of log[standard deviation (SD)] on log[mean] was
0.46, and the relation was linear and highly significant). This information was helpful in
selecting a suitable variance model with high confidence (i.e., variance constant, for square-root
transformed data). Thus, the square root was taken of the original individual count data, and the
mean and variance of the transformed count data were used in the BMD modeling.
The high-dose group was dropped due to supra-linear dose-response shapes in two cases:
fetal cardiac malformations from Johnson et al. (2003) and decreased PFC response from
Woolhiser et al. (2006). Johnson et al. (2003) is discussed in more detail below (see
Section F.4.2.1). For Woolhiser et al. (2006), model fit near the BMD and the lower doses as
well as the model fit to the variance were improved by dropping the highest dose, a procedure
suggested in U.S. EPA (2000a).
In some cases, the supralinear dose-response shape could not be accommodated by these
measures, and a LOAEL or no-observed-adverse-effect level (NOAEL) was used instead. These
include NCI (1976) (toxic nephrosis, >90% response at lowest dose), Keil (2009) (autoimmune
markers and decreased thymus weight, only two dose groups in addition to controls), and Peden-
Adams et al. (2006) (developmental immunotoxicity, only two dose groups in addition to
controls).
F.4. DOSE-RESPONSE MODELING RESULTS
F.4.1. Quantal Dichotomous and Continuous Modeling Results
The documents and show the fitted model curves. The graphics include observations
(group means or proportions), the estimated model curve (solid red line) and estimated BMD,
with a BMDL. Vertical bars show 95% confidence intervals for the observed means. Printed
above each plot are some key statistics (necessarily rounded) for model goodness of fit and
estimated parameters. Printed in the plots in the upper left are the BMD and BMDL for the
rodent data, in the same units as the rodent dose.
More detailed results, including alternative BMRs, alternative dose metrics, quantal
analyses for endpoints for which nested analyses were performed, etc. are documented in the
several spreadsheets. Input data for the analyses are in the following documents: and . The
documents and present the data and model summary statistics, including goodness-of-fit
measures (Chi-square goodness-of-fit />value, AIC), parameter estimates, BMD, and BMDL.
The group numbers "GRP" are arbitrary and are the same as GRP in the plots. Finally, note that
This document is a draft for review purposes only and does not constitute Agency policy.
F-10 DRAFT—DO NOT CITE OR QUOTE
-------
1	not all plots are shown in the documents above, since these spreadsheets include many
2	"alternative" analyses.
3
F.4.2. Nested Dichotomous Modeling Results
F.4.2.1. Johnson et al. (2003) Fetal Cardiac Defects
F.4.2.1.1. Results using applied dose. The biological endpoint was frequency of rat fetuses
having cardiac defects, as shown in Table F-4. Individual animal data were kindly
provided by Dr. Johnson (personal communication from Paula Johnson, University of
Arizona, to Susan Makris, U.S. EPA, 26 August 2009). Cochran-Armitage trend tests using
number of fetuses and number of litters indicated significant increases in response with
dose (with or without including the highest dose).
4	One suitable candidate for a LSC was available: female weight gain during pregnancy.
5	Based on goodness of fit, this covariate did not contribute to better fit and was not used. Some
6	ICs were significant and these parameters were included in the model.
7
This document is a draft for review purposes only and does not constitute Agency policy.
F-11 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table F-4. Data on fetuses and litters with abnormal hearts from Johnson et
2	al. (2003)
3
Dose group
(mg/kg/d):
0
0.00045
0.048
0.218
129
Fetuses
Number of pups:
606
144
110
181
105
Abnormal heart:
13
0
5
9
11
Litters
Number of litters:
55
12
9
13
9
Abnormal heart:
9
0
4
5
6
4
5
6	With the high dose included, the chi-square goodness of fit was acceptable, but some
7	residuals were large (1.5 to 2) for the control and two lower doses. Therefore, models were also
8	fitted after dropping the highest dose. For these, goodness of fit was adequate, and scaled
9	residuals were smaller for the low doses and control. Predicted expected response values were
10	closer to observed when the high dose was dropped, as shown in Table F-5:
11
12	Table F-5. Comparison of observed and predicted numbers of fetuses with
13	abnormal hearts from Johnson et al. (2003), with and without the high-dose
14	group, using a nested model
15
Dose group (mg/kg/d):
Abnormal hearts (pups)
0
0.00045
0.048
0.218
129
Observed:
13
0
5
9
11
Predicted expected:
With high dose
19.3
4.5
3.5
5.7
11
Without high dose
13.9
3.3
3.4
10
—
16
17
18	Accuracy in the low-dose range is especially important because the BMD is based upon
19	the predicted responses at the control and the lower doses. Based on the foregoing measures of
20	goodness of fit, the model based on dropping the high dose was used.
This document is a draft for review purposes only and does not constitute Agency policy.
F-12 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
The nested log-logistic and Rai-VanRyzin models were fitted; these gave essentially the
same predicted responses and POD. The former model was used as the basis for a POD; results
are in Table F-6 and Figure F-2.
Table F-6. Results of nested log-logistic model for fetal cardiac anomalies
from Johnson et al. (2003) without the high-dose group, on the basis of
applied dose (mg/kg/d in drinking water)
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
246.877
NA (df = 0)
0.01
0.252433
0.03776
NLOG
Y
N
251.203
0.0112
0.01
0.238776
0.039285
NLOG
N
N
248.853
0.0098
0.01
0.057807
0.028977
NLOG
N
Y
243.815
0.0128
0.1
0.71114
0.227675
NLOG
N
Y
243.815
0.0128
0.05
0.336856
0.107846
NLOG*
N
Y
243.815
0.0128
0.01
0.064649
0.020698
* Indicates model selected (Rai-VanRyzin model fits are essentially the same).
NLOG = "nested log-logistic" model.
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
BMDL
0.15 0.2
dose
0.25 0.3
13:36 08/27 2008
LSC analyzed was female weight gain during pregnancy.
This document is a draft for review purposes only and does not constitute Agency policy.
F-13 DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Nested Logistic Model with 0.95 Confidence Level
0.12
0.1
0.08
0.06
0.04
0.02
0
13:37
Figure F-2. BMD modeling of Johnson et al. (2003) using nested log-logistic
model, with applied dose, without LSC, with IC, and without the high-dose
group, using a BMR of 0.05 extra risk (top panel) or 0.01 extra risk (bottom
panel).
F.4.2.1.2. Chi-square Goodness of Fit Test for nested log-logistic. The BMDS choice of
subgroups did not seem appropriate given the data. The high-dose group of 13 litters was
subdivided into three subgroups having sums of expected counts 3, 3, and 2. However, the
control group of 55 litters could have been subdivided because expected response rates for
controls were relatively high. There was also concern that the goodness of fit might change with
alternative choices of subgroupings.
An R program was written to read the BMDS output, reading parameters and the table of
litter-specific results (dose, covariate, estimated probability of response, litter size, expected
response count, observed response count, scaled chi-square residual). The control group of
55 litters was subdivided into three subgroups of 18, 18, and 19 litters. Control litters were
sampled randomly without replacement 100 times, each time creating 3 subgroups—i.e.,
100 random assignments of the 55 control litters to three subgroups were made. For each of
these, the goodness-of-fit calculation was made and the /;-value saved. Within these
100 /^-values, >75% were >0.05, and >50% had /-values >0.11, this indicated that the model is
acceptable based on goodness-of-fit criteria.
This document is a draft for review purposes only and does not constitute Agency policy.
F-14 DRAFT—DO NOT CITE OR QUOTE
Nested Logistic
BMDL
0
BMD
0.05
0.1
0.15
0.2
dose
08/27 2008
-------
F.4.2.1.3. Results using physiologically based pharmacokinetic (PBPK) model-based dose
metrics. The nested log-logistic model was also run using the dose metrics in the dams of
total oxidative metabolism scaled by body weight to the %-power (TotOxMetabBW34) and
the area-under-the-curve of TCE in blood (AUCCBld). As with the applied dose modeling,
LSC (maternal weight gain) was not included, but IC was included, based on the criteria
outlined previously (see Section F.3.1.2). The results are summarized in Table F-7 and
Figure F-3 for TotOxMetabBW34 and Table F-8 and Figure F-4 for AUCCBld.
1
F.4.2.2. Narotsky et al. (1995)
2	Data were combined for the high doses in the single-agent experiment and the lower
3	doses in the 'five-cube' experiment. Individual animal data were kindly provided by Dr.
4	Narotsky (personal communications from Michael Narotsky, U.S. EPA, to John Fox, U.S. EPA,
5	19 June 2008, and to Jennifer Jinot, U.S. EPA, 10 June 2008). Two endpoints were examined:
6	frequency of eye defects in rat pups and prenatal loss (number of implantation sites minus
7	number of live pups on postnatal day 1).
8
9
This document is a draft for review purposes only and does not constitute Agency policy.
F-15 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table F-7. Results of nested log-logistic model for fetal cardiac anomalies
2	from Johnson et al. (2003) without the high-dose group, using the
3	TotOxMetabBW34 dose metric
4
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
246.877
NA (df = 0)
0.01
0.174253
0.0259884
NLOG
Y
N
251.203
0.0112
0.01
0.164902
0.0270378
NLOG
N
Y
243.815
0.0128
0.1
0.489442
0.156698
NLOG*
N
Y
243.815
0.0128
0.01
0.0444948
0.0142453
NLOG
N
N
248.853
0.0098
0.01
0.0397876
0.0199438
5
6	* Indicates model selected. BMDS failed with the Rai-VanRyzin and NCTR models.
7
8	NLOG = "nested log-logistic" model.
9	LSC analyzed was female weight gain during pregnancy.
10
11
Nested Logistic Model with 0.95 Confidence Level
°-|4l£sted Logistic
0.08
0.06
0.04
0.02
BMDL
0 0.02
BMP
0.04 0.06 0.08 0.1 0.12 0.14
dose
12:44 02/06 2009
12
13	Figure F-3. BMD modeling of Johnson et al. (2003) using nested log-logistic
14	model, with TotOxMetabBW34 dose metric, without LSC, with IC, and
15	without the high-dose group, using a BMR of 0.01 extra risk.
16
This document is a draft for review purposes only and does not constitute Agency policy.
F-16 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table F-8. Results of nested log-logistic model for fetal cardiac anomalies
2	from Johnson et al. (2003) without the high-dose group, using the AUCCBld
3	dose metric
4
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
246.877
NA (df = 0)
0.01
0.00793783
0.00118286
NLOG
Y
N
251.203
0.0112
0.01
0.00750874
0.00123047
NLOG*
N
Y
243.816
0.0128
0.1
0.0222789
0.00712997
NLOG*
N
Y
243.816
0.0128
0.01
0.00202535
0.000648179
NLOG
N
N
248.853
0.0098
0.01
0.00181058
0.000907513
5
6	* Indicates model selected. BMDS failed with the Rai-VanRyzin and NCTR models.
7
8	NLOG = "nested log-logistic" model.
9	LSC analyzed was female weight gain during pregnancy.
10
11
Nested Logistic Model with 0.95 Confidence Level
01Rlested Logistic
0.08
0.06
0.04
0.02
BMDL
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007
BMP
002 0
dose
12:42 02/06 2009
12
13	Figure F-4. BMD modeling of Johnson et al. (2003) using nested log-logistic
14	model, with AUCCBld dose metric, without LSC, with IC, and without the
15	high-dose group, using a BMR of 0.01 extra risk.
16
This document is a draft for review purposes only and does not constitute Agency policy.
F-17 DRAFT—DO NOT CITE OR QUOTE
-------
1	Two LSCs were considered, with analyses summarized in Table F-9. The number of implants is
2	unrelated to dose, as inferred from regression and analysis of variance, and was considered as a
3	LSC for eye defects. As number of implants is part of the definition for the endpoint of prenatal
4	loss, it is not considered as a LSC for prenatal loss. A second LSC, the dam body weight on
5	gestation day (GD) 6 (damBW6) was significantly related to dose and is unsuitable as a litter-
6	specific covariate.
7
8	Table F-9. Analysis of LSCs with respect to dose from Narotsky et al. (1995)
9
Relation of litter-specific covariates to dose
Implants:
none


damBW6:
significant




Mean
Mean

TCE
Implants
damBW6

0
9.5
176.0

10.1
10.1
180.9

32
9.1
174.9

101
7.8
170.1

320
10.4
174.5

475
9.7
182.4

633
9.6
185.3

844
8.9
182.9

1,125
9.6
184.2
Using expt as covariate, e.g., damBW6 ~ TCE.mg.kgd + expt
Linear regression
p = 0.7486
p = 0.0069
AoV (ordered factor)
p = 0.1782
p = 0.0927
10
11
12	Two LSCs were considered, with analyses summarized in Table F-9. The number of
13	implants is unrelated to dose, as inferred from regression and analysis of variance, and was
14	considered as a LSC for eye defects. As number of implants is part of the definition for the
15	endpoint of prenatal loss, it is not considered as a LSC for prenatal loss. A second LSC, the dam
16	body weight on GD 6 (damBW6) was significantly related to dose and is unsuitable as a litter-
17	specific covariate.
This document is a draft for review purposes only and does not constitute Agency policy.
F-18 DRAFT—DO NOT CITE OR QUOTE
-------
F.4.2.2.1. Fetal eye defects. The nested log-logistic and Rai-VanRyzin models were fitted to
the number of pups with eye defects reported by Narotsky et al. (1995), with the results
summarized in Table F-10.
1
2	Table F-10. Results of nested log-logistic and Rai-VanRyzin model for fetal
3	eye defects from Narotsky et al. (1995), on the basis of applied dose (mg/kg/d
4	in drinking water)
5
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
255.771
0.3489
0.05
875.347
737.328a
NLOG
Y
N
259.024
0.0445
0.05
830.511
661.629
NLOG
N
Y
270.407
0.2281
0.05
622.342
206.460
NLOG
N
N
262.784
0.0529
0.10
691.93
542.101
NLOG
N
N
262.784
0.0529
0.05
427.389
264.386
NLOG
N
N
262.784
0.0529
0.01
147.41
38.7117b
RAI
Y
Y
274.339
0.1047
0.05
619.849
309.925
RAI
Y
N
264.899
0.0577
0.05
404.788
354.961
RAI
N
Y
270.339
0.2309
0.05
619.882
309.941
RAI
N
N
262.481
0.0619
0.10
693.04
346.52
RAI
N
N
262.481
0.0619
0.05
429.686
214.843
RAI
N
N
262.481
0.0619
0.01
145.563
130.938b
6
7	" Graphical fit at the origin exceeds observed control and low dose responses and slope is quite flat (see Figure F-5),
8	fitted curve does not represent the data well.
9	b Indicates model selected.
10
11	NLOG = "nested log-logistic" model; RAI = Rai-VanRyzin model.
12	LSC analyzed was implants.
13
14
15	Results for the nested log-logistic model suggested a better model fit with the inclusion of
16	the LSC and IC, based on AIC. However, the graphical fit (see Figure F-5) is strongly sublinear
17	and high at the origin where the fitted response exceeds the observed low-dose responses for the
18	control group and two low-dose groups. An alternative nested log-logistic model without either
19	LSC or IC (see Figure F-6), which fits the low-dose responses better, was selected. Given that
20	this model had no LSC and no IC, the nested log-logistic model reduces to a quantal log-logistic
21	model. Parameter estimates and the ^-values were essentially the same for the two models (see
22	Table F-l 1). A similar model selection can be justified for the Rai-Van Ryzin model (see
This document is a draft for review purposes only and does not constitute Agency policy.
F-19	DRAFT—DO NOT CITE OR QUOTE
-------
1	Figure F-7). Because no LSC and no IC were needed, this endpoint was modeled with quantal
2	models, using totals of implants and losses for each dose group, which allowed choice from a
3	wider range of models (those results appear with quantal model results in this appendix).
4
5
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
0.5
0.4
0.3
0.2
0.1
0
BMDL
BMD
0
200
400
600
800
1000
dose
17:27 08/04 2008
6	Figure F-5. BMD modeling of fetal eye defects from Narotsky et al. (1995)
7	using nested log-logistic model, with applied dose, with both LSC and IC,
8	using a BMR of 0.05 extra risk.
9
This document is a draft for review purposes only and does not constitute Agency policy.
F-20 DRAFT—DO NOT CITE OR QUOTE
-------
Nested Logistic Model with 0.95 Confidence Level
0.5
0.4
0.3
0.2
0.1
0
0	200 400 600 800 1000
dose
17:28 08/04 2008
1	Figure F-6. BMD modeling of fetal eye defects from Narotsky et al. (1995)
2	using nested log-logistic model, with applied dose, without either LSC or IC,
3	using a BMR of 0.05 extra risk.
4
5
6	Table F-ll. Comparison of results of nested log-logistic (without LSC or IC)
7	and quantal log-logistic model for fetal eye defects from Narotsky et al.
8	(1995)
9
Model
Parameter
BMD05
BMDLos
Alpha
Beta
Rho
Nested
0.00550062
-12.3392
1.55088
427.4
264.4
Quantal
0.00549976
-12.3386
1.55079
427.4
260.2
10
11
12
Nested Logistic
BMDL
This document is a draft for review purposes only and does not constitute Agency policy.
F-21 DRAFT—DO NOT CITE OR QUOTE
-------
RaiVR Model with 0.95 Confidence Level
0.5
0.4
0.3
0.2
0.1
0
0	200 400 600 800 1000
dose
17:25 08/04 2008
1
2	Figure F-7. BMD modeling of fetal eye defects from Narotsky et al. (1995)
3	using nested Rai-VanRyzin model, with applied dose, without either LSC or
4	IC, using a BMR of 0.05 extra risk.
5
6
F.4.2.2.2. Narotsky et al. (1995) prenatal loss. The nested log-logistic and Rai-VanRyzin
models were fitted to prenatal loss reported by Narotsky et al. (1995), with the results
summarized in Table F-12.
7	The BMDS nested models require a LSC, so dam body weight on GD6 ("damBW6") was
8	used as the LSC. However, damBW6 is significantly related to dose and, so, is not a reliable
9	LSC. Number of implants could not be used as a LSC because it was identified as number at risk
10	in the BMDS models. These issues were obviated because the model selected did not employ
11	the LSC.
12
RaiVR
BMDL
BMD
This document is a draft for review purposes only and does not constitute Agency policy.
F-22 DRAFT—DO NOT CITE OR QUOTE
-------
1	Table F-12. Results of nested log-logistic and Rai-VanRyzin model for
2	prenatal loss from Narotsky et al. (1995), on the basis of applied dose
3	(mg/kg/d in drinking water)
4
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
494.489
0.2314
0.10
799.723
539.094
NLOG
Y
N
627.341
0.0000
0.10
790.96
694.673
NLOG
N
N
628.158
0.0000
0.10
812.92
725.928
NLOG
N
Y
490.766
0.2509
0.10
814.781
572.057
NLOG
N
Y
490.766
0.2509
0.05
738.749
447.077
NLOG
N
Y
490.766
0.2509
0.01
594.995
252.437 *
RAI
Y
Y
491.859
0.3044
0.10
802.871
669.059
RAI
Y
N
626.776
0.0000
0.10
819.972
683.31
RAI
N
N
626.456
0.0000
0.10
814.98
424.469
RAI
N
Y
488.856
0.2983
0.10
814.048
678.373
RAI
N
Y
488.856
0.2983
0.05
726.882
605.735
RAI
N
Y
488.856
0.2983
0.01
562.455
468.713 *
5
6	* Indicates model selected.
7
8	NLOG = "nested log-logistic" model; RAI = Rai-VanRyzin model.
9	LSC analyzed was dam body weight on GD6.
10
11
12	For the nested log-logistic models, the AIC is much larger when the IC is dropped, so the
13	IC is needed in the model. The LSC can be dropped (and is also suspect because it is correlated
14	with dose). The model with IC and without LSC was selected on the basis of AIC (shown in
15	Figure F-8). For the Rai-VanRyzin models, the model selection was similar to that for the nested
16	log-logistic, leading to a model with IC and without LSC, which had the lowest AIC (shown in
17	Figure F-9).
18
F.4.3. Model Selections and Results
19	The final model selections and results for noncancer dose-response modeling are
20	presented in Table F-l3.
21
This document is a draft for review purposes only and does not constitute Agency policy.
F-23 DRAFT—DO NOT CITE OR QUOTE
-------
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
BMDL
0 200
16:44 08/20 2008
400
800
1000
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
0
BMDL
200
400
16:45 08/20 2008
Figure F-8. BMD modeling of prenatal loss reported in Narotsky et al.
(1995) using nested log-logistic model, with applied dose, without LSC, with
IC, using a BMR of 0.05 extra risk (top panel) or 0.01 extra risk (bottom
panel).
This document is a draft for review purposes only and does not constitute Agency policy.
F-24 DRAFT—DO NOT CITE OR QUOTE
-------
RaiVR Model with 0.95 Confidence Level
0 9 feaiVR
0.8 f-
0.7
0.6
0.5
0.4
0.3
0.2
BMDL
600
BMP
800
0
200
400
1000
dose
16:46 08/20 2008
RaiVR Model with 0.95 Confidence Level
0.9
RaiVR
0.8
0.7
0.6
0.5
0.4
0.3
0.2
BMDL
400
BMD
0
200
600
800
1000
dose
16:46 08/20 2008
2
3	Figure F-9. BMD modeling of prenatal loss reported in Narotsky et al.
4	(1995) using nested Rai-VanRyzin model, with applied dose, without LSC,
5	with IC, using a BMR of 0.05 extra risk (top panel) or 0.01 extra risk (bottom
6	panel).
This document is a draft for review purposes only and does not constitute Agency policy.
F-25 DRAFT—DO NOT CITE OR QUOTE
-------
Table F-13. Model selections and results for noncancer dose-response analyses
GRP
Study/run
abbrev.
Specie
s
Sex
Strain
Exp.
route
End point
Dose metric
BMR
type
BMR
BMD/
BMDL
BMDL
Model
Rep.
BMD
Notes
Dichotomous models
3
Chia et al.
(1996)
human
M
workers.elec.factory
inhal
N.hyperzoospermia
appl.dose
extra
0.1
2.14
1.43
loglogistic.1
3.06

7
Narotsky et al.
(1995)
rat

F344
oral.gav
N.pups.eye.defects
appl.dose
extra
0.01
1.46
60.1
multistage
806
a
13
Narotsky et al.
(1995).sa
rat

F344
oral.gav
N.dams.w.resorbed. litters
appl.dose
extra
0.01
5.47
32.2
multistage.2
570

13
Narotsky et al.
(1995).sa
rat

F344
oral.gav
N.dams.w.resorbed. litters
AUCCBId
extra
0.01
5.77
17.5
multistage.2
327

13
Narotsky et al.
(1995).sa
rat

F344
oral.gav
N.dams.w.resorbed. litters
TotMetabBW34
extra
0.01
1.77
77.5
weibull
156

14
Johnson et al.
(2003).drophi
rat

Sprague.Dawley
oral.dw
N.litters.abnormal.hearts
appl.dose
extra
0.1
2.78
0.0146
loglogistic.1
0.0406
b
36
Griffin et al.
(2000)
mice

MRL++
oral.dw
portal.infiltration
appl.dose
extra
0.1
2.67
13.4
loglogistic.1
35.8

38
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
appl.dose
extra
0.1
1.22
40.2
multistage
49.2
c
38
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
ABioactDCVCBW34
extra
0.1
1.18
0.0888
loglogistic
0.105

38
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
AMetGSHBW34
extra
0.1
1.19
0.086
loglogistic
0.102

38
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
TotMetabBW34
extra
0.1
1.13
53.8
weibull
61
d
39
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
oral.gav
megalonucleocytosis
appl.dose
extra
0.1
1.53
33.8
multistage.2
51.8
e
39
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
oral.gav
megalonucleocytosis
ABioactDCVCBW34
extra
0.1
1.60
0.0594
multistage.2
0.0948

39
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
oral.gav
megalonucleocytosis
AMetGSHBW34
extra
0.1
1.65
0.0605
multistage.2
0.0977

39
Maltoni et al.
(1986)
rat
M
Sprague.Dawley
oral.gav
megalonucleocytosis
TotMetabBW34
extra
0.1
1.41
20.5
multistage.2
29
e
49
NTP (1988)
rat
F
Marshall
oral.gav
toxic nephropathy
appl.dose
extra
0.05
1.45
9.45
loglogistic.1
28.9

49
NTP (1988)
rat
F
Marshall
oral.gav
toxic nephropathy
ABioactDCVCBW34
extra
0.05
1.45
0.0132
loglogistic.1
0.0404

49
NTP (1988)
rat
F
Marshall
oral.gav
toxic nephropathy
AMetGSHBW34
extra
0.05
1.46
0.0129
loglogistic.1
0.0397

49
NTP (1988)
rat
F
Marshall
oral.gav
toxic nephropathy
TotMetabBW34
extra
0.05
1.45
2.13
loglogistic.1
6.5

to
On
-------
^3
"3	i?
o	^
&•
£	8
2	s
* v	§
s »	^
s s	^
5 a,^'
a §-	a
S, ^
"*	^
5	>
«s	^
^	•<
^	^3.
S?	1
S?"
Table F-13. Model selections and results for noncancer dose-response analyses (continued)
GRP
Study/run
abbrev.
Species
Sex
Strain
Exp.
route
End point
Dose metric
BMR
type
BMR
BMD1
BMDL
BMDL
Model
Rep.
BMD
Notes
Nested dichotomous models
NA
Johnson
et al.
(2003).drop
hi
rat
F
Sprague.Dawley
oral.dw
N.pups.abnormal.hearts
appl.dose
extra
0.01
3.12
0.0207
loglogistic.lC
0.711
b














NA
Johnson
et al.
(2003).drop
hi
rat
F
Sprague.Dawley
oral.dw
N.pups.abnormal.hearts
TotOxMetab BW34
extra
0.01
3.12
0.0142
loglogistic.lC

b














NA
Johnson
et al.
(2003).drop
hi
rat
F
Sprague.Dawley
oral.dw
N.pups.abnormal.hearts
AUCCBId
extra
0.01
3.12
0.000648
loglogistic.lC

b














NA
Narotsky
et al. (1995)
rat
F
F344
oral.gav
N.prenatal.loss
appl.dose
extra
0.01
1.2
469
RAI.IC
814

Continuous models
2
Land et al.
(1981)
mouse
M
(C57B1xC3H)F1
inhal
pet.abnormal.sperm
appl.dose
standard
0.5
1.33
46.9
polynomial.constvar
125

6
Carney
et al. (2006)
rat
F
Sprague-Dawley
(Crl:CD)
inhal
gm.wgt.gain.GD6.9
appl.dose
relative
0.1
2.5
10.5
hill
62.3

8
Narotsky
et al. (1995)
rat
F
F344
oral.gav
gm.wgt.gain.GD6.20
appl.dose
relative
0.1
1.11
108
polynomial.constvar
312

19
Crofton and
Zhao
(1997)
rat
M
Long-Evans
inhal
dB.auditory .threshold^ 6kHz)
appl.dose
absolute
10
1.11
274
polynomial.constvar
330

21
George
et al. (1986)
rat
F
F344
oral .food
litters
appl.dose
standard
0.5
1.69
179
polynomial.constvar
604

23
George
et al. (1986)
rat
F
F344
oral .food
live.pups
appl.dose
standard
0.5
1.55
152
polynomial.constvar
470

26
George
et al. (1986)
rat
F
F344
oral .food
Foffspring. BWgm ,day21
appl.dose
relative
0.05
1.41
79.7
polynomial.constvar
225

34sq
Moser et al.
(1995)+pers
com
rat
F
F344
oral.gav
no. rears
appl.dose
standard
1
1.64
248
polynomial.constvar
406
b,f
49
George
et al. (1986)
rat
F
F344
oral .food
traverse.time.21 do
appl.dose
relative
1
1.98
72.6
power
84.9

51
Buben and
O'Flaherty
(1985)
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
appl.dose
relative
0.1
1.26
81.5
hill.constvar
92.8

r1
to
^1
-------
^3
"3	i?
o	^
&•
£	8
2	s
* v	§
s »	^
s s	^
5 a,^'
a §-	a
S, ^
"*	^
5	>
«s	^
^	•<
^	^3.
S?	1
S?"
Table F-13. Model selections and results for noncancer dose-response analyses (continued)
GRP
Study/run
abbrev.
Species
Sex
Strain
Exp.
route
End point
Dose metric
BMR
type
BMR
BMD1
BMDL
BMDL
Model
Rep.
BMD
Notes
51
Buben and
O'Flaherty
(1985)
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
AMetLivl BW34
relative
0.1
1.08
28.6
polynomial.constvar
28.4

51
Buben and
O'Flaherty
(1985)
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
TotOxMetab BW34
relative
0.1
1.08
37
polynomial.constvar
36.7

58
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Liverwt.pctBW
appl.dose
relative
0.1
1.36
21.6
hill
30.4















58
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Liverwt.pctBW
AMetLivl BW34
relative
0.1
1.4
22.7
hill
32.9















58
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Liverwt.pctBW
TotOxMetab BW34
relative
0.1
1.3
73.4
hill
97.7















60. Rp
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Kidneywt.pctBW
appl.dose
relative
0.1
1.17
34.7
polynomial
47.1















60. Rp
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Kidneywt.pctBW
AMetGSHBW34
relative
0.1
1.18
0.17
polynomial
0.236















60. Rp
Kjellstrand
et al.
(1983b)
mouse
M
NMRI
inhal
Kidneywt.pctBW
TotMetabBW34
relative
0.1
1.17
71
polynomial
95.2















63
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
Antibody.Forming Cells
appl.dose
standard
1
1.94
31.2
power.constvar
60.6
b
62
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
Antibody.Forming Cells
AUCCBId
standard
1
1.44
149
polynomial
214

62
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
Antibody.Forming Cells
TotMetabBW34
standard
1
1.5
40.8
polynomial
61.3

65
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
kidney, wt.perl OOgm
appl.dose
relative
0.1
4.29
15.7
hill.constvar
54.3

65
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
kidney.wt.perl OOgm
ABioactDCVCBW34
relative
0.1
4.27
0.0309
hill.constvar
0.103

65
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
kidney.wt.perl OOgm
AMetGSHBW34
relative
0.1
4.28
0.032
hill.constvar
0.107

65
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
kidney.wt.perl OOgm
TotMetabBW34
relative
0.1
1.47
40.8
polynomial.constvar
52.3

r1
to
00
-------
s
s
TO
Co
Table F-13. Model selections and results for noncancer dose-response analyses (continued)
S"4
§•
to
GRP
Study/run
abbrev.
Species
Sex
Strain
Exp.
route
End point
Dose metric
BMR
type
BMR
BMD1
BMDL
BMDL
Model
Rep.
BMD
Notes
67
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
liver.wt.perlOOgm
appl.dose
relative
0.1
4.13
25.2
hill.constvar
70.3

67
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
liver.wt.perlOOgm
AMetLivl BW34
relative
0.1
1.53
46
polynomial.constvar
56.1

67
Woolhiser
et al. (2006)
rat
F
CD (Sprague-
Dawley)
inhal
liver.wt.perlOOgm
TotOxMetab BW34
relative
0.1
1.53
48.9
polynomial.constvar
59.8

o
* v §
s » ^
a, Co'
TO Sj-
§ ^
o
>3
*
Table F-13. Model selections and results for noncancer dose-response analyses (continued)
aEight-stage multistage model.
'Dropped highest dose.
'Three-stage multistage model.
''Weibull selected over log-logistic with the same AIC on basis of visual fit (less extreme curvature).
'Second-order MS selected on basis of visual fit (less extreme curvature).
fSquare-root transformation of original individual count data.
Applied dose BMDLs are in units of ppm in air for inhalation exposures and mg/kg/d for oral exposures. Internal dose BMDLs are in dose metric units. Reporting BMD is BMD using a BMR of 0.1
extra risk for dichotomous models, and 1 control SD for continuous models.
Log-logistic = unconstrained log-logistic; log-logistic. 1 = constrained log-logistic; multistage = multistage with #stages=dose groups-1; multistage.n = n-stage multistage; log-logistic.IC = nested log-
logistic with IC, without LSC; RAI.IC = Rai-VanRyzin model with IC, without LSC; zzz.constvar = continuous model zzz with constant variance (otherwise variance is modeled).
Rep. = reporting, Exp. = exposure, Abbrev. = abbreviation.
1
r1
to
VO
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
F.5. DERIVATION OF POINTS OF DEPARTURE
F. 5.1. Applied Dose Points of Departure
For oral studies in rodents, the POD on the basis of applied dose in mg/kg/d was taken to
be the BMDL, NOAEL, or LOAEL. NOAELs and LOAELs were adjusted for intermittent
exposure to their equivalent continuous average daily exposure (for BMDLs, the adjustments
were already performed prior to BMD modeling).
For inhalation studies in rodents, the POD on the basis of applied dose in ppm was taken
to be the BMDL, NOAEL, or LOAEL. NOAELs and LOAELs were adjusted for intermittent
exposure to their equivalent continuous average daily exposure (for BMDLs, the adjustments
were already performed prior to BMD modeling). These adjusted concentrations are considered
human equivalent concentrations, in accordance with U.S. EPA (1994a), as TCE is considered a
Category 3 gas (systemically acting) and has a blood-air partition coefficient in rodents greater
than that in humans (see Section 3.1).
F.5.2. Physiologically Based Pharmacokinetic (PBPK) Model-Based Human Points of
Departure
As discussed in Section 5.1.3, the PBPK model was used for simultaneous interspecies
(for endpoints in rodent studies), intraspecies, and route-to-route extrapolation based on the
estimates from the PBPK model of the internal dose points of departure (idPOD) for each
candidate critical study/endpoints. The following documents contain figures showing the
derivation of the human equivalent doses and concentrations (human equivalent doses [HEDs]
and human equivalent concentrations [HECs]) for the median (50th percentile) and sensitive (99th
percentile) individual from the (rodent or human) study idPOD. In each case, for a specific
study/endpoint(s)/sex/species (in the figure main title), and for a particular dose metric (Y-axis
label), the horizontal line shows the original study idPOD (a BMDL, NOAEL, or LOAEL as
noted) and where it intersects with the human 99th percentile (open square) or median (closed
square) exposure-internal-dose relationship:
This document is a draft for review purposes only and does not constitute Agency policy.
10120109	F-30
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
The original study internal doses are based on the median estimates from about 2,000
"study groups" (for rodent studies) or "individuals" (for human studies), and corresponding
exposures for the human median and 99th percentiles were derived from a distribution of 2,000
"individuals." In both cases, the distributions reflect combined uncertainty (in the population
means and variances) and population variability.
In addition, as part of the uncertainty/variability analysis described in Section 5.1.4.2, the
POD for studies/endpoints for which BMD modeling was done was replaced by the LOAEL or
NOAEL. This was done to because there was no available tested software for performing BMD
modeling in such a context and because of limitations in time and resources to develop such
software. However, the relative degree of uncertainty/variability should be adequately captured
in the use of the LOAEL or NOAEL. The graphical depiction of the HEC99 or HED99 using
these alternative PODs is shown in the following files:
F.6. SUMMARY OF POINTS OF DEPARTURE (PODs) FOR STUDIES AND EFFECTS
SUPPORTING THE INHALATION REFERENCE CONCENTRATION (RfC) AND
ORAL REFERENCE DOSE (RfD)
This section summarizes the selection and/or derivation of PODs from the critical and
supporting) studies and effects that support the inhalation reference concentration (RfC) and oral
reference dose (RfD). In particular, for each endpoint, the following are described: dosimetry
(adjustments of continuous exposure, PBPK dose metrics), selection of BMR and BMD model
(if BMD modeling was performed), and derivation of the human equivalent concentration or
dose for a sensitive individual (if PBPK modeling was used). Section 5.1.3.1 discusses the dose
metric selection for different endpoints.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-31
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
F.6.1. National Toxicology Program (NTP, 1988)—Benchmark Dose (BMD) Modeling of
Toxic Nephropathy in Rats
The supporting endpoint here is toxic nephropathy in female Marshall rats (NTP, 1988),
which was the most sensitive sex/strain in this study, although the differences among different
sex/strain combinations was not large (BMDLs differed by <3-fold).
F.6.1.1. Dosimetry and Benchmark Dose (BMD) Modeling
Rats were exposed to 500 or 1,000 day, 5 days/week, for 104 weeks. The primary dose
metric was selected to be average amount of dichlorovinyl cysteine (DCVC)
bioactivated/kgyYday, with median estimates from the PBPK model for the female Marshall rats
in this study of 0.47 and 1.1.
Figure F-10 shows BMD modeling for the dichotomous models used (see Section F.5.1,
above). The log-logistic model with slope constrained to >1 was selected because (1) the log-
logistic model with unconstrained slope yielded a slope estimate <1 and (2) it had the lowest
AIC.
The idPOD of 0.0132 mg DCVC bioactivated/kgyYday was a BMDL for a BMR of 5%
extra risk. This BMR was selected because toxic nephropathy is a clear toxic effect. This BMR
required substantial extrapolation below the observed responses (about 60%); however, the
response level seemed warranted for this type of effect and the ratio of the BMD to the BMDL
was not large (1.56 for the selected model).
F.6.1.2. Derivation of HEC99 and HED99
The HEC99 and HED99 are the lower 99th percentiles for the continuous human exposure
concentration and continuous human ingestion dose that lead to a human internal dose equal to
the rodent idPOD. The derivation of the HEC99 of 0.0056 ppm and HED99 of 0.00338 mg/kg/d
for the 99th percentile for uncertainty and variability are shown in Figure F-l 1. These values are
used as this supporting effect's POD to which additional uncertainty factors (UFs) are applied.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-32
-------
NTP.1988 kidney toxic nephropathy rat Marshall F oral.gav(GRP 49)
BMR: 0.05 extra
loglogistic, Pval = 1, AIC = 123
background 0, intercept 0.74, slope 0.31
"a u.o

| 0.6
L
BMDand BMDL, 0.0
1191, 0.0132
0.4
0.0 -i
0.6 0.
0.0 0.2 0.4
ABioactDCVCBW34
0.8 -
~o
0)
1 0.6
y=
<
c
o
£ 0.2
0.4 -
0.0 -i
multistage-1, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4, Beta(2) 0
BMDand BMDL. 0.0658. 0.0288
0.0 0.2 0.4 0.6 0.8 1.0
ABioactDCVCBW34
"O

a> 0.6
y=
<
c
o
£ 0.2
0.4 -
0.0 -i
multistage-2, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4, Beta(2) 0
0.8 -
BMDand BMDL. 0.0
358. 0.0288
0.6 0
0.0 0.2 0.4
ABioactDCVCBW34
0.8 "
"O
a;
| 0.6
<
£=
0
1	°-2
^ 0.0
0.4 "
multistage-1, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4
BMDand BMDL. 0.0358. 0.0288
0.6 0.
0.0 0.2 0.4
ABioactDCVCBW34
-o 0.8 -
a;
| 0.6
<
£=
O
0.4 -
£ °-2
^ 0.0
weibull, Pval = 1, AIC = 123
Background 0, Slope 1.1, Power 0.19
BMDand BMDL, 9.1
Be-08, NA
0.0
T
T
0.2 0.4 0.6 0.8
ABioactDCVCBW34
1.0
This document is a draft for review purposes only and does not constitute Agency policy.
10120109	' '	F-33
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Figure F-10. BMD modeling of NTP (1988) toxic nephropathy in female
Marshall rats.
NTP. 1988
BMDL for systemic kidney toxic.nephropathy in
F Marshall rat
Human
99%
0.0056
BMDL
Rodent
0.013
Human
median%
0.0422
rnnn mnnn i i urn
NTP. 1988
BMDL for systemic kidney toxic.nephropathy in
F Marshall rat
CD
O
>
o
Q
-I—'
o
CO
O
CO
<
Human
BMDL
Roden
Human
median%
10 4 10 3 10 2 10 1
101 102 103 104
10 4 10 3 10 2 10 1
101 102 103 104
TCE inhalation (ppm)
TCE oral (mg/kg-d)
Figure F-ll. Derivation of HEC99 and HED99 corresponding to the rodent
idPOD from NTP (1988) toxic nephropathy in rats.
F.6.2. Woolhiser et al. (2006)—Benchmark Dose (BMD) Modeling of Increased Kidney
Weight in Rats
The endpoint here is increased kidney weights in female Sprague-Dawley (S-D) rats
(Woolhiser et al., 2006), which was considered a supporting effect for the RfD.
F.6.2.1. Dosimetry and Benchmark Dose (BMD) Modeling
Rats were exposed to 100, 300, and 1000, 6 hours/day, 5 days/week, for 4 weeks. The
primary dose metric was selected to be average amount of DCVC bioactivated/kgyYday, with
median estimates from the PBPK model for this study of 0.038, 0.10, and 0.51.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-34
-------
1	Figure F-12 shows BMD modeling for the continuous models used (see Section F.5.2,
2	above). The Hill model with constant variance was selected because it had the lowest AIC and
3	because other models with the same AIC either were a power model with power parameter <1 or
4	had poor fits to the control data set.
5
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-35
-------
Woolhiser.etal.2006 Kidney kidney.wt.per1 OOgm rat CD (Sprague-Dawley) F inhal (GRP 65)
BMR: 0.1 relative
power, P(V) = 0.81, P(M) = 0.92, AIC = -128
lalpha -5, rho 2, control 0.81, slope 0.19, power 0.44
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.148, 0.0146
, I
i	1	r
0.0 0.1 0.2 0.3 0.4
ABioactDCVCBW34
0.5
power, P(V) = 0.89, P(M) = 0.87, AC = -130
alpha 0.0049, rho NA control 0.81, slope 0.19, power 0.44
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.146, 0.0126
, I
T
T
0.0 0.1 0.2 0.3 0.4 0.5
ABioactDCVCBW34
power, P(V) = 0.81, P(M) = 0.38, AC = -128
lalpha -5.1, rho 1.4, control 0.83, slope 0.23, power 1
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.356, 0.234
I
1—
0.0
~~r
0.1
T
—I—
0.3
0.2
ABioactDCVCBW34
~l—
0.4
~T~
0.5
power, P(V) = 0.89, P(M) = 0.4, AC = -130
alpha 0.0052, rho NA control 0.83, slope 0.23, power 1
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.36, 0.243
I
1—
0.0
—r
0.1
i—
0.2
T"
0.3
ABioactDCVCBW34
~l—
0.4
—T"
0.5
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
polyn, P(V) = 0.81, P(M) = 0.38, AC =-128
lalpha-5.1, rho 1.4, beta0 0.83, betal 0.23
BMD and BMDL, 0.356, 0.234
I
r~
o.o
"T"
0.1
0.2
I
0.3
I
0.4
ABioactDCVCBW34
I
0.5
polyn, P(V) = 0.89, P(M) = 0.4, AIC = -130
alpha 0.0052, rho NA, betaO 0.83, betal 0.23
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.36, 0.243
I
r~
o.o
~~r~
0.1
i
0.2
0.3
I
0.4
ABioactDCVCBW34
r~
0.5
hill, P(V) = 0.81, P(M) = 0.65, AIC=-128
lalpha -5, rho 2.2, Intercept 0.81, v 0.18, n 1, k 0.15
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.129, 0.0323
1	1	1	1	1	T"
0.0 0.1 0.2 0.3 0.4 0.5
ABioactDCVCBW34
hill, P(V) = 0.89, P(M) = 0.6, AIC = -130
alpha 0.005, rho NA intercept 0.81, v 0.18, n 1, k 0.15
1.00
c 0.95
S 0.90
E 0.85
0.80
0.75
BMD and BMDL, 0.132, 0.0309
	1	1	1	1	r
0.0 0.1 0.2 0.3 0.4 0.5
ABioactDCVCBW34
Figure F-122. BMD modeling of Woolhiser et al. (2006) for increased kidney
weight in female S-D rats.
This document is a draft for review purposes only and does not constitute Agency policy.
10120109	' '	F-36
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
The idPOD of 0.0309 mg DCVC bioactivated/kgyYday was a BMDL for a BMR of 10%
weight change, which is the BMR typically used by U.S. EPA for body weight and organ weight
changes. The response used in each case was the organ weight as a percentage of body weight,
to account for any commensurate decreases in body weight, although the results did not differ
much when absolute weights were used instead.
F.6.2.2. Derivation of HEC99 and HED99
The HEC99 and HED99 are the lower 99th percentiles for the continuous human exposure
concentration and continuous human ingestion dose that lead to a human internal dose equal to
the rodent idPOD. The derivation of the HEC99 of 0.0131 ppm and HED99 of 0.00791 mg/kg/d
for the 99th percentile for uncertainty and variability are shown in Figure F-13. These values are
used as this effect's POD to which additional UFs are applied, and the resulting candidate RfD
value is supportive of the RfD.
F.6.3. Keil et al. (2009)—Lowest-Observed-Adverse-Effect Level (LOAEL) for Decreased
Thymus Weight in Mice
The critical endpoint here is decreased thymus weight female B6C3F1 mice (Keil et al.,
2009)
F.6.3.1. Dosimetry
Mice were exposed to 1400 and 14000 ppb of TCE in drinking water, with an average
dose estimated by the authors to be 0.35 and 3.5 mg/kg/d, for 30 weeks. The dose-response
relationships were sufficiently supralinear that BMD modeling failed to produce an adequate fit.
The primary dose metric was selected to be the average amount of TCE metabolized/kgyYday.
The lower dose group was the LOAEL, and the median estimate from the PBPK model at that
exposure level was 0.139 mg TCE metabolized/kgyYday, which is used as the rodent idPOD.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-37
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Woolhiser.etal.2006
BMDL for systemic kidney weight.increased in
F Spraaue-Daw lev rat
o —
o —
Human
99%
0.0131
BMDL
Rodent
0.0309
Human
rredian%
0.0987
mimi i iiiiii i 11 iiiiii iiiiiiii i iiiiiiin
Wooihiser.etai.2006
BMDL for systemic kidney weight.increased in
F Spraaue-Daw lev rat
O -
CD
O
>
o
Q
-I—>
o
(0
o
CD
<
Human
99%:
0.00791
BMDL
Rodent
0.0309
Human
rredian%
0.078
inn i 111iiiii i 1111iii i 11 iiiii i 11 iiiii i 111iiiii i
10 4 10 3 10 2 10 1 1 101 102 103 104
10 4 10 3 10 2 10 1 1 101 102 103 104
TCE inhalation (ppm)
TCE oral (mg/kg-d)
Figure F-133. Derivation of HEC99 and HED99 corresponding to the rodent
idPOD from Woolhiser et al. (2006) for increased kidney weight in rats.
F.6.3.2. Derivation of HEC99 and HED99
The HEC99 and HED99 are the lower 99th percentiles for the continuous human exposure
concentration and continuous human ingestion dose that lead to a human internal dose equal to
the rodent idPOD. The derivation of the HEC99 of 0.0332 ppm and HED99 of 0.0482 mg/kg/d for
the 99th percentile for uncertainty and variability are shown in Figure F-14. These values are
used as this critical effect's POD to which additional UFs are applied.
F.6.4. Johnson et al. (2003)—Benchmark Dose (BMD) Modeling of Fetal Heart
Malformations in Rats
The critical endpoint here is increased fetal heart malformations in female S-D rats
(Johnson et al., 2003).
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-38
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Keil.etal.2009
mouse Exp 2
F R6C3F1 mouse RxnnsRd tn 0 35 rtm knrt 74 hr/rt 7 c
Hurren
0.0332
LOAEL
Rodent
0.139
Hurren
rredian%
0.092
ran i i iiiiiiii i mill i 11iiiii i iiiiiiirrnnmn
Keil.etal.2009
mouse E*p 2
F B6C3F1 mouse exposed to 0 35 ma.kad 24 hr/d 7 d/w kfor 30 w k
O -
CD
_Q
£
a>
Human
99%
0.0482
LOAEL
Rodent
0.139
Human
rredian%
0.0489
mn i i iiiiiii i 1111iii i 11iiiii i iiiiiiii i iiiiiiin
10 4 10 3 10 2 10 1 1
101 102 103 104
10 4 10 3 10 2 10 1 1 101 102 103 104
TCE inhalation (ppm)
TCE oral (mg/kg-d)
Figure F-144. Derivation of HEC99 and HED99 corresponding to the rodent
idPOD from Keil et al. (2009) for decreased thymus weight in mice.
F.6.4.1. Dosimetry and Benchmark Dose (BMD) Modeling
Rats were exposed to 2.5, 250, 1.5, or 1,100 ppm TCE in drinking water for 22 days
(GD 1-22). The primary dose metric was selected to be average amount of TCE metabolized by
oxidation/kgyYday, with median estimates from the PBPK model for this study of 0.00031, 0.033,
0.15, and 88.
As discussed previously in Section F.4.2.1, from results of nested log-logistic modeling
of these data, with the highest dose group dropped, the idPOD of 0.0142 mg TCE metabolized
by oxidation/kgyYday was a BMDL for a BMR of 1% increased in incidence in pups. A 1%
extra risk of a pup having a heart malformation was used as the BMR because of the severity of
the effect; some of the types of malformations observed could have been fatal.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-39
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
F.6.4.2. Derivation of HEC99 and HED99
The HEC99 and HED99 are the lower 99th percentiles for the continuous human exposure
concentration and continuous human ingestion dose that lead to a human internal dose equal to
the rodent idPOD. The derivation of the HEC99 of 0.00365 ppm and HED99 of 0.00515 mg/kg/d
for the 99th percentile for uncertainty and variability are shown in Figure F-15. These values are
used as this critical effect's POD to which additional UFs are applied.
Johnson.etal.2003
BMDL for developmental heart malformations in
F Soraaue-Dawlev rat
Human
99%:
0.00365
BMDL
Rodent
0.014
Human
rredian%
0.0116
im|—1 1 iiiiii|—1 1 inn |—111 inij—1 111 iiii|—1 111 iiii|—n
10 410 3 10 210 1 1 101 102 103 104
CD
_Q
£
a)
x
O
Johnson.etal.2003
BMDL for developmental heart malformations in
	F Spraque-Daw lev rat
Human
99%
0.00515
BMDL
Rodent
0.0142
Human
median%
0.00576
1 iiiij 1 1 iiiii| 1111 ni| iiiuiij 111 niij 11 iiiii| 11 iiii>| 1 111 ui| 1
10 510 410 310 210 1 1 101 102 103 104
TCE inhalation (ppm)
TCE oral (mg/kg-d)
Figure F-155. Derivation of HEC99 and HED99 corresponding to the rodent
idPOD from Johnson et al. (2003) for increased fetal cardiac malformations
in female S-D rats using the total oxidative metabolism dose metric.
F.6.5. Peden-Adams et al. (2006)—Lowest-Observed-Adverse-Effect Level (LOAEL) for
Decreased PFC Response and Increased Delayed-Type Hypersensitivity in Mice
The critical endpoints here are decreased PFC response and increased delayed-type
hypersensitivity in mice exposed pre- and postnatally (Peden-Adams et al., 2006).
Mice were exposed to 1400 and 14,000 ppb in drinking water, with an average dose in
the dams estimated by the authors to be 0.37 and 3.7 mg/kg/d, from GD0 to postnatal ages of 3
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-40
-------
1	or 8 weeks. The dose-response relationships were sufficiently supralinear that BMD modeling
2	failed to produce an adequate fit. In addition, because of the lack of an appropriate PBPK model
3	and parameters to estimate internal doses given the complex exposure pattern (placental and
4	lactational transfer, and pup ingestion postweaning), no internal dose estimates were made.
5	Therefore, the LOAEL of 0.37 mg/kg/d on the basis of applied dose was used as the critical
6	effect's POD to which additional UFs are applied.
7
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	F-41
-------
1
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	42
-------
APPENDIX G
TCE Cancer Dose-Response Analyses with
Rodent Cancer Bioassay Data
This document is a draft for review purposes only and does not constitute Agency policy.
G-i	DRAFT-DO NOT CITE OR QUOTE
-------
CONTENTS—Appendix G: TCE Cancer Dose-Response Analyses With Rodent Cancer
Bioassay Data
LIST OF TABLES	G-iii
LIST OF FIGURES	G-xix
APPENDIX G: TCE CANCER DOSE-RESPONSE ANALYSES WITH RODENT
CANCER BIOASSAY DATA	Error! Bookmark not defined.
1	G.l. DATA SOURCES	G-34
G. 1.1. Numbers at Risk	G-34
G. 1.2. Cumulative Incidence	G-35
2	G.2. INTERNAL DOSE METRICS AND DOSE ADJUSTMENTS	G-35
3	G.3. DOSE ADJUSTMENTS FOR INTERMITTENT EXPOSURE	G-37
4	G.4. RODENT TO HUMAN DOSE EXTRAPOLATION	G-37
5	G. 5. COMBINING DATA FROM RELATED EXPERIMENTS IN MALTONI
6	ETAL. (1986)	G-39
7	G.6. DOSE-RESPONSE MODELING RESULTS	G-40
8	G.7. MODELING TO ACCOUNT FOR DOSE GROUPS DIFFERING IN
9	SURVIVAL TIMES	G-45
G.7.1. Time-to-Tumor Modeling	G-46
G.l2. Poly-3 Calculation of Adjusted Number at Risk	G-47
10	G.8. COMBINED RISK FROM MULTIPLE TUMOR SITES	G-48
G.S.I. Methods	G-49
G.8.1.1. Single Tumor Sites	G-49
G.8.1.2. Combined Risk From Multiple Tumor Sites	G-49
G.S.2. Results	G-50
11	G.9. PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK)-MODEL
12	UNCERTAINTY ANALYSIS OF UNIT RISK ESTIMATES	G-74
13
This document is a draft for review purposes only and does not constitute Agency policy.
G-ii	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
LIST OF TABLES
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Table 2-1.	TCE metabolites and related parent compounds 	Error
Table 2-2.	Chemical properties of TCE	Error
Table 2-3.	Properties and uses of TCE related compounds	Error
Table 2-4.	TRI releases of TCE (pounds/year)	Error
Table 2-5.	Concentrations of trichloroethylene in ambient air	Error
-3
Table 2-6. TCE ambient air monitoring data ((J,g/m )	Error
Table 2-7. Mean TCE air levels across monitors by land setting and use (1985-1998)	Error!
Bookmark not defined.
Table 2-8. Concentrations of trichloroethylene in water based on pre-1990 studies	Error!
Bookmark not defined.
Table 2-9. Levels in food	Error! Bookmark not defined.
Table 2-10. TCE levels in whole blood by population percentile Error! Bookmark not defined.
"3
Table 2-11. Modeled 1999 annual exposure concentrations ((J,g/m ) for trichloroethylene. Error!
Bookmark not defined.
Table 2-12. Preliminary estimates of TCE intake from food ingestion	Error! Bookmark not
defined.
Table 2-13. Preliminary intake estimates of TCE and TCE-related chemicals .Error! Bookmark
not defined.
Table 2-14. Years of solvent use in industrial degreasing and cleaning operations	Error!
Bookmark not defined.
Table 2-15. TCE standards	Error! Bookmark not defined.
Table 3-1. Blood:air PC values for humans	Error! Bookmark not defined.
Table 3-2. Blood:air PC values for rats and mice	Error! Bookmark not defined.
Table 3-3. Air and blood concentrations during exposure to TCE in humans (Astrand and
Ovrum, 1976)	Error! Bookmark not defined.
Table 3-4. Retention of inhaled TCE vapor in humans (Jakubowski and Wieczorek, 1988)
	Error! Bookmark not defined.
Table 3-5. Uptake of TCE in human volunteers following 4 hour exposure to 70 ppm (Monster
et al., 1979)	Error! Bookmark not defined.
Table 3-6. Concentrations of TCE in maternal and fetal blood at birth	Error! Bookmark not
defined.
Table 3-7. Distribution of TCE to rat tissues" following inhalation exposure (Savolainen et al.,
1977)	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-iii	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 3-8. Tissue:blood partition coefficient values for TCE	Error! Bookmark not defined.
Table 3-9. Age-dependence of tissue:air partition coefficients in rats	Error! Bookmark not
defined.
Table 3-10. Predicted maximal concentrations of TCE in rat blood following a 6-hour inhalation
exposure (Rodriguez et al., 2007)	Error! Bookmark not defined.
Table 3-11. Tissue distribution of TCE metabolites following inhalation exposure	Error!
Bookmark not defined.
Table 3-12. Binding of 14C from [14C]TCE in rat liver and kidney at 72 hours after oral
administration of 200 mg/kg [14C]TCE (Dekant et al., 1986b)	Error! Bookmark not defined.
Table 3-13. In vitro TCE oxidative metabolism in hepatocytes and microsomal fractions . Error!
Bookmark not defined.
Table 3-14. In vitro kinetics of trichloroethanol and trichloroacetic acid formation from chloral
hydrate in rat, mouse, and human liver homogenates	Error! Bookmark not defined.
Table 3-15. In vitro kinetics of DCA metabolism in hepatic cytosol of mice, rats, and humans
	Error! Bookmark not defined.
Table 3-16. TCOH and TCA formed from CH in vitro in lysed whole blood of rats and mice or
fractionated blood of humans (nmoles formed in 400 [j,L samples over 30 minutes)	Error!
Bookmark not defined.
Table 3-17. Reported TCA plasma binding parameters	Error! Bookmark not defined.
Table 3-18. Partition coefficients for TCE oxidative metabolites Error! Bookmark not defined.
Table 3-19. Urinary excretion of trichloroacetic acid by various species exposed to
trichloroethylene [based on data reviewed in (Fisher et al., 1991)]	Error! Bookmark not
defined.
Table 3-20. P450 isoform kinetics for metabolism of TCE to CH in human, rat, and mouse
recombinant P450s	Error! Bookmark not defined.
Table 3-21. P450 isoform activities in human liver microsomes exhibiting different affinities for
TCE	Error! Bookmark not defined.
Table 3-22. Comparison of peak blood concentrations in humans exposed to 100 ppm (537
mg/m3) TCE for 4 hours (Fisher et al., 1998; Lash et al., 1999a)..Error! Bookmark not defined.
Table 3-23. GSH conjugation of TCE (at 1-2 mM) in liver and kidney cellular fractions in
humans, male F344 rats, and male B6C3F1 mice from Lash laboratory .... Error! Bookmark not
defined.
Table 3-24. Kinetics of TCE metabolism via GSH conjugation in male F344 rat kidney and
human liver and kidney cellular and subcellular fractions from Lash laboratory	Error!
Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-iv	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 3-25. GSH conjugation of TCE (at 1.4-4 mM) in liver and kidney cellular fractions in
humans, male F344 rats, and male B6C3F1 mice from Green and Dekant laboratories	Error!
Bookmark not defined.
Table 3-26. GGT activity in liver and kidney subcellular fractions of mice, rats, and humans
	Error! Bookmark not defined.
Table 3-27. Multispecies comparison of whole-organ activity levels of GGT and dipeptidase
	Error! Bookmark not defined.
Table 3-28. Comparison of hepatic in vitro oxidation and conjugation of TCEError! Bookmark
not defined.
Table 3-29. Estimates of DCVG in blood relative to inhaled TCE dose in humans exposed to 50
and 100 ppm (269 and 537 mg/m3) (Fisher et al., 1998; Lash et al., 1999b)	Error! Bookmark
not defined.
Table 3-30. Concentrations of TCE in expired breath from inhalation-exposed humans (Astrand
and Ovrum, 1976)	Error! Bookmark not defined.
Table 3-31. Conclusions from evaluation of Hack et al. (2006), and implications for PBPK
model development	Error! Bookmark not defined.
Table 3-32. Discussion of changes to the Hack et al. (2006) PBPK model implemented for this
assessment	Error! Bookmark not defined.
Table 3-33. PBPK model-based dose-metrics	Error! Bookmark not defined.
Table 3-34. Rodent studies with pharmacokinetic data considered for analysisError! Bookmark
not defined.
Table 3-35. Human studies with pharmacokinetic data considered for analysis	Error!
Bookmark not defined.
Table 3-36. Parameters for which scaling from mouse to rat, or from mouse and rat to human,
was used to update the prior distributions	Error! Bookmark not defined.
Table 3-37. Prior and posterior uncertainty and variability in mouse PBPK model parameters
	Error! Bookmark not defined.
Table 3-38. Prior and posterior uncertainty and variability in rat PBPK model parameters Error!
Bookmark not defined.
Table 3-39. Prior and posterior uncertainty and variability in human PBPK model parameters
	Error! Bookmark not defined.
Table 3-40. Confidence interval (CI) widths (ratio of 97.5% to 2.5% estimates) and fold-shift in
median estimate for the PBPK model population median parameters, sorted in order of
decreasing CI width. Shifts in the median estimate greater than threefold are in bold to denote
larger shifts between the prior and posterior distributions	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-v	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 3-41. Estimates of the residual-error	Error! Bookmark not defined.
Table 3-42. Summary comparison of updated PBPK model predictions and in vivo data in mice
	Error! Bookmark not defined.
Table 3-43. Summary comparison of updated PBPK model predictions and in vivo data used for
"calibration" in rats	Error! Bookmark not defined.
Table 3-44. Summary comparison of updated PBPK model predictions and in vivo data used for
"out-of-sample" evaluation in rats	Error! Bookmark not defined.
Table 3-45. Summary comparison of updated PBPK model predictions and in vivo data used for
"calibration" in humans	Error! Bookmark not defined.
Table 3-46. Summary comparison of updated PBPK model predictions and in vivo data used for
"out-of-sample" evaluation in humans	Error! Bookmark not defined.
Table 3-47. Summary of scaling parameters ordered by fraction of calibration data of moderate
or high sensitivity	Error! Bookmark not defined.
Table 3-48. Posterior predictions for representative internal doses: mouse Error! Bookmark not
defined.
Table 3-49. Posterior predictions for representative internal doses: rat	Error! Bookmark not
defined.
Table 3-50. Posterior predictions for representative internal doses: humanError! Bookmark not
defined.
Table 3-51. Degree of variance in dose-metric predictions due to incomplete convergence
(columns 2-4), combined uncertainty and population variability (columns 5-7), uncertainty in
particular human population percentiles (columns 8-10), model fits to in vivo data (column 11).
The GSD is the geometric standard deviation, which is a "fold-change" from the central
tendency	Error! Bookmark not defined.
Table 4-1. Description of epidemiologic cohort and proportionate mortality ratio (PMR) studies
assessing cancer and TCE exposure	Error! Bookmark not defined.
Table 4-2. Case-control epidemiologic studies examining cancer and TCE exposure	Error!
Bookmark not defined.
Table 4-3. Geographic-based studies assessing cancer and TCE exposure Error! Bookmark not
defined.
Table 4-4. Standards of epidemiologic study design and analysis use for identifying cancer
hazard and TCE exposure	Error! Bookmark not defined.
Table 4-5. Summary of criteria for meta-analysis study selectionError! Bookmark not defined.
Table 4-6. TCE genotoxicity: bacterial assays	Error! Bookmark not defined.
Table 4-7. TCE genotoxicity: fungal and yeast systems	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-vi	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 4-8. TCE genotoxicity: mammalian systems—gene mutations and chromosome
aberrations	Error! Bookmark not defined.
Table 4-9. TCE genotoxicity: mammalian systems—micronucleus, sister chromatic exchanges
	Error! Bookmark not defined.
Table 4-10. TCE genotoxicity: mammalian systems—unscheduled DNA synthesis, DNA strand
breaks/protein crosslinks, cell transformation	Error! Bookmark not defined.
Table 4-11. Genotoxicity of trichloroacetic acid—bacterial systems	Error! Bookmark not
defined.
Table 4-12. TCA Genotoxicity—mammalian systems (both in vitro and in vivo)	Error!
Bookmark not defined.
Table 4-13. Genotoxicity of dichloroacetic acid (bacterial systems)	Error! Bookmark not
defined.
Table 4-14. Genotoxicity of dichloroacetic acid—mammalian systems.... Error! Bookmark not
defined.
Table 4-15. Chloral hydrate genotoxicity: bacterial, yeast and fungal systems Error! Bookmark
not defined.
Table 4-16.. Chloral hydrate genotoxicity: mammalian systems—all genetic endpoints, in vitro
	Error! Bookmark not defined.
Table 4-17.. Chloral hydrate genotoxicity: mammalian systems—all genetic damage, in vivo
	Error! Bookmark not defined.
Table 4-18. TCE GSH conjugation metabolites genotoxicity	Error! Bookmark not defined.
Table 4-19. Genotoxicity of trichloroethanol	Error! Bookmark not defined.
Table 4-20. Summary of human trigeminal nerve and nerve conduction velocity studies... Error!
Bookmark not defined.
Table 4-21.	Summary of animal trigeminal nerve studies	Error
Table 4-22.	Summary of human auditory function studies	Error
Table 4-23.	Summary of animal auditory function studies	Error
Table 4-26.	Summary of animal visual system studies	Error
Table 4-27.	Summary of human cognition effect studies	Error
Table 4-28.	Summary of animal cognition effect studies	Error
Table 4-29.	Summary of human choice reaction time studies	Error
Table 4-30.	Summary of animal psychomotor function and reaction time studies	Error!
Bookmark	not defined.
Table 4-31.	Summary of animal locomotor activity studies	Error! Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-vii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 4-32. Summary of animal mood effect and sleep disorder studies... Error! Bookmark not
defined.
Table 4-33. Summary of human developmental neurotoxicity associated with TCE exposures
	Error! Bookmark not defined.
Table 4-34. Summary of mammalian in vivo developmental neurotoxicity studies—oral
exposures	Error! Bookmark not defined.
Table 4-35. Summary of animal dopamine neuronal studies	Error! Bookmark not defined.
Table 4-36. Summary of neurophysiological, neurochemical, and neuropathological effects with
TCE exposure	Error! Bookmark not defined.
Table 4-37. Summary of in vitro ion channel effects with TCE exposure . Error! Bookmark not
defined.
Table 4-38. Summary of human kidney toxicity studies	Error! Bookmark not defined.
Table 4-39. Summary of human studies on TCE exposure and kidney cancer. Error! Bookmark
not defined.
Table 4-40. Summary of case-control studies on kidney cancer and occupation or job titleError!
Bookmark not defined.
Table 4-41. Summary of lung and kidney cancer risks in active smokers (from IARC, 2004b)
	Error! Bookmark not defined.
Table 4-42. Summary of human studies on somatic mutations of the VHL genea	Error!
Bookmark not defined.
Table 4-43. Inhalation studies of kidney noncancer toxicity in laboratory animals	Error!
Bookmark not defined.
Table 4-44. Oral and i.p. studies of kidney noncancer toxicity in laboratory animals	Error!
Bookmark not defined.
Table 4-45. Summary of renal toxicity and tumor findings in gavage studies of trichloroethylene
by NTP (1990)	Error! Bookmark not defined.
Table 4-46. Summary of renal toxicity and tumor findings in gavage studies of trichloroethylene
by NCI (1976)	Error! Bookmark not defined.
Table 4-47. Summary of renal toxicity findings in gavage studies of trichloroethylene by
Maltoni et al. (1988)	Error! Bookmark not defined.
Table 4-48. Summary of renal toxicity and tumor incidence in gavage studies of
trichloroethylene by NTP (1988)	Error! Bookmark not defined.
Table 4-49. Summary of renal toxicity and tumor findings in inhalation studies of
trichloroethylene by Maltoni et al. (1988)a	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-viii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 4-50. Summary of renal tumor findings in inhalation studies of trichloroethylene by
Henschler et al. (1980)a and Fukuda et al. (1983)b	Error! Bookmark not defined.
Table 4-51. Summary of renal tumor findings in gavage studies of trichloroethylene by
Henschler et al. (1984)a and Van Duuren et al. (1979)b	Error! Bookmark not defined.
Table 4-52. Laboratory animal studies of kidney noncancer toxicity of TCE metabolites.. Error!
Bookmark not defined.
Table 4-53. Summary of histological changes in renal proximal tubular cells induced by chronic
exposure to TCE, DCVC, and TCOH	Error! Bookmark not defined.
Table 4-54. Summary of major mode of action conclusions for TCE kidney carcinogenesis
	Error! Bookmark not defined.
Table 4-55. Summary of human liver toxicity studies	Error! Bookmark not defined.
Table 4-56. Selected results from epidemiologic studies of TCE exposure and cirrhosis ... Error!
Bookmark not defined.
Table 4-57. Selected results from epidemiologic studies of TCE exposure and liver cancer
	Error! Bookmark not defined.
Table 4-58. Oral studies of TCE-induced liver effects in mice and rats	Error! Bookmark not
defined.
Table 4-59. Inhalation and i.p. studies of TCE-induced liver effects in mice and rats	Error!
Bookmark not defined.
Table 4-60. Summary of liver tumor findings in gavage studies of trichloroethylene by NTP
(1990)a	Error! Bookmark not defined.
Table 4-61. Summary of liver tumor findings in gavage studies of trichloroethylene by NCI
(1976)	Error! Bookmark not defined.
Table 4-62. Summary of liver tumor incidence in gavage studies of trichloroethylene by NTP
(1988)	Error! Bookmark not defined.
Table 4-63. Summary of liver tumor findings in inhalation studies of trichloroethylene by
Maltoni et al. (1988)a	Error! Bookmark not defined.
Table 4-64. Summary of liver tumor findings in inhalation studies of trichloroethylene by
Henschler et al. (1980)a and Fukuda et al. (1983)	Error! Bookmark not defined.
Table 4-65. Summary of liver tumor findings in gavage studies of trichloroethylene by
Henschler et al. (1984)a	Error! Bookmark not defined.
Table 4-66. Potency indicators for mouse hepatocarcinogenicity and in vitro transactivation of
mouse PPARa for four PPARa agonists	Error! Bookmark not defined.
Table 4-67. Potency indicators for rat hepatocarcinogenicity and common short-term markers of
PPARa activation for four PPARa agonists	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-ix	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 4-68. Summary of mode of action conclusions for TCE-induced liver carcinogenesis
	Error! Bookmark not defined.
Table 4-69. Studies of immune parameters (IgE antibodies and cytokines) and trichloroethylene
in humans	Error! Bookmark not defined.
Table 4-70. Case-control studies of autoimmune diseases with measures of trichloroethylene
exposure	Error! Bookmark not defined.
Table 4-71. Incidence cohort studies of TCE exposure and lymphopoietic and hematopoietic
cancer risk	Error! Bookmark not defined.
Table 4-72. Mortality cohort and PMR studies of TCE exposure and lymphopoietic and
hematopoietic cancer risk	Error! Bookmark not defined.
Table 4-73. Case-control studies of TCE exposure and lymphopoietic cancer, leukemia or
multiple myeloma	Error! Bookmark not defined.
Table 4-74. Geographic-based studies of TCE and non-Hodgkin lymphoma or leukemia in
adults	Error! Bookmark not defined.
Table 4-75. Selected results from epidemiologic studies of TCE exposure and childhood
leukemia	Error! Bookmark not defined.
Table 4-76. Summary of TCE immunosuppression studies	Error! Bookmark not defined.
Table 4-77. Summary of TCE hypersensitivity studies	Error! Bookmark not defined.
Table 4-78. Summary of autoimmune-related studies of TCE and metabolites in mice and rats
(by sex, strain, and route of exposure)51	Error! Bookmark not defined.
Table 4-79. Malignant lymphomas incidence in mice exposed to TCE in gavage and inhalation
exposure studies	Error! Bookmark not defined.
Table 4-80. Leukemia incidence in rats exposed to TCE in gavage and inhalation exposure
studies	Error! Bookmark not defined.
Table 4-81. Selected results from epidemiologic studies of TCE exposure and lung cancer
	Error! Bookmark not defined.
Table 4-82. Selected results from epidemiologic studies of TCE exposure and laryngeal cancer
	Error! Bookmark not defined.
Table 4-83. Animal toxicity studies of trichloroethylene	Error! Bookmark not defined.
Table 4-84. Animal carcinogenicity studies of trichloroethylene.Error! Bookmark not defined.
Table 4-85. Human reproductive effects	Error! Bookmark not defined.
Table 4-86. Summary of mammalian in vivo reproductive toxicity studies—inhalation exposures
	Error! Bookmark not defined.
Table 4-87. Summary of mammalian in vivo reproductive toxicity studies—oral exposures
	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-x	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 4-88. Summary of adverse female reproductive outcomes associated with TCE exposures
	Error! Bookmark not defined.
Table 4-89. Summary of adverse male reproductive outcomes associated with TCE exposures
	Error! Bookmark not defined.
Table 4-90. Summary of human studies on TCE exposure and prostate cancerError! Bookmark
not defined.
Table 4-91. Summary of human studies on TCE exposure and breast cancer ..Error! Bookmark
not defined.
Table 4-92. Summary of human studies on TCE exposure and cervical cancerError! Bookmark
not defined.
Table 4-93. Histopathology findings in reproductive organs	Error! Bookmark not defined.
Table 4-94. Testicular tumors in male rats exposed to TCE, adjusted for reduced survival51
	Error! Bookmark not defined.
Table 4-95. Developmental studies in humans	Error! Bookmark not defined.
Table 4-96. Summary of mammalian in vivo developmental toxicity studies—inhalation
exposures	Error! Bookmark not defined.
Table 4-97. Ocular defects observed (Narotsky et al., 1995)	Error! Bookmark not defined.
Table 4-98. Summary of mammalian in vivo developmental toxicity studies—oral exposures
	Error! Bookmark not defined.
Table 4-99. Types of congenital cardiac defects observed in TCE-exposed fetuses (Dawson et
al., 1993, Table 3)	Error! Bookmark not defined.
Table 4-100. Types of heart malformations per 100 fetuses (Johnson et al., 2003, Table 2, p.
290)	Error! Bookmark not defined.
Table 4-101. Congenital cardiac malformations (Johnson et al., 1998a, Table 2, p. 997)... Error!
Bookmark not defined.
Table 4-102. Summary of adverse fetal and early neonatal outcomes associated with TCE
exposures	Error! Bookmark not defined.
Table 4-103. Summary of studies that identified cardiac malformations associated with TCE
exposures	Error! Bookmark not defined.
Table 4-104. Events in cardiac valve formation in mammals and birdsa ... Error! Bookmark not
defined.
Table 4-105. Summary of other structural developmental outcomes associated with TCE
exposures	Error! Bookmark not defined.
Table 4-106. Summary of developmental neurotoxicity associated with TCE exposures ... Error!
Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xi	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 4-107. Summary of developmental immunotoxicity associated with TCE exposuresError!
Bookmark not defined.
Table 4-108. Summary of childhood cancers associated with TCE exposures. Error! Bookmark
not defined.
Table 4-109. Selected observations from case-control studies of TCE exposure and esophageal
cancer	Error! Bookmark not defined.
Table 4-110. Summary of human studies on TCE exposure and esophageal cancer	Error!
Bookmark not defined.
Table 4-111. Summary of human studies on TCE exposure and bladder cancer	Error!
Bookmark not defined.
Table 4-112. Summary of human studies on TCE exposure and brain cancer.. Error! Bookmark
not defined.
Table 4-113. Estimated lifestage-specific daily doses for TCE in water3.. Error! Bookmark not
defined.
Table. 5-1 Summary of studies of neurological effects suitable for dose-response assessment
	Error! Bookmark not defined.
Table 5-2. Neurological effects in studies suitable for dose-response assessment, and
corresponding cRfCs and cRfDs	Error! Bookmark not defined.
Table 5-3. Summary of studies of kidney, liver, and body weight effects suitable for
dose-response assessment	Error! Bookmark not defined.
Table 5-4. Kidney, liver, and body weight effects in studies suitable for dose-response
assessment, and corresponding cRfCs and cRfDs	Error! Bookmark not defined.
Table 5-5. Summary of studies of immunological effects suitable for dose-response assessment
	Error! Bookmark not defined.
Table 5-6. Immunological effects in studies suitable for dose-response assessment, and
corresponding cRfCs and cRfDs	Error! Bookmark not defined.
Table 5-7. Summary of studies of reproductive effects suitable for dose-response assessment
	Error! Bookmark not defined.
Table 5-8. Reproductive effects in studies suitable for dose-response assessment, and
corresponding cRfCs and cRfDs	Error! Bookmark not defined.
Table 5-9. Summary of studies of developmental effects suitable for dose-response assessment
	Error! Bookmark not defined.
Table 5-10. Developmental effects in studies suitable for dose-response assessment, and
corresponding cRfCs and cRfDs	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xii	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 5-11. Ranges of cRfCs based on applied dose for various noncancer effects associated
with inhalation TCE exposure	Error! Bookmark not defined.
Table 5-12. Ranges of cRfDs based on applied dose for various noncancer effects associated
with oral TCE exposure	Error! Bookmark not defined.
Table 5-13. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical neurological effects	Error!
Bookmark not defined.
Table 5-14. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical kidney effects Error! Bookmark not
defined.
Table 5-15. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical liver effects... Error! Bookmark not
defined.
Table 5-16. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical immunological effects	Error!
Bookmark not defined.
Table 5-17. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical reproductive effects	Error!
Bookmark not defined.
Table 5-18. cRfCs and cRfDs (based on applied dose) and p-cRfCs and p-cRfDs (based on
PBPK modeled internal dose-metrics) for candidate critical developmental effects	Error!
Bookmark not defined.
Table 5-19. Comparison of "sensitive individual" HECs or HEDs for neurological effects based
on PBPK modeled internal dose-metrics at different levels of confidence and sensitivity, at the
NOAEL or LOAEL	Error! Bookmark not defined.
Table 5-20. Comparison of "sensitive individual" HECs or HEDs for kidney and liver effects
based on PBPK modeled internal dose-metrics at different levels of confidence and sensitivity, at
the NOAEL or LOAEL	Error! Bookmark not defined.
Table 5-21. Comparison of "sensitive individual" HECs or HEDs for immunological effects
based on PBPK modeled internal dose-metrics at different levels of confidence and sensitivity, at
the NOAEL or LOAEL	Error! Bookmark not defined.
Table 5-22. Comparison of "sensitive individual" HECs or HEDs for reproductive effects based
on PBPK modeled internal dose-metrics at different levels of confidence and sensitivity, at the
NOAEL or LOAEL	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xiii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table 5-23. Comparison of "sensitive individual" HECs or HEDs for developmental effects
based on PBPK modeled internal dose-metrics at different levels of confidence and sensitivity, at
the NOAEL or LOAEL	Error! Bookmark not defined.
Table 5-24. Lowest p-cRfCs or cRfCs for different effect domains	Error! Bookmark not
defined.
Table 5-25. Lowest p-cRfDs or cRfDs for different effect domains	Error! Bookmark not
defined.
Table 5-26. Lowest p-cRfCs for candidate critical effects for different types of effect based on
primary dose-metric	Error! Bookmark not defined.
Table 5-27. Lowest p-cRfDs for candidate critical effects for different types of effect based on
primary dose-metric	Error! Bookmark not defined.
Table 5-28. Summary of critical studies, effects, PODs, and UFs used to derive the RfC ..Error!
Bookmark not defined.
Table 5-29. Summary of supporting studies, effects, PODs, and UFs for the RfC	Error!
Bookmark not defined.
Table 5-30. Summary of critical studies, effects, PODs, and UFs used to derive the RfD ..Error!
Bookmark not defined.
Table 5-31. Summary of supporting studies, effects, PODs, and UFs for the RfD	Error!
Bookmark not defined.
Table 5-32. Inhalation bioassays	Error! Bookmark not defined.
Table 5-33. Oral bioassays	Error! Bookmark not defined.
Table 5-34. Specific dose-response analyses performed and dose-metrics used	Error!
Bookmark not defined.
Table 5-35. Mean PBPK model predictions for weekly internal dose in humans exposed
continuously to low levels of TCE via inhalation (ppm) or orally (mg/kg-day) Error! Bookmark
not defined.
Table 5-36. Summary of PODs and unit risk estimates for each sex/species/bioassay/tumor type
(inhalation)	Error! Bookmark not defined.
Table 5-37. Summary of PODs and unit risk estimates for each sex/species/bioassay/tumor type
(oral)	Error! Bookmark not defined.
Table 5-38. Comparison of survival-adjusted results for 3 oral male rat data sets3	Error!
Bookmark not defined.
Table 5-39. Inhalation: most sensitive bioassay for each sex/species combination3	Error!
Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xiv DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table 5-40. Oral: most sensitive bioassay for each sex/species combination51 ..Error! Bookmark
not defined.
Table 5-41. Summary of PBPK model-based uncertainty analysis of unit risk estimates for each
sex/species/bioassay/tumor type (inhalation)	Error! Bookmark not defined.
Table 5-42. Summary of PBPK model-based uncertainty analysis of unit risk estimates for each
sex/species/bioassay/tumor type (oral)	Error! Bookmark not defined.
Table 5-43. Results from Charbotel et al. (2006) on relationship between TCE exposure and
RCC	Error! Bookmark not defined.
Table 5-44. Extra risk estimates for RCC incidence from various levels of lifetime exposure to
TCE, using linear cumulative exposure model	Error! Bookmark not defined.
Table 5-45. ECoi, LECoi, and unit risk estimates for RCC incidence, using linear cumulative
exposure model	Error! Bookmark not defined.
Table 5-46. Relative contributions to extra risk for cancer incidence from TCE exposure for
multiple tumor types	Error! Bookmark not defined.
Table 5-47. Route-to-route extrapolation of site-specific inhalation unit risks to oral slope
factors	Error! Bookmark not defined.
Table 5-48. Sample calculation for total lifetime cancer risk based on the kidney unit risk
estimate, potential risk for NHL and liver cancer, and potential increased early-life susceptibility,
-3
assuming a constant lifetime exposure to 1 (J,g/m of TCE in air ..Error! Bookmark not defined.
Table 5-49. Sample calculation for total lifetime cancer risk based on the kidney cancer unit risk
estimate, potential risk for NHL and liver cancer, and potential increased early-life susceptibility,
assuming a constant lifetime exposure to 1 [j,g/L of TCE in drinking water Error! Bookmark not
defined.
Table A-l. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in mice A-
11
Table A-2. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in rats .. A-
18
Table A-3. Evaluation of Hack et al. (2006) PBPK model predictions for in vivo data in humans
	A-27
Table A-4. PBPK model parameters, baseline values, and scaling relationships	A-51
Table A-5. Uncertainty distributions for the population mean of the PBPK model parameters . A-
89
Table A-6. Updated prior distributions for selected parameters in the rat and human	A-95
Table A-7. Uncertainty distributions for the population variance of the PBPK model parameters
	A-100
This document is a draft for review purposes only and does not constitute Agency policy.
G-xv	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table A-8. Measurements used for calibration	A-104
•	Table A-9. Posterior distributions for mouse PBPK model population parameters
A-110
Table A-10. Posterior distributions for mouse residual errors	A-l 12
Table A-l 1. Posterior correlations for mouse population mean parameters	A-l 13
Table A-12. Posterior distributions for rat PBPK model population parameters	A-120
Table A-13. Posterior distributions for rat residual errors	A-123
Table A-14. Posterior correlations for rat population mean parameters	A-127
Table A-15. Posterior distributions for human PBPK model population parameters	A-134
Table A-16. Posterior distributions for human residual errors	A-138
Table A-17. Posterior correlations for human population mean parameters	A-140
Table A-18. Summary characteristics of model runs	A-224
Table B-l. Description of epidemiologic cohort and PMR studies assessing cancer and TCE
exposure	B-3
Table B-2. Case-control epidemiologic studies examining cancer and TCE exposure	B-12
Table B-3. Geographic-based studies assessing cancer and TCE exposure	B-25
Table B-4. Approximate statistical power (%) in cohort and geographic-based studies to detect
an RR 2	B-47
Table B-5. Summary of rationale for study selection for meta-analysis	B-72
Table B-6. Characteristics of epidemiologic investigations of Rocketdyne workers	B-94
Table C-l. Selected RR estimates for NHL associated with TCE exposure (overall effect) from
cohort studies	C-l
Table C-2. Selected RR estimates for NHL associated with TCE exposure from case-control
studies3	C-9
Table C-3. Summary of some meta-analysis results for TCE (overall) and NHL	C-14
Table C-4. Selected RR estimates for NHL risk in highest TCE exposure groups	C-23
Table C-5. Summary of some meta-analysis results for TCE (highest exposure groups) and NHL
	C-27
Table C-6. Selected RR estimates for kidney cancer associated with TCE exposure (overall
effect) from cohort studies	C-3 5
Table C-l. Selected RR estimates for renal cell carcinoma associated with TCE exposure from
case-control studies3	C-37
Table C-8. Summary of some meta-analysis results for TCE (overall) and kidney cancer	C-42
Table C-9. Selected RR estimates for kidney cancer risk in highest TCE exposure groups... C-47
This document is a draft for review purposes only and does not constitute Agency policy.
G-xvi DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table C-10. Summary of some meta-analysis results for TCE (highest exposure groups) and
kidney cancer	C-55
Table C-l 1. Selected RR estimates for liver cancer associated with TCE exposure (overall
effect) from cohort studies	C-61
Table C-12. Summary of some meta-analysis results for TCE and liver cancer	C-65
Table C-13. Selected RR estimates for liver cancer risk in highest TCE exposure groups	C-69
Table C-14. Selected RR estimates for lung (& bronchus) cancer associated with TCE exposure
(overall effect) from cohort studies	C-77
Table C-15. Summary of some meta-analysis results for TCE and lung cancer	C-83
Table C-16. Selected RR estimates for lung cancer risk in highest TCE exposure groups	C-86
Table C-17. Summary of some meta-analysis results for TCE (highest exposure groups) and
lung cancer	C-89
Table D-l. Epidemiological studies: Neurological effects of trichloroethylene	D-30
Table D-2. Epidemiological studies: Neurological effects of trichloroethylene/mixed solvents D-
76
Table D-3. Literature review of studies of TCE and domains assessed with
neurobehavioral/neurological methods	D-l05
Table D-4. Summary of mammalian in vivo trigeminal nerve studies	D-130
Table D-5. Summary of mammalian in vivo ototoxicity studies	D-131
Table D-6. Summary of mammalian sensory studies—vestibular and visual systems	D-134
Table D-7. Summary of mammalian cognition studies	D-136
Table D-8. Summary of mammalian psychomotor function, locomotor activity, and reaction
time studies	D-138
Table D-9. Summary of mammalian in vivo dopamine neuronal studies	D-141
Table D-10. Summary of neurochemical effects with TCE exposure	D-142
Table D-l 1. Summary of in vitro ion channel effects with TCE exposure	D-145
Table D-12. Summary of mammalian in vivo developmental neurotoxicity studies—oral
exposures	D-l 46
Table E-l. Mice data for 13 weeks: mean body and liver weights	Error! Bookmark not
defined.
Table E-2. Prevalence and Multiplicity data from DeAngelo et al. (1999) Error! Bookmark not
defined.
Table E-3. Difference in pathology by inclusion of unscheduled deaths from DeAngelo et al.
(1999)	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xvii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Table E-4. Comparison of data from Carter et al. (2003) and DeAngelo et al. (1999)	Error!
Bookmark not defined.
Table E-5. Prevalence of foci and tumors in mice administered NaCl, DC A, or TCA from
Pereira (1996)	Error! Bookmark not defined.
Table E-6. Multiplicity of foci and tumors in mice administered NaCl, DC A, or TCA from
Pereira (1996)	Error! Bookmark not defined.
Table E-7. Phenotype of foci reported in mice exposed to NaCl, DC A, or TCA by Pereira (1996)
	Error! Bookmark not defined.
Table E-8. Phenotype of tumors reported in mice exposed NaCl, DC A, or TCA by Pereira
(1996)	Error! Bookmark not defined.
Table E-9. Multiplicity and incidence data (31 week treatment) from Pereira and Phelps (1996)
	Error! Bookmark not defined.
Table E-10. Comparison of descriptions of control data between George et al. (2000) and
DeAngelo et al. (2008)	Error! Bookmark not defined.
Table E-l 1. TCA-induced increases in liver tumor occurrence and other parameter over control
after 60 weeks (Study #1)	Error! Bookmark not defined.
Table E-12. TCA-induced increases in liver tumor occurrence after 104 wks (Studies #2 and #3)
	Error! Bookmark not defined.
Table E-13. Comparison of liver effects from TCE, TCA, and DCA (10-day exposures in mice)
	Error! Bookmark not defined.
Table E-14. Liver weight induction as percent liver/body weight fold-of-control in male
B6C3F1 mice from DCA or TCA drinking water studies	Error! Bookmark not defined.
Table E-l5. Liver weight induction as percent liver/body weight fold-of-control in male
B6C3F1 or Swiss mice from TCE gavage studies	Error! Bookmark not defined.
Table E-16. B6C3F1 and Swiss (data sets combined)	Error! Bookmark not defined.
Table E-17. Power calculations11 for experimental design described in text, using Pereira et al. as
an example	Error! Bookmark not defined.
Table E-18. Comparison between results for Yang et al. (2007) and Cheung et al. (2004)aError!
Bookmark not defined.
Table F-l. Dose-response data from Chia et al. (1996)	F-l
Table F-2. Data on TCE in air (ppm) and urinary metabolite concentrations in workers reported
by Ikeda et al. (1972)	F-2
Table F-3. Estimated urinary metabolite and TCE air concentrations in dose groups from Chia et
al. (1996)	1-5
Table F-4. Data on fetuses and litters with abnormal hearts from Johnson et al. (2003)	F-12
This document is a draft for review purposes only and does not constitute Agency policy.
G-xviii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Table F-5. Comparison of observed and predicted numbers of fetuses with abnormal hearts from
Johnson et al. (2003), with and without the high-dose group, using a nested model	F-12
Table F-6. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, on the basis of applied dose (mg/kg/d in drinking water)... F-
13
Table F-7. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, using the TotOxMetabBW34 dose metric	F-16
Table F-8. Results of nested log-logistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high-dose group, using the AUCCBld dose metric	F-17
Table F-9. Analysis of LSCs with respect to dose from Narotsky et al. (1995)	F-18
Table F-10. Results of nested log-logistic and Rai-VanRyzin model for fetal eye defects from
Narotsky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water)	F-19
Table F-l 1. Comparison of results of nested log-logistic (without LSC or IC) and quantal log-
logistic model for fetal eye defects from Narotsky et al. (1995)	F-21
Table F-12. Results of nested log-logistic and Rai-VanRyzin model for prenatal loss from
Narotsky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water)	F-23
Table F-13. Model selections and results for noncancer dose-response analyses	F-26
Table G-l. Internal dose metrics used in dose-response analyses, identified by "X"	G-35
Table G-2. Experiments BT304 and BT304bis, female Sprague-Dawley rats, Maltoni et al.
(1986). Number alive is reported for week of first tumor observation in either males or females.a
These data were not used for dose-response modeling because there is no consistent trend (for
the combined data, there is no significant trend by the Cochran-Armitage test, and no significant
differences between control and dose groups by Fisher's exact test)	G-41
Table G-3. Experiments BT304 and BT304bis, male Sprague-Dawley rats, Maltoni et al. (1986):
leukemias. Number alive is reported for week of first tumor observation in either males or
females.a	G-42
Table G-4. Experiments BT304 and BT304bis, male Sprague-Dawley rats, Maltoni et al. (1986):
kidney adenomas + carcinomas. Number alive is reported for week of first tumor observation in
either males or females.a	G-43
Table G-5. Experiments BT304 and BT304bis, male Sprague-Dawley rats, Maltoni et al. (1986):
testis, Leydig cell tumors. Number alive is reported for week of first tumor observation.51.... G-44
Table G-6. Rodent to human conversions for internal dose metric TotOxMetabBW34	G-50
Table G-l. Rodent to human conversions for internal dose metric TotMetabBW34	G-50
Table G-8. Female B6C3F1 mice—applied doses: data	G-52
This document is a draft for review purposes only and does not constitute Agency policy.
G-xix DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
Table G-9. Female B6C3F1 mice—applied doses: model selection comparison of model fit
statistics for multistage models of increasing order	G-52
Table G-10. Female B6C3F1 mice—applied doses: BMD and risk estimates (inferences for
BMR of 0.05 extra risk at 95% confidence level)	G-53
Table G-l 1. B6C3F1 female mice inhalation exposure—applied doses	G-56
Table G-12. B6C3F1 female mice—applied doses: model selection comparison of model fit
statistics for multistage models of increasing order	G-56
Table G-13. B6C3F1 female mice inhalation exposure—applied doses (inferences for 0.05 extra
risk at 95% confidence level)	G-57
Table G-14. Maltoni Sprague-Dawley male rats—applied doses	G-60
Table G-15. Maltoni Sprague-Dawley male rats—applied doses: model selection comparison of
model fit statistics for multistage models of increasing order	G-60
Table G-16. Maltoni Sprague-Dawley male rats—applied doses	G-61
Table G-17. Female B6C3F1 mice—internal dose metric (total oxidative metabolism): dataG-64
Table G-18. Female B6C3F1 mice—internal dose: model selection comparison of model fit
statistics for multistage models of increasing order	G-64
Table G-19. Female B6C3F1 mice—internal dose metric (total oxidative metabolism): BMD
and risk estimates (values rounded to 4 significant figures) (inferences for BMR of 0.05 extra
risk at 95% confidence level)	G-65
Table G-20. B6C3F1 female mice inhalation exposure—internal dose metric (total oxidative
metabolism)	G-67
Table G-21. B6C3F1 female mice—internal dose: model selection comparison of model fit
statistics for multistage models of increasing order	G-67
Table G-22. B6C3F1 female mice inhalation exposure—internal dose metric (total oxidative
metabolism) (inferences for 0.05 extra risk at 95% confidence level)	G-68
Table G-23. Maltoni Sprague-Dawley male rats—internal dose metric (total metabolism)... G-71
Table G-24. Maltoni Sprague-Dawley male rats—internal dose model selection comparison of
model fit statistics for multistage models of increasing order	G-71
Table G-25. Maltoni Sprague-Dawley male rats—internal dose metric (total metabolism)
(inferences for 0.01 extra risk at 95% confidence level)	G-72
This document is a draft for review purposes only and does not constitute Agency policy.
G-xx	DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
LIST OF FIGURES
Figure 2-1. Molecular structure of TCE	Error! Bookmark not defined.
Figure 2-2. Source contribution to TCE emissions	Error! Bookmark not defined.
Figure 2-3. Annual emissions of TCE	Error! Bookmark not defined.
Figure 2-4. Modeled ambient air concentrations of TCE	Error! Bookmark not defined.
Figure 3-1. Gas uptake data from closed-chamber exposure of rats to TCE. Symbols represent
measured chamber concentrations. Source: Simmons et al. (2002)	Error! Bookmark not
defined.
Figure 3-2. Disposition of [14C]TCE administered by oral gavage in mice (Dekant et al., 1984;
Dekant et al., 1986b; Green and Prout, 1985; Prout et al., 1985). .Error! Bookmark not defined.
Figure 3-3. Disposition of [14C]TCE administered by oral gavage in rats (Dekant et al., 1984;
Dekant et al., 1986b; Green and Prout, 1985; Prout et al., 1985). .Error! Bookmark not defined.
Figure 3-4. Scheme for the oxidative metabolism of TCE	Error! Bookmark not defined.
Figure 3-5. Scheme for GSH-dependent metabolism of TCE	Error! Bookmark not defined.
Figure 3-6. Interorgan TCE transport and metabolism via the GSH pathway. See Figure 3-5 for
enzymes involved in metabolic steps. Source: Lash et al.(2000a; 2000b); NRC (2006)	Error!
Bookmark not defined.
Figure 3-7. Overall structure of PBPK model for TCE and metabolites used in this assessment.
Boxes with underlined labels are additions or modifications of the Hack et al. (2006) model,
which are discussed in Table 3-32	Error! Bookmark not defined.
Figure 3-8. Schematic of how posterior predictions were generated for comparison with
experimental data. Two sets of posterior predictions were generated: population predictions
(diagonal hashing) and subject-specific predictions (vertical hashing). (Same as Figure A-2 in
Appendix A)	Error! Bookmark not defined.
Figure 3-9. Comparison of mouse data and PBPK model predictions from a random posterior
sample. Each panel shows results for a different measurement. The solid line represents
prediction = data, and the grey dotted lines show prediction = data x GSDerr and data ^ GSDerr,
where GSDerr is the median estimate of the residual-error GSD shown in Table 3-41	Error!
Bookmark not defined.
Figure 3-10. Comparison of rat data and PBPK model predictions from a random posterior
sample. Each panel shows results for a different measurement. The solid line represents
prediction = data, and the grey lines show prediction = data x GSDerr and data ^ GSDerr, where
GSDerr is the lowest (dotted) and highest (dashed) median estimate of the residual-error GSD
shown in Table 3-41	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxi DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Figure 3-11. Comparison of urinary excretion data for NAcDCVC and predictions from the
Hack et al. (2006) and the updated PBPK models. Data are from Bernauer et al. (1996) for (A
and B) rats or (C and D) humans exposed for 6 hour to 40 (o), 80 (A), or 160 (+) ppm in air
(thick horizontal line denotes the exposure period). Predictions from Hack et al. (2006) and the
corresponding data (A and C) are only for the 1,2 isomer, whereas those from the updated model
(B and D) are for both isomers combined. Parameter values used for each prediction are a
random sample from the subject-specific parameters from the rat and human MCMC chains (the
last iteration of the first chain was used in each case). Note that in the Hack et al. (2006) model,
each dose group had different model parameters, whereas in the updated model, all dose groups
are required to have the same model parameters. See files linked to Appendix A for comparisons
with the full distribution of predictions	Error! Bookmark not defined.
Figure 3-12. Comparison of human data and PBPK model predictions from a random posterior
sample. Each panel shows results for a different measurement. The solid line represents
prediction = data, and the grey lines show prediction = data x GSDerr and data ^ GSDerr, where
GSDerr is the lowest (dotted) and highest (dashed) median estimate of the residual-error GSD
shown in Table 3-41	Error! Bookmark not defined.
Figure 3-13. Comparison of DCVG concentrations in human blood and predictions from the
updated model. Data are mean concentrations for males (A) and females (o) reported in Lash
et al. (1999a) for humans exposed for 4 hours to 100 ppm TCE in air (thick horizontal line
denotes the exposure period). Data for oxidative metabolites from the same individuals were
reported in Fisher et al. (1998) but could not be matched with the individual DCVG data (Lash
2007, personal communication). The vertical error bars are standard errors of the mean as
reported in Lash et al. (1999a) (n = 8, so standard deviation is 80.5-fold larger). Lines are PBPK
model predictions for individual male (solid) and female (dashed) subjects. Parameter values
used for each prediction are a random sample from the individual-specific parameters from the
human MCMC chains (the last iteration of the 1st chain was used). See files linked to Appendix
A for comparisons with the full distribution of predictions	Error! Bookmark not defined.
Figure 3-14. Sensitivity analysis results: Number of mouse calibration data points with SC in
various categories for each scaling parameter	Error! Bookmark not defined.
Figure 3-15. Sensitivity analysis results: Number of rat calibration data points with SC in
various categories for each scaling parameter	Error! Bookmark not defined.
Figure 3-16. Sensitivity analysis results: Number of human calibration data points with SC in
various categories for each scaling parameter	Error! Bookmark not defined.
Figure 3-17. PBPK model predictions for the fraction of intake that is metabolized under
continuous inhalation (A) and oral (B) exposure conditions in mice (white), rats (diagonal
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
hashing), and humans (horizontal hashing). Bars and thin error bars represent the median
estimate and 95% CI for a random subject, and reflect combined uncertainty and variability.
Circles and thick error bars represent the median estimate and 95% CI for the population mean,
and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-18. PBPK model predictions for the fraction of intake that is metabolized by oxidation
(in the liver and lung) under continuous inhalation (A) and oral (B) exposure conditions in mice
(white), rats (diagonal hashing), and humans (horizontal hashing). Bars and thin error bars
represent the median estimate and 95% CI for a random subject, and reflect combined
uncertainty and variability. Circles and thick error bars represent the median estimate and 95%
CI for the population mean, and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-19. PBPK model predictions for the fraction of intake that is metabolized by GSH
conjugation (in the liver and kidney) under continuous inhalation (A) and oral (B) exposure
conditions in mice (dotted line), rats (dashed line), and humans (solid line). X-values are slightly
offset for clarity. Open circles (connected by lines) and thin error bars represent the median
estimate and 95% CI for a random subject, and reflect combined uncertainty and variability.
Filled circles and thick error bars represent the median estimate and 95% CI for the population
mean, and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-20. PBPK model predictions for the fraction of intake that is bioactivated DCVC in the
kidney under continuous inhalation (A) and oral (B) exposure conditions in rats (dashed line) and
humans (solid line). X-values are slightly offset for clarity. Open circles (connected by lines)
and thin error bars represent the median estimate and 95% CI for a random subject, and reflect
combined uncertainty and variability. Filled circles and thick error bars represent the median
estimate and 95% CI for the population mean, and reflect uncertainty only	Error! Bookmark
not defined.
Figure 3-21. PBPK model predictions for fraction of intake that is oxidized in the respiratory
tract under continuous inhalation (A) and oral (B) exposure conditions in mice (dotted line), rats
(dashed line), and humans (solid line). X-values are slightly offset for clarity. Open circles
(connected by lines) and thin error bars represent the median estimate and 95% CI for a random
subject, and reflect combined uncertainty and variability. Filled circles and thick error bars
represent the median estimate and 95% CI for the population mean, and reflect uncertainty only.
	Error! Bookmark not defined.
Figure 3-22. PBPK model predictions for the fraction of intake that is "untracked" oxidation of
TCE in the liver under continuous inhalation (A) and oral (B) exposure conditions in mice
(dotted line), rats (dashed line), and humans (solid line) X-values are slightly offset for clarity.
Open circles (connected by lines) and thin error bars represent the median estimate and 95% CI
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxiii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
for a random subject, and reflect combined uncertainty and variability. Filled circles and thick
error bars represent the median estimate and 95% CI for the population mean, and reflect
uncertainty only	Error! Bookmark not defined.
Figure 3-23. PBPK model predictions for the weekly AUC of TCE in venous blood
(mg-hour/L-week) per unit exposure (ppm or mg/kg-day) under continuous inhalation (A) and
oral (B) exposure conditions in mice (dotted line), rats (dashed line), and humans (solid line).
X-values are slightly offset for clarity. Open circles (connected by lines) and thin error bars
represent the median estimate and 95% CI for a random subject, and reflect combined
uncertainty and variability. Filled circles and thick error bars represent the median estimate and
95% CI for the population mean, and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-24. PBPK model predictions for the weekly AUC of TCOH in blood
(mg-hour/L-week) per unit exposure (ppm or mg/kg-day) under continuous inhalation (A) and
oral (B) exposure conditions in mice (dotted line), rats (dashed line), and humans (solid line).
X-values are slightly offset for clarity. Open circles (connected by lines) and thin error bars
represent the median estimate and 95% CI for a random subject, and reflect combined
uncertainty and variability. Filled circles and thick error bars represent the median estimate and
95% CI for the population mean, and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-25. PBPK model predictions for the weekly AUC of TCA in the liver
(mg-hour/L-week) per unit exposure (ppm or mg/kg-day) under continuous inhalation (A) and
oral (B) exposure conditions in mice (dotted line), rats (dashed line), and humans (solid line).
X-values are slightly offset for clarity. Open circles (connected by lines) and thin error bars
represent the median estimate and 95% CI for a random subject, and reflect combined
uncertainty and variability. Filled circles and thick error bars represent the median estimate and
95% CI for the population mean, and reflect uncertainty only	Error! Bookmark not defined.
Figure 3-26. Sensitivity analysis results: SC for mouse scaling parameters with respect to
dose-metrics following 100 ppm (light bars) and 600 ppm (dark bars), 7 h/day, 5 day/wk
inhalation exposures	Error! Bookmark not defined.
Figure 3-27. Sensitivity analysis results: SC for mouse scaling parameters with respect to
dose-metrics following 300 mg/kg-day (light bars) and 1,000 mg/kg-day (dark bars), 5 day/wk
oral gavage exposures	Error! Bookmark not defined.
Figure 3-28. Sensitivity analysis results: SC for rat scaling parameters with respect to
dose-metrics following 100 ppm (light bars) and 600 ppm (dark bars), 7 h/day, 5 day/wk
inhalation exposures	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxiv DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Figure 3-29. Sensitivity analysis results: SC for rat scaling parameters with respect to
dose-metrics following 300 mg/kg-day (light bars) and 1,000 mg/kg-day (dark bars), 5 day/wk
oral gavage exposures	Error! Bookmark not defined.
Figure 3-30. Sensitivity analysis results: SC for female (light bars) and male (dark bars) human
scaling parameters with respect to dose-metrics following 0.001 ppm continuous inhalation
exposures	Error! Bookmark not defined.
Figure 3-31. Sensitivity analysis results: SC for female (light bars) and male (dark bars) human
scaling parameters with respect to dose-metrics following 0.001 mg/kg-day continuous oral
exposures	Error! Bookmark not defined.
Figure 4-1. Meta-analysis of kidney cancer and overall TCE exposure (the summary estimate is
in the bottom row, represented by the diamond). Random effects model; fixed effect model
same. Symbol sizes reflect relative weights of the studies	Error! Bookmark not defined.
Figure 4-2. Meta-analysis of kidney cancer and TCE exposure—highest exposure groups. With
assumed null RR estimates for Antilla, Axelson, and Hansen (see Appendix C text). Random
effects model; fixed effect model same. The summary estimate is in the bottom row, represented
by the diamond. Symbol sizes reflect relative weights of the studies	Error! Bookmark not
defined.
Figure 4-3. Relative risk estimates of liver and biliary tract cancer and overall TCE exposure.
Random effects model; fixed effect model same. The summary estimate is in the bottom row,
represented by the diamond. Symbol sizes reflect relative weights of the studies	Error!
Bookmark not defined.
Figure 4-4. Meta-analysis of liver cancer and TCE exposure—highest exposure groups, with
assumed null RR estimates for Hansen and Zhao (see Appendix C text). Random effects model;
fixed effect model same. The summary estimate is in the bottom row, represented by the
diamond. Symbol sizes reflect relative weights of the studies	Error! Bookmark not defined.
Figure 4-5. Comparison of average fold-changes in relative liver weight to control and exposure
concentrations of 2 g/L or less in drinking water for TCA and DC A in male B6C3F1 mice for
14-30 days (Carter et al., 1995; DeAngelo et al., 1989; 2008; Kato-Weinstein et al., 2001;
Parrish et al., 1996; Sanchez and Bull, 1990)	Error! Bookmark not defined.
Figure 4-6. Comparisons of fold-changes in average relative liver weight and gavage dose of
(top panel) male B6C3F1 mice for 10-28 days of exposure (Dees and Travis, 1993; Elcombe et
al., 1985; Goldsworthy and Popp, 1987; Merrick et al., 1989) and (bottom panel) in male
B6C3F1 and Swiss mice	Error! Bookmark not defined.
Figure 4-7. Comparison of fold-changes in relative liver weight for data sets in male B6C3F1,
Swiss, and NRMI mice between TCE studies Kjellstrand et al., 1983b (Buben and O'Flaherty,
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxv DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
1985; Goel et al., 1992; Merrick et al., 1989) [duration 28-42 days]) and studies of direct oral
TCA administration to B6C3F1 mice (DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-
Weinstein et al., 2001; Parrish et al., 1996)[duration 14-28 days]). Abscissa for TCE studies
consists of the median estimates of the internal dose of TCA predicted from metabolism of TCE
using the PBPK model described in Section 3.5 of the TCE risk assessment. Lines show linear
regression with intercept fixed at unity. All data were reported fold-change in mean liver
weight/body weight ratios, except for Kjellstrand et al. (1983b), with were the fold-change in the
ratio of mean liver weight to mean body weight. In addition, in Kjellstrand et al. (1983b), some
systemic toxicity as evidence by decreased total body weight was reported in the highest-dose
group	Error! Bookmark not defined.
Figure 4-8. Comparison of hepatomegaly as a function of AUC of TCA in liver, using values for
the TCA drinking water fractional absorption (Fabs). Fold-changes in relative liver weight for
data sets in male B6C3F1, Swiss, and NRMI mice between TCE studies Kjellstrand et al., 1983
(Buben and O'Flaherty, 1985; Goel et al., 1992; Merrick et al., 1989) [duration 28-42 days] and
studies of direct oral TCA administration to B6C3F1 mice (DeAngelo et al., 1989; DeAngelo et
al., 2008; Kato-Weinstein et al., 2001; Parrish et al., 1996) Green, 2003b [duration 14-28 days].
Linear regressions were compared using ANOVA to assess whether the TCE studies were
consistent with the TCA studies, using TCA as the dose-metric. For each analysis of drinking
water fraction absorption, ANOVA /^-values were <10 4 when comparing the assumption that all
the data had a common slope with the assumption that TCE and TCA data had different slopes.
	Error! Bookmark not defined.
Figure 4-9. Fold-changes in relative liver weight for data sets in male B6C3F1, Swiss, and
NRMI mice reported by TCE studies of duration 28-42 days (Kjellstrand et al., 1983b (Buben
and O'Flaherty, 1985; Goel et al., 1992; Merrick et al., 1989) using internal dose-metrics
predicted by the PBPK model described in Section 3.5: (A) dose-metric is the median estimate of
the daily AUC of TCE in blood, (B) dose-metric is the median estimate of the total daily rate of
TCE oxidation. Lines show linear regression. Use of liver oxidative metabolism as a
dose-metric gives results qualitatively similar to (B), with R =0.86	Error! Bookmark not
defined.
Figure 4-10. Dose-response relationship, expressed as (A) percentage incidence and
(B) fold-increase over controls, for TCE hepatocarcinogenicity in NCI (1976). For comparison,
incidences of carcinomas for NTP (1990), Anna et al. (1994), and Bull et al. (2002) are included,
but without connecting lines since they are not appropriate for assessing the shape of the
dose-response relationship	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxvi DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure 4-11. Dose-response relationship, expressed as (A) incidence and (B) fold-increase over
controls, for TCE hepatocarcinogenicity in Maltoni et al. (1986). Note that the BT306
experiment reported excessive mortality due to fighting, and so the paradigm was repeated in
experiment BT306bis using mice from a different source	Error! Bookmark not defined.
Figure 4-12. Dose-response data for hepatocellular carcinomas (HCC) (A) incidence and
(B) multiplicity, induced by DCA from DeAngelo et al. (1999). Drinking water concentrations
were 0, 0.05, 0.5, 1, 2, and 3.5 g/L, from which daily average doses were calculated using
observed water consumption in the study	Error! Bookmark not defined.
Figure 4-13. Reported incidences of hepatocellular carcinomas (HCC) and hepatocellular
adenomas plus carcinomas (HCA + HCC) in various studies in B6C3F1 mice (DeAngelo et al.,
2008; Pereira, 1996). Combined HCA + HCC were not reported in Pereira (1996)	Error!
Bookmark not defined.
Figure 4-14. Reported incidence of hepatocellular carcinomas induced by DCA and TCA in
104-week studies (DeAngelo et al., 2008; DeAngelo et al., 1999). Only carcinomas were
reported in DeAngelo et al. (1999), so combined adenomas and carcinomas could not be
compared	Error! Bookmark not defined.
Figure 4-15. Effects of dietary control on the dose-response curves for changes in liver tumor
incidences induced by CH in diet (Leakey et al., 2003a)	Error! Bookmark not defined.
Figure 4-17. Meta-analysis of NHL and TCE exposure—highest exposure groups. The
summary estimate is in the bottom row. Symbol sizes reflect relative weights of the studies. The
horizontal midpoint of the bottom diamond represents the RRm estimate. Error! Bookmark not
defined.
Figure 5-1. Flow-chart of the process used to derive the RfD and RfC for noncancer effects.
	Error! Bookmark not defined.
Figure 5-2. Flow-chart for dose-response analyses of rodent noncancer effects using PBPK
model-based dose-metrics. Square nodes indicate point values, circle nodes indicate
distributions, and the inverted triangle indicates a (deterministic) functional relationship. ..Error!
Bookmark not defined.
Figure 5-3. Schematic of combined interspecies, intraspecies, and route-to-route extrapolation
from a rodent study LOAEL or NOAEL. In the case where BMD modeling is performed, the
applied dose values are replaced by the corresponding median internal dose estimate, and the
idPOD is the modeled BMDL in internal dose units	Error! Bookmark not defined.
Figure 5-4. Flow-chart for uncertainty analysis of HECs and HEDs derived using PBPK
model-based dose-metrics. Square nodes indicate point values, circle nodes indicate
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxvii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
distributions, and the inverted triangle indicates a (deterministic) functional relationship. ..Error!
Bookmark not defined.
Figure 5-5. Flow-chart for dose-response analyses of rodent bioassays using PBPK model-based
dose-metrics. Square nodes indicate point values, circular nodes indicate distributions, and the
inverted triangles indicate a (deterministic) functional relationship	Error! Bookmark not
defined.
Figure 5-6. Flow-chart for uncertainty analysis of dose-response analyses of rodent bioassays
using PBPK model-based dose-metrics. Square nodes indicate point values, circular nodes
indicate distributions, and the inverted triangles indicate a (deterministic) functional relationship.
	Error! Bookmark not defined.
Figure 5-7. Flow-chart for route-to-route extrapolation of human site-specific cancer inhalation
unit risks to oral slope factors. Square nodes indicate point values, circle nodes indicate
distributions, and the inverted triangle indicates a (deterministic) functional relationship... Error!
Bookmark not defined.
Figure A-l. Hierarchical population statistical model for PBPK model parameter uncertainty
and variability (see Gelman et al., 1996). Square nodes denote fixed or observed quantities;
circle notes represent uncertain or unobserved quantities, and the nonlinear model outputs are
denoted by the inverted triangle. Solid arrows denote a stochastic relationship represented by a
conditional distribution [A^B means B ~ P{B\A)\ while dashed arrows represent a function
relationship [B =f(A)~\. The population consists of subjects z, each of which undergoes one or
more experiments j with exposure parameters Ey with data yl:lki collected at times where k
denotes different types of outputs and / denotes the different time points. The PBPK model
produces outputs /,/ / for comparison with the data>',,/./. The difference between them
("measurement error") has variance a & with a fixed prior distribution Pr, which in this case is
the same for the entire population. The PBPK model also depends on measured covariates (j),
(e.g., body weight) and unobserved model parameters 9/ (e.g., Vmax)- The parameters 9/ are
drawn from a population with mean |j, and variance Z , each of which is uncertain and has a prior
distribution assigned to it	A-2
Figure A-2. Schematic of how posterior predictions were generated for comparison with
experimental data. Two sets of posterior predictions were generated: population predictions
(diagonal hashing) and subject-specific predictions (vertical hashing)	A-8
Figure A-3. Limited optimization results for male closed-chamber data from Fisher et al. (1991)
without (top) and with (bottom) respiratory metabolism	A-45
Figure A-4. Limited optimization results for female closed-chamber data from Fisher et al.
(1991) without (top) and with (bottom) respiratory metabolism	A-46
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxviii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure A-5. Respiratory metabolism model for updated PBPK model	
Figure A-6. Sub-model for TCE gas exchange, respiratory metabolism, and arterial blood
concentration	
A-49
A-67
Figure A-7 Sub-model for TCE oral absorption, tissue distribution, and metabolism
A-69
A-76
A-78
A-79
A-84
Figure A-8. Submodel for TCOH.
Figure A-9. Submodel for TCOG.
Figure A-10. Submodel for TCA.
Figure A-l 1. Submodel for TCE GSH conjugation metabolites
Figure A-12. Updated hierarchical structure for rat and human models. Symbols have the same
meaning as Figure A-l, with modifications for the rat and human. In particular, in the rat, each
"subject" consists of animals (usually comprising multiple dose groups) of the same sex, species,
and strain within a study (possibly reported in more than one publication, but reasonably
presumed to be of animals in the same "lot"). Animals within each subject are presumed to be
"identical," with the same PBPK model parameters, and each such subject is assigned its own set
of "residual" error variances o In humans, each "subject" is a single person, possibly exposed
in multiple experiments, and each subject is assigned a set of PBPK model parameters drawn
from the population. However, in humans, "residual" error variances are assigned at an
intermediate level of the hierarchy—the "study" level, o km—rather than the subject or the
population level	A-107
Figure A-13. Prior and posterior mouse population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 15
Figure A-14. Prior and posterior mouse population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 16
Figure A-l5. Prior and posterior mouse population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 16
Figure A-16. Prior and posterior mouse population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 17
Figure A-17. Prior and posterior mouse population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 18
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxix DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Figure A-18. Prior and posterior mouse population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-l 19
Figure A-19. Prior and posterior rat population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions,, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-128
Figure A-20. Prior and posterior rat population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-129
Figure A-21. Prior and posterior rat population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-130
Figure A-22. Prior and posterior rat population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-131
Figure A-23. Prior and posterior rat population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-132
Figure A-24. Prior and posterior rat population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-133
Figure A-25. Prior and posterior human population mean parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-142
Figure A-26. Prior and posterior human population mean parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-143
Figure A-27. Prior and posterior human population mean parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-144
Figure A-28. Prior and posterior human population variance parameters (Part 1). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-145
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxx DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure A-29. Prior and posterior human population variance parameters (Part 2). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-146
Figure A-30. Prior and posterior human population variance parameters (Part 3). Thick lines are
medians, boxes are interquartile regions, and error bars are (2.5, 97.5%) confidence intervals.
Parameters labeled with have nonoverlapping interquartile regions	A-147
Figure A-31. Comparison of mouse calibration data (boxes) and PBPK model predictions (red
line: using the posterior mean of the subject-specific parameters; + with error bars: single data
points; or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-149
Figure A-32. Comparison of rat calibration data (boxes) and PBPK model predictions (red line:
using the posterior mean of the subject-specific parameters; + with error bars: single data points;
or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-162
Figure A-33. Comparison of rat evaluation data (boxes) and PBPK model predictions (+ with
error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-174
Figure A-34. Comparison of human calibration data (boxes) and PBPK model predictions (red
line: using the posterior mean of the subject-specific parameters; + with error bars: single data
points; or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based predictions)	A-178
Figure A-3 5. Comparison of human evaluation data (boxes) and PBPK model predictions (+
with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-203
Figure A-36. Comparison of Kim et al. (2009) mouse data (boxes) and PBPK model predictions
(+ with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and
97.5% population-based predictions)	A-212
Figure A-3 7. Comparison of best-fitting (out of 50,000 posterior samples) PBPK model
prediction and Kim et al. (2009) TCA blood concentration data for mice gavaged with
2,140 mg/kg TCE	A-213
Figure A-38. Comparison of best-fitting (out of 50,000 posterior samples) PBPK model
prediction and Kim et al. (2009) DCVG blood concentration data for mice gavaged with
2,140 mg/kg TCE	A-214
Figure A-39. PBPK model predictions for the fraction of intake undergoing GSH conjugation in
mice continuously exposed orally to TCE. Lines and error bars represent the median and 95th
percentile confidence interval for the posterior predictions, respectively (also reported in
Section 3.5.7.2.1). Filled circles represent the predictions from the sample (out of 50,000 total
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxi DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
posterior samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG blood
concentration data for mice gavaged with 2,140 mg/kg TCE	A-215
Figure A-40. PBPK model predictions for the fraction of intake undergoing GSH conjugation in
mice continuously exposed via inhalation to TCE. Lines and error bars represent the median and
95th percentile confidence interval for the posterior predictions, respectively (also reported in
Section 3.5.7.2.1). Filled circles represent the predictions from the sample (out of 50,000 total
posterior samples) which provides the best fit to the Kim et al. (2009) TCA and DCVG blood
concentration data for mice gavaged with 2,140 mg/kg TCE	A-216
Figure A-41. Comparison of Liu et al. (2009) rat data (boxes) and PBPK model predictions (+
with error bars: single data points or shaded regions: 2.5, 25, 50, 75, and 97.5% population-based
predictions)	A-219
Figure A-42. Assumed drinking water patterns as a function of time since beginning of
exposure. The upper left panel (LH) assumes that t = 0 is at the beginning of the "light" part of
the "light/dark" cycle (light is dashed grey line at the bottom, dark is thick black line at the
bottom). The upper right panel (LHL) assumes that t = 0 is in the middle of the "light" part of
the cycle. The lower left panel (HL) assumes that t = 0 is at the end of the "light" part of the
cycle	A-223
Figure A-43. PBPK model predictions for TCA in blood and liver of male B6C3Fi mice from
Mahle et al. (2001). Three- and 14-day exposures to 0.08 (data: open circles, predictions: solid
line), 0.8 (data: open triangle, predictions: dashed line), and 2 g/L TCA in drinking water (data:
crosses, predictions: dotted line). Predictions use a representative parameter sample from the
converged MCMC chain for the LHL drinking water intake pattern	A-225
Figure A-44. PBPK model predictions for TCA in blood and liver of male B6C3F1 mice from
Green (2003a, 2003b). Green (2003a): 5-day drinking water exposures to 0.5 (data: open circle;
predictions: solid line), 1 (data: open triangle; predictions: dashed line), and 2.5 g/L TCA (data:
crosses; predictions: dotted lines). Green (2003b): 5- and 14-day drinking water exposures to 1
(data: open circle; predictions: solid line) and 2.5 g/L TCA (data: open triangle; predictions:
dashed line). Predictions use a representative parameter sample from the converged MCMC
chain for the LHL drinking water intake pattern	A-226
Figure A-45. Distribution of fractional absorption fit to each TCA drinking water kinetic study
group in mice, using LHL drinking water intake patterns. Fits are to a Michaelis-Menten
function for "effective" concentration Ceff = Cmax x C/(Cy2 + C), so that the fractional absorption
Fabs = Ceff/C = Cmax/(Ci/2 + C). Sweeney et al. (2009) estimates of Fabs, along with a Michaelis-
Menten fit, are included for comparison. The ratio Cmax/Cu gives the fractional uptake at low
concentrations	A-228
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
Figure C-l. Meta-analysis of NHL and overall TCE exposure. The summary estimate is in the
bottom row. Symbol sizes reflect relative weights of the studies. The horizontal midpoint of the
bottom diamond represents the summary RR estimate	C-16
Figure C-3. Cumulative meta-analysis of TCE and lymphomaNHL studies, progressively
including studies with increasing SEs	C-20
Figure C-4. Meta-analysis of NHL and TCE exposure—highest exposure groups. The summary
estimate is in the bottom row. Symbol sizes reflect relative weights of the studies. The
horizontal midpoint of the bottom diamond represents the summary RR estimate	C-28
Figure C-5. Meta-analysis of kidney cancer and overall TCE exposure. Random-effects model;
fixed-effect model same. The summary estimate is in the bottom row, represented by the
diamond. Symbol sizes reflect relative weights of the studies	C-44
Figure C-8. Meta-analysis of liver cancer and TCE exposure. Random-effects model; fixed-
effect model same. The summary estimate is in the bottom row, represented by the diamond.
Symbol sizes reflect relative weights of the studies	C-66
Figure C-9. Funnel plot of SE by log RR estimate for TCE and liver cancer studies	C-68
Figure C-10. Meta-analysis of liver cancer and TCE exposure—highest exposure groups, with
assumed null RR estimates for Hansen and Zhao (see text). Random-effects model; fixed-effect
model same. The summary estimate is in the bottom row, represented by the diamond. Symbol
sizes reflect relative weights of the studies	C-73
Figure E-l. Comparison of average fold-changes in relative liver weight to control and exposure
concentrations of 2 g/L or less in drinking water for TCA and DC A in male B6C3F1 mice for
14-30 days (Carter et al., 1995; DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-Weinstein et
al., 2001; Parrish et al., 1996; Sanchez and Bull, 1990)\. (Reproduced from Section 4.5.). Error!
Bookmark not defined.
Figure E-2. Comparisons of fold-changes in average relative liver weight and gavage dose of
(top panel) male B6C3F1 mice for 10-28 days of exposure (Dees and Travis, 1993; Elcombe et
al., 1985; Goldsworthy and Popp, 1987; Merrick et al., 1989) and (bottom panel) in male
B6C3F1 and Swiss mice. (Reproduced from Section 4.5.)	Error! Bookmark not defined.
Figure E-3. Comparison of fold-changes in relative liver weight for data sets in male B6C3F1,
Swiss, and NRMI mice between TCE studies (Buben and O'Flaherty, 1985; Goel et al., 1992;
Kjellstrand et al., 1983a; Merrick et al., 1989) [duration 28-42 days] and studies of direct oral
TCA administration to B6C3 F1 mice (DeAngelo et al., 1989; DeAngelo et al., 2008; Kato-
Weinstein et al., 2001; Parrish et al., 1996) [duration 14-28 days]. Abscissa for TCE studies
consists of the median estimates of the internal dose of TCA predicted from metabolism of TCE
using the PBPK model described in Section 3.5 of the TCE risk assessment. Lines show linear
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxiii DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
regression with intercept fixed at 1. All data were reported fold-change in mean liver
weight/body weight ratios, except for Kjellstrand et al. (1983a), with were the fold-change in the
ratio of mean liver weight to mean body weight. In addition, in Kjellstrand et al. (1983a), some
systemic toxicity as evidence by decreased total body weight was reported in the highest dose
group. (Reproduced from Section 4.5.)	Error! Bookmark not defined.
Figure E-4. Fold-changes in relative liver weight for data sets in male B6C3F1, Swiss, and
NRMI mice reported by TCE studies of duration 28-42 days (Buben and O'Flaherty, 1985; Goel
et al., 1992; Kjellstrand et al., 1983a; Merrick et al., 1989) using internal dose metrics predicted
by the PBPK model described in Section E.3.5: (A) dose metric is the median estimate of the
daily AUC of TCE in blood, (B) dose metric is the median estimate of the total daily rate of TCE
oxidation. Lines show linear regression. Use of liver oxidative metabolism as a dose metric
gives results qualitatively similar to (B), with R = 0.86. (Reproduced from Section 4.5.). Error!
Bookmark not defined.
Figure E-5. Comparison of Ito et al. and David et al. data for DEHP tumor induction from
(Guyton et al., 2009)	Error! Bookmark not defined.
Figure F-l. Regression of TCE in air (ppm) and TCA in urine (mg/g creatinine) based on data
from Ikeda et al. (1972)	F-4
Figure F-2. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
applied dose, without LSC, with IC, and without the high-dose group, using a BMR of 0.05 extra
risk (top panel) or 0.01 extra risk (bottom panel)	F-14
Figure F-3. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
TotOxMetabBW34 dose metric, without LSC, with IC, and without the high-dose group, using a
BMR of 0.01 extra risk	F-16
Figure F-4. BMD modeling of Johnson et al. (2003) using nested log-logistic model, with
AUCCBld dose metric, without LSC, with IC, and without the high-dose group, using a BMR of
0.01 extra risk	F-17
Figure F-5. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested log-
logistic model, with applied dose, with both LSC and IC, using a BMR of 0.05 extra risk	F-20
Figure F-6. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested log-
logistic model, with applied dose, without either LSC or IC, using a BMR of 0.05 extra risk. F-21
Figure F-7. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested Rai-
VanRyzin model, with applied dose, without either LSC or IC, using a BMR of 0.05 extra risk.F-
22
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxiv DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure F-8. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested log-
logistic model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
panel) or 0.01 extra risk (bottom panel)	F-24
Figure F-9. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested Rai-
VanRyzin model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
panel) or 0.01 extra risk (bottom panel)	F-25
Figure F-10. BMD modeling of NTP (1988) toxic nephropathy in female Marshall rats	F-34
Figure F-l 1. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from NTP
(1988) toxic nephropathy in rats	F-34
Figure F-12. BMD modeling of Woolhiser et al. (2006) for increased kidney weight in female S-
D rats	F-36
Figure F-13. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from
Woolhiser et al. (2006) for increased kidney weight in rats	F-38
Figure F-14. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from
Keil et al. (2009) for decreased thymus weight in mice	F-39
Figure F-15. Derivation of HEC99 and HED99 corresponding to the rodent idPOD from Johnson
et al. (2003) for increased fetal cardiac malformations in female S-D rats using the total oxidative
metabolism dose metric	F-40
Figure G-l. Female B6C3F1 mice—applied doses: combined and individual tumor extra-risk
functions	G-55
Figure G-2. Female B6C3F1 mice—applied doses: posterior distribution of BMDc for combined
risk	G-55
Figure G-3. B6C3F1 female mice inhalation exposure—applied doses: combined and individual
tumor extra-risk functions	G-59
Figure G-4. B6C3F1 female mice inhalation exposure—applied doses: posterior distribution of
BMDc for combined risk	G-59
Figure G-5. Maltoni Sprague-Dawley male rats—applied doses: combined and individual tumor
extra-risk functions	G-63
Figure G-6. Maltoni Sprague-Dawley male rats—applied doses: posterior distribution of BMDc
for combined risk	G-63
Figure G-l. Female B6C3F1 mice—internal dose metric (total oxidative metabolism): combined
and individual tumor extra-risk functions	G-66
Figure G-8. Female B6C3F1 mice—internal dose metric (total oxidative metabolism): posterior
distribution of BMDc for combined risk	G-66
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxv DRAFT-DO NOT CITE OR QUOTE
-------
1	Figure G-9. B6C3F1 female mice inhalation exposure—internal dose metric: combined and
2	individual tumor extra-risk functions	G-70
3	Figure G-10. B6C3F1 female mice inhalation exposure—internal dose metric: posterior
4	distribution of BMDc for combined risk	G-70
5	Figure G-l 1. Maltoni Sprague-Dawley male rats—internal dose metric: combined and
6	individual tumor extra-risk functions	G-74
7	Figure G-12. Maltoni Sprague-Dawley male rats—internal dose metric: posterior distribution of
8	BMDc for combined risk	G-74
This document is a draft for review purposes only and does not constitute Agency policy.
G-xxxvi DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
G.l. DATA SOURCES
Trichloroethylene (TCE) cancer endpoints were identified in Maltoni et al. (1986),
National Cancer Institute (1976), National Toxicology Program (1988, 1990), Fukuda et al.
(1983), and Henschler et al. (1980). These data were reviewed and tabulated in spreadsheets,
and the numbers were verified. All endpoint data identified by authors as having a statistically
significant response to dose were tabulated, and data that had marginally significant trends with
dose were also reviewed. For all endpoints for which dose-response model estimates were
presented, trends were verified using the Cochran-Armitage or the Poly-3 test.
G.l .l. Numbers at Risk
The numbers of animals at risk are not necessarily those used by the authors; instead, as
the number at risk, the number alive at 52 weeks was used (if the first cancer of the type of
interest was observed at later than 52 weeks) or the number alive at the week when the first
cancer of the type of interest was observed. In general, the data of Maltoni et al. (1986) were
presented in this way, in their tables titled "Incidence of the different types of tumors referred to
specific corrected numbers." In a few cases in Maltoni et al. (1986), the time of first occurrence
was later than 52 weeks, so an alternative number at risk was used from another column (for
another cancer) in the same table having a first occurrence close to 52 weeks. For NTP (1988,
1990) and for NCI (1976), the week of the first observation and the numbers alive at that week
were determined from the appendix tables. For Fukuda et al. (1983), the reported "effective
number of mice" in their Table 2 was used, which is consistent with numbers alive at
40-42 weeks (when the first tumor, a thymic lymphoma, was observed) in their mortality curve.
For Henschler et al. (1980), the number of mice alive at Week 36 (from their Figure 1), which is
when the first tumor was observed (according to their Figure 2), was used.
In cases in which there is high early mortality or differential mortality across dose groups
and the individual animal data are available, a more involved analysis which takes into account
animals at risk at different times (ages) is preferred (e.g., the poly-3 approach or time-to-tumor
modeling; see G.7). The more rudimentary approach of adjusting the denominator to account for
animals alive at the time of the first tumor entails some inaccuracy (bias) in estimating the
animals at risk compared to a more involved analysis accounting more completely for time.
This document is a draft for review purposes only and does not constitute Agency policy.
G-3 7 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
However, it is generally agreed that is is better to use such an adjustment than to use no
adjustment at all (Gart et al., 1979; Haseman et al., 1984; Hoel and Walburg, 1972).
G.1.2. Cumulative Incidence
Maltoni et al. (1986) conducted a lifetime study, in which rodents were exposed for
104 weeks (rats) or 78 weeks (mice), and allowed to live until they died "naturally." Maltoni
et al. (1986) reported cumulative incidence on this basis, and it was not possible for us to
determine incidence at any fixed time such as 104 weeks on study. For Henschler et al. (1980),
the number of mice with tumors observed by Week 104 (their Figure 2) was used. The
cumulative incidence reported by Fukuda et al. (1983) at 107 weeks (after 104 weeks of
exposure) was used. For the NCI (1976) and NTP (1988, 1990) studies, the reported cumulative
incidence at 103 to 107 weeks (study time varied by study and species) was used.
G.2. INTERNAL DOSE METRICS AND DOSE ADJUSTMENTS
Physiologically based pharmacokinetic (PBPK) modeling was used to estimate levels of
dose metrics corresponding to different exposure scenarios in rodents and humans (see
Section 3.5). The selection of dose metrics for specific organs and endpoints is discussed under
Section 5.2. Internal dose metrics were selected based on applicability to each major affected
organ. The dose metrics used with our cancer dose-response analyses are shown in Table G-l.
Table G-l. Internal dose metrics used in dose-response analyses, identified
by "X"
Dose metric units
Liver
Lung
Kidney
Other
ABioactDCVCBW34 (mg/wk-kg3/4)
0
0
X
0
AMetGSHBW3 4 (mg/wk-kg3/4)
0
0
X
0
AMetLivlBW34 (mg/wk-kg3/4)
X
0
0
0
AMetLngBW34 (mg/wk-kg3 4)
0
X
0
0
AUCCBld (mg-hr/L-wk)
0
X
0
X
TotMetabBW34 (mg/wk-kg3/4)
0
0
X
X
TotOxMetabBW34 (mg/wk-kg3/4)
X
X
0
0
This document is a draft for review purposes only and does not constitute Agency policy.
G-3 8 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
The PBPK model requires the rodent body weight as an input. For most of the studies,
central estimates specific to each species, strain, and sex (and substudy) were used. These were
estimated by medians of body weights digitized from graphics in Maltoni et al. (1986), by
medians of weekly averages in NTP (1988, 1990), and by averages over the study duration of
weekly mean body weights tabulated in NCI (1976).
For the studies by Fukuda et al. (1983) and Henschler et al. (1980), mouse body weights
were not available. After reviewing body weights reported for similar strains by two
laboratories 14 and in the other studies reported for TCE, it was concluded that a plausible range
for lifetime average body weight is 20-35 g, with a median near 28 g. For these two studies,
internal dose metrics for these three average body weights (20, 28, and 35 g) were computed.
The percentage differences between the internal dose metrics for the intermediate body weight
(BW) of 28 g and the low and high average BW of 20 gm and 35 g were then evaluated. Internal
dose metrics were little affected by choice of body weight. For all dose metrics, the differences
were less than ±13%. A body weight of 28 g was used for these two studies.
The medians (from the Markov chain Monte Carlo posterior distribution) for each of the
dose metrics for the rodent were used in quantal dose-response analyses. The median is probably
the most appropriate posterior parameter to use as a dose metric, as it identifies a "central"
measure and it is also a quantile, making it more useful in nonlinear modeling. The "multistage"
dose-response functions are nonlinear. One is interested in estimating the expected response.
The expected value of a nonlinear function of dose is under- or overestimated when the mean
(expected value) of the dose is used, depending on whether the function is concave or convex.
(This is Jensen's Inequality: for a real convex function f(X), f[E(X)] < E[f(X)].) For the
dose-response function, one is interested in E[f(X)], so using E(X) (estimated by the posterior
mean) as the dose metric will not necessarily predict the mean response. Using the posterior
median rather than the mean as the dose metric should lead to a response function that is closer
to the median response. However, if the estimated dose-response function is close to linear, this
source of distortion may be small, and the mean response might be predicted reasonably well by
using the posterior mean as the dose metric. The mean and median are expected to be rather
different because the posterior distributions are skewed and approximately lognormal.
Therefore, results based on the posterior median and the posterior mean dose metric were
compared before deciding to use the median.
14http ://phenome .i ax. org/pub -
cgi/phenome/mpdcgi?rtn=meas%2Fdatalister&rea=Cbodv+weight&pan=2&noomit=&datamode=measavg.
http://www.hilltoplabs.com/public/growth.html.
This document is a draft for review purposes only and does not constitute Agency policy.
G-3 9 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
G.3. DOSE ADJUSTMENTS FOR INTERMITTENT EXPOSURE
The nominal applied dose was adjusted for exposure discontinuity (e.g., exposure for
5 days per week and 6 hours per day reduced the dose by the factor [(5/7) * (6/24)]), and for
exposure durations less than full study time (up to 2 years) (e.g., the dose might be reduced by a
factor [78 wk/104 wk]). The PBPK dose metrics took into account the daily and weekly
discontinuity to produce an equivalent dose for continuous exposure. The NCI (1976) gavage
study applied one dose for weeks 1-12 and another, slightly different dose for weeks 13-78;
PBPK dose metrics were produced for both dose regimes and then time-averaged (e.g., average
dose = (12/78) x D1 + (66/78) x D2). For Henschler et al. (1980), Maltoni et al. (1986), and NCI
(1976), a further adjustment of (exposure duration/study duration) was made to account for the
fact that exposures ended prior to terminal sacrifice, so that the dose metrics reflect average
weekly values over the exposure period. Finally, for NCI (1976), the dose metrics were then
adjusted for early sacrificel5 (at 91 weeks rather than 104 weeks) by a factor of (91 wk/104
wk)3.16
G.4. RODENT TO HUMAN DOSE EXTRAPOLATION
Adjustments for rodent-to-human extrapolation were applied to the final results—the
benchmark dose (BMD), benchmark dose lower bound (BMDL), and cancer slope factor
(potency), which is calculated as benchmark response (BMR)/BMDL, e.g., 0.10/BMDLio.
For the PBPK dose metrics, a ratio between human and laboratory animal internal dose
was determined by methods described in Section 3.5. The cancer slope factor is relevant only for
very low extra risk (typically on the order of 10 4 to 10 6), thus very low dose, and it was
determined that the relation between human and animal internal dose was linear in the low-dose
range for each of the dose metrics used, hence this ratio was multiplied by the animal dose (or
divided into the cancer slope factor) to extrapolate animal to human dose or concentration.
For the experimentally applied dose, default interspecies extrapolation approaches were
used. These are provided for comparison to results based on PBPK metrics. To extrapolate
animal inhalation exposure to human inhalation exposure, the "equivalent" human exposure
concentration (i.e., the exposure concentration in humans that is expected to give the same level
15For studies of less than 2 years (i.e., with terminal kills before 2 years), the doses are generally adjusted by the
study length ratio to a power of three (i.e., a factor [length of study in wk/104 wk]3) to reflect the fact that the
animals were not observed for the full standard lifetime (1980).
16For studies of less than 2 years (i.e., with terminal kills before 2 years), the doses are generally adjusted by the
study length ratio to a power of three (i.e., a factor [length of study in wk/104 wk]3) to reflect the fact that the
animals were not observed for the full standard lifetime (1980).
This document is a draft for review purposes only and does not constitute Agency policy.
G-40 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
of response that was observed in the test species) was assumed to be identical to the animal
inhalation exposure concentration, i.e., "ppm equivalence." This assumption is consistent with
U.S. Environmental Protection Agency recommendations (U.S. EPA, 1984) for deriving a
human equivalent concentration for a Category 3 gas for which the blood:air partition coefficient
in laboratory animals is greater than that in humans (see Section 3.1 for discussion of the TCE
blood:air partition coefficient). To extrapolate animal oral exposure to equivalent human oral
exposure, animal dose was scaled up by body weight to the 3/4-power using the factor
0 75
(BWHumar/BWAnimal) • To extrapolate animal inhalation exposure to human oral exposure, the
following equation (Eq. G-l) was used; 17
Animal, equivalent oral intake, mg/kg/d =
ppm * [MJFtc£/24.45]18 * MV* (60 min/hr) * (103 mg/g) * [24 hrIBWkg\ (Eq. G-l)
with units
ppm * [g/mol ^ L/mol ] * L/min * (min/hr) * (mg/g) * [hr/day ^ kg]	(Eq. G-2)
which reduces to
ppm * [7.738307 * MV/BWkg]	(Eq. G-3)
where
ppm = animal inhalation concentration, 1/106, unitless
MV = minute volume (breathing rate) at rest, L/minute.
Minute volume (MV) was estimated using equations from U.S. EPA (1994, p. 4-27),
Mouse ln(MV) = 0.326 + 1.05 * ln(BWkg)
Rat ln(MV) = -0.578 + 0.821 * ln(BWkg)
Animal equivalent oral intake was converted to human equivalent oral intake by
multiplying by the rodent to human ratio of body weights to the power +0.25.19
17ToxRisk version 5.3, © 2000-2001 by the KS Crump Group, Inc.
18Molecular weight of TCE is 131.39; there are 24.45 L of perfect gas per g-mol at standard temperature and
pressure (U.S. EPA, 1994b).
This document is a draft for review purposes only and does not constitute Agency policy.
G-41 DRAFT-DO NOT CITE OR QUOTE
(Eq. G-4)
(Eq. G-5)
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
To extrapolate animal oral exposure to equivalent human inhalation exposure, the
calculation above was reversed to extrapolate the animal inhalation exposure.
G.5. COMBINING DATA FROM RELATED EXPERIMENTS IN MALTONI ET AL.
(1986)
Data from Maltoni et al. (1986) required decisions by us regarding whether to combine
related experiments for certain species and cancers.
In experiment BT306, which used B6C3F1 mice, males experienced unusually low
survival, reportedly because of the age of the mice at the outset and resulting aggression. The
protocol was repeated (for males only), with an earlier starting age, as experiment BT306bis, and
male survival was higher (and typical for such studies). The rapid male mortality in experiment
BT306 apparently censored later-developing cancers, as suggested by the low frequency of liver
cancers for males in BT306 as compared to BT306bis. Data for the two experiments clearly
cannot legitimately be combined. Therefore only experiment BT306bis males were used in the
analyses.
Experiments BT304 and BT304bis, on rats, provide evidence in male rats of leukemia,
carcinomas of the kidney, and testicular (Leydig cell) tumors, and provide evidence in female
rats for leukemia. Maltoni et al. (1986) stated "Since experiments BT 304 and BT 304bis on
Sprague-Dawley rats were performed at the same time, exactly in the same way, on animals of
the same breed, divided by litter distribution within the two experiments, they have been
evaluated separately and comprehensively." The data were also analyzed separately and in
combination.
The data and modeling results for these tumors in the BT304 and BT304bis experiments
are tabulated in Tables G-2 through G-5, below. It was decided that it was best to combine the
data for the two experiments. There were no consistent differences between experiments, and no
firm basis for selecting one of them. Our final analyses are, therefore, based on the combined
numbers and tumor responses for these two experiments.
19Find whole animal intake from mg/kg/d * BW^i. Scale this allometrically by (BWH,mn/BWjn_i)07S to
extrapolate whole human intake. Divide by human body weight to find mg/kg/d for the human. The net effect is
Animal mg/kg/d * (BWAnimai/BWHuman)A0.25 = Human mg/kg/d.
This document is a draft for review purposes only and does not constitute Agency policy.
G-42 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
G.6. DOSE-RESPONSE MODELING RESULTS
Using BenchMark Dose Software (BMDS), the multistage quantal model was fitted using
the applicable dose metrics for each combination of study, species, strain, sex, organ, and BMR
(extra risk) value under consideration. A multistage model of order one less than the number of
dose groups (g) was fitted. This means that in some cases the fitted model could be strictly
nonlinear at low dose (estimated coefficient "bl" was zero), and in other cases, higher-order
coefficients might be estimated as zero so the resulting model would not necessarily have order
(#groups-l). Because more parsimonious, lst-order models often fit such data well, based on our
extensive experience and that of others (Nitcheva et al., 2007), if the resulting model was not a
lst-order multistage, then lower-order models were also fitted, down to a lst-order multistage
model. This permitted us to screen results efficiently.
This document is a draft for review purposes only and does not constitute Agency policy.
G-43 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-2. Experiments BT304 and BT304bis, female Sprague-Dawley rats,
2	Maltoni et al. (1986). Number alive is reported for week of first tumor
3	observation in either males or females.3 These data were not used for
4	dose-response modeling because there is no consistent trend (for the combined
5	data, there is no significant trend by the Cochran-Armitage test, and no significant
6	differences between control and dose groups by Fisher's exact test).
7
Exposure
Concen.
(ppm)
No.
alive
No. rats
with this
cancer
Proportion
with
cancer
Multistage model fit statisticsb
Model
order
p-Value
AIC
BMDio
BMDL10

Experiment BT304, female rats, leukemias, N alive at 7 weeks
0
105
7
0.067
No adequately fitting model
100
90
6
0.067





300
90
0
0.000





600
90
7
0.078






Experiment BT304bis, female rats, leukemias, N alive at 7 weeks
0
40
0
0.000
1
0.202
70.4
127
58.7
100
40
3
0.075





300
40
2
0.050





600
40
4
0.100






Experiments BT304 and BT304bis, female rats, leukemias, combined data
0
145
7
0.048
3
0.081
227
180
134
100
130
9
0.069





300
130
2
0.015





600
130
11
0.085





9	11 First tumor occurrences were not reported separately by sex.
10	b Models of orders 3 were fitted; the highest-order nonzero coefficient is reported in column "Model order."
11	BMDL was estimated for extra risk of 0.10 and confidence level 0.95. Exposure concentrations were multiplied
12	by (7/24) * (5/7) = 0.20833 before fitting the models, to adjust for exposure periodicity (i.e., the time-averaged
13	concentrations were about 20% of the nominal concentrations).
14
15	AIC - Akaike Information Criteria.
16
This document is a draft for review purposes only and does not constitute Agency policy.
G-44 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-3. Experiments BT304 and BT304bis, male Sprague-Dawley rats,
2	Maltoni et al. (1986): leukemias. Number alive is reported for week of first
3	tumor observation in either males or females.a
4
Exposure
concen.
(ppm)
No.
alive
No. rats
with this
cancer
Proportion
with
cancer
Multistage model fit statisticsb
Model
order
/j-Value
AIC
BMDio
BMDL10

Experiment BT304, male rats, leukemias, Nalive at 7 weeks
0
95
6
0.063
1
0.429
238
NA
NA
100
90
10
0.111





300
90
11
0.122





600
89
9
0.101






Experiment BT304bis, male rats, leukemias, Nalive at 7 weeks
0
39
3
0.077
3
0.979
102
143
71.9
100
40
3
0.075





300
40
3
0.075





600
40
6
0.150






Combined data for BT304 and BT304bis, male rats, leukemias
0
134
9
0.067
1
0.715
337
269
111
100
130
13
0.100





300
130
14
0.108





600
129
15
0.116





5
6	aFirst tumor occurrences were not reported separately by sex.
7	bModels of orders 3 were fitted; the highest-order nonzero coefficient is reported in column "Model order." BMDL
8	was estimated for extra risk of 0.10 and confidence level 0.95. Exposure concentrations were multiplied by
9	(7/24)*(5/7) = 0.20833 before fitting the models, to adjust for exposure periodicity (i.e., the time-averaged
10	concentrations were about 20% of the nominal concentrations). "NA" indicates the BMD or BMDL could not be
11	solved because it exceeded the highest dose.
12
13	AIC—Akaike Information Criteria.
14
This document is a draft for review purposes only and does not constitute Agency policy.
G-45 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-4. Experiments BT304 and BT304bis, male Sprague-Dawley rats,
2	Maltoni et al. (1986): kidney adenomas + carcinomas. Number alive is
3	reported for week of first tumor observation in either males or females.3
4
Exposure
concen.
(ppm)
No.
alive
No. rats
with this
cancer
Proportion
with cancer
Multistage model fit statisticsb
Model
order
p-Value
AIC
BMDio
BMDL10

Experiment BT304 male rats, kidney adenomas + carcinomas, N alive at 47 weeks
0
87
0
0.000
3
0.318
50.1
173
134
100
86
1
0.012





300
80
0
0.000





600
85
4
0.047






Experiment BT304bis, male rats, kidney adenomas + carcinomas, N alive at 53
weeks
0
34
0
0.000
3
0.988
13.0
266
173
100
32
0
0.000





300
36
0
0.000





600
38
1
0.027






Combined data for BT304 and BT304bis, male rats, kidney adenomas + carcinomas
0
121
0
0.000
3
0.292
60.5
181
144
100
118
1
0.008





300
116
0
0.000





600
123
5
0.041





5
6	a First tumor occurrences were not reported separately by sex.
7	b Models of orders three were fitted; the highest-order nonzero coefficient is reported in column "Model order."
8	BMDL was estimated for extra risk of 0.10 and confidence level 0.95. Exposure concentrations were multiplied by
9	(7/24)*(5/7) = 0.20833 before fitting the models, to adjust for exposure periodicity (i.e., the time-averaged
10	concentrations were about 20% of the nominal concentrations). "NA" indicates the BMD or BMDL could not be
11	solved because it exceeded the highest dose.
12
13	AIC - Akaike Information Criteria.
14
This document is a draft for review purposes only and does not constitute Agency policy.
G-46 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-5. Experiments BT304 and BT304bis, male Sprague-Dawley rats,
2	Maltoni et al. (1986): testis, Leydig cell tumors. Number alive is reported for
3	week of first tumor observation.3
4
Exposure
concen.
(ppm)
No.
alive
No. rats
with this
cancer
Proportion
with
cancer
Multistage model fit statisticsb
Model
order
p-Value
AIC
BMDio
BMDL10

Experiment BT304, male rats, Leydig cell tumors, N alive at 47 weeks
0
87
5
0.057
1
0.0494
309
41.5
29.2
100
86
11
0.128





300
80
24
0.300





600
85
22
0.259






Experiment BT304bis, male rats, Leydig cell tumors, Nalive at 53 weeks
0
34
1
0.029
1
0.369
117
54.5
30.9
100
32
5
0.156





300
36
6
0.167





600
38
9
0.237






Combined data for BT304 and BT304bis, male rats, Leydig cell tumors
0
121
6
0.050
1
0.0566
421
44.7
32.7
100
116
16
0.138





300
116
30
0.259





600
122
31
0.254





5
6	" Numbers alive reported for weeks as close as possible to Week 52 (first tumors observed at weeks 81, 62,
7	respectively, for the two experiments).
8	b Models of orders three were fitted; the highest-order nonzero coefficient is reported in column "Model order."
9	BMDL was estimated for extra risk of 0.10 and confidence level 0.95. Exposure concentrations were multiplied
10	by (7/24)*(5/7) = 0.20833 before fitting the models, to adjust for exposure periodicity (i.e., the time-averaged
11	concentrations were about 20% of the nominal concentrations). "NA" indicates the BMD or BMDL could not be
12	solved because it exceeded the highest dose.
13
14	AIC - Akaike Information Criteria.
15
This document is a draft for review purposes only and does not constitute Agency policy.
G-47 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
The document shows the fitted model curves. The graphics include observations (as
proportions, i.e., cumulative incidence divided by number at risk), the estimated multistage curve
(solid red line) and estimated BMD, with a BMDL. Vertical bars show 95% confidence intervals
for the observed proportions. Printed above each plot are some key statistics (necessarily
rounded) for model goodness of fit and estimated parameters. Printed in the plots at upper left
are the BMD and BMDL for the rodent data, in the same units as the rodent dose. Within the
plot at lower right are human exposure values (BMDL and cancer slope factor for continuous
inhalation and oral exposures) corresponding to the rodent BMDL. For applied doses, the human
equivalent
values were calculated by "default" methods,20 as discussed above, and then only for the same
route of exposure as the rodent, and they are in units of rodent dose. For internal dose metrics,
the human values are based upon the PBPK rodent-to-human extrapolation, as discussed in
Section 5.2.1.2.
The document presents the data and model summary statistics, including goodness-of-fit
measures (Chi-square goodness-of-fit/> value, Akaike Information Criteria [AIC]), parameter
estimates, BMD, BMDL, and "cancer slope factor" ("CSF"), which is the extra risk divided by
the BMDL. Much more descriptive information appears also, including the adjustment terms for
intermittent exposure, and the doses before applying those adjustments. The group "GRP"
numbers are arbitrary, and are the same as GRP numbers in the plots. There is one line in this
table for each dose-response graph in the preceding document. Input data for the analyses are in
the file . Finally, the values and model selections for the results used in Section 5.2 are
summarized in the file (primary dose metrics in bold).
G.7. MODELING TO ACCOUNT FOR DOSE GROUPS DIFFERING IN SURVIVAL
TIMES
Differential mortality among dose groups can potentially interfere with (i.e., censor) the
occurrence of late-appearing cancers. Usually the situation is one of greater mortality rates at
higher doses, caused by toxic effects, or, sometimes, by cancers other than the cancer of interest.
Statistical methods of estimation (for the cancer of interest) in the presence of competing risks
assume uninformative censoring.
For bioassays with differential early mortality occurring primarily before the time of the
1st tumor or 52 weeks (whichever came first), the effects of early mortality were largely
20For oral intake, dose (BMDL) is multiplied by the ratio of animal to human body weight (60 kg female, 70 kg
male) taken to the 'A power. For inhalation exposures, ppm equivalence is assumed.
This document is a draft for review purposes only and does not constitute Agency policy.
G-48 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
accounted for by adjusting the tumor incidence for animals at risk, as described above, and the
dose-response data were modeled using the multistage model.
If, however, there was substantial overlap between the appearances of cancers and
progressively differential mortality among dose groups, it was necessary to apply methods that
take into account individual animal survival times. Two such methods were used here:
time-to-tumor modeling and the poly-3 method of adjusting numbers at risk. Three such studies
were identified, all with male rats (see Table 5-27). Using both survival-adjustment approaches,
BMDs and BMDLs were obtained and unit risks derived. Section 5.2.1.3 presents a comparison
of the results for the three data sets and for various dose metrics.
G.7.1. Time-to-Tumor Modeling
The first approach used to take into account individual survival times was application of
the multistage Weibull (MSW) time-to-tumor model. This model has the general form
P(d,t) = 1 - exp[-(<7o + q\d + q2d1 + ... + q\J) * (t- /0)z],	(Eq. G-6)
where P(dj) represents the probability of a tumor by age t for dose and parameters z > 1,
to > 0, and q, > 0 for i = 0,1,...,A:, where k = the number of dose groups; the parameter /0
represents the time between when a potentially fatal tumor becomes observable and when it
causes death. The MSW model likelihood accounts for the left-censoring inherent in
"Incidental" observations of nonfatal tumors discovered upon necropsy and the right-censoring
inherent in deaths not caused by fatal tumors. All of our analyses used the model for incidental
tumors, which has no to term, and which assumes that the tumors are nonfatal (or effectively so,
to a reasonable approximation). This seems reasonable because the tumors of concern appeared
relatively late in life and there were multiple competing probable causes of death (especially
toxic effects) operating in these studies (also note that cause of death was not reported by the
studies used). It is difficult to formally evaluate model fit with this model because there is no
applicable goodness-of-fit statistic with a well-defined asymptotic distribution. However, plots
of fitted vs. observed responses were examined.
A computer program ("MSW") to implement the multistage Weibull time-to-tumor
model was designed, developed and tested for U.S. EPA by Battelle Columbus (Ohio). The
MSW program obtains maximum likelihood estimates for model parameters and solves for the
BMDL (lower confidence limit for BMD) using the profile-likelihood method. The model, with
documentation for methodology (statistical theory and estimation, and numerical algorithms) and
This document is a draft for review purposes only and does not constitute Agency policy.
G-49 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
testing, was externally reviewed by experts in June 2007. Reviews were generally positive and
confirmed that the functioning of the computer code has been rigorously tested. (U.S. EPA and
Battelle confirmed that MSW gave results essentially identical to those of "ToxRisk," a program
no longer commercially issued or supported.) U.S. EPA's BMDS Web site provided reviewers'
comments and U.S. EPA's responses.21 The MSW program and reports on statistical and
computational methodology and model testing will be made available in 2009 (after
implementing some changes to reporting features and error-handling).
Results of this modeling are shown in the file .
G.7.2. Poly-3 Calculation of Adjusted Number at Risk
To obtain an independent estimate of a point of departure using different assumptions, it
was thought desirable to compare time-to-tumor modeling to an alternative survival-adjustment
technique, "poly-3 adjustment" (Portier and Bailer, 1989), applied to the same data. This
technique was used to adjust the tumor incidence denominators based on the individual animal
survival times. The adjusted incidence data then served as inputs for U.S. EPA's BMDS
multistage model, and multistage model selection was conducted as described in Section 5.2.
A detailed exposition is given by (Piegorsch and Bailer, 1997), Section 6.3.2. Each
tumor-less animal is weighted by its fractional survival time (survival time divided by the
duration of the bioassay) raised to the power of 3 to reflect the fact that animals are at greater
risk of cancer at older ages. Animals with tumors are given a weight of 1. The sum of the
weights of all the animals in an exposure group yields the effective survival-adjusted
denominator. The "default" power of 3 (thus, "poly-3") was assumed, which was found to be
representative for a large number of cancer types (Portier et al., 1986). Algebraically,
Nadj = Yji W'	(Etl- G"7)
21 At http://www. epa. gov/ncea/bmds/response .html under title "2007 External Review of New Quantal Models;"
use links to comments and responses.
This document is a draft for review purposes only and does not constitute Agency policy.
G-50 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
where
W; = 1 if tumor is present
Wj = (tjl'f if tumor is absent at time of death (/,)
T = duration of study. N was rounded to the nearest integer.22
Calculations are reproduced in the spreadsheets linked above.
G.8. COMBINED RISK FROM MULTIPLE TUMOR SITES
For bioassays that exhibited more than one type of tumor response in the same sex and
species (these studies have a row for "combined risk" in the "Endpoint" column of Table 5-27,
Section 5.2), the cancer potency for the different tumor types combined was estimated. The
combined tumor risk estimate describes the risk of developing tumors for any (not all together)
of the tumor types that exhibited a TCE-associated tumor response; this estimate then represents
the total excess cancer risk. The model for the combined tumor risk is also multistage, with the
sum of the stage-specific multistage coefficients from the individual tumor models serving as the
stage-specific coefficients for the combined risk model (i.e., for each
q ,q r , „ = Q, + Q ¦¦¦ + Q,, where the q s are the coefficients for the powers of dose and k is
i[combined] 1i\ 1i2	1ik' l	'
the number of tumor types being combined) (Bogen, 1990; NRC, 1994). This model assumes
that the occurrences of two or more tumor types are independent. The resulting model equation
can be readily solved for a given BMR to obtain a maximum likelihood estimate (BMD) for the
combined risk. However, the confidence bounds for the combined risk estimate are not
calculated by available modeling software. Therefore, a Bayesian approach was used to estimate
confidence bounds on the combined BMD. This approach was implemented using the freely
available WinBUGS software (Spiegelhalter et al., 2003), which applies Markov chain Monte
Carlo computations. Use of WinBUGS has been demonstrated for derivation of a distribution of
BMDs for a single multistage model (Kopylev et al., 2007) and can be straightforwardly
generalized to derive the distribution of BMDs for the combined tumor load.
22Notice that the assumptions required for significance testing and estimating variances of parameters are changed
by this procedure. The Williams-Bieler variance estimator is described by (Piegorsch and Bailer, 1997). Our
multistage modeling did not take this into account, so the resulting BMDL may be somewhat lower than could be
obtained by more laborious calculations.
This document is a draft for review purposes only and does not constitute Agency policy.
G-51 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
G.8.1. Methods
G.8.1.1. Single Tumor Sites
Cancer dose-response models were fitted to data using BMDS. These were multistage
models with coefficients constrained to be non-negative. The order of model fitted was (g - 1),
where g is the number of dose groups. For internal dose metrics, the values shown in tables
above were used.
The multistage model was modified for U.S. EPA NCEA by Battelle (under contract
EPC04027) to provide model-based estimates of extra risk at a user-specified dose and
profile-likelihood confidence intervals for that risk. Thus, confidence intervals for extra risk in
addition to BMDs could be reported.
G.8.1.2. Combined Risk From Multiple Tumor Sites
The multistage model identified by BMDS23 was used in a WinBUGS script to generate
posterior distributions for model parameters, the BMD and extra risk at the same dose specified
for the BMDS estimates. The prior used for multistage parameters was the positive half of a
normal distribution having a mean of zero and a variance of 10,000, effectively a very flat prior.
The burn-in was of length 10,000, then 100,000 updates were made and thinned to every 10th
update for sample monitoring. From a WinBUGS run, the sample histories, posterior
distribution plots, summary statistics, and codas were archived.
Codas were then imported to R and processed using R programs to compute BMD and
the extra risk at a specific dose for each tumor type. BMD and extra risk for the combined risk
function (assuming independence) were also computed following Bogen (1990, Chapter IV;
NRC, 1994, Chapter 11, Appendix 1-1, Appendix 1-2). Results were summarized as percentiles,
means, and modes (modes were based upon the smoothed posterior distributions). The extra
risks across tumor types at a specific dose (10 or 100 was used) were also summed.
BMDLs for rodent internal doses, reported below, were converted to human external
doses using the conversion factors in Tables G-6 and G-7 (based on PBPK model described in
Section 3.5).
23The highest-order model was used, e.g., if BMDS estimates were gamma = 0, beta.l > 0, beta.2 = 0, beta.3 > 0,
the model in WinBUGS allowed beta.2 to be estimated (rather than being fixed at zero).
This document is a draft for review purposes only and does not constitute Agency policy.
G-52 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-6. Rodent to human conversions for internal dose metric
2	TotOxMetabBW34
3
Route
Sex
Human (mean)
Inhalation, ppm
F
9.843477
M
9.702822
Oral, mg/kg/d
F
15.72291
M
16.4192
4
5	Table G-7. Rodent to human conversions for internal dose metric
6	TotMetabBW34
7
Route
Sex
Human (mean)
Inhalation, ppm
F
11.84204
M
11.69996
Oral, mg/kg/d
F
18.76327
M
19.6
8	The application of rodent to human conversion factors is as follows:
9
10	Given rodent internal dose D in some units of TotOxMetabBW34, divide by tabled value Y
11	above to find human exposure in ppm or mg/kg/d.
12
13	Example: ppm (human) = Z)(rodent)/7
14	ppm (human female mean) = 500 (internal units)/9.843477
15	= 50.80 ppm	(Eq. G-8)
16
G.8.2. Results
17	The results follow in this order:
18
19	Applied doses
20	NCI (1976), Female B6C3F1 mice, oral gavage, liver and lung tumors and lymphomas
21	(see Tables G-8 through G-10 and Figures G-l and G-2)
This document is a draft for review purposes only and does not constitute Agency policy.
G-53 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Maltoni (1986), Female B6C3F1 mice, inhalation (expt. BT306), liver and lung tumors
(see Tables G-l 1 through G-13 and Figures G-3 and G-4)
Maltoni (1986), Male Sprague-Dawley rats, inhalation (expt. BT304), kidney tumors,
testis Ley dig Cell tumors, and lymphomas (see Tables G-l 4 through G-l 6 and
Figures G-5 and G-6)
Internal Doses
NCI (1976) Female B6C3F1 mice, oral gavage, liver and lung tumors and lymphomas
(see Tables G-l7 through G-l9 and Figures G-7 and G-8)
Maltoni (1986), Female B6C3F1 mice, inhalation (expt. BT306), liver and lung tumors
(see Tables G-20 through G-22 and Figures G-9 and G-10)
Maltoni (1986), Male Sprague-Dawley rats, inhalation (expt. BT304), kidney tumors,
Testis Leydig Cell tumors, and lymphomas (see Tables G-23 through G-25 and
Figures G-l 1 and G-l2)
This document is a draft for review purposes only and does not constitute Agency policy.
G-54 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Table G-8. Female B6C3F1 mice—applied doses: data


Liver
Lung
Hematopoietic


hepatocellular
adenomas +
lymphomas +
Dosea
Nh
carcinomas
carcinomas
sarcomas
0
18
0
1
1
356.4
45
4
4
5
713.3
41
11
7
6
a Doses were adjusted by a factor 0.41015625, accounting for exposure 5/7 days/week, exposure duration 78/91
weeks, and duration of study (91/104)A3. Averaged applied gavage exposures were low-dose 869 mg/kg/d,
high dose 1,739 mg/kg/d.
b Numbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study.
Source: NCI (1976).
Table G-9. Female B6C3F1 mice—applied doses: model selection
comparison of model fit statistics for multistage models of increasing order
Tumor site
Model
order,
^selected
Coeff.
estimates
equal
zero
AIC
Largest*
scaled
residual
Goodness
of fit
/7-value
Liver
2
y
78.68
0
1

1*
y
77.52
-0.711
0.6698
Lung
2
NA
78.20
0
1

1*
NA
76.74
-0.551
0.4649
Lymphomas + sarcomas
2
P2
77.28
0.113
0.8812

1*
NA
77.28
0.113
0.8812
* Largest in absolute value.
Source: NCI (1976).
This document is a draft for review purposes only and does not constitute Agency policy.
G-55 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-10. Female B6C3F1 mice—applied doses: BMD and risk estimates
2	(inferences for BMR of 0.05 extra risk at 95% confidence level)
3

Liver
hepatocellular
carcinomas
Lung
adenomas +
carcinomas
Hematopoietic
lymphomas +
sarcomas
Parameters used in model
q0,ql
q0,ql
q0,ql
p- Value for BMDS model
0.6698
0.6611
0.8812
BMDos (from BMDS)
138.4
295.2
358.8
BMD05 (median, mode—WinBUGS)
155.5, 135.4
314.5, 212.7
352.3, 231.7
BMDL (BMDS)*
92.95
144.3
151.4
BMDL (5th percentile, WinBUGS)
97.48
150.7
157.7
BMD05 for combined risk (median,
mode, from WinBUGS)
84.99, 78.95
BMDL for combined risk (5th
percentile, WinBUGS)
53.61
BMDS maximum likelihood risk estimates
Risk at dose 100
0.03640
0.01722
0.01419
Upper 95% CL
0.05749
0.03849
0.03699
Sum of risks at dose 100
0.06781
WinBUGS Bayes risk estimates
Risk at dose 100: mean, median
0.0327, 0.0324
0.0168, 0.0161
0.0152, 0.0143
Upper 95% CL
0.0513
0.0334
0.0319
Comb, risk at dose 100 mean, median
0.06337, 0.0629
Comb, risk at dose 100, upper 95% CL
0.09124
4
5	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
6
7	Source: NCI (1976).
This document is a draft for review purposes only and does not constitute Agency policy.
G-56 DRAFT-DO NOT CITE OR QUOTE
-------
CvJ
O
vertical solid, BMDc
vertical dash, BMDLc
oo
o
o
o
o
o
o
o
0	50	100 150 200
Dose
1
This document is a draft for review purposes only and does not constitute Agency policy.
G-57 DRAFT-DO NOT CITE OR QUOTE
-------
1	Figure G-l. Female B6C3F1 mice—applied doses: combined and individual
2	tumor extra-risk functions.
3
o
CN
CD
O
LO
O
O
&
w
c
CD
Q
o
o
o
LO
o
p
d
o
o
p
d
0
200 400 600 800 1000
N = 300000 Bandwidth = 1.602
4
5	Figure G-2. Female B6C3F1 mice—applied doses: posterior distribution of
6	BMDc for combined risk.
7
This document is a draft for review purposes only and does not constitute Agency policy.
G-58 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Table G-ll. B6C3F1 female mice inhalation exposure—applied doses
Dosea

Liver hepatomas/iVb
Lung adenomas +
carcinomas/iVb
0

3/88
2/90
15.6

4/89
6/90
46.9

4/88
7/89
93.8

9/85
14/87
a Doses adjusted by a factor 0.133928571, accounting for exposure 7/24 hours/day x 5/7 days/week, and
exposure duration 78/104 weeks. Applied doses were 100, 300, and 600 ppm.
b Numbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study.
Source: Maltoni (1986).
Table G-12. B6C3F1 female mice—applied doses: model selection
comparison of model fit statistics for multistage models of increasing order
Tumor Site
Model
order,
^selected
Coeff.
estimates
equal zero
AIC
Largest*
scaled
residual
Goodness
of fit
/7-value
Liver
3
P2
154.91
0.289
0.7129
2
PI
153.02
0.330
0.8868
1*
NA
153.47
-0.678
0.7223
Lung
3
P2
195.91
0.741
0.3509
2
P2
193.91
0.714
0.6471
1*
NA
193.91
0.714
0.6471
*Largest in absolute value.
Source: Maltoni (1986).
This document is a draft for review purposes only and does not constitute Agency policy.
G-59 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-13. B6C3F1 female mice inhalation exposure—applied doses
2	(inferences for 0.05 extra risk at 95% confidence level)
3

Liver hepatomas
Lung adenomas +
carcinomas
Parameters used in model
q0,ql
q0,ql
p- Value for BMDS model
0.7223
0.06471
BMD05 (from BMDS)
72.73
33.81
BMD05 (median, mode—WinBUGS)
71.55, 56.79
34.49, 31.65
BMDL (BMDS)*
37.13
21.73
ms combo.exe BMD05C, BMDLc
32.12, 16.22
BMDos (5th percentile, WinBUGS)
37.03
22.07
BMDo5 for combined risk (median,
mode, from WinBUGS)
23.07, 20.39
BMDL for combined risk (5th percentile,
WinBUGS)
15.67
BMDS maximum likelihood risk estimates
Risk at dose 10
0.0070281
0.0150572
Upper 95% CL
0.0151186
0.0250168
Sum of risks at dose 10
0.0220853
WinBUGS Bayes risk estimates: means (medians)
Risk at dose 10: mean, median
0.007377, 0.007138
0.01489, 0.01476
Upper 95% CL
0.01374
0.02
Comb, risk at dose 10: mean, median
0.02216, 0.02198
Comb, risk at dose 10: upper 95% CL
0.03220
4
5	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
6
7	Source: Maltoni (1986).
This document is a draft for review purposes only and does not constitute Agency policy.
G-60 DRAFT-DO NOT CITE OR QUOTE
-------
vertical solid, BMDc
vertical dash, BMDLc
O
oo
o
o
CM
o
o
o
o
O
o
0	50 100 150 200
Dose
1
This document is a draft for review purposes only and does not constitute Agency policy.
G-61 DRAFT-DO NOT CITE OR QUOTE
-------
1	Figure G-3. B6C3F1 female mice inhalation exposure—applied doses:
2	combined and individual tumor extra-risk functions.
3
CO
O
O _
(/) O
§ <=> -
Q
CM
O
d
0	100 200 300
N = 300000 Bandwidth = 0.4731
4
5	Figure G-4. B6C3F1 female mice inhalation exposure—applied doses:
6	posterior distribution of BMDc for combined risk.
7
This document is a draft for review purposes only and does not constitute Agency policy.
G-62 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-14. Maltoni Sprague-Dawley male rats—applied doses
2
Dosea

Kidney adenomas
+ carcinomas/iVb
Leukemias/iVb
Testis, Leydig
cell tumors/iV
0

0/121
9/134
6/121
20.8

1/118
13/130
16/116
62.5

0/116
14/130
30/116
125

5/123
15/129
31/122
3
4	11 Doses adjusted by a factor 0.208333333, accounting for exposure 7 hours/day x 5/7 days/week. Applied doses
5	were 100, 300, and 600 ppm.
6	b Numbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study.
7
8
9	Table G-15. Maltoni Sprague-Dawley male rats—applied doses: model
10	selection comparison of model fit statistics for multistage models of
11	increasing order
12
Tumor site
Model
order*
Coeff.
estimates
equal
zero
AIC
Largest+
scaled
residual
Goodness
of fit
/j-value
Kidney
3
P1,P2
60.55
1.115
0.292

2
y
61.16
-1.207
0.253

1*
y
59.55
-1.331
0.4669
Leukemia
3
P2, P3
336.8
0.537
0.715

2
P2
336.8
0.537
0.715

1
NA
336.8
0.537
0.715
Dropping high dose
2
P2
243.7
0.512
0.529

1*
NA
243.7
0.512
0.529
Testis
3
P2, P3
421.4
-1.293
0.057

2
P2
421.4
-1.293
0.057

1
NA
421.4
-1.293
0.057
Dropping high dose
2
P2
277.6
0.291
0.728

1*
NA
277.6
0.291
0.728
13
14	* Model order selected + largest in absolute value
15
This document is a draft for review purposes only and does not constitute Agency policy.
G-63 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-16. Maltoni Sprague-Dawley male rats—applied doses
2

Kidney
adenomas +
carcinomas
Leukemia
(high dose
dropped)
Testis, Leydig
cell tumors (high
dose dropped)
Parameters used in models
q0,ql
q0,ql
q0,ql
p- Value for BMDS model
0.4669
0.5290
0.7277
BMDoi (from BMDS)
41.47
14.5854
2.46989
BMDoi (median, mode—WinBUGS)
46.00, 35.71
12.32, 8.021
2.497, 2.309
BMDL (BMDS)*
22.66
5.52597
1.77697
BMDL (5th percentile, WinBUGS)
23.23
5.362
1.789
BMDoi for combined risk (median,
mode, from WinBUGS)
1.960, 1.826
BMDL for combined risk (5th
percentile, WinBUGS)
1.437
BMDS maximum likelihood risk estimates
Risk at dose 10
0.0024208
0.0068670
0.0398747
Upper 95% CL
0.0048995
0.0202747
0.0641010
Sum of risks at dose 10

Risk at dose 1
0.0002423
0.0006888
0.0040609
Upper 95% CL
0.0004911
0.0020462
0.0066029
Sum of risks at dose 1

WinBUGS Bayes risk estimates: means (medians)
Risk at dose 10: mean, median
0.002302,
0.002182
0.008752,
0.008120
0.03961, 0.03945
Upper 95% CL
0.004316
0.01860
0.05462
Comb, risk at dose 10, mean, median
0.05020, 0.04998
Comb, risk at dose 10, upper 95% CL
0.06757
Risk at dose 1: mean, median
2.305e-04,
2.184e-04
8.800e-04,
8.150e04
0.004037,
0.004017
Upper 95% CL
4.325e-04
1.876e-03
0.005601
Comb, risk at dose 1, mean, median
0.005143, 0.005114
Comb, risk at dose 1, upper 95% CL
0.006971
3
4	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
This document is a draft for review purposes only and does not constitute Agency policy.
G-64 DRAFT-DO NOT CITE OR QUOTE
-------
o
cn o
ha
2
X CN
lu d
	 vertical solid, BMDc
-- vertical dash, BMDLc
O
o
40 60
80 100
This document is a draft for review purposes only and does not constitute Agency policy.
G-65 DRAFT-DO NOT CITE OR QUOTE
-------
1	Figure G-5. Maltoni Sprague-Dawley male rats—applied doses: combined
2	and individual tumor extra-risk functions.
3
c
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Table G-17. Female B6C3F1 mice—internal dose metric (total oxidative
metabolism): data


Liver
Lung
Hematopoietic

JVb
hepatocellular
adenomas +
lymphomas +
Internal dosea
carcinomas
carcinomas
sarcomas
0
18
0
1
1
549.8
45
4
4
5
813.4
41
11
7
6
internal dose, Total Oxidative Metabolism, adjusted for body weight, units |mg/(wk-kgvl)|. Internal doses were
adjusted by a factor 0.574219, accounting for exposure duration 78/91 weeks, and duration of study
(91/104)A3. Before adjustment, the median internal doses were 95748 and 1416.55 (mg/wk-kg3 4).
bNumbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study.
Source: NCI (1976).
Table G-18. Female B6C3F1 mice—internal dose: model selection
comparison of model fit statistics for multistage models of increasing order
Tumor site
BMD,
BMDL
Model
order*
Coeff.
estimates
equal zero
AIC
Largest+
scaled
residual
Goodness
of fit
/7-value
Liver
505, 284
2*
Y> P1
77.25
-0.594
0.7618

367, 245
1
y
78.86
-1.083
0.3542
Lung
742, 396
2*
pi
76.33
-0.274
0.7197

780, 380
1
NA
76.74
-0.551
0.4649
Lymphomas + sarcomas
870, 389
2
NA
79.26
0
1

839, 390
1*
NA
77.27
-0.081
0.9140
* Model order selected + largest in absolute value.
Source: NCI (1976).
This document is a draft for review purposes only and does not constitute Agency policy.
G-67 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-19. Female B6C3F1 mice—internal dose metric (total oxidative
2	metabolism): BMD and risk estimates (values rounded to 4 significant
3	figures) (inferences for BMR of 0.05 extra risk at 95% confidence level)
4

Liver
hepatocellular
carcinomas
Lung
adenomas +
carcinomas
Hematopoietic
lymphomas +
sarcomas
Parameters used in models
q0, ql, q2
q0, ql, q2
q0,ql
p- Value for BMDS model
0.7618
0.7197
0.9140
BMD05 (from BMDS)
352.4
517.8
423.8
BMD05 (median, mode from WinBUGS)
284.8, 292.5
414.3, 299.9
409.8, 382.6
BMDL (BMDS)*
138.1
193.0
189.5
BMDL (5th percentile, WinBUGS)
162.6
195.4
226.2
BMDo5 for Combined Risk (median,
mode, from WinBUGS)
136.1, 121.1
BMDL for Combined Risk (5th percentile,
WinBUGS)
85.65
BMDS maximum likelihood risk estimates
Risk at dose 100
0.004123
0.001912
0.0120315
Upper 95% CL
0.04039
0.02919
0.0295375
Sum of risks at dose 100

WinBUGS Bayes risk estimates
Risk at dose 100: mean, median
0.01468,
0.01311
0.01284,
0.01226
0.009552,
0.008286
Upper 95% CL
0.03032
0.02590
0.021410
Comb, risk at dose 100 mean, median
0.03663, 0.03572
Comb, risk at dose 100, upper 95% CL
0.05847
5
6	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
7
8	Source: NCI (1976).
9
This document is a draft for review purposes only and does not constitute Agency policy.
G-68 DRAFT-DO NOT CITE OR QUOTE
-------
o
— vertical solid, BMDc
-vertical dash, BMDLc
CO
d
CNJ
d
d
. /¦"
p
d
0
200
400
600
800
Dose
Figure G-7. Female B6C3F1 mice—internal dose metric (total oxidative
metabolism): combined and individual tumor extra-risk functions.
cxd
o
o
d
o
o
d
o
o
p
d
100 200 300 400 500 600
N = 300000 Bandwidth = 3.023
Figure G-8. Female B6C3F1 mice—internal dose metric (total oxidative
metabolism): posterior distribution of BMDc for combined risk.
This document is a draft for review purposes only and does not constitute Agency policy.
G-69 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-20. B6C3F1 female mice inhalation exposure—internal dose metric
2	(total oxidative metabolism)
3
Internal dosea
Liver
hepatomas/A^
Lung adenomas +
carcinomas/A^
0
3/88
2/90
280.946
4/89
6/90
622.530
4/88
7/89
939.105
9/85
14/87
4
5	a Internal dose, Total Oxidative Metabolism, adjusted for body weight, units (mg/[wk-kg3/4]).
6	Internal doses were adjusted by a factor 0.75, accounting for exposure duration 78/104 weeks.
7	Before adjustment, median internal doses were 374.5945, 830.0405, 1252.14 (mg/[wk-kg3/4]).
8	b Numbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study
9
10	Source: Maltoni (1986).
11
12
13	Table G-21. B6C3F1 female mice—internal dose: model selection
14	comparison of model fit statistics for multistage models of increasing order
15
Tumor site
Model
order,
*selected
Coeff.
estimates
equal
zero
AIC
Largest+
scaled
residual
Goodness
of fit
/j-value
Liver
3*
P1,P2
153.1
-0.410
0.8511
2
PI
153.4
-0.625
0.7541
1
NA
154
-0.816
0.5571
Lung
3
P2
195.8
-0.571
0.3995
2
NA
195.9
-0.671
0.3666
1*
NA
194
-0.776
0.6325
16
17	* Model order selected + largest in absolute value.
18
19	Source: Maltoni (1986).
20
This document is a draft for review purposes only and does not constitute Agency policy.
G-70 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-22. B6C3F1 female mice inhalation exposure—internal dose metric
2	(total oxidative metabolism) (inferences for 0.05 extra risk at 95% confidence
3	level)
4

Liver
hepatomas
Lung adenomas +
carcinomas
Parameters used in models
qO, ql, q2, q3
q0,ql
p- Value for BMDS model
0.5571
0.6325
BMD05 (from BMDS)
813.7
366.7
BMD05 (median, mode—WinBUGS)
672.9, 648.0
382.8, 372.1
BMDL (BMDS)*
419.7
244.6
ms combo BMD05C, BMDLc
412.76, 189.23
BMDL (5th percentile, WinBUGS)
482.7
251.1
BMD05 for combined risk (median, mode, from
WinBUGS)
286.7, 263.1
BMDL for combined risk (5th percentile,
WinBUGS)
199.5
BMDS maximum likelihood risk estimates
Risk at dose 100
0.006284
0.01389
Upper 95% CL
0.01335
0.02215
Sum of risks at dose 100
0.02017
WinBUGS Bayes risk estimates: means (medians)
Risk at dose 100: mean, median
0.003482,
0.002906
0.01337,
0.01331
Upper 95% CL,
0.008279
0.02022
Comb, risk at dose 100 mean, median
0.01637, 0.01621
Comb, risk at dose 100, upper 95% CL
0.02455
5
6	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
7
8	Source: Maltoni (1986).
This document is a draft for review purposes only and does not constitute Agency policy.
G-71 DRAFT-DO NOT CITE OR QUOTE
-------
vertical solid, BMDc
vertical dash, BMDLc
ID
O
O
O _
O
O
O
d
0	200	400	600	800
Dose
1
This document is a draft for review purposes only and does not constitute Agency policy.
G-72 DRAFT-DO NOT CITE OR QUOTE
-------
Figure G-9. B6C3F1 female mice inhalation exposure—internal dose metric:
combined and individual tumor extra-risk functions.
cr>
c
0)
Q
200 400 600 800 1000 1400
N = 300000 Bandwidth = 5.053
Figure G-10. B6C3F1 female mice inhalation exposure—internal dose
metric: posterior distribution of BMDc for combined risk.
This document is a draft for review purposes only and does not constitute Agency policy.
G-73 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-23. Maltoni Sprague-Dawley male rats—internal dose metric (total
2	metabolism)
3
Internal dosea
Kidney adenomas +
carcinomas/A^
Leukemias/A^
Testis, Leydig
cell tumors/A^
0
0/121
9/134
6/121
214.6540
1/118
13/130
16/116
507.0845
0/116
14/130
30/116
764.4790
5/123
15/129
31/122
4
5	a Internal dose, Total Oxidative Metabolism, adjusted for body weight, units |mg/(\vk-kg3 4)|.
6	b Numbers at risk are the smaller of (a) time of first tumor observation or (b) 52 weeks on study.
7
8
9	Table G-24. Maltoni Sprague-Dawley male rats—internal dose model
10	selection comparison of model fit statistics for multistage models of
11	increasing order
12
Tumor site
Model
order,
*selected
Coeff.
estimates
equal zero
AIC
Largest*
scaled
residual
Goodness
of fit
/7-value
Kidney
3
y, P2
61.35
-1.264
0.262
2
y
61.75
-1.343
0.246
1*
y
60.32
-1.422
0.370
Leukemias
3
P2, P3
336.5
0.479
0.828
2
P2
336.5
0.479
0.828
1*
NA
336.5
0.479
0.828
Testis, Leydig cell tumors
3
P2, P3
417.7
1.008
0.363
2
P2
417.7
1.008
0.363
1*
NA
417.7
1.008
0.363
13
14	* Largest in absolute value.
15
This document is a draft for review purposes only and does not constitute Agency policy.
G-74 DRAFT-DO NOT CITE OR QUOTE
-------
1	Table G-25. Maltoni Sprague-Dawley male rats—internal dose metric (total
2	metabolism) (inferences for 0.01 extra risk at 95% confidence level)
3

Kidney adenomas
+ carcinomas
Leukemias
Testis, Leydig
cell tumors
Parameters used in models
q0,ql
q0,ql
q0,ql
p- Value for BMDS model
0.3703
0.8285
0.3626
BMDoi (from BMDS)
295.1
145.8
26.65
BMDoi (median, mode—WinBUGS)



BMDL (BMDS)*
161.3
65.29
20.32
BMDL (5th percentile, WinBUGS)



BMDoi for combined risk (median,
mode, from WinBUGS)
20.97, 19.73
BMDL for combined risk (5th
percentile, WinBUGS)
16.14
BMDS maximum likelihood risk estimates
Risk at dose 100
0.003400
0.0068694
0.0370162
Upper 95% CL
0.0068784
0.0169134
0.0504547
Sum of risks at dose 100
0.04729
Risk at dose 10
0.0003406
0.0006891
0.0037648
Upper 95% CL
0.0006900
0.0017044
0.0051638
Sum of risks at dose 10
0.004795
WinBUGS Bayes risk estimates: means (medians)
Risk at dose 100: mean, median
0.003191,
0.003028
7.691e-03,
7.351e-03
0.03641,
0.03641
Upper 95% CL
0.006044
1.539e-02
0.04769
Comb, risk at dose 100—mean, median
0.04688, 0.04680
Comb, risk at dose 100, upper 95% CL
0.060380
Risk at dose 100—mean, median
3.196e-04,
3.032e04
7.726e-04,
7.376e04
0.003705,
0.003703
Upper 95% CL
6.060000e-04
1.550000e-03
0.004874000
Comb, risk at dose 10—mean, median
0.004793, 0.0047820
Comb, risk at dose 10, upper 95% CL
0.006208
4
5	* All confidence intervals are at 5% (lower) or 95% (upper) level, one-sided.
This document is a draft for review purposes only and does not constitute Agency policy.
G-75 DRAFT-DO NOT CITE OR QUOTE
-------
	 vertical solid, BMDc
- - vertical dash, BMDLc
400 500
Dose
This document is a draft for review purposes only and does not constitute Agency policy.
G-76 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Figure G-ll. Maltoni Sprague-Dawley male rats—internal dose metric:
combined and individual tumor extra-risk functions.
Distribution of BMDc for combined risk
o
o
o
o
o
20
40
60
80
100
N = 300000 Bandwidth = 0.2732
Figure G-12. Maltoni Sprague-Dawley male rats—internal dose metric:
posterior distribution of BMDc for combined risk.
G.9. PHYSIOLOGICALLY BASED PHARMACOKINETIC (PBPK)-MODEL
UNCERTAINTY ANALYSIS OF UNIT RISK ESTIMATES
As discussed in Section 5.2, an uncertainty analysis was performed on the unit risk
estimates derived from rodent bioassays to characterize the impact of pharmacokinetic
uncertainty. In particular, two sources of uncertainty are incorporated: (a) uncertainty in the
rodent internal doses for each dose group in each chronic bioassay and (b) uncertainty in the
relationship between exposure and the human population mean internal dose at low exposure
levels.
A Bayesian approach provided the statistical framework for this uncertainty analysis.
Rodent bioassay internal dose-response relationships were modeled with the multistage model,
with general form
P(id) = 1 - exp[-(7/o + cjiid + q2id2 + ... + qiAcl)\,	(Eq. G-9)
This document is a draft for review purposes only and does not constitute Agency policy.
G-ll DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
where P(id) represents the lifetime risk (probability) of cancer at internal dose id, and multistage
parameters q, > 0, for i = 0, 1, ...,&. Since the BMD (in internal dose units) for a given BMR can
be derived from the multistage model parameters qu it is sufficient to estimate the posterior
distribution of q, given the combined bioassay data (for each dose group j, the number
responding^, the number at risk nh and the administered dose dj) and the rodent
pharmacokinetic data, for which the posterior distribution can be derived using the Bayesian
analysis of the PBPK model described in Section 3.5. In particular, the posterior distribution of
qi can be expressed as
P{fl[{\\Dbioassay Dpk) ^ P(fl[{\) P'CV[/]| *7[i] n\j]) P(}d\j^\d\j], Dpk)	(EC]. G-10)
Here, the first term after the proportionality P{q[i\) is the prior distribution of the multistage
model parameters (assumed to be noninformative), the second term P(y\j]\q[q n\i\) is the likelihood
of observing the bioassay response given a particular set of multistage parameters and the
number at risk (the product of binomial distributions for each dose group), and PQd^d^ I)p/C) is
the posterior distribution of the rodent internal doses id\j\, given the bioassay doses and the
pharmacokinetic data used to estimate the PBPK model parameters.
The distribution of unit risk (l/Ruj = BMR/BMD) estimates in units of "per internal dose"
is then derived deterministically from the distribution of multistage model parameters:
P( URiJ\Dbioassay Dpk-rodent) \P(q[2] \Dbioassay Dpk-rodent) 5[UR - BMR/dqH (Eq. G-l 1)
Here S is the Dirac delta-function. Then, the distribution of unit risk estimates in units of "per
human exposure" (per mg/kg/d ingested or per continuous ppm exposure) is derived by
converting the unit risk estimate in internal dose units:
P(URhuman\Dbioassay Dpk-rodent) \P(URic^Pbioassay 1 ^pk-iv>dcnl) P(jdconversion\Dpk-hUman)
£}(URhuman URid x idconversiOK;) d\ACOnversion	(EC[. G-l2)
Here, idconversion is the population mean of the ratio between internal dose and administered
exposure at low dose (0.001 ppm or 0.001 mg/kg/d), and P(idconversion\Dpk-human) is its posterior
distribution from the Bayesian analysis of the human PBPK model.
This statistical model was implemented via Monte Carlo as follows. For each bioassay,
for a particular iteration r (r = 1... //,.),
This document is a draft for review purposes only and does not constitute Agency policy.
G-78 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
(1)	A sample of rodent PBPK model population parameters (\a£)rodent,r was drawn from the
posterior distribution. Using these population parameters, a single set of group rodent
PBPK model parameters Brodent,r was drawn from the population distribution. As
discussed in Section 3.5, for rodents, the population model describes the variability
among groups of rodents, and the group-level parameters represent the "average"
toxicokinetics for that group.
(2)	Using Qrodent.r, the rodent PBPK model was run to generate a set of internal doses for
the bioassay.
(3)	Using this set of internal doses id^r, a sample q\,yr was selected from the distribution
(conditional on /t%r) of multistage model parameters, generated using the WinBUGS,
following the methodology of Kopylev et al. (2007).
(4)	The unit risk in internal dose units IIR,j,r = BMR/BMD{q\iir) was calculated based on the
multistage model parameters.
(5)	A sample of human PBPK model population parameters {\iX)human,r was drawn from the
posterior distribution. Using these population parameters, multiple sets of individual
human PBPK model parameters Qhuman,r,[s] (v = 1 ...ns) were generated. A continuous
exposure scenario at low exposure was run for each individual, and the population mean
internal dose conversion was derived by taking the arithmetic mean of the internal dose
conversion for each individual: idconversio„tr = Sum(idconversionrrtS)/ns.
(6)	The sample for the unit risk in units per human exposure was calculated by multiplying
the sample for the unit risk in internal dose units by the sample for the population internal
dose Conversion. URjquman,r URid,r ^ idconrcrsion.r-
In practice, samples for each of the above distributions were "precalculated," and
inferences were performed by re-sampling (with replacement) according to the scheme above.
For the results described in Section 5.2, a total of nr = 15,000 samples was used for deriving
summary statistics. For calculating the unit risks in units of internal dose, the BMDs were
derived by re-sampling from a total of 4.5 x 106 multistage model parameter values (1,500 rodent
PBPK model parameters from the Bayesian analysis described in Section 3.5, for each of which
there were conditional distributions of multistage model parameters of length 3,000 derived
using WinBUGS). The conversion to unit risks in units of human exposure was re-sampled from
500 population mean values, each of which was estimated from 500 sampled individuals.
The file contains summary statistics (mean, and selected quantiles from 0.01 to 0.99)
from these analyses, and is the source for the results presented in Chapter 5 (see Tables 5-34 and
5-35). Histograms of the distribution of unit risks in per unit human exposure are in the file for
the rodent inhalation bioassays and for the rodent oral bioassays. Route-to-route extrapolated
unit risks are in the files (inhalation unit risks extrapolated from oral bioassays) and (oral unit
risks extrapolated from inhalation bioassays). Each figure shows the uncertainty distribution for
This document is a draft for review purposes only and does not constitute Agency policy.
G-79 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
1
the male and female combined population risk per unit exposure (transformed to base-10
logarithm), with the exception of testicular tumors, for which only the population risk per unit
exposure for males is shown.
This document is a draft for review purposes only and does not constitute Agency policy.
G-80 DRAFT-DO NOT CITE OR QUOTE
-------
APPENDIX H
Lifetable Analysis and Weighted Linear
Regression based on Results from
Charbotel et al. (2006)
This document is a draft for review purposes only and does not constitute Agency policy.
H-i DRAFT—DO NOT CITE OR QUOTE
-------
CONTENTS—Appendix H: Lifetable Analysis and Weighted Linear Regression based on
Results from Charbotel et al. (2006)
APPENDIX H: LIFETABLE ANALYSIS AND WEIGHTED LINEAR REGRESSION
BASED ON RESULTS FROM CHARBOTEL ET AL. (2006)	H-i
II. 1 LIFETABLE ANALYSIS	Error! Bookmark not defined.
H.2. EQUATIONS USED FOR WEIGHTED LINEAR REGRESSION OF
RESULTS FROM CHARBOTEL ET AL. (2006)	Error! Bookmark not defined.
This document is a draft for review purposes only and does not constitute Agency policy.
H-ii DRAFT—DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
H.l. LIFETABLE ANALYSIS
A spreadsheet illustrating the extra-risk calculation for the derivation of the lower 95%
bound on the effective concentration associated with a 1% extra risk (LECoi) for renal cell
carcinoma (RCC) incidence is presented in Table H-l.
H.2. EQUATIONS USED FOR WEIGHTED LINEAR REGRESSION OF RESULTS
FROM CHARBOTEL ET AL. (2006) [source: Rothman (1986), p. 343-344]
Linear model: RR = 1 + bX
where RR = risk ratio, X= exposure, and b = slope
b can be estimated from the following equation:
2>VVM Z
b = ^	
wjxj
j= 2
(Eq H-l)
I
j= 2
WJXJ
where j specifies the exposure category level and the reference category (j = 1) is ignored.
The standard error of the slope can be estimated as follows:
SE(b) ¦¦
1
(Eq. H-2)
S
j= 2
WJXJ
The weights, Wj, are estimated from the confidence intervals (as the inverse of the variance):
VariRRj) ~ i^/Far[ln(i^,)] ~ RR,
\n(RRj) - \n(RRj)
2 x 1.96
(Eq. H-3)
where RR , is the 95% upper bound on the RR, estimate (for the /th exposure category) and RR, is
the 95% lower bound on the RR, estimate.
This document is a draft for review purposes only and does not constitute Agency policy.
10/20/09	H-1 DRAFT—DO NOT CITE OR QUOTE
-------
s
s
TO
Co
S"4
si
Table H-l. Extra-risk calculation" for environmental exposure to 1.82 ppm TCE (the LECoi for RCC
incidence)b using a linear exposure-response model based on the categorical cumulative exposure results of
Charbotel et al. (2006), as described in Section 5.2.2.1.2.
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P








Cond.





Exposed







Prob. of

prob. of
Exp.


Exposed
Exposed
prob. of
Exposed




All cause
Prob. of
surviving
RCC
RCC
duration
Cum.
Exposed
all cause
prob. of
surviving cond. prob.
Interval

All cause
RCC
hazard
surviving
up to
cancer
incidence
mid
exp. mid
RCC
hazard
surviving
up to
of RCC in
number
Age
mortality
incidence
rate
interval
interval
hazard rate in interval
interval
interval hazard rate
rate
interval
interval
interval
(0
interval
(xl05/yr)
(xl05/yr)
(A*)
(?)
(S)
0h)
(Ro)
(xtime)
(xdose)
(lix)
(h Xx)
(qx)
(Sx)
(Rx)
1
<1
685.2
0
0.0069
0.9932
1.0000
0.000000
0.000000
0.5
2.77
0.000000
0.0069
0.9932
1.0000
0.000000
2
1-4
29.9
0
0.0012
0.9988
0.9932
0.000000
0.000000
3
16.61
0.000000
0.0012
0.9988
0.9932
0.000000
3
5-9
14.7
0
0.0007
0.9993
0.9920
0.000000
0.000000
7.5
41.52
0.000000
0.0007
0.9993
0.9920
0.000000
4
10-14
18.7
0.1
0.0009
0.9991
0.9913
0.000005
0.000005
12.5
69.20
0.000006
0.0009
0.9991
0.9913
0.000006
5
15-19
66.1
0.1
0.0033
0.9967
0.9903
0.000005
0.000005
17.5
96.88
0.000006
0.0033
0.9967
0.9903
0.000006
6
20-24
94
0.2
0.0047
0.9953
0.9871
0.000010
0.000010
22.5
124.56
0.000013
0.0047
0.9953
0.9871
0.000013
7
25-29
96
0.7
0.0048
0.9952
0.9824
0.000035
0.000034
27.5
152.24
0.000049
0.0048
0.9952
0.9824
0.000048
8
30-34
107.9
1.6
0.0054
0.9946
0.9777
0.000080
0.000078
32.5
179.91
0.000117
0.0054
0.9946
0.9777
0.000114
9
35-39
151.7
3.2
0.0076
0.9924
0.9725
0.000160
0.000155
37.5
207.59
0.000245
0.0077
0.9924
0.9724
0.000237
10
40-44
231.7
6.3
0.0116
0.9885
0.9651
0.000315
0.000302
42.5
235.27
0.000504
0.0118
0.9883
0.9650
0.000484
11
45-49
352.3
11
0.0176
0.9825
0.9540
0.000550
0.000520
47.5
262.95
0.000919
0.0180
0.9822
0.9537
0.000869
12
50-54
511.7
17.3
0.0256
0.9747
0.9373
0.000865
0.000801
52.5
290.63
0.001507
0.0262
0.9741
0.9367
0.001393
13
55-59
734.8
26.2
0.0367
0.9639
0.9137
0.001310
0.001175
57.5
318.31
0.002375
0.0378
0.9629
0.9124
0.002127
14
60-64
1140.1
36.2
0.0570
0.9446
0.8807
0.001810
0.001549
62.5
345.99
0.003409
0.0586
0.9431
0.8786
0.002909
15
65-69
1727.4
44.6
0.0864
0.9173
0.8319
0.002230
0.001777
67.5
373.67
0.004358
0.0885
0.9153
0.8286
0.003456
16
70-74
2676.4
49
0.1338
0.8747
0.7631
0.002450
0.001750
72.5
401.35
0.004961
0.1363
0.8726
0.7584
0.003518
17
75-59
4193.2
51.6
0.2097
0.8109
0.6675
0.002580
0.001554
77.5
429.03
0.005407
0.2125
0.8086
0.6617
0.003223
18
80-84
6717.2
44.4
0.3359
0.7147
0.5412
0.002220
0.001021
82.5
456.71
0.004809
0.3384
0.7129
0.5351
0.002183







Ro =
0.010736





Rx =
0.020586
Extra risk = (Rx
-Ro)/( 1-
Ro) = 0.00996











0
* v §
S » si
a, Co'
TO Sj-
1	I»
O
S
>S
TO
TO'
*
o
VO
to
o
to
-------
Column A: interval index number (/').
^3	Column B: 5-year age interval (except <1 and 1-4) up to age 85.
^	^	Column C: all-cause mortality rate for interval i (x 105/year) [2004 data from CDC (2007)].
g	?	Column D: RCC incidence rate for interval i (x 105/vear) (2001-2005 SEER data [http://seer.cancer.govl).
>	o	Column E: all-cause hazard rate for interval/'(h*,) [= all-cause mortality rate x number of years in age interval].0
S	s	Column F: probability of surviving interval /' without being diagnosed with RCC ((/,) [= e\p(-/?*,)|.
\>	j?	Column G: probability of surviving up to interval /' without having been diagnosed with RCC (.V,) |.V, = 1; ,V, = .V, i x q, for i > 1],
J ;	i	Column H: RCC incidence hazard rate for interval /' (/?,) [= RCC incidence rate x number of years in interval].
3^5'	Column I: conditional probability of being diagnosed with RCC in interval /' [= (/?,//?*,) x ,S', x (1-g,)], i.e., conditional upon surviving up to interval /' without
q	^	having been diagnosed with RCC [Ro, the background lifetime probability of being diagnosed with RCC = the sum of the conditional probabilities
o ^	§-
55? to
Cl
>r	across the intervals].
S ^ Column J: exposure duration (in years) at mid-interval (xtime).
Column K: cumulative exposure mid-interval (xdose) [= exposure level (i.e., 1.82 ppm) x 365/240 x 20/10 x xtime] (365/240 x 20/10 converts continuous
o	environmental exposures to corresponding occupational exposures).
§ ^ Column L: RCC incidence hazard rate in exposed people for interval i (hx,) [= ht x (1 + (3 x xdose), where (3 = 0.001205 + (1.645 x 0.0008195) = 0.002554]
to
*
[0.001205 per ppm x year is the regression coefficient obtained from the weighted linear regression of the categorical results (see Section 5.2.2.1.2).
To estimate the LEC0i, i.e., the 95% lower bound on the continuous exposure giving an extra risk of 1%, the 95% upper bound on the regression
coefficient is used, i.e., MLE + 1.645 x SE],
(-5 ^ Column M: all-cause hazard rate in exposed people for interval /' (h*x,) [= h*{ + (Ax, - /?,)].
° ,3 Column N: probability of surviving interval /' without being diagnosed with RCC for exposed people (qx,) [= cxp(-/?*x,)|.
Column O: probability of surviving up to interval /' without having been diagnosed with RCC for exposed people (.S'x,) |,S'x; = 1; ,S'x, = ,S'x, , x qxh]. for /> 1],
Column P: conditional probability of being diagnosed with RCC in interval /' for exposed people [= (/?x,//?*x,) x Sxt x (1 ~qxt)\ (Rx, the lifetime probability of
being diagnosed with RCC for exposed people = the sum of the conditional probabilities across the intervals).
o
a Using the methodology of BEIRIV (1988).
b The estimated 95% lower bound on the continuous exposure level of TCE that gives a 1% extra lifetime risk of RCC.
0 For the cancer incidence calculation, the all-cause hazard rate for interval /' should technically be the rate of either dying of any cause or being diagnosed with
the specific cancer during the interval, i.e., (the all-cause mortality rate for the interval + the cancer-specific incidence rate for the interval—the cancer-specific
mortality rate for the interval [so that a cancer case isn't counted twice, i.e., upon diagnosis and upon death]) x number of years in interval. This adjustment
was ignored here because the RCC incidence rates are small compared with the all-cause mortality rates.
MLE = maximum likelihood estimate, SE = standard error.
LtJ
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1
2
3
H-4
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
APPENDIX I
EPA Response to Major Peer Review and
Public Comments
5
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.1. PBPK Modeling (SAB Report Section 1): Comments and EPA Response
1.1.1.	SAB Overall Comments:
The Panel commended the updated physiologically-based pharmacokinetic (PBPK)
model (Chiu et al., 2009; Evans et al., 2009) for dose-response assessment. The Panel found that
while the PBPK model was generally well presented, its description was incomplete in that
mass-balance equations were not presented. The Panel provided suggestions to improve model
documentation and clarity, including clearer descriptions of the strategy behind the model
structure and the biological relevance of each model equation. Model assumptions need to be
more clearly described and the consequences of potential violations of these assumptions should
be discussed. In addition, a more detailed justification was needed for the handling of between-
animal variability in the model. The Panel agreed that use of the Bayesian framework for
estimation and characterization of the PBPK model parameter uncertainties was appropriate.
However, a more thorough description was needed for the choice of prior distributions, the
Bayesian fitting methodology, and the fit of the posterior distribution for each model parameter.
The Panel also generally endorsed the hierarchical calibration approach that uses the posterior
results in mice to establish the rat priors, and the rat posterior results to set the human priors.
The Panel also recommended performance of a local sensitivity analysis to identify key model
parameters that drive changes in modeling results.
1.1.2.	Major SAB Recommendations and EPA Response:
1.1.2.1. PBPK model structure (SAB Report Section la)
•	Provide a better description of the final model structure and, in particular, provide a
revised model structure diagram that identifies model parameters with model states and
pathways (flows).
EPA response: EPA accepts this recommendation and has provided revised model structure
diagrams in Appendix A, Section A.4.1.
•	Clarify the strategy behind the model structure and describe the biological relevance of
each model equation.
EPA response: EPA accepts this recommendation and has clarified the model structure and
equations, and their biological relevance, in Appendix A, Section A.4.1.
•	Document model assumptions and discuss the consequences of potential violations of
these assumptions (e.g. impacts on bias and accuracy).
EPA response: EPA accepts this recommendation and has expanded the discussion of
limitations of the model to include added discussion of model assumptions and the consequences
of potential violations in Section 3.5.7.4.
6
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
•	Provide a more detailed justification for how between animal variability is accounted for
in the model.
EPA response: EPA accepts this recommendation and has expanded the discussion of how
between animal variability is addressed in the model in Section 3.5.5.2.
1.1.2.2.	Bayesian statistical approach (SAB Report Section lb)
•	Present better descriptions and/or details on the choice of prior distributions, the Bayesian
fitting methodology and fit of the posterior distribution for each model parameter.
EPA response: EPA accepts this recommendation and has added a description of the choice of
prior distribution functions in Section 3.5.5.2; presented a description of the overall Bayesian
posterior distribution function used in the parameter fitting in Section A.4.4.; and added
graphical presentation to Section A. 5.1 of the posterior distributions, in comparison with the
prior distribution, for each model parameter. In addition, the use of the terms "population" and
"group" have been clarified throughout Chapter 3 and Appendix A.
•	Provide some information on correlations around posterior medians for species-specific
parameters.
EPA response: EPA accepts this recommendation and provided tables of correlation coefficients
in Appendix A, Section A.5.1.
•	Supply more information on the model ordinary differential equations and on the
likelihood function used in the Bayesian estimation.
EPA response: EPA accepts this recommendation and has supplied more information on the
model ordinary differential equations in Appendix A, Section A.4.1, and more information on
the likelihood function in Appendix A, Section A.4.3.4.
1.1.2.3.	Parameter Calibration (SAB Report Section lc)
•	Improve the quality and the description of the assumptions underlying the use of the
hierarchical approach to parameter calibration. Help the reader to understand the extent to
which these assumptions are used consistently throughout the parameter calibration
process.
EPA response: EPA accepts this recommendation and revised Table A-4 to clarify the scaling
assumptions consistently used throughout the parameter calibration process, and revised Section
3.5.5.3 to clarify the description of the assumptions underlying the hierarchical approach.
1.1.2.4.	Model fit assessment and dose metric projections (SAB Report Section Id)
•	Move some graphical presentations from the linked graphics documents into the body of
the report or into Appendix A.
7
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA response: EPA accepts this recommendation and has moved (in a more condensed form)
graphical presentations of the PBPK model predictions as compared to the in vivo data to the
body of Appendix A.
•	Incorporate more discussion on model fit and in particular indicate areas where the model
fits well and areas where it did not fit well. Tie this discussion somehow to Table 3-41.
EPA response: EPA accepts this recommendation and has incorporated more discussion of
model fit in Section 3.5.6.3 indicating areas where the model fits well and areas where it did not
fit well. This discussion is tied to the Table previously labeled "3-41," as recommended. In
addition, the interpretation of the residual error GSD is more closely tied to this revised
discussion.
•	Include graphs that show predicted versus observed values for all data points used in the
analysis (one graph per endpoint).
EPA response: EPA accepts this recommendation and has added graphics showing predicted
versus observed values for all data points used in the analysis (one graph per endpoint) to Section
3.5.6.3. The width of the residual error GSDs are also included on these graphs for comparison.
In addition, this is tied to the revised discussion on model fit and the Table previously labeled "3-
41."
•	To help readers identify which parameters are better specified than others, provide a table
of model parameters listed in reverse order by the width of their posterior variability
(width of the IQR or width of 95% CI).
EPA response: EPA accepts this recommendation and has added a table to Section 3.5.6.2 of
model parameters listed in reverse order by the width of their posterior variability, indicated by
the width of 95% CI.
•	Identify those parameters with very different prior and posterior distributions and discuss
why this might be a reasonable result of the parameter calibration process. An alternative
would be to provide a table where parameters are ranked based on the percent change of
the posterior from the prior.
EPA response: EPA accepts this recommendation and has included a table in Section 3.5.6.2
that indicates the fold-change between the prior and posterior medians. This table is already
sorted by reverse order of the width of the posterior variability (see previous recommendation).
In order to identify those parameter with more different priors and posteriors, the fold-change
was bolded if the change was greater than 3-fold. It is noted in the revised text for Section
3.5.6.2 that those parameters with shifts greater than 3-fold had prior confidence intervals greater
(sometimes substantially) than 100-fold, so that such shifts are reasonable in that context.
8
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
•	Clarify which parameters are related to variability and which address parameter
uncertainty. Separate the discussion of the two types of parameters.
EPA response: EPA accepts this recommendation and has replaced the tables in Section 3.5.6.2
that previously showed combined uncertainty and variability with tables that separately
summarize parameter uncertainty and variability. This separation of uncertainty and variability
has the added benefit of removing the appearance that posterior parameter distributions appear
flatter than prior distributions, since posterior parameter uncertainty should always be less than
or equal to prior parameter uncertainty. In addition, the text of Section 3.5.6.2 has been revised
to discuss separately estimates of the central tendency of the population from estimates of
population variability.
1.1.2.5. Lack of adequate sensitivity analysis (SAB Report Section le)
•	Perform a local sensitivity analysis, starting from the final fitted PBPK model, to assess
how small changes in model parameter estimates impact predictions. Provide graphical
presentations of the sensitivity of the model to changes in key model parameters in the
final documentation.
EPA response: EPA accepts this recommendation and has conducted a local sensitivity analysis
starting from the final fitted PBPK model, and assessing how small changes (5% increase or
decrease) in model parameter estimates impact predictions. Two types of model predictions are
analyzed. First, in Section 3.5.6.4, the sensitivity of predictions of calibration data is assessed,
including a graphical presentation of the number of data points that are sensitive to each
parameter. Second, in Section 3.5.7.2, the sensitivity of prediction of dose metrics is assessed,
including a graphical presentation of the sensitivity coefficient for each parameter and dose
metric.
1.1.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with the extent and degree of variability of GSH
conjugation in humans predicted by the PBPK model.
EPA response: In accordance with SAB recommendations (see response below in Section
1.5.2.3), EPA has revised the discussions in Sections 3.3 and 3.5 to reflect the uncertainty in GSH
conjugation predictions in humans.
•	Some public commenters disagreed with the extent of population variability predicted by
the PBPK model for some parameters.
EPA response: The External Review Draft reported posterior distributions as lumped
uncertainty and variability. For the parameters raised as a concern in the comments, the high
apparent variability is actually predominantly uncertainty, so the extent of population variability
9
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
is not exceedingly high. In accordance with SAB recommendations (see response above in
Section 1.1.2.4), EPA has revised the description of posterior parameters to separate uncertainty
and variability, providing additional clarity the posterior predictions.
•	Some public commenters recommended that EPA perform a sensitivity analysis on the
PBPK model.
EPA response: In accordance with SAB recommendations (see response above in Section
1.1.2.5), EPA has conducted a local sensitivity analysis of the PBPK model.
•	Some public commenters recommended that EPA incorporate additional data in its PBPK
model.
EPA response: In accordance with SAB recommendations (see response below in Section
1.5.2.2), EPA incorporated additional data on TCA bioavailability in the TCA sub-model of the
PBPK model. Additional data were evaluated in Appendix A for the purposes of additional
validation, but were not directly incorporated in the PBPK model.
1.2. Meta-Analyses of Cancer Epidemiology (SAB Report Section 2): Comments and EPA
Response
1.2.1.	SAB Overall Comments:
The Panel agreed that EPA's updated meta-analyses for kidney cancer, lymphoma and
liver cancer followed the National Research Council (2006) recommendations. The Panel agreed
with EPA's conclusions that TCE increased the risk for the three cancers studied, based on
appropriate inclusion criteria for studies, the methods of conducting the meta-analysis that
included consideration of bias and confounding, and the robustness of the findings based on the
tests for heterogeneity and sensitivity. The Panel also suggested performing a meta-analysis for
lung cancer to further support the absence of smoking as a possible confounder.
1.2.2.	Major SAB Recommendations and EPA Response:
•	Provide a rationale for the three cancer sites selected for the meta-analysis. The rationale
could be nicely summarized in a table.
EPA Response: EPA accepts this recommendation and has added text to Section 4.1 and
Appendix C.
•	Consider including meta-analysis for lung cancer for confounding purposes or other sites
for comparison for which some association with TCE exposure has been reported in
epidemiologic studies, such as childhood leukemia and cervical cancer. It might also be
possible to provide this information without a formal meta-analysis.
10
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA Response: EPA accepts this recommendation and has included a meta-analysis for lung
cancer in Appendix C. Additionally, in the discussion in Chapter 4 of the possible role of
smoking in confounding the association between TCE exposure and kidney cancer, EPA
compares the relative risk estimates for lung and kidney cancers in five smoking cohorts and
discusses the expected contribution by smoking to kidney cancer in Raaschou-Nielsen et al.
(2003), which was estimated as 1 - 6%, far smaller than the 20 - 40% excess reported in this
study. Meta-analyses were not conducted for other cancer types for which there may have been
suggestive associations because there was inadequate reporting in the cohort studies, and, for
childhood leukemia, there were too few studies of sufficient quality.
•	Provide measures of heterogeneity such as the I statistic for each meta-analysis.
Although this information was provided and accurately explained in Appendix C, it was
mischaracterized at several points in the primary document. For example, the summary
of the kidney cancer meta-analysis on p. 4-167 of the primary document states that "there
was no observable heterogeneity across the studies for any of the meta-analyses," but
Appendix C indicates "the I value of 38% suggested the extent of the heterogeneity was
low-to-moderate." Non-significant heterogeneity is indeed observed heterogeneity.
EPA Response: EPA accepts this recommendation and has provided measures of heterogeneity
in the primary document. EPA has also corrected this sentence in 4.4.2.5; it now reads "there
was no observable heterogeneity for any of the meta-analyses of the 15 studies and no indication
of publication bias."
•	Evaluate the likely impact of converting odds ratios to relative risk estimates (i.e., using
the method of Greenland (2004) or Zhang and Yu (1998)), and decide if necessary to
perform these conversions for the meta-analysis.
EPA Response: The papers cited by the SAB describe methods for correcting odds ratios (ORs)
in studies of common outcomes. Each of the cancer types for which EPA did meta-analyses has
a background incidence <10% and is thus considered a rare disease, so no correction should be
necessary. In the case of rare diseases, only high ORs might notably overestimate RRs. In the
TCE studies, only Hardell et al. (1994) reported an OR high enough to be of potential concern, a
Mantel-Haenszel-adjusted OR of 7.2 for non-Hodgkin lymphoma. According to Zhang and Yu
(1998), the Mantel-Haenszel adjustment is a suitable way to estimate the RR; in fact, in the
example they provide, the Mantel-Haenszel adjustment outperforms the adjustment they are
proposing. Furthermore, according to McNutt et al. (2003), the Zhang and Yu method is
incorrect when applied to an adjusted OR and will produce a biased estimate when confounding
is present. Additionally, the model-based methods for estimating a RR from a case-control study
described by Greenland (2004) are only applicable when one has the raw data. Thus, neither of
the papers cited by the SAB provides an appropriate way to convert the Hardell et al. OR. In any
11
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
event, any overestimation that might occur by treating the Hardell et al. OR as a RR estimate is
negligible in the overall analysis. Removing the study all together only decreases the RRm from
1.23 to 1.21, and the latter result is still statistically significant (p = 0.004).
•	Change the terminology regarding the meta-analysis results for 'lymphoma' to 'non-
Hodgkin lymphoma' throughout the document.
EPA Response: EPA accepts this recommendation and has revised the terminology throughout
the document.
1.2.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters requested a glossary of epidemiology terms be included in the
document.
EPA response: EPA did not implement this recommendation, as definitions of epidemiologic
terms can be easily found from authoritative sources on the internet.
•	Some public commenters suggested that EPA examine the TCE subregistry for
information about the association between TCE and cancer.
EPA response: EPA did not implement this recommendation with respect to cancer, as the
ATSDR TCE subregistry provides only limited information on cancer outcomes as analyses are
for total cancers and less informative than for cancer types. EPA did consider, however,
observations on neurotoxicity and other noncancer outcomes.
•	Some public commenters disagreed with the meta-analysis conclusions from the External
Review Draft, noting heterogeneity of findings, lack of consistent exposure-response, and
other methodological problems. These comments noted that while EPA's meta-analysis
methods and summaries are generally consistent with recent published summaries of this
literature, the commenters did not agree with EPA's interpretation. These comments
asserted that it is more accurate to report the epidemiologic evidence as "mixed" rather
than "consistent" or "robust."
Other public commenters agreed with the meta-analysis conclusions from the External
Review Draft, noting that epidemiologic studies are usually biased towards the null,
making it harder to detect a true causal relationship.
EPA response: In accordance with the SAB review, EPA maintains its meta-analysis
conclusions. EPA agrees with the public commenters that the characterization of the general
association between overall TCE exposure and cancer is "modest"; this was one of the points
explicitly brought out in the discussion in Section 4.11.2.1.2 concerning the strength of the
association. EPA also carefully considered the questions raised by the public commenters
regarding consistency of the results and regarding alternative explanations for these findings.
12
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
This consideration is discussed in detail in Section 4.11.2.1. A strength of the meta-analytic
approach is its ability to assess heterogeneity among studies, which is of particular importance in
situations in which the overall relative risk estimate is modest and in situations in which results
from individual studies may be quite imprecise because of sample size limitations. In reviewing
the available data, including the results of the meta-analyses, EPA noted that chance was not
supported as an explanation for the findings, nor was there support for confounding by other
known or suspected risk factors as an explanation for the results.
1.3. Non-Cancer Hazard Assessment (SAB Report Section 3): Comments and EPA
Response
1.3.1.	SAB Overall Comments:
EPA has provided a comprehensive synthesis of the available evidence regarding the
effects of TCE and its major metabolites on the central nervous system, the kidney, the liver, the
immune system, the male reproductive system, and the developing fetus. One issue of concern
was the inconsistencies between reported levels of glutathione conjugation pathway metabolites.
The Panel recommended that the impact of these divergent levels be more transparently
presented. The Panel recommended inclusion of the potential for TCE-induced immune
dysfunctions (i.e., immunosuppression, autoimmunity, inappropriate and/or excessive
inflammation) to mechanistically underlie other adverse health endpoints.
1.3.2.	Major SAB Recommendations and EPA Response:
•	If additional endpoints of renal dysfunction (e.g. diuresis, increased glucose excretion)
were present in the reported studies, they should be included in the report. Often only
one or two parameters of renal function and histopathology were presented. A better
overall description of renal dysfunction should be presented if available (especially for
animal studies).
EPA Response: EPA accepts this recommendation, and has added the information to all studies
where such data are available.
•	There should be a better description of the location of the renal lesion, including nephron
segment, if known. For example, TCE and DCVC appeared to affect the proximal tubule
at the level of the outer stripe of the medulla (S3 segment of proximal tubule). Is this the
site of lesions seen with other TCE metabolites? Explaining the role (or lack of a role) of
any other TCE metabolites in TCE nephrotoxicity could be strengthened by comparing
the sites of the renal lesion.
EPA Response: EPA accepts this recommendation, and has added the information to all studies
where such data are available.
13
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
•	On page 4-338, please clarify the use of the phrase, "subpopulation levels", on lines 31
and 33.
EPA Response: EPA accepts this recommendation, and has clarified the use of
"subpopulations."
•	A statement should be added that the spectrum of TCE-induced immune dysfunctions
(immunosuppression, autoimmunity, inappropriate and/or excessive inflammation)
included in this EPA draft report has the potential to produce adverse effects that are seen
well beyond lymphoid organs and involving several other physiological tissues and
systems. The types of immune-inflammatory dysfunctions described in this report have
been observed to affect function and risk of disease in the nervous system (e.g., loss of
hearing), the skin, the respiratory system, the liver, the kidney, the reproductive system
(e.g., male sterility), and the cardiovascular system (e.g., heart disease, atherosclerosis).
EPA Response: EPA accepts this recommendation, and has added statements to Sections 4.6
and 4.6.3.1 that immume-related and inflammatory effects, particularly cell-mediated immunity,
may influence a broader range of disease, including cancer.
•	A statement should be added to emphasize the cell-mediated immune effects of TCE as
some of this has been supported by the human epidemiology data and the issue is
pertinent to risk of cancer.
EPA Response: See previous response.
1.3 .3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with EPA's draft conclusion that TCE poses a human
health hazard for developmental cardiac effects due to limitations in the available data.
EPA response: In accordance with the SAB review, EPA acknowledges the limitations of the
available data, but maintains its conclusion that TCE poses a human health hazard for
developmental cardiac effects.
•	Some public commenters disagreed with EPA's draft conclusion TCE poses a human
health hazard for immunotoxicity because additional confirmatory studies are needed.
EPA response: In accordance with the SAB review, EPA concludes that adequate data are
available to conclude that TCE poses a human health hazard for immunotoxicity.
14
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.4. Carcinogenic Weight of Evidence (SAB Report Section 4): Comments and EPA
Response
1.4.1.	SAB Overall Comments:
The Panel agreed with EPA's conclusion that TCE is "Carcinogenic to Humans" by all
routes of exposure. This is based on convincing evidence of a causal association between TCE
exposure and kidney cancer, compelling evidence for lymphoma, and more limited evidence for
liver cancer as presented in the draft document. The epidemiologic data, in the aggregate, were
quite strong. The summary risk estimates from the meta-analyses provided a clear indication of
a cancer hazard from TCE. In addition, both animal data and toxicokinetic information provide
biological plausibility and support the epidemiologic data.
1.4.2.	Major SAB Recommendations and EPA Response:
•	The immune effects as highlighted in the hazard assessment should be referred to in the
conclusion especially in the criteria of biological plausibility and coherence because of
the relationship between immune system dysfunction and cancer risk.
EPA Response: EPA accepts this recommendation, and has added a statement to Section
4.11.2.1.6 that immune-related effects should also be considered in the biologic plausibility of
TCE carcinogenicity.
•	Although the summary evaluation focused on the scientific evidence and meta-analysis
for kidney, lymphoma and liver cancers, there is also some suggestive evidence for TCE
as a risk factor for cancer at other sites including bladder, esophagus, prostate, cervix,
breast and childhood leukemia. This evidence that also supports the conclusion should be
mentioned in the summary evaluation (section 4.11.2.1).
EPA Response: EPA accepts this recommendation, and has added a statement mentioning the
suggestive evidence of association between TCE and other types of cancer in Section
4.11.2.1.10.
•	Add a paragraph describing the definition of lymphoma as used in IRIS. Change the
terminology regarding the meta-analysis to 'non-Hodgkin lymphoma' instead of
'lymphoma', throughout the document. The term 'NHL' more accurately describes the
intent of the analysis as well as the overwhelming majority of cases in the analysis,
despite changing classification schemes. The focus of the meta-analysis on NHL and the
exact classifications the meta-analysis includes where it may diverge from classical NHL
(as in studies that included chronic lymphocytic leukemia) should be clearly explained in
both Appendix C and in the Hazard Characterization document (section 4.6.1.2.2).
15
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA Response: EPA accepts this recommendation and has added text describing the definition
of NHL as used in the assessment, in addition to revising the terminology and indicating
divergent case definitions in both Appendix C and Section 4.6.1.2.2.
•	To assist the reader, please include references in the summary section (section 4.11.2).
For example, "The other 13 high-quality studies [note: besides Hardell and Hansen]
reported elevated Relative Risk estimates with overall TCE exposure that were not
statistically significant." References for statements like this would be helpful. The Panel
counted fewer than 13 studies in the meta-analysis after subtracting out Hardell and
Hansen, and not all of these showed elevated risk estimates, so it would be helpful for the
reader to know which 13 studies this statement refers to.
EPA Response: EPA accepts this recommendation and has added references to conclusions in
section 4.11.2.1.
1.4.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with EPA's draft conclusion in the External Review
Draft that TCE is "Carcinogenic to humans," judging the epidemiologic evidence to be
inadequate due to limitations of the body of evidence. Limitations cited by these
comments include exposure assessment limitations, potential unmeasured confounding,
potential selection biases, and inconsistent findings across groups of studies. Comments
also cited limitations in the experimental animal data, such as conflicting or negative
experimental animal data for kidney and immune tumors. These comments suggested
that a classification of "likely to be carcinogenic in humans" or "suggestive evidence of
carcinogenicity" would be more appropriate. Some of these comments cited the National
Academy of Sciences (NAS)/National Research Council (NRC) Contaminated Water
Supplies at Camp Lejeune: Assessing Potential Health Effects (NRC, 2009) as support.
Other public commenters supported EPA's draft conclusion in the External Review Draft
that TCE is "Carcinogenic to humans."
EPA response: In accordance with the SAB review, EPA concludes that TCE is "Carcinogenic
to humans." EPA also notes that the NRC (2009) Camp Lejeune report was discussed during the
SAB review meetings. The meeting minutes from the June 24, 2010 teleconference call state
that "Panelists discussed the extent to which the EPA draft risk assessment document should
discuss or compare its findings and conclusions to those of the 2009 NAS Report on Camp
Lejuene. It was generally agreed that it was not necessary to compare EPA conclusions to all the
other reviews, particularly in view of the different criteria applied across reviews, different
studies used across assessments and different scopes of each review and the fact that the current
draft risk assessments carries out a meta-analysis that was not considered in the 2009 NAS
16
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
review." (SAB, 2010a) In the meeting minutes from the December 15, 2010 SAB quality review
teleconference call, the Panel chair stated that "the material reviewed by the Panel was different
from what the NAS had reviewed" (SAB, 2010b).
1.5. Role of Metabolism (SAB Report Section 5): Comments and EPA Response
1.5.1.	SAB Overall Comments:
The Panel agreed with EPA's conclusion that oxidative metabolites of TCE were likely
responsible for mediating the liver effects. The Panel recommended that EPA examine studies
that provided quantitative assessment of trichloroacetic acid (TCA) and dichloroacetic acid
(DCA) formation after TCE exposure. Dose-response modeling, similar to that performed for
tetrachloroethylene, may be considered by EPA to provide scientifically-based information on
relative contribution, or lack thereof, of TCA and/or DCA to the liver carcinogenesis effect of
TCE.
EPA has provided a clear and comprehensive summary of the available evidence that
metabolites derived from glutathione (GSH) conjugation of TCE mediate kidney effects. The
Panel noted that uncertainties exist with regard to the extent of formation of the dichlorovinyl
metabolites of TCE between humans and rodents. The issue of quantitative assessment of the
metabolic flux of TCE through the GSH pathway vs. the oxidative metabolism pathway needs to
be considered carefully. A more complete discussion of the strengths and limitations of the
analytical methodologies used should be provided to address the large discrepancies in estimates
of S-dichlorovinyl glutathione (DCVG) formation.
1.5.2.	Major SAB Recommendations and EPA Response:
1.5.2.1.	Mediation of TCE-Induced Liver Effects by Oxidative Metabolism (SAB
Report Section 5a)
•	No major recommendations in this section.
1.5.2.2.	Contribution of TCA to Adverse effects on the Liver (SAB Report Section 5b)
•	The EPA should examine studies that provide quantitative assessment of TCA and DCA
formation after TCE exposure in vivo and draw conclusions with regards to the relative
amount and kinetics of the oxidative metabolites of interest for liver toxicity.
EPA response: Most studies of TCA following TCE exposure have already been incorporated
into the PBPK model-based analyses, and previous studies of DCA following TCE exposure are
limited by the rapid clearance of DCA at low concentrations and analytical artifacts in DCA
detection. Section 4.5.6.1 has been revised to include discussion of the studies by Delinksy et al.
(2005) and Kim et al. (2009), which use more reliable analytic methods to quantify DCA
formation after TCE exposure in vivo.
17
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
•	A careful evaluation of the concentration-time kinetics is needed to achieve certainty in
the comparisons of liver effects and the conclusions drawn by the EPA which suggest
that TCA-induced adverse liver effects do not explain those observed with TCE. Equally
important is to fully consider the bioavailability of TCE itself with regards to the vehicle
effects between studies.
EPA response: EPA assumes that the first part of this comment refers to the issue of TCA
bioavailability, which is mentioned in the narrative text preceding this recommendations. EPA
has incorporated into Section 4.5.6.2.1 an updated analysis of TCA bioavailability and its impact
on conclusions regarding the role of TCA in TCE-induced hepatomegaly (Chiu, In Press). With
respect to TCE vehicle effects, there are not enough kinetic data using different vehicles to
quantitatively address vehicle effects. However, it is noted that if they are important, they may
be a significant contributor to the residual variability in the combined analysis of TCE-induced
hepatomegaly.
•	The body of the document could be further strengthened by reporting EPA's evaluation
on the strength of the specific criteria used for phenotypic classification described in each
study discussed, and noting where specific criteria were not reported. While most of this
information was included in the appendix, the EPA may consider constructing a summary
table for Section 4.5.6.
EPA response: Section 4.5.6.3.3.1 has been revised to note that no specific criteria are usually
given as to what constitutes "basophilic" or "eosinophilic," with the one exception of Carter et
al. (2003) noted. For immunochemical staining, it is noted that some studies use negative
controls as a comparison.
•	Dose-response modeling, similar to that performed for PERC, may be considered by the
EPA to provide science-based information on relative contribution, or lack thereof, of
TCA and/or DCA to the apical liver carcinogenesis effect of TCE. While data gaps exist
and there are limitations in the comparisons between independent cancer bioassays, the
document should clearly state what the limitations are should such analysis be deemed
futile.
EPA response: EPA agrees that a quantitative analysis of the relative contributions of TCA
and/or DCA to TCE-induced liver carcinogenesis would be useful if feasible. However, as noted
in the revised Section 4.5.6.3.2.5, such an analysis is precluded by the high degree of
heterogeneity both within and across the databases for TCE and its oxidative metabolites. The
revised section gives several examples of this substantial variability in cancer bioassay data.
•	The draft assessment may be strengthened by including information from human use of
DCA in clinical practice.
18
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA response: Human data on use of DCA in clinical practice is summarized in EPA's
Toxicological Review of Dichloroacetic Acid (2003c), and reference has been made to this
document in Section 4.5.6. In particular, it is noted that data on DCA in humans are scarce and
complicated by the fact that available studies have predominantly focused on individuals who
have a pre-existing (usually severe) disease.
1.5.2.3. Role of GSH-conjugation pathway on TCE-induced kidney effects (SAB
Report Section 5c)
•	The issue of quantitative assessment of the metabolic flux of TCE through the GSH
pathway vs. the oxidative metabolism pathway should be considered carefully since
uncertainties exist with regard to the extent of formation of the dichlorovinyl metabolites
of TCE between humans and rodents. EPA may need to provide appropriate reservations
to the conclusions based on the limited data for GSH metabolites.
•	The discussion of how each of the in vitro and in vivo data sets were used to estimate
DCVG formation parameters for the PBPK model should be more transparent indicating
strengths and weaknesses in the database.
EPA responses: EPA accepts these two related recommendations. EPA has revised Section
3.3.3.2 to articulate additional reservations as to its conclusions regarding the except of
formation of dichlorovinyl metabolites of TCE between rodents and humans, and to be more
transparent regarding the strengths and weaknesses in vitro and in vivo datasets. In addition,
cross-references to this discussion have been added in the context of the PBPK model parameters
and predictions to Section 3.5.4.3, 3.5.5.1, 3.5.6.3.3, 3.5.7.3.1, 3.5.7.3.2, 3.5.7.4, and 3.5.7.5.
1.5.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with EPA's conclusion that DCA may play a role in
TCE-induced liver effects. These comments also recommended EPA take into account
the bioavailability of TCA in its evaluation of liver effects.
EPA response: In accordance with SAB recommendations, EPA has re-examined the data on
the contributions of TCA and/or DCA to TCE-induced liver effects, including incorporation of
data on TCA bioavailability, in Section 4.4. However, EPA's conclusion remains that TCA
cannot adequately account for account the liver effects of TCE.
•	Some public commenters disagreed with EPA's conclusion that GSH-conjugation-
derived metabolites play the primary role in TCE-induced nephrotoxicity and
nephrocarcinogeni city.
EPA response: EPA maintains its conclusions, and notes that both the SAB review and the NRC
(2006) report support the conclusion that the GSH pathway is primarily responsible for TCE-
induced kidney effects.
19
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.6. Mode of Action (MOA) (SAB Report Section 6): Comments and EPA Response
1.6.1.	SAB Overall Comments:
The Panel agreed that the weight of evidence supports a mutagenic MOA for TCE-
induced kidney tumors. However, the Panel concluded that the weight of evidence also
supported an MOA involving cytotoxicity and compensatory cell proliferation and including
these may more accurately reflect kidney tumor formation than does a mutagenic mechanism
alone. The combination of cytotoxicity, proliferation and DNA damage together may be a much
stronger MOA than any individual components.
The Panel agreed that the data are inadequate to conclude that any of the TCE-induced
cancer and non-cancer effects in rodents are not relevant to humans.
The Panel agreed that there is inadequate support for peroxisome proliferator activated
receptor alpha (PPARa) agonism and its sequellae being key events in TCE-induced human liver
carcinogenesis. Recent data from animal models (Yang et al., 2007) suggest that activation of
PPARa is an important but not limiting factor for the development of mouse liver tumors, and
additional molecular events may be involved. The Panel viewed the mode of action (MOA) for
liver carcinogenicity in rodents as complex rather than unknown. It is likely that key events from
several pathways may operate leading to acute, subchronic and chronic liver toxicity of TCE.
1.6.2.	Major SAB Recommendations and EPA Response:
1.6.2.1. Hazard Assessment and Mode of Action (SAB Report Section 6a)
•	The impact of the inconsistencies in data on the quantity of GSH pathway metabolites
formed in humans and rodents should be presented more transparently.
EPA Response: EPA accepts this recommendation, and has added discussion and/or mention of
the quantitative uncertainties with respect to GSH conjugation wherever relevant throughout the
document.
•	In the body of the document, MOA information should be systematized and broken down
into key events for each proposed MOA. The EPA may consider using a tabular format to
facilitate the ease of evaluation. Information on supporting/refuting (if any) evidence
(with appropriate references indicated), human relevance (if available), and "strength" of
each line of evidence/study should be included.
EPA Response: EPA accepts this recommendation, and has added tables summarizing the
proposed MO As and conclusions for kidney and liver carcinogenesis.
•	EPA should consider tabular summaries by specific metabolites when studies used
metabolite exposure rather than the parent compound.
20
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA Response: EPA considered this recommendation, but decided against adding the tables for
the metabolites because in most cases (TCA, DC A, and CH), those studies are described and
tabulated in detail in other toxicological reviews.
•	Data gaps should be clearly identified to help guide future research.
EPA Response: EPA considered this recommendation, and decided to focus on data gaps related
to dose-response, as these will have the greatest impact on any future revision to the
Toxicological Review. These research needs are now included as a separate section at the end of
Chapter 5.
•	Key conclusions supporting/refuting each key event should be presented in bullet form
indicating where in the document a more detailed narrative/tables can be found.
EPA Response: EPA accepts this recommendation, and has included key conclusions in the
summary MOA tables described above for kidney and liver carcinogenesis.
1.6.2.2.	MOA for TCE-Induced Kidney Tumors (SAB Report Section 6b)
•	Modify the relevant text to reflect that the available data do, in fact, provide support for
TCE-induced kidney tumors involving cytotoxicity and compensatory cell proliferation,
possibly in combination with a mutagenic MOA, although not to the extent that support
for a mutagenic MOA was provided.
EPA Response: EPA accepts this recommendation and has included additional discussion along
the lines suggested to the section on kidney tumor MOA.
1.6.2.3.	Inadequate Support for PPARa agonism and its sequellae being key events in
TCE-induced liver carcinogenesis (SAB Report Section 6c)
•	Graphical or tabular presentation of these data to strengthen the comparative analysis
between metabolites and chemicals.
EPA Response: EPA accepts this recommendation, and has added the tabular presentation of
quantitative differences among PPARa agonists from Guyton et al. (2009) to strengthen the
comparative analysis.
•	Including some of the analyses which compare the receptor transactivation potency and
the carcinogenic potential of TCA, DCA and other model peroxisome proliferators from
Guyton et al (2009) to strengthen the arguments.
EPA Response: EPA accepts this recommendation, and has added the quantitative analyses
comparing carcinogenic potential and the receptor transactivation potency or other short-term
markers of PPARa activation from Guyton et al (2009) to strengthen the arguments.
21
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.6.2.4.	Inadequate Data to specify Key Events and MOAs involved in other TCE-
Induced Cancer and Non-Cancer Effects (SAB Report Section 6d)
•	No major recommendations in this section.
1.6.2.5.	Human Relevance of TCE-Induced Cancer and Non-Cancer Effects in
Rodents (SAB Report Section 6e)
•	The impact of potential overestimation of the extent of the GSH pathway in humans in
Section 4.4.7 (Kidney) must be transparent
EPA Response: EPA accepts this recommendation, and has added discussion and/or mention of
the quantitative uncertainties with respect to GSH conjugation wherever relevant throughout the
document.
•	The MOA for carcinogenicity should be described as complex rather than unknown in
Section 4.5.7.4. Mode of Action (MOA). With respect to conclusions regarding the liver,
while the complete MOA in animals may not be clear at this time, complex is a more
appropriate descriptor since it is likely that key events from several pathways may
operate leading to acute, sub-chronic and chronic liver toxicity of TCE.
EPA Response: EPA accepts this recommendation, and has rephrased the liver MOA
conclusions in Section 4.5.7.4 along the lines suggested.
1.6.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with EPA's conclusion that a mutagenic MOA is
operative for TCE-induced kidney tumors, recommending instead that a MOA involving
cytotoxicity is involved.
EPA response: EPA maintains its conclusion, in accordance with the SAB review (see Section
1.6.1, above), that a mutagenic MOA is operative for TCE-induced kidney tumors. However, in
accordance with the SAB recommendations (see Section 1.6.2.2, above) and in partial response
to this public comment, EPA has added additional discussion of the data supporting a MOA
involving cytotoxicity.
•	Some public commenters disagreed with EPA's conclusion that there is inadequate
support for PPARa agonism and its sequellae being key events in TCE-induced
hepatocarcinogenesis. Other public commenters agreed with EPA's conclusions.
EPA response: In accordance with the SAB recommendations (see Section 1.6.2.3, above), EPA
has provided additional analysis to support its conclusions.
•	Some public commenters disagreed with EPA's conclusion that a cytotoxic MOA was
inadequately supported for TCE-induced lung tumors, citing analogies to other chemicals
and other indirect data.
22
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA response: EPA has added discussion of data from other compounds hypothesized to have
the same MOA for inducing mouse lung tumors. However, in accordance with the SAB review,
EPA still concludes that there are inadequate data to specify the key events and MO As involved
in TCE-induced lung cancer and non-cancer effects.
1.7. Susceptible Populations (SAB Report Section 7): Comments and EPA Response
1.7.1.	SAB Overall Comment:
The Panel found EPA's hazard assessment provided a good review of potentially
susceptible populations, and identified factors (genetics, lifestage, background, co-exposures and
pre-existing conditions) that may modulate susceptibility to TCE carcinogenicity and non-cancer
effects. However, the Panel disagreed with EPA's conclusion that toxicokinetic variability can
be adequately quantified using existing data. The Panel recommended that exposure to solvent
mixtures should be considered for potential co-exposures, since exposure to more than one
chemical with the same target organ likely increases risk.
1.7.2.	Major SAB Recommendations and EPA Response:
•	The Panel disagreed with the statement that "toxicokinetic variability in adults can be
quantified given the existing data," as the main study characterizing toxicokinetic
variability in adults was small (n<100) and was composed of subjects selected non-
randomly. The Hazard Assessment document should note the limitations of the adult
data for toxicokinetic modeling in terms of uncertainty and possible bias in section
4.10.3, and elsewhere in the document where these data are used for hazard
characterization modeling.
EPA response: EPA accepts this recommendation and has added a statement in Section 4.10.3
noting the limitations of the toxicokinetic database.
•	Section 4.10 of the Hazard Assessment should discuss explicitly the lack of data
demonstrating modulation of health effects from TCE by the identified factors (genetics,
lifestage, background, co-exposures, and pre-existing conditions), and the need for such
data in risk assessment.
EPA response: A statement has been added to the introduction of Section 4.10 noting the lack of
data on susceptible populations and the need for such data. A statement on the need for
additional data to address uncertainties regarding susceptible populations has been added to
Section 4.10.3. The title of Section 4.10.3 has been amended to now read "Uncertainty of
Database and Research Needs for Susceptible Populations."
•	EPA should make specific recommendations for studies that would fill the data gap for
susceptible groups. For example, epidemiologic studies in which TCE exposure is well-
23
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
characterized and in which internal comparisons can be made to determine whether there
is effect modification, and animal studies comparing subgroups (e.g., based on genetics,
obesity, multiple solvent exposures).
EPA response: Where appropriate, statements on the need for additional research to fill data
gaps regarding susceptible populations have been added where appropriate throughout Section
4.10.
•	Modulation of TCE exposure-related hypersensitivity dermatitis by genetic variation may
be relevant for future study, given results of the study of hypersensitivity dermatitis in
Asian workers reported in Li et al. (2007) and increasing industrial chemical exposures in
China.
EPA response: The need for future research on the relationship between genetic variation and
generalized hypersensitivity skin diseases is now highlighted in Section 4.10.3.
•	The wording in Section 4.10 was often not clear about whether it was describing results
for a study that looked at effect modification of the TCE effect or not, as opposed to
direct effects of age, gender, etc. Also, the draft document needs to state explicitly where
effects of TCE within one subgroup were stated, whether the other subgroup was also
examined in the same study.
EPA response: Additional clarification was added throughout Section 4.10 where necessary to
address when the results were unrelated to TCE exposure or related to TCE exposure.
Additional information was also added regarding the comparison group.
•	The Panel recommended that exposure to solvent mixtures should be added as a potential
susceptibility factor (co-exposures) to Section 4.10, since exposure to more than one
chemical to the same target organ likely increases risk.
EPA response: A new Section 4.10.2.6 has been added on mixtures. This text is broader than
solvent mixtures, as there are available studies that address exposure to TCE together with non-
solvents.
•	Section 4.10.2.4.1 (page 4-585) should be more accurately titled 'Obesity', rather than
'Obesity and metabolic syndrome'. As presently written, Section 4.10.2.4.1 gives no
clear message as to how obesity affected the kinetics of TCE, and the section should be
revised to provide clarification.
EPA response: As recommended, Section 4.10.2.4.1 has been retitled as "Obesity", and the text
has been amended to more clearly present the data on toxicokinetics of TCE as it relates to
obesity.
24
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.7.3. Summary of Major Public Comments and EPA Responses:
•	Some comments noted that there is widespread exposure to TCE, including potentially
vulnerable subpopulations.
EPA response: No response needed.
•	Some comments questioned why EPA was not basing its assessment on in utero
exposures.
EPA response: For non-cancer effects, studies with in utero exposures were considered and, in
one case used, for the basis of the RfC or RfD. No data on in utero exposures and cancer effects
were located that were adequate for dose-response analysis.
1.8. Non-Cancer Dose-Response Assessment (SAB Report Section 8): Comments and EPA
Response
1.8.1. SAB Overall Comments:
1.8.1.1. Selection of Critical Studies and Effects
The Panel supported the selection of a RfC and RfD based on multiple candidate
reference values that lie within a narrow range at the low end of the full range of candidate
reference values developed, rather than basing these values on the single most sensitive critical
endpoint. The Panel expressed concerns about the use of several candidate critical studies and
effects, specifically National Toxicology Program (1988) [toxic nephropathy], National Cancer
Institute (1976) [toxic nephrosis], and Woolhiser et al. (2006) [increased kidney weights].
However, the Panel noted that uncertainties about the quantitative risk assessment based on
kidney effects in NTP (1988), NCI (1976) and Woolhiser et al. (2006) did not indicate that there
was uncertainty that TCE caused renal toxicity. As discussed previously, the three PBPK model-
based candidate RfCs/RfDs (p-cRfCs/RfDs) for renal endpoints were based on an uncertain dose
metric, especially in regard to the relative rate of formation of the toxic metabolite in humans
and animals. Additional issues related to choice of toxic nephropathy in female Marshall rats
from NTP (1988) included excessive mortality due to dosing errors and possibly other causes,
and a high level of uncertainty in the extrapolation to the benchmark dose (BMD) due to the use
of very high doses and a high incidence (>60%) of toxic nephropathy at both dose levels used.
With respect to toxic nephrosis in mice from NCI (1976), the BMD analysis was not supported
because the effect occurred in nearly 100% of animals in both dose groups, and because a high
level of uncertainty is associated with extrapolation from the Lowest Adverse Effect Level
(LOAEL) at which nearly 100% animals were affected. Renal cytomegaly and toxic
nephropathy, which were not selected as critical effects, occurred at high frequency in all treated
groups.
25
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
The Panel recommended that the two endpoints for immune effects from Keil et al.
(2009) and the cardiac malformations from Johnson et al. (2003) be considered the principal
studies supporting the RfC. The Panel also recommended that the endpoints for immune effects
from Keil et al. (2009) and Peden-Adams et al. (2008) and the cardiac malformations from
Johnson et al. (2003) be considered as the principal studies supporting the RfD.
1.8.1.2.	Derivation of RfD and RfC
The screening, evaluation, and selection of candidate critical studies and effects used for
the development of the RfC and RfD were sound. The derivation of the points of departure
(PODs) was generally appropriate. However, the BMD modeling results were uncertain for
some datasets. For example, the log-logistic BMD analysis for toxic nephropathy in female
Marshall rats in NTP (1988), shown in Figure F-10 in Appendix F, may greatly overestimate the
risks at low doses. As discussed above, this modeling involved extrapolation from a high
LOAEL at which a high percentage of the animals were affected.
EPA used PBPK-based dose metrics for interspecies, intraspecies, and route-to-route
extrapolation. The Panel supported this approach for development of the RfC and RfD. The
Panel noted that the candidate RfDs /RfCs for kidney endpoints were highly sensitive to the rate
of renal bioactivation of the cysteine conjugate, S-dichlorovinyl glutathione (DCVC), in humans
relative to rodents. Candidate RfDs/RfCs developed using this dose-metric were several
hundred-fold lower than RfD/RfCs for the same endpoints based on applied dose with standard
uncertainty factors. The Panel noted that the uncertainties about the in vitro and in vivo data
used to estimate the rate of renal bioactivation of DCVC were much greater than for other dose
metrics [e.g. there are large discrepancies in the rates of human glutathione conjugation reported
by Lash et al. (1999b) and Green et al. (1997a)]. These uncertainties should be clarified and
should be the basis of a sensitivity analysis in the next update of the TCE draft risk assessment.
The Panel also recommended that the rationale for scaling the dose metric to body weight3 4, in
conjunction with the interspecies extrapolation based on PBPK modeling, should be presented in
a clearer and more transparent way.
1.8.1.3.	Uncertainty Factors
The Panel agreed that, in general, the selection of uncertainty factors was clearly and
transparently described and appropriate. EPA developed equivalent doses and concentrations for
sensitive humans to replace standard uncertainty factors for inter- and intra-species
toxicokinetics. The Panel concluded that the approach used, including the selections of PODs
and the extrapolations from rodent to human, followed by consideration of the 99th percentile
human estimates, was acceptable to address the sensitive population. In future work, the
26
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
variability and uncertainty could be better characterized by considering other quantiles of the
distribution.
1.8.2. Major SAB Recommendations and EPA Response:
1.8.2.1.	The screening, evaluation, and selection of candidate critical studies and
effects (SAB Report Section 8a)
•	Chapter 5 should include a list of all non-cancer health effects and studies discussed in
Chapter 4, noting those which were considered candidate critical effects and studies.
EPA Response: EPA considered this recommendation and concluded that a list of all the non-
cancer health effects and studies discussed in Chapter 4 would be overly long and redundant. As
an alternative, first, EPA has ensured that each section of Chapter 4 includes tables of the
relevant non-cancer health effects and studies discussed, with studies and effects in bold
designating those considered in Chapter 5. Second, EPA has added to Chapter 5 tables with the
experimental details (e.g., which species, doses, effects) of the candidate studies for each
endpoint type, with cross-references back to the tables in Chapter 4 that contain all the studies
for each type of effect. Therefore, there is now a transparent trace-back from the PODs used in
Chapter 5 (tables in the external review draft), to the experimental details from which the POD
was derived (new tables in Chapter 5), to the larger set of studies considered for each effect type
(tables in Chapter 4).
•	Tables 5.1-5.5 should provide cross-references to the table or page in Chapter 4 and/or to
the Appendices (such as Appendix E for hepatic studies) where the listed study was
discussed, and should include more details (e.g. gender, strain, duration) of the studies
selected as the basis for cRfDs and cRfCs when these details were needed to prevent
ambiguity.
EPA Response: EPA accepts this recommendation and has addressed it as part of its response to
the previous recommendation for a table in Chapter 5 listing all the studies.
•	Consistent dose units should be used in discussing the same study in different places in
the document.
EPA Response: EPA accepts this recommendation and has checked the dose units used as it
developed the new tables for Chapter 5.
1.8.2.2.	The points of departure, including those derived from benchmark dose
modeling (e.g., selection of dose-response models, benchmark response levels)
(SAB Report Section 8b)
•	Chapter 5 should include the information on POD derivation from Table F-13 of
Appendix F, including approach, selection criterion and decision points.
27
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA Response: After reviewing Chapter 5, EPA did not implement this suggestion. Chapter 5
describes the modeling approaches and selection criteria and important decisions in sufficient
detail, and on page 5-3 the reader is directed to Appendix F for further details. The succeeding
pages of Chapter 5 describe studies and effects by effect domain quite extensively, and the tables
and footnotes contain sufficient detail on BMRs, PODs, and reasons for study selection. We
think it is appropriate to provide the mass of numerical modeling details in Appendix F, and that
the modeling decisions are well described therein. Integrating this material into Chapter 5 would
greatly increase the length of Chapter 5 and make it unwieldy for the reader.
1.8.2.3. The selected PBPK-based dose metrics for inter-species, intra-species, and
route-to-route extrapolation, including the use of body weight to the % power
scaling for some dose metrics (SAB Report Section 8c)
•	The uncertainty about the rate of human glutathione conjugation found in Lash et al.
(1999b) versus Green et al. (1997a) should be highlighted in the current assessment.
EPA Response: EPA accepts this recommendation and has added discussion and/or mention of
the quantitative uncertainties with respect to GSH conjugation wherever relevant throughout the
document.
•	The basis for the renal bioactivation dose metric should be more clearly and transparently
presented and discussed in Chapter 3 and other appropriate sections. If this dose metric
was derived indirectly from data on other metabolic pathways leading to and/or
competing with bioactivation, this should be more clearly discussed.
EPA Response: EPA accepts this recommendation and has revised section 3.5.7.3.1 to more
clearly discuss the basis of the renal bioactivation dose metric. In other sections of the document
where the dose metric is discussed, reference is made to section 3.5.7.3.1.
•	The rationale for scaling the dose metric to body weight3 4, in conjunction with the
interspecies extrapolation based on PBPK modeling, should be presented in a clearer and
more transparent way (e.g. on pp. 5-33 - 5-36).
EPA Response: EPA accepts this recommendation and has revised the discussion of this
rationale substantially.
•	The discussion of "empirical dosimetry" vs. "concentration equivalence dosimetry"
should be made clearer and more transparent (pp. 5-33 - 5-36).
EPA Response: As noted by the SAB in the narrative preceding this recommendation, it is not
necessary to include an extensive discussion of the two dosimetry approaches in these sections.
EPA accepts this recommendation and has replaced this discussion with a clearer and more
transparent rationale for the body weight3 4 scaling.
28
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.8.2.4. Uncertainty factors (SAB Report Section 8d)
•	The definitions of chronic and subchronic studies should be provided in the document
and a citation given.
EPA Response: EPA accepts this recommendation and has added a footnote with this
information on page 5-6 in the paragraph describing uncertainty factors for sub chronic-to-
chronic extrapolation.
•	The discussion of the subchronic to chronic uncertainty factor on p. 5-6 should be
clarified as far as durations of studies considered suitable as the basis of a chronic risk
assessment.
EPA Response: There is no hard and fast rule in this area. Longer studies are generally
preferred as the basis for a chronic risk assessment; however, in any given case, the basis of an
RfC or RfD, or whether one is derived at all, will depend on the studies available and an
assessment of their relevance for extrapolation to longer durations.
•	The draft document should include discussion of whether studies in the lower end of the
range defined as subchronic (e.g. 4 weeks) are of sufficient duration to be used as the
basis for a chronic (lifetime) risk assessment.
EPA Response: EPA notes that studies of this duration have been evaluated for other risk
assessments. For any study and endpoint that is used as a basis for a POD in this and previous
assessments, EPA has explained its applicability in the light of alternative studies of the same
endpoint having longer and shorter duration and alternative studies and endpoints within the
same domain having various durations.
•	Studies only slightly longer than the minimum needed to be considered chronic should be
noted as such, and the use of an uncertainty factor to account for less than lifetime
exposure (of less than the full uncertainty factor of 10) could be considered for studies of
such durations, especially for endpoints thought to progress in incidence or severity with
time.
EPA Response: If there is evidence suggesting there might be further progression with increased
exposure duration, a subchronic-to-chronic UF of 3 might be considered for a nominally chronic
study. The example given by the panel could merit special justification of an UF of 3 if there
were evidence that the response continued to increase with exposure durations longer than 18
weeks. No such evidence was found. For the study of Kulig et al. (1987), severity didn't
progress beyond week 9 for the two-choice response, and, in the 1000 ppm exposure group, it
didn't progress much in those first 9 weeks; thus, it is not anticipated that the 500 ppm response,
which was flat over the 18 weeks, would become significant over an extended duration of
exposure.
29
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.8.2.5.	The equivalent doses and concentrations for sensitive humans developed from
PBPK modeling to replace standard uncertainty factors for inter- and intra-
species toxicokinetics, including selection of the 99th percentile for overall
uncertainty and variability to represent the toxicokinetically-sensitive
individual (SAB Report Section 8e)
•	The Panel noted variability/uncertainty for the toxicokinetically-sensitive individual
could be quantified in future work by considering distributions in addition to the
distribution of the 99th percentile, such as the 95th percentile.
EPA Response: EPA agrees that this could be the subject of future work.
•	A quantile regression looking simultaneously at several quantiles could be developed in
the future and presented in future refinements of this assessment.
EPA Response: EPA agrees that this could be developed in the future and presented in future
refinements of this assessment.
1.8.2.6.	The qualitative and quantitative characterization of uncertainty and
variability (SAB Report Section 8f)
•	The quantitative uncertainty analysis of PBPK model-based dose metrics for LOAEL or
NOAEL based PODs (Section 5.1.4.2) should be revised to clarify the objective of this 2-
D type analysis, as well as the methodology used.
EPA Response: EPA accepts this recommendation and has revised the discussion, clarifying its
objective and methodology.
•	In future work, EPA could develop an approach using distribution to characterize
uncertainty in a Bayesian framework.
EPA Response: EPA agrees that such an approach could be developed in future work.
1.8.2.7.	The selection of NTP (1988) [toxic nephropathy], NCI (1976) [toxic nephrosis],
Woolhiser et al. (2006) [increased kidney weights], Keil et al. (2009)
[decreased thymus weights and increased anti-dsDNA and anti-ssDNA
antibodies], Peden-Adams et al. (2006 [developmental immunotoxicity], and
Johnson et al. (2003) [fetal heart malformations] as the critical studies and
effects for non-cancer dose-response assessment (SAB Report Section 8g)
EPA Response: See recommendation in Section 1.8.2.8, below.
1.8.2.8.	The selection of the draft RfC and RfD on the basis of multiple critical effects
for which candidate reference values are in a narrow range at the low end of
the full range of candidate critical effects, rather than on the basis of the
single most sensitive critical effect. (SAB Report Section 8h)
•	The two endpoints for immune effects from Keil et al. (2009) and the cardiac
malformations from Johnson et al. (2003) should be considered the principal studies
supporting the RfC.
30
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
EPA Response: EPA accepts this recommendation and has revised Chapter 5 accordingly.
•	The endpoints for immune effects from Keil et al. (2009) and Peden-Adams et al. (2008)
and the cardiac malformations from Johnson et al. (2003) should be considered as the
principal studies supporting the RfD.
EPA Response: EPA accepts this recommendation and has revised Chapter 5 accordingly.
1.8.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters disagreed with the choices of critical studies for dose-response
analyses of non-cancer endpoints.
EPA response: In accordance with SAB recommendations (see Section 1.8.2.8, above), EPA has
selected the immune effects from Keil et al. (2009) and the cardiac malformations from Johnson
et al. (2003) as the principal studies supporting the RfC, and the immune effects from Keil et al.
(2009) and Peden-Adams et al. (2008) and the cardiac malformations from Johnson et al. (2003)
as the principal studies supporting the RfD.
•	Some public commenters recommended that EPA not rely on PBPK model-based
estimates of DCVC bioactivation in conducting dose-response analysis for kidney
endpoints.
EPA response: In accordance with SAB recommendations (see Section 1.8.2.3, above), EPA has
noted the uncertainties in the PBPK model-based DCVC bioactivation dose metrics and
considers the kidney effects as supporting, rather than as principal bases for, the RfC and RfD.
•	Some public commenters recommended that EPA provide a more concise and
consolidated characterization of the RfC and RfD determination, particularly in the
context of kidney effects.
EPA response: A concise and consolidated characterization of the RfC and RfD determination
appears in Section 5.1.5.2 and 5.1.5.3. EPA has added discussion of the uncertainties related in
kidney effects to these summary characterizations.
•	Some public commenters recommended that EPA provide more discussion of the
proportionality between applied and internal dose and its impact on the quantitative
analysis.
EPA response: The impact of the proportionality of applied and internal dose, as well as its
impact both dose-response analysis, is discussed in Section 5.1.3.3 and shown graphically in
Appendix F.
31
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
• Some public commenters viewed the use of PBPK modeling as "double counting"
variability, based on the idea that the observed dose-response is in part due to
pharmacokinetic variability.
EPA response: In accordance with the SAB review, the methodology EPA used is consistent
with existing practice in the derivation of RfDs and RfCs. The methodology used is also
consistent with previous applications of PBPK modeling in non-cancer risk assessment. The
comments highlight some uncertainties and ambiguities inherent in the RfD/RfC methodology,
but disaggregating the multiple contributions to dose-response assessment— including effects of
TK variation, TD variation, experimental error, stochasticity, and other factors in both the
experimental animal and human population— requires development of new approaches that are
beyond the scope of the assessment. While some published literature have addressed some of
these issues, further research and development are needed, as no alternative approach has been
generally accepted at the current time.
1.9. Cancer Dose-Response Assessment (Inhalation Unit Risk and Oral Unit Risk) (SAB
Report Section 9): Comments and EPA Response
1.9.1. SAB Overall Comment:
In this assessment, EPA developed an inhalation unit risk and oral unit risk for the
carcinogenic potency of TCE in accordance with the approach outlined in the U.S. EPA Cancer
Guidelines (2005c, 2005d). The unit risks for renal cell carcinoma were based on a case control
study published by Charbotel et al. (2006). The Panel found that the analysis of the Charbotel et
al. (2006) data was well described and that the selection of this study to estimate unit risks was
appropriate. However, more discussion is needed on whether or not it is necessary to adjust for
exposure to cutting oils when computing an odds ratio or relative risk relating TCE exposure to
kidney cancer. The Panel recommended that EPA take a closer look at the literature to
determine if there are other studies which suggest that exposure to cutting oils is a risk factor for
kidney cancer. EPA should also provide a more detailed discussion on the implication of
assumptions made in their analysis. In addition, background kidney cancer rates in the United
States were used in constructing the life table, although the Charbotel et al. (2006) data was
based on a French cohort. A comparison of background cancer rates in France and the United
States would be helpful in supporting their conclusions. The Panel supported the adjustment of
the renal cell carcinoma unit risks to account for the added risk of other cancers, using the meta-
analysis results and Raaschou-Nielsen et al. (2003).
The Panel agreed that human data, when available, should be preferred over rodent data
when estimating unit risk since within species uncertainty is easier to address than between
species uncertainty. The Panel supported the use of linear extrapolation from the POD for cancer
32
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
dose-response assessment of TCE as a default approach. The Panel agreed that characterization
of uncertainty and variability was appropriate, and was exceptionally strong in the PBPK
models.
1.9.2. Major SAB Recommendations and EPA Response:
1.9.2.1. Estimation of Unit Risks for Renal Cell Carcinoma (SAB Report Section 9a)
•	The Panel believed more discussion was needed on whether or not it is necessary to
adjust for exposure to cutting oils when computing an odds ratio or relative risk relating
TCE exposure to kidney cancer. The Panel recommended that EPA take a closer look at
the literature to determine if there are other studies which suggest that exposure to cutting
oils is a risk factor for kidney cancer.
EPA Response: EPA accepts this recommendation and has discussed other studies examining
cutting fluids (Section 4.4.2.3). These studies suggest that potential confounding by cutting
fluids is of minimal concern, and thus including these exposures in the logistic regression may
over-adjust because of the correlation with TCE exposure. Nonetheless, EPA has included, as a
sensitivity analysis, the derivation of a unit risk estimate based on the Charbotel et al. RCC ORs
further adjusted for cutting fluids and petroleum oils, and this estimate is essentially the same as
the original estimate (Section 5.2.2.1.3).
•	The Panel believed that the EPA should provide a more detailed discussion of the
limitations of their analysis. In particular, the model described on p. 5-131 made some
very restrictive assumptions: linear dose-response and exposure was measured without
error. In addition, the life table analysis applied the same estimated RR to each age
interval; another restrictive assumption. While the Panel understood that these
assumptions were necessary due to limited data, there was inadequate discussion of how
violations of these assumptions may affect the results.
EPA Response: EPA accepts the recommendation and has added text pertaining to these
assumptions. Note, too, that the uncertainties in the unit risk estimate, including uncertainties
about the exposure assessment, are discussed in some detail in the uncertainty section (Section
5.2.2.1.3).
•	Finally, in constructing the life table, the EPA used background kidney cancer rates in the
US though the Charbotel et al. (2006) data were based on a French cohort. Hence, a
comparison of background cancer rates in France and the U.S. would be helpful in
supporting their conclusions.
EPA Response: EPA accepts this recommendation, and has added additional information to
Section 5.2.2.1.2. In particular, this section now notes that the usual assumption is that relative
33
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
risk transfers across populations independent of background rates. In addition, this section now
contains information comparing background kidney cancer rates in France and the U.S..
1.9.2.2.	Adjustment of Renal Cell Carcinoma Unit Risks (SAB Report Section 9b)
•	No major recommendations in this section.
1.9.2.3.	Estimation of Human Unit Risks from Rodent Bioassays (SAB Report Section
9c)
•	The Panel agreed that the analysis and results were appropriate but recommended that the
EPA provide more details about their implementation and potential biases. For instance,
in bioassays in which mortality occurred before time to first tumor, the authors simply
adjusted their denominators to equal the number alive at time to first tumor. This
approach assumed that drop-out prior to time to first tumor was unrelated to future risk of
a tumor which could result in biased estimates.
EPA Response: EPA accepts this recommendation and has added a paragraph discussing the
potential biases of this approach, along with citations to relevant literature, to Section G.l.l.
•	In addition, more information was needed on the priors used in their Bayesian analysis of
combined risk across tumor types.
EPA Response: EPA accepts this recommendation and has added this information to Section
G.8.1.2.
1.9.2.4.	Use of Linear Extrapolation for Cancer Dose-Response Assessment (SAB
Report Section 9d)
•	No major recommendations in this section.
1.9.2.5.	Application of PBPK Modeling (SAB Report Section 9e)
•	No major recommendations in this section.
1.9.2.6.	Qualitative and Quantitative Characterization of Uncertainty and Variability
(SAB Report Section 9f)
•	No major recommendations in this section.
1.9.2.7.	Conclusion on the Consistency of Unit Risk Estimates Based on Human
Epidemiologic Data and Rodent Bioassay Data (SAB Report Section 9g)
•	No major recommendations in this section.
1.9.2.8.	Preference for the Unit Risk Estimates based on Human Epidemiologic Data
(SAB Report Section 9h)
•	No major recommendations in this section.
34
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.9.3. Summary of Major Public Comments and EPA Responses:
•	Some public commenters stated that the time courses of kidney cancer, liver and biliary
cancer, and NHL do not support the hypothesis that TCE poses a great risk of cancer in
the human population. These comments recommended that EPA perform a "validation"
exercise to determine if the draft cancer classification and quantitative risk estimates are
consistent with the observable facts concerning human cancer rates and other known risk
factors for the tumor types listed.
EPA response: The analysis suggested by this comment is beyond the scope of the
Toxicological Review. Moreover, such an analysis would require data that do not currently
exist, including detailed historical population estimates not only of TCE exposure, but also of all
other exposures and risk factors associated with each cancer, as well as quantitative estimates as
to how each risk factor modulates the risk of cancer. It is noted, however, that limited
"validation" was performed by comparing qualitative and quantitative inferences based on
epidemiologic data to those based on animal bioassay data. Further quantitative "validation"
may be possible in the future if epidemiologic studies with quantitative exposure information are
conducted.
•	Some public commenters disagreed with the use of epidemiologic data as the primary
basis for the cancer dose-response analysis.
EPA response: EPA maintains its conclusion, in accordance with the SAB review (see Section
1.9.1, above), that the epidemiological data are appropriate to use for estimating cancer risks. In
response to recommendations by the SAB, EPA has provided more detailed discussions as to the
limitations of the analysis.
•	Some public commenters disagreed with the use of linear low-dose extrapolation for
estimating cancer risks at levels below the point of departure, recommending instead the
use of non-linear extrapolation.
EPA response: EPA maintains its conclusion, in accordance with the SAB review (see Section
1.9.1, above), that the linear low-dose extrapolation is appropriate to use given the available data.
1.10. Age-Dependent Adjustment Factors (ADAFs) (SAB Report Section 10): Comments
and EPA Response
1.10.1. SAB Overall Comment:
The Panel agreed that application of age-dependent adjustment factors (ADAFs) in the
TCE analysis consistently followed recommendations in the U.S. EPA Cancer Guidelines
(2005c). All of the steps were clearly presented for inhalation exposure. However, the
discussion for the oral exposure route was shortened and referred back to the inhalation section,
35
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
making understanding of the example difficult to follow. Currently, EPA's IRIS assessment
provides lifetime cancer risk drinking water concentrations for adults only. The Panel
recommended that drinking water concentrations for specified cancer risk levels should also be
derived for various age groups.
1.10.2.	Major SAB Recommendations and EPA Response:
•	The Panel recommended that the statement on page 5-151, lines 14-18, be expanded to
better explain why age-dependent adjustment factors were used for <16 years of age, but
not for the elderly, and why EPA did not directly produce age dependent unit risks per
mg/kg/d.
EPA Response: EPA accepts these recommendations. Section 5.2.3.3 notes that due to lack of
appropriate data, no ADAFs are used for other life-stages, such as the elderly. ADAF-adjusted
unit risks per ppm and per mg/kg/d are now presented in each sample calculation table in
Sections 5.2.3.3.1 and 5.2.3.3.2.
•	Include all details presented for the inhalation sample calculations as was done for the
oral exposure sample calculations.
EPA Response: EPA accepts this recommendation and has revised Sections 5.2.3.3.1 and
5.2.3.3.2 to include all the details for each sample calculation.
•	IRIS assessments in which ADAFs are applied, such as TCE, should include estimated
drinking water concentrations for specified lifetime cancer risk levels (1CT4, 10"5, 10"6),
using representative drinking water intakes for various age groups, while noting that
other drinking water estimates may be used if preferred.
EPA Response: EPA accepts this recommendation and has added drinking water concentrations
for specified lifetime cancer risks under the assumptions used in the drinking water example
calculation to Section 5.1.3.3.2. Similarly, EPA has added air concentrations for specified
lifetime cancer risks under the assumptions used in the inhalation example calculation to Section
5.1.3.3.1.
•	Include in the documentation a discussion of the perceived conflict between the use of
ADAFs and the assumptions underlying the life table analysis of the Charbotel et al.
(2006) data.
EPA Response: EPA accepts this recommendation and has added a discussion addressing the
use of the ADAFs and the assumptions underlying the life table analysis.
1.10.3.	Summary of Major Public Comments and EPA Responses:
•	None
36
-------
DRAFT - EPA DELIBERATIVE - DO NOT QUOTE OR CITE
1.11. Additional key studies (SAB Report Section 11) and editorial comments: Comments
and EPA Response
• The Panel has identified additional studies to be considered in the assessment, as well as
a number of editorial comments.
EPA Response: EPA has incorporated the additional studies in the appropriate sections, and
addressed the editorial comments.
37
-------