1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F. Appendix: TCE Non-Cancer Dose-Response Analyses
F. Appendix: TCE Non-Cancer Dose-Response Analyses	F-l
F.l. Data Sources	F-2
F.2. Dosimetry	F-2
F.2.1. Estimates of TCE in air from urinary metabolite data using Ikeda et al. (1972) .F-2
F.2.2. Dose Adjustments to applied doses for intermittent exposure	F-5
F.2.3. PBPK model-based internal dose metrics	F-6
F.3. Dose-Response Modeling Procedures	F-6
F.3.1. Models for dichotomous response data	F-6
F.3.2. Models for continuous response data	F-7
F.3.3. Model selection	F-7
F.3.4. Additional adjustments for selected datasets	F-8
F.4. Dose-Response Modeling Results	F-9
F.4.1. Quantal dichotomous and continuous modeling results	F-9
F.4.2. Nested dichotomous modeling results	F-10
F.4.3. Model selections and results	F-23
F.5. Derivation of points of departure	F-26
F.5.1. Applied dose points of departure	F-26
F.5.2. PBPK model-based human points of departure	F-26
F.6. Summary of PODs for critical studies and effects supporting the RfC and RfD	F-27
F.6.1. NTP (1988) -BMD modeling of toxic nephropathy in rats	F-27
F.6.2. NCI (1976) -LOAEL for toxic nephrosis in mice	F-30
F.6.3. Woolhiser et al. (2006) - BMD modeling of increased kidney weight in rats ..F-31
F.6.4. Keil et al. (2009) - LOAEL for decreased thymus weight and increased anti-
sdDNA and anti-ssDNA antibodies in mice	F-34
F.6.5. Johnson et al. (2003) - BMD modeling of fetal heart malformations in rats ....F-35
F.6.6. Peden-Adams et al. (2006) - LOAEL for decreased PFC response and increased
delayed-type hypersensitivity in mice	F-37
F.7. References	F-37
6/8/2009
F-l

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.l. Data Sources
Data sources are cited in the body of this report in the section describing dose-response
analyses (Section 5).
F.2. Dosimetry
This section describes some of the more detailed dosimetry calculations and adjustments
used in Section 5.1.
F.2.1. Estimates of TCE in air from urinary metabolite data using Ikeda et al. (1972)
F.2.1.1. Results for Chia et al. (1996)
Chia et al. (1996) demonstrated a dose-related effect on hyperzoospermia in male
workers exposed to TCE, lumping subjects into four groups based on range of TCA in urine.
Table F.l. Dose-response data from Chia et al. (1996)
TCA, mg pergm. no. of no. with
creatinine	subjects hyperzoospermia
Minimum and maximum TCA levels are reported in
the text of Chia et al. (1996), the other data, in their
Table 5.	
We used data from Ikeda et al. (1972) to estimate the TCE exposure concentrations
corresponding to the urinary TCA levels reported by Chia et al. (1996). Ikeda et al. (1972)
studied ten workshops, in each of which TCE vapour concentration was "relatively constant."
They measured atmospheric concentrations of TCE and concentrations in workers' urine of total
tri-chloro compounds (TTC), TCA, and creatinine, and demonstrated a linear relation between
TTC/creatinine (mg/gm) in urine and TCE in the work atmosphere. Their data are tabulated as
geometric means (the last column was calculated by us, as described below).
0.8 to <25
50 to <75
75 to <100
37
18
8
5
6
8
4
3
>= 100 to 136.4
6/8/2009
F-2

-------
1
2
3
4
5
6
7
8
9
10
11
12
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Table F.2. Data on TCE in air (ppm) and urinary metabolite concentrations in workers reported
by Ikeda et al. (1972).
TCE TTC TCA TTC (mg/g TCA (mg/g
n	(ppm) (mg/L) (mg/L) creatinine) 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
We used these data to construct the last column "TCA.cr.mg.gm" (mg TCA/gm
creatinine), as follows: TCA (mg/g creatinine) = TCA (mg/L) x TTC (mg/g creatinine) / TTC
(mg/L). We then evaluated the regression relation between TCE (ppm) and TCA (mg/gm
creatinine) 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, we used the regression relation between loglO(TCA (mg/g
creatinine)) and loglO(TCE (ppm)), assuming constant variances and using number of subjects
"n" as weights. The results are shown in Figure F. 1.
6/8/2009
F-3

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
log10(TCA, mg/g.creatinine in urine) = 0.7098 + 0.7218 * log10(TCE, ppm)
o
o
o
io
5
10
50
100
TCE, ppm
Coefficients:


Value
Std. Error
t value Pr(> | 11 )
(Intercept) 0.7098
0.1132
6.2688 0.0002
loglO(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. 1. Regression of TCE in air (ppm) and TCA in urine (mg/g creatinine) based on data
from Ikeda et al. (1972).
Next, we assumed a Berkson setting for linear calibration, in which we wanted to predict
X (TCE, ppm) from means for Y (TCA, mg/gm creatinine), with substantial error in Y (Snedecor
and Cochran, 1980). Thus we used inverse prediction for the data of Chia et al. (1996) to infer
their mean TCE exposures. The relation based on data from Ikeda et al. (1972) is
loglO(TCA, mg/g.creatinine) = 0.7098 + 0.7218*logl0(TCE, ppm)
and the inverse prediction is
6/8/2009
F-4

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
loglO(TCE) = (loglO(TCA) - 0.7098)/0.7218
TCE, ppm = 10A( (loglO(TCA) - 0 . 7098)/0 . 7218 )
Because of the lognormality of data reported by Ikeda et al. (1972), we used the means of
the logarithms of the ranges for TCA (mg/gm creatinine) in Chia et al. (1996), which is an
estimate of the median for the group. The results are shown in Table F.3.
Table F.3. Estimated urinary metabolite and TCE air concentrations in dose groups from Chia et
al. (1996)
TCA, mgpergm. Estim. TCA
Creatinine
0.8 to <25
50 to <75
75 to <100
>= 100 to 136.4
median
4.47
61.2
86.6
117
log10(TCA
median)
0.650515
1.787016
1.937531
2.067407
10A( mean(log10(TCA limits in first column)))
10A( (log10(TCA median)) - 0.7098)/0.7218
Estim. ppm TCE
b
0.827685
31.074370
50.226119
76.008668
We modeled dose-response relations for the data of Chia et al. (1996) using both the
estimated medians for TCA (mg/gm 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 BMD. As noted below (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 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 the above "inverse prediction" equation gives an occupational exposure level =
10A((logl0(32.6) - 0.7098)/0.7218) = 12.97404 ppm. As noted below (Section F.2.2), the
occupational exposure levels are further adjusted to equivalent continuous exposure for deriving
the point of departure (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 PBPK dose
metrics took into account the daily and weekly discontinuity to produce an equivalent average
6/8/2009	F-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
27
28
29
30
31
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
dose for continuous exposure. No dose 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 sub chronic-to-chronic exposure was
applied where appropriate.
For human occupational studies, inhalation exposures (air concentrations) were adjusted
by the number of work (versus non-work) days and the amount of air intake during working
hours as a fraction of the entire day (10 m3 during work/20 m3 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. 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 (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, we used
averages specific to each species, strain, and sex. 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, 1990, 1988). Where
necessary, we used default adult body weights specific to the strain (USEPA, 1994).
F.3. Dose-Response Modeling Procedures
Where adequate dose-response data were available, we fitted models with the BMDS
software (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 BMR
under consideration.
F.3.1. Models for dichotomous response data
F.3.1.1. Quantal models
For dichotomous responses, we fitted the loglogistic, multistage, and Weibull models.
These models adequately describe the dose-response relationship for the great majority of
datasets, specifically in past TCE studies (Filipsson and Victorin, 2003). If the slope parameter
of the loglogistic model was less than 1, indicating a supralinear dose-response shape, we also
fitted the model with the slope constrained to 1 for comparison. For the multistage model, we
used an order one less than the number of dose groups, in addition to the 2nd-order multistage
6/8/2009
F-6

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
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). We also fitted the
Weibull model with the power parameter unconstrained.
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 intra-litter correlation.
The available nested models in BMDS are the nested loglogistic 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 AIC for models with and without a LSC. The need to estimate
intra-litter 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, we fitted the distinct models available in BMDS: 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 p-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 order
model (up to order n-1) was accepted if AIC decreased more than 3 units and the p-value for the
mean did not decrease.
F.3.3. Model selection
After fitting these models to the datasets, we applied the recommendations for model
selection set out in EPA's Benchmark Dose Technical Guidance Document (Inter-Agency
Review Draft, US EPA, 2008b). First, models were generally rejected if the p-value 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 BMDL estimates from the remaining models were "sufficiently close" (we used a criterion of
6/8/2009
F-7

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
within 2-fold for "sufficiently close"), then the model with the lowest Akaike Information
Criteria (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 2 or more models share the lowest AIC, the
BMD Technical Guidance Document (US EPA, 2008b) 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 datasets
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 8 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
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(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 (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 (2008b).
In some cases, the supralinear dose-response shape could not be accommodated by these
measures, and a LOAEL or NOAEL was used instead. These include NCI (1976) (toxic
nephrosis, >90% response at lowest dose), Keil et al. (2009) (autoimmune markers and decreased
6/8/2009
F-8

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
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 Appendix.linked.files\AppF.Non-cancer.Plots.TCE.contin.DRAFT.pdf
and Appendix.linked.files\AppF.Non-cancer.Plots.TCE.dichot.DRAFT.pdf 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:
Appendix.linked.files\AppF.Non-cancer.Input.Data.TCE.contin.DRAFT.pdf and
Appendix.linked.files\AppF.Non-cancer.Input.Data.TCE.dichot.DRAFT.pdf. The documents
Appendix.linked.files\AppF.Non-cancer.Results.TCE.contin.DRAFT.pdf and
Appendix.linked.files\AppF.Non-cancer.Results.TCE.dichot.DRAFT.pdf present the data and
model summary statistics, including goodness-of-fit measures (Chi-square goodness-of-fit P-
value, AIC), parameter estimates, BMD, and BMDL. The group numbers "GRP" are arbitrary
and are the same as GRP numbers in the plots. Note finally that not all plots are shown in the
documents above, since these spreadsheets include many "alternative" analyses.
6/8/2009
F-9

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
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).
Table F.4. Data on fetuses and litters with abnormal hearts from Johnson et al. (2003)
Dose group
(mg/kg/day):
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
One suitable candidate for a LSC was available: female weight gain during pregnancy.
Based on goodness of fit, this covariate did not contribute to better fit and was not used. Some
ICs were significant and these parameters were included in the model.
With the high dose included, the chi-square goodness of fit was acceptable, but some
residuals were large (1.5 to 2) for the control and two lower doses. Therefore, models were also
fitted after dropping the highest dose. For these, goodness of fit was adequate and scaled
residuals were smaller for the low doses and control. Predicted expected response values were
closer to observed when the high dose was dropped, as shown in Table F.5:
6/8/2009
F-10

-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1
2	Table F.5. Comparison of observed and predicted numbers of fetuses with abnormal hearts from
3	Johnson et al. (2003), with and without the high dose group, using a nested model.
Abnormal Hearts (pups)
Dose group
(mg/kg/day):
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
--
4
5	Accuracy in the low-dose range is especially important because the BMD is based upon
6	the predicted responses at the control and the lower doses. Based on the foregoing measures of
7	goodness of fit, we used the model based on dropping the high dose.
8	The Nested LogLogistic and Rai-VanRyzin models were fitted; these gave essentially the
9	same predicted responses and POD. The former model was used as the basis for a POD; results
10	are in Table F.6 and Figure F.2.
11
12	Table F.6. Results of nested loglogistic model for fetal cardiac anomalies from Johnson et al.
13	(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
14	NLOG = "nested loglogistic" model
15	LSC analyzed was female weight gain during pregnancy.
16	* Indicates model selected (Rai-VanRyzin model fits are essentially the same)
17
6/8/2009
F-ll

-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Figure F.2. BMD modeling of Johnson et al. (2003) using nested loglogistic 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).
Nested Logistic Model with 0.95 Confidence Level
0.12
Nested Logistic
< 0.06
t5 0.04
03
BMDL
0.05 0.1
0.15 0.2
dose
0.25 0.3
13:36 08/27 2008
Nested Logistic Model with 0.95 Confidence Level
0.12
Nested Logistic
® 0.08
< 0.06
BMDL
13:37 08/27 2008
6/8/2009
F-12

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.4.2.1.2. Chi-square Goodness of Fit Test for Nested Loglogistic
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. We were also concerned 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 p-value saved. Within these 100 p-values, >75%
were >0.05, and >50% had p-values >0.11, this indicated that the model is acceptable based on
goodness-of-fit criteria.
F.4.2.1.3. Results using PBPK model-based dose metrics
The nested loglogistic model was also run using the dose metrics in the dams of total
oxidative metabolism scaled by body weight to the 3/4-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 (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.
6/8/2009
F-13

-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1
2	Table F.7. Results of nested loglogistic model for fetal cardiac anomalies from Johnson et al.
3	(2003) without the high dose group, using the TotOxMetabBW34 dose metric	
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
4	NLOG = "nested loglogistic" model
5	LSC analyzed was female weight gain during pregnancy.
6	* Indicates model selected. BMDS failed with the Rai-VanRyzin and NCTR models.
7
8	Figure F.3. BMD modeling of Johnson et al. (2003) using nested loglogistic model, with
9	TotOxMetabBW34 dose metric, without LSC, with IC, and without the high dose group, using a
10	BMR of 0.01 extra risk.
Nested Logistic Model with 0.95 Confidence Level
°-«
0.1
§).08
it
<0.06
c
•¦§>.04
2
4).02
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
dose
11	12:44 02/06 2009
sted Logistic
BMDL
6/8/2009
F-14

-------
1
2
3
4
5
6
7
8
9
10
11
12
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Table F.8. Results of nested loglogistic model for fetal cardiac anomalies from Johnson et al.
(2003) without the high dose group, using the AUCCBld dose metric
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
NLOG = "nested loglogistic" model
LSC analyzed was female weight gain during pregnancy.
* Indicates model selected. BMDS failed with the Rai-VanRyzin and NCTR models.
Figure F.4. BMD modeling of Johnson et al. (2003) using nested loglogistic model, with
AUCCBld dose metric, without LSC, with IC, and without the high dose group, using a BMR of
0.01 extra risk.
0.1
§0.08
o
£
<0.06
c
¦§0.04
cc
lJ"0.02
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007
dose
12:42 02/06 2009
Nested Logistic Model with 0.95 Confidence Level
ested Logistic
BMDL
6/8/2009
F-15

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.4.2.2. Narotsky et al. (1995)
Data were combined for the high doses in the single-agent experiment and the lower
doses in the 'five-cube' experiment. Individual animal data were kindly provided by Dr.
Narotsky (personal communications from Michael Narotsky, U.S. EPA, to John Fox, U.S. EPA,
19 June 2008, and to Jennifer Jinot, U.S. EPA, 10 June 2008). Two endpoints were examined:
frequency of eye defects in rat pups and prenatal loss (number of implantation sites minus
number of live pups on postnatal day 1).
Two LSCs were considered, with analyses summarized in Table F.9. The number of
implants is unrelated to dose, as inferred from regression and analysis of variance, and was
considered as a LSC for eye defects. As number of implants is part of the definition for the
endpoint of prenatal loss, it is not considered as a LSC for prenatal loss. A second LSC, the dam
body weight on GD6 (damBW6) was significantly related to dose and is unsuitable as a litter-
specific covariate.
Table F.9. Analysis of LSCs with respect to dose from Narotsky et al. (1995)
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
1125
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
6/8/2009
F-16

-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.4.2.2.1. Fetal eye defects
The Nested LogLogistic 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.
Table F. 10. Results of nested loglogistic and Rai-VanRyzin model for fetal eye defects from
Nartosky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water).	
Model
LSC?
IC?
AIC
Pval
BMR
BMD
BMDL
NLOG
Y
Y
255.771
0.3489
0.05
875.347
737.328 **
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
38.7117*
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
130.938*
NLOG = "nested loglogistic" model; RAI = Rai-VanRyzin model
LSC analyzed was implants.
* Indicates model selected.
** Graphical fit at the origin exceeds observed control and low dose responses and slope is quite
flat (Figure F.5), fitted curve does not represent the data well
Results for Nested Loglogistic model suggested a better model fit with the inclusion of
the LSC and IC, based on AIC. However, the graphical fit (Figure F.5) is strongly sublinear and
high at the origin where the fitted response exceeds the observed low-dose responses for the
control group and two low-dose groups. We selected an alternative Nested Loglogistic model
without either LSC or IC (Figure F.6), which fits the low-dose responses better. Given that this
model had no LSC and no IC, the nested loglogistic model reduces to a quantal loglogistic
model. Parameter estimates and the p-values were essentially the same for the two models
(Table F. 11). A similar model selection can be justified for the Rai-Van Ryzin model. Because
no LSC and no IC were needed (Figure F.12), we modeled this endpoint using quantal models,
using totals of implants and losses for each dose group, which allowed choice from a wider range
of models (those results appear with quantal model results in this appendix).
6/8/2009
F-17

-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Figure F.5. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested
loglogistic model, with applied dose, with both LSC and IC, using a BMR of 0.05 extra risk.
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
0.5
0.4
T3
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1
2	Table F. 11. Comparison of results of nested loglogistic (without LSC or IC) and quantal
3	loglogistic model for fetal eye defects from Nartosky et al. (1995).

Parameter


BMD05
BMDL05
Model
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
4
5
6
7	Figure F.7. BMD modeling of fetal eye defects from Narotsky et al. (1995) using nested Rai-
8	VanRyzin model model, with applied dose, without either LSC or IC, using a BMR of 0.05 extra
9	risk.
RaiVR Model with 0.95 Confidence Level
RaiVR
BMDL
0	200 400 600 800 1000
dose
10 17:25 08/04 2008
6/8/2009
F-19
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.4.2.2.2. Narotsky et al. (1995) prenatal loss
The Nested LogLogistic and Rai-VanRyzin models were fitted to prenatal loss reported
by Narotsky et al. (1995), with the results summarized in Table F. 12.
Table F. 12. Results of nested loglogistic and Rai-VanRyzin model for prenatal loss from
Nartosky et al. (1995), on the basis of applied dose (mg/kg/d in drinking water).
Model LSC? IC?	AIC	Pval	BMR	BMP	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	O05	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	Ol	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*
NLOG = "nested loglogistic" model; RAI = Rai-VanRyzin model
LSC analyzed was dam body	weight on GD6.
* Indicates model selected.
The BMDS nested models require a LSC, so we used dam body weight on GD6
("damBW6") as the LSC. However, damBW6 is significantly related to dose and, so, is not a
reliable LSC. Number of implants could not be used as a LSC because it was identified as
number at risk in the BMDS models. These issues were obviated because the model selected did
not employ the LSC.
For the nested loglogistic models, the AIC is much larger when the IC is dropped, so the
IC is needed in the model. The LSC can be dropped (and is also suspect because it is correlated
with dose). The model with IC and without LSC was selected on the basis of AIC (shown in
Figure F.8). For the Rai-VanRyzin models, the model selection was similar to that for the nested
loglogistic, leading to a model with IC and without LSC, which had the lowest AIC (shown in
Figure F.9).
6/8/2009
F-20
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1	Figure F.8. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested
2	loglogistic model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
3	panel) or 0.01 extra risk (bottom panel).
Nested Logistic Model with 0.95 Confidence Level
0.9
0.8
0.7
"S 0.6
o
£ 0.5
<
C 0.4
o
'o 0.3
CO
^ 0.2
0.1
0
0 200 400 600 800 1000
dose
4	16:44 08/20 2008
0.9
0.8
0.7
£ 0.6
o
£ 0.5
c 0.4
o
T5 0.3
CO
£ 0.2
0.1
0
0 200 400 600 800 1000
dose
5	16:45 08/20 2008
6
7
Nested Logistic
BMP.L
BMP
Nested Logistic Model with 0.95 Confidence Level
Nested Logistic
BMDL
BMD
6/8/2009
F-21
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1
2	Figure F.9. BMD modeling of prenatal loss reported in Narotsky et al. (1995) using nested Rai-
3	VanRyzin model, with applied dose, without LSC, with IC, using a BMR of 0.05 extra risk (top
4	panel) or 0.01 extra risk (bottom panel).
RaiVR Model with 0.95 Confidence Level
0.7
0 0.6
T5
sg 0.5
c 0.4
o
0.3
o
cc
£ 0.2
BMD.L
600
BMP
800
0
200
400
1000
dose
5 16:46 08/20 2008
RaiVR Model with 0.95 Confidence Level
0.9
RaiVR
0.8
0.7
"8 0.6
"8
a 0.5
4—
<
c 0.4
o
0.3
CO
^ 0.2
BMDL
BMD
0
200
400
600
800
1000
dose
6 16:46 08/20 2008
6/8/2009
F-22
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1
2	F.4.3. Model selections and results
3	The final model selections and results for non-cancer dose-response modeling are presented in Table F.13.
4
6/8/2009
F-23
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Table F. 13. Model selections and results for non-cancer dose-response analyses.
GRP
Study
Species
Sex
Strain
Exposure
route
Endpoint
dose metric
BMRtype
BMR
BMD /
BMDL
BMDL
model
reporting
BMD
Notes
Dichotomous models













3
Chia.etal.1996
human
M
wo rkers. e lec. facto ry
inhal
N.hyperzoospermia
appl.dose
extra
0.1
2.14
1.43
loglogistic.1
3.06

7
Narotsky.etal.1995
rat
F
F344
oral.gav
N.pups.eye.defects
appl.dose
extra
0.01
1.46
60.1
multistage
806
a
13
Narotsky. etal. 1995. sa
rat
F
F344
oral.gav
N.dams.w. resorbed. litters
appl.dose
extra
0.01
5.47
32.2
multistage.2
570

13
Narotsky. etal. 1995. sa
rat
F
F344
oral.gav
N.dams.w.resorbed. litters
AUCCBId
extra
0.01
5.77
17.5
multistage.2
327

13
Narotsky.etal.1995.sa
rat
F
F344
oral.gav
N.dams.w. resorbed. litters
TotMetabBW34
extra
0.01
1.77
77.5
weibull
156

14
Johnson, etal. 2003. drophi
rat
F
Sprague.Dawley
oral.dw
N. litters, abnormal, hearts
appl.dose
extra
0.1
2.78
0.0146
loglogistic.1
0.0406
b
36
Griffin etal.2000
mice
F
MRL++
oral.dw
portal.infiltration
appl.dose
extra
0.1
2.67
13.4
loglogistic.1
35.8

38
Maltoni.etal.1986
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
appl.dose
extra
0.1
1.22
40.2
multistage
49.2
c
38
Maltoni.etal.1986
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
ABioactDCVCBW34
extra
0.1
1.18
0.0888
loglogistic
0.105

38
Maltoni.etal.1986
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
AMetGSHBW34
extra
0.1
1.19
0.086
loglogistic
0.102

38
Maltoni.etal.1986
rat
M
Sprague.Dawley
inhal
megalonucleocytosis
TotMetabBW34
extra
0.1
1.13
53.8
weibull
61
d
39
Maltoni.etal.1986
rat
M
Sprague.Dawley
oral.gav
megalonucleocytosis
appl.dose
extra
0.1
1.53
33.8
multistage.2
51.8
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

Nested dichotomous models













NA
Johnson.etal. 2003. drophi
rat
F
Sprague.Dawley
oral.dw
N. pups, abnormal, hearts
appl.dose
extra
0.01
3.12
0.0207
loglogistic. IC
0.711
b
NA
Johnson, etal. 2003. drophi
rat
F
Sprague.Dawley
oral.dw
N. pups, abnormal, hearts
T otOxM eta bBW34
extra
0.01
3.12
0.0142
loglogistic. IC

b
NA
Johnson.etal. 2003. drophi
rat
F
Sprague.Dawley
oral.dw
N. pups, abnormal, hearts
AUCCBId
extra
0.01
3.12
0.000648
loglogistic. IC

b
NA
Narotsky.etal.1995
rat
F
F344
oral.gav
N. prenatal, loss
appl.dose
extra
0.01
1.2
469
RAI.IC
814

Continuous models













2
Land.etal.1981
mouse
M
(C57B1xC3H)F1
inhal
pet.abnormal.sperm
appl.dose
standard
0.5
1.33
46.9
polynomial.constvar
125

6
Carney.etal.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.etal.1995
rat
F
F344
oral.gav
gm.wgt.gain.GD6.20
appl.dose
relative
0.1
1.11
108
polynomial.constvar
312

19
Crofton.etal.97
rat
M
Long-Evans
inhal
dB.auditory.threshold(16kHz)
appl.dose
absolute
10
1.11
274
polynomial.constvar
330

21
George.etal.1986
rat
F
F344
oral.food
litters
appl.dose
standard
0.5
1.69
179
polynomial.constvar
604

23
George.etal.1986
rat
F
F344
oral.food
live.pups
appl.dose
standard
0.5
1.55
152
polynomial.constvar
470

26
George.etal.1986
rat
F
F344
oral.food
Foffspring.BWgm.day21
appl.dose
relative
0.05
1.41
79.7
polynomial.constvar
225

34sq
Moser.etal. 1995+perscom
rat
F
F344
oral.gav
no. rears
appl.dose
standard

1.64
248
polynomial.constvar
406
b,f
49
George.etal.1986
rat
F
F344
oral.food
traverse.time.21 do
appl.dose
relative

1.98
72.6
power
84.9

51
Buben.O'Flaherty.85
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
appl.dose
relative
0.1
1.26
81.5
hill.constvar
92.8

51
Buben.O'Flaherty.85
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
AMetLiv1BW34
relative
0.1
1.08
28.6
polynomial.constvar
28.4

51
Buben.O'Flaherty.85
mouse
M
SwissCox
oral.gav
Liverwt.pctBW
TotOxM eta bBW34
relative
0.1
1.08
37
polynomial.constvar
36.7

58
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Liverwt.pctBW
appl.dose
relative
0.1
1.36
21.6
h
II
30.4

58
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Liverwt.pctBW
AMetLiv1BW34
relative
0.1
1.4
22.7
h
II
32.9

58
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Liverwt.pctBW
TotOxM eta bBW34
relative
0.1
1.3
73.4
h
II
97.7

60.Rp
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Kidneywt. pctBW
appl.dose
relative
0.1
1.17
34.7
polynomial
47.1

60.Rp
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Kidneywt. pctBW
AMetGSHBW34
relative
0.1
1.18
0.17
polynomial
0.236

60.Rp
Kjellstrand.etal. 1983b
mouse
M
NMRI
inhal
Kidneywt. pctBW
TotMetabBW34
relative
0.1
1.17
71
polynomial
95.2

63
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
Antibody. Forming Cells
appl.dose
standard

1.94
31.2
power.constvar
60.6
b
62
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
Antibody. Forming Cells
AUCCBId
standard

1.44
149
polynomial
214

62
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
Antibody. Forming Cells
TotMetabBW34
standard

1.5
40.8
polynomial
61.3

65
Woolhiser.etal.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.etal.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.etal.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.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
kidney, wt. perl OOgm
TotMetabBW34
relative
0.1
1.47
40.8
polynomial.constvar
52.3

67
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
liver.wt. perl OOgm
appl.dose
relative
0.1
4.13
25.2
hill.constvar
70.3

67
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
liver, wt. perl OOgm
AMetLiv1BW34
relative
0.1
1.53
46
polynomial.constvar
56.1

67
Woolhiser.etal.2006
rat
F
CD (Sprague-Dawley)
inhal
liver, wt. perl OOgm
TotOxM eta bBW34
relative
0.1
1.53
48.9
polynomial.constvar
59.8

6/8/2009
F-24
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
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.
loglogistic = unconstrained loglogistic; loglogistic. 1 = constrained loglogistic; multistage = multistage with #stages=dose groups-1; multistage.n = n-
stage multistage; loglogistic.IC = nested loglogistic 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)
Notes:
a	Eight-stage multistage model,
b	Dropped highest dose,
c	Three-stage multistage model.
d	Weibull selected over loglogistic w/same AIC on basis of visual fit (less extreme curvature),
e	Second-order MS selected on basis of visual fit (less extreme curvature),
f	Square-root transformation of original individual count data.
6/8/2009
F-25
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.5. Derivation of points of departure
F.5.1. Applied dose points of departure
For oral studies in rodents, the point of departure (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 (1994), 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. PBPK model-based human points of departure
As discussed in Section 5.1.3, the PBPK model was used for simultaneous inter-species
(for endpoints in rodent studies), intra-species, and route-to-route extrapolation (rtr) based on the
estimates from the PBPK model of the internal dose points of departure (iPOD) for each
candidate critical study/endpoints. The following documents contain figures showing the
derivation of the human equivalent doses and concentrations (HEDs and HECs) for the median
(50th percentile) and sensitive (99th percentile) individual from the (rodent or human) study
iPOD. 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 iPOD (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:
Appendix.linked.files\AppF.Non-cancer.HECs.Plots.human.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF.Non-cancer.HECs.Plots.rodent.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF.Non-cancer.HECs.Plots.rodent.oral.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF.Non-cancer.HEDs.Plots.human.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF.Non-cancer.HEDs.Plots.rodent.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF.Non-cancer.HEDs.Plots.rodent.oral.studies.TCE.DRAFT.pdf
6/8/2009
F-26
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
The original study internal doses are based on the median estimates from about 2000
"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 2000
"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:
Appendix.linked.files\AppF .Non-
cancer.HECs.AltPOD.Plots.rodent.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF .Non-
cancer.HECs.AltPOD.Plots.rodent.oral.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF .Non-
cancer.HEDs.AltPOD.Plots.rodent.inhalation.studies.TCE.DRAFT.pdf
Appendix.linked.files\AppF .Non-
cancer.HEDs.AltPOD.Plots.rodent.oral.studies.TCE.DRAFT.pdf
F.6. Summary of PODs for critical studies and effects supporting the RfC and RfD
This section summarizes the selection and/or derivation of PODs from the critical studies
and effects that support the RfC and RfD. In particular, for each endpoint, the following are
described: the 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). The
dose metric selection for different endpoints is discussed in Section 5.1.3.1.
F.6.1. NTP (1988) - BMD modeling of toxic nephropathy in rats
The critical 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).
6/8/2009
F-27
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.6.1.1. Dosimetry and BMD modeling
Rats were exposed to 500 or 1000 mg/kg/d, 5 d/wk, for 104 weeks. The primary dose
metric was selected to be average amount of DCVC bioactivated/kgyYd, 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 F.5.1, above).
The log-logistic model with slope constrained to > 1 was selected because (i) the log-logistic
model with unconstrained slope yielded a slope estimate < 1 and (ii) it had the lowest AIC.
The iPOD of 0.0132 mg DCVC bioactivated/kgyYd 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 iPOD. 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. 11. These values are
used as this critical effect's POD to which additional UFs are applied.
6/8/2009
F-28
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
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
BMDand BMDL. 6.70e-O6. NA
(D 0.6 _
£ 02 "
"i	1	1	1	r
0.0 0.2 0.4 0.6 0.8 1.0
ABioactDCVCBW34
loglogistic, Pval = 0.44, AIC = 122
background 0, intercept 1, slope 1
BMDand BMDL. 0.0fl91. 0.0132
0 0.6 ~~
0.0 0.2 0.4 0.6 0.8 1.0
ABioactDCVCBW34
0.8 -
"O
a;
| 0.6
<
£=
0
1	02
^ 0.0
0.4 -
multistage-1, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4, Beta(2) 0
BMDand BMDL. 0.0358. 0.0288
0.0 0.2 0.4 0.6 0.8 1.0
ABioactDCVCBW34
"O

0.8 -
a> 0.6
!t=
<
£=
O
0.4 -
£ 02
^ 0.0
multistage-2, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4, Beta(2) 0
BMDand BMDL 0.0 358, 0.0288
0.0 0.2 0.4 0.6 0.8 1.0
ABioactDCVCBW34
S 0 2
05
0.0 H
multistage-1, Pval = 0.05, AIC = 126
Background 0, Beta(1) 1.4
BMDand BMDL. 0.0658. 0.0288
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F Marshall rat
Human
BMDL
Rodent:
0.0132/
F Marshall rat
Human
0.00338
BMDL
Rodent
0.0132
NTP.1988
BMDL for systemic kidney toxic.nephropathy in
NTP.1988
BMDL for systemic kidney toxic.nephropathy in
Human
median%:
0.0422
Human
media n%:
0.0333
TCE inhalation (ppm)	TCE oral (mg/kg-d)
Figure F. 11. Derivation of HEC99 and HED99 corresponding to the rodent iPOD from NTP
(1988) toxic nephropathy in rats.
F.6.2. NCI (1976) - LOAEL for toxic nephrosis in mice
The critical endpoint here is toxic nephrosis in female B6C3F1 mice (NCI, 1976), which
was the most sensitive sex in this study, although the LOAEL for males differed by less than
50%.
F.6.2.1. Dosimetry
Mice were exposed to a time-weighted average of 869 and 1739 mg/kg/d, 5 d/wk, for 78
weeks. BMD modeling was not performed because the response at the LOAEL was > 90%. The
primary dose metric was selected to be average amount of TCE conjugated with GSH/kg Yd. In
this study, the lower dose group was exposed to two different dose levels (700 mg/kg/d for 12
weeks and 900 mg/kg/d for 66 weeks). The median estimates from the PBPK model for the two
dose levels were 0.583 and 0.762 mg TCE conjugation with GSH/kg Yd. Applying the same
time-weighted averaging gives an iPOD LOAEL of 0.735 mg TCE conjugation with GSH/kg Yd.
6/8/2009
F-30
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.6.2.2. Derivation of HECgg 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 iPOD. The derivation of the HEC99 of 0.50 ppm and HED99 of 0.30 mg/kg/d for the
99th percentile for uncertainty and variability are shown in Figure F. 12. These values are used as
this critical effect's POD to which additional UFs are applied.
NCI. 1976
TWA-LOAEL for systemic kidney toxic.nephropathy in
	FB5C3F1 mouse
o -
CD
X
CO
CD
-1—¦
0
o -=
Human
Human
median%
3
10 10 10 1 10 10 10 10
: TWA-LOAEL
¦	Rodent:
¦	0.735
TCE inhalation (ppm)
CD
X
(f)
(3
-*->
0
NCI. 1976
TWA-LOAEL for systemic kidney toxic.nephropathy in
	FB5C3F1 mouse
o -
: TWA-LOAEL
¦	Rodent:
¦	0.735
Human
99%:
0.304
Human
media n%
2
rTTTTTTTf I I I lllll|—I 11IIII^—I I 11IIll|—I I mHIJ—I I 11IIll|
^	. «-1 . .«1 . _.2 . _3 . _4
10 10 10 1 10 10 10 10
TCE oral (mg/kg-d)
Figure F. 12. Derivation of HEC99 and HED99 corresponding to the rodent iPOD from NTP
(1988) toxic nephrosis in mice.
F.6.3. Woolhiser et al. (2006) - BMD modeling of increased kidney weight in rats
The critical endpoint here is increased kidney weights in female SD rats (Woolhiser et al.,
2006).
F.6.3.1. Dosimetry and BMD modeling
Rats were exposed to 100, 300, and 1000, 6 hr/d, 5 d/wk, for 4 weeks. The primary dose
metric was selected to be average amount of DCVC bioactivated/kg Yd, with median estimates
from the PBPK model for this study of 0.038, 0.10, and 0.51.
6/8/2009
F-31
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Figure F. 13 shows BMD modeling for the continuous models used (see F.5.2, above).
The Hill model with constant variance was selected because it had the lowest AIC and because
other models with the same AIC either were a power model with power parameter <1 or had
poor fits to the control dataset.
The iPOD of 0.0309 mg DCVC bioactivated/kg/4/d was a BMDL for a BMR of 10%
weight change, which is the BMR typically used by 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.6.3.2. 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 iPOD. 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. 14. These values are
used as this critical effect's POD to which additional UFs are applied.
6/8/2009
F-32
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Woolhiser.etal.2006 Kidney kidney.wt.perlOOgm 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
	1	1	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
0.0
T
-r~
0.2
0.1 0.2 0.3 0.4
ABioactDCVCBW34
0.5
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
—r~
0.2
—I—
0.3
I—
0.4
ABioactDCVCBW34
I
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
0.1
—I—
0.2
I—
0.3
I—
0.4
ABioactDCVCBW34
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, betaO 0.83, betal 0.23
BMD and BMDL, 0.356, 0.234
I
0.0 0.1
0.2
-r~
0.3
0.4 0.5
ABioactDCVCBW34
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
	T"
0.0 0.1
-r~
0.2
0.3 0.4
ABioactDCVCBW34
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
0.0 0.1
0.2
~~r~
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
0.0
T
~~r~
0.2
0.1 0.2 0.3 0.4
ABioactDCVCBW34
0.5
Figure F.13. BMD modeling of Woolhiser et al. (2006) for increased kidney weight in female
SD rats.
6/8/2009
F-33
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Woolhiser.etal.2006	Woolhiser.etal.2006
F Snranue-Daw lev rat
Human
0.00791
BMDL
Rodent:
0.0309
F Snranue-raw lev rat
BMDL for systemic kidney weight.increased in
BMDL for systemic kidney weight.increased in
rnnr]—i i iiini|—11 iiiiij—11 iinij—i 11 mii|—11 iiini|—i i inii
10 4 10 3 10 2 10 1 1 101 102 103 104
rnnr]—11 iiuii|—11 miij—i i iinij—i iiiini|—i i iiini|—11 mil
10 4 10 3 10 2 10 1 1 101 102 103 104
Human
median%:
0.078
Human
99%:
BMDL
Rodent:
0.0309
Human
median%:
0.0987
TCE inhalation (ppm)	TCE oral (mg/kg-d)
Figure F. 14. Derivation of HEC99 and HED99 corresponding to the rodent iPOD from
Woolhiser et al. (2006) for increased kidney weight in rats.
F.6.4. Keil et al. (2009) - LOAEL for decreased thymus weight and increased anti-dsDNA
and anti-ssDNA antibodies in mice
The critical endpoints here are decreased thymus weight and increased anti-dsDNA and
anti-ssDNA antibodies in female B6C3F1 mice (Keil et al., 2009).
F.6.4.1. Dosimetry
Mice were exposed to 1400 and 14000 ppb 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/kg Yd. The
lower dose group was the LOAEL for both effects, and the median estimate from the PBPK
model at that exposure level was 0.139 mg TCE metabolized/kg Yd, which is used as the rodent
iPOD.
6/8/2009
F-34
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.6.4.2. Derivation of HECgg 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 iPOD. 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. 15. These values are
used as this critical effect's POD to which additional UFs are applied.
Keil.elal.2009
mouse B
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
metabolized by oxidation/kg3 Yd, 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 loglogistic modeling of
these data, with the highest dose group dropped, the iPOD of 0.0142 mg TCE metabolized by
oxidation/kg3 Yd 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.
F.6.5.2. 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 iPOD. 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. 16. These values are
used as this critical effect's POD to which additional UFs are applied.
0.00515
Human
99%:
0.00365
BMDL
Rodent:
0.0142/
Johnson.etal.2003
BMDL for developmental heart malformations in
Johnson.etal.2003
BMDL for developmental heart malformations in
Human
median%:
0.0116
I 11 Illll|—I I tlllll|—I I lllllj—I I Mill]—I 11 tllll|—I I IIllll|—I I Mill
10 4 10 3 10 2 10 1 1 101 102 103 104
BMDL
Rodent:
0.0142
Human
99%:
Human
median%:
0.00576
TCE inhalation (ppm)	TCE oral (mg/kg-d)
Figure F.16. Derivation of HEC99 and HED99 corresponding to the rodent iPOD from Johnson
et al. (2003) for increased fetal cardiac malformations in female SD rats using the total oxidative
metabolism dose metric.
6/8/2009
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
24
25
26
27
28
29
30
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
F.6.6. Peden-Adams et al. (2006) - 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 post-natally (Peden-Adams et al., 2006).
Mice were exposed to 1400 and 14000 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 post-natal ages of 3 or 8
weeks. The dose-response relationships were sufficiently supralinear that BMD modeling failed
to produce an adequate fit. In addition, because of the lack of an appropriate PBPK model and
parameters to estimate internal doses given the complex exposure pattern (placental and
lactational transfer, and pup ingestion post-weaning), no internal dose estimates were made.
Therefore, the LOAEL of 0.37 mg/kg/d on the basis of applied dose was used as the critical
effect's POD to which additional UFs are applied.
F.7. References
Box GEP, Hunter WG, Hunter JS. (1978). Statistics for Experimenters, New York: John Wiley
& Sons.
Buben, JA; O'Flaherty, EJ. (1985) Delineation of the role of metabolism in the hepatotoxicity of
trichloroethylene and perchloroethylene: a dose-effect study. Toxicol Appl Pharmacol
78:105-122.
Carney, EW; Thorsrud, BA; Dugard, PH; Zablotny, CL. (2006) Developmental toxicity studies
in Crl:Cd (SD) rats following inhalation exposure to trichloroethylene and
perchloroethylene. Birth Defects Research (Part B) 77:405-412.
Chia SE, Ong CN, Tsakok MF, Ho A. (1996) Semen parameters in workers exposed to
trichloroethylene. Reprod Toxicol 10(4):295-299.
Crofton, KM; Zhao, X. (1997) The ototoxicity of trichloroethylene: extrapolation and relevance
of high-concentration, short-duration animal exposure data. Fundam Appl Toxicol
38(1): 101—106.
Filipsson, A.F., K. Victorin. (2003). Comparison of available benchmark dose softwares and
models using trichloroethylene as a model substance. Regulatory Toxicology and
Pharmacology 37:343-355
6/8/2009
F-37
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
George, JD; Reel, JR; Myers, CB; Lawton, AD; Lamb, JC. (1986) Trichloroethylene:
reproduction and fertility assessment in F344 rats when administered in the feed. RTI
Project No. 310-2344, NTP-86-085. National Institute of Environmental Health Sciences,
National Toxicology Program, RTP, NC.
Griffin JM, Gilbert KM, Lamps LW, Pumford NR. 2000b. CD4(+) T-cell activation and
induction of autoimmune hepatitis following trichloroethylene treatment in MRL+/+
mice. Toxicol Sci 57:345-352.
Johnson, PD; Goldberg, SJ; Mays, MZ; Dawson, BV. (2003) Threshold of trichloroethylene
contamination in maternal drinking waters affecting fetal heart development in the rat.
Environ Health Perspect 111(3):289-292.
Keil, DE; Peden-Adams, MM; Wallace, S; Ruiz, P; Gilkeson, GS. (2009) Assessment of
trichloroethylene (TCE) exposure in murine strains genetically-prone and non-prone to
develop autoimmune disease. Journal of Environmental Science and Health, Part A 44:
443-453.
Kjellstrand P, Holmquist B, Aim P, Kanje M, Romare S, Jonsson I, Mansson L, Bjerkemo M.
(1983a) Trichloroethylene: further studies of the effects on body and organ weights and
plasma butyrylcholinesterase activity in mice. Acta Pharmacol Toxicol (Copenh)
53(5):375-84.
Kjellstrand P, Holmquist B, Mandahl N, Bjerkemo M. (1983b) Effects of continuous
trichloroethylene inhalation on different strains of mice. Acta Pharmacol Toxicol
(Copenh) 53(5):369-74.
Land, PC; Owen, EL; Linde, HW. (1981) Morphologic changes in mouse spermatozoa after
exposure to inhalational anesthetics during early spermatogenesis. Anesthesiology 54:53-
56.
Maltoni, C; Lefemine, G; Cotti, G. (1986) Experimental research on trichloroethylene
carcinogenesis. In: Maltoni, C; Mehlman MA., eds. Vol. 5. Archives of research on
industrial carcinogenesis. Princeton, NJ: Princeton Scientific Publishing;
Mhiri, C; Choyakh, F; Ben, HM; et al. (2004) Trigeminal somatosensory evoked potentials in
trichloroethylene-exposed workers. Neurosciences 9(2): 102-107.
6/8/2009
F-38
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
Moser, Cheek & MacPhail. A multidisciplinary approach to toxicological screening III.
Neurotbehavioral toxicity. J Toxicol. Environ. Hlth., 1995, 45, 173-210.
Narotsky, MG; Weller, EA; Chinchilli, VM; Kavlock, RJ. (1995) Nonadditive developmental
toxicity in mixtures of trichloroethylene, di(2-ethylhexyl) phthalate, and heptachlor in a 5
x 5 x 5 design. Fundam Appl Toxicol 27:203-216.
NCI (National Cancer Institute). (1976) Carcinogenesis bioassay of trichloroethylene. Division
of Cancer Cause and Prevention, National Cancer Institute, U.S. Department of Health,
Education, and Welfare, DHEW Publication No. (NIH) 76-802, Technical Report Series
No. 2, 218 pages; NCI-CG-TR-2; NTIS PB254122.
http://ntp.niehs.nih.gov/ntp/htdocs/LT_rpts/tr002.pdf.
NTP (National Toxicology Program). (1988) Toxicology and carcinogenesis studies of
trichloroethylene (CAS no. 79-01-6) in four strains of rats (ACI, August, Marshall,
Osborne-Mendel) (gavage studies). Public Health Service, U.S. Department of Health
and Human Services; NTP TR-273; NIH Publication No. 88-2529. Available from the
National Institute of Environmental Health Sciences, Research Triangle Park, NC, and
the National Technical Information Service, Springfield, VA; PB88-218896.
http://ntp.niehs.nih.gov/ntp/htdocs/LT_rpts/tr273.pdf.
NTP (National Toxicology Program). (1990) Carcinogenesis Studies of Trichloroethylene
(Without Epichlorhydrin) (CAS No. 79-01-6) in F344/N Rats and B6C3F1 Mice (Gavage
Study). NTP TR 243. Research Triangle Park, NC: U.S Department of Health and
Human Services.
Peden-Adams MM, Eudaly JG, Heesemann LM, Smythe J, Miller J, Gilkeson GS, et al. (2006).
Developmental immunotoxicity of trichloroethylene (TCE): studies in B6C3F1 mice. J
Environ Sci Health A Tox Hazard Subst Environ Eng 41:249-271.
Snedecor GW, Cochran WG. (1980). Statistical Methods (7th ed.), Ch. 9.12 and Ch 9.14
(pp.169-172)
U.S. EPA (Environmental Protection Agency) (2008b). Benchmark Dose Technical Guidance
(Inter-Agency Review Draft).
U.S. EPA (Environmental Protection Agency). (1994) Methods for derivation of inhalation
reference concentrations and application of inhalation dosimetry. Environmental Criteria
and Assessment Office, Office of Health and Environmental Assessment, Washington,
6/8/2009
F-39
-------
INTER-AGENCY REVIEW DRAFT - DO NOT QUOTE OR CITE
1	Washington, DC; EPA/600/8-90/066F. Available from: National Technical Information
2	Service, Springfield, VA; PB2000-500023.
3	Woolhiser, MR; Krieger, SM; Thomas, J; Hotchkiss, JA. (2006) Trichloroethylene (TCE):
4	Immunotoxicity potential in CD rats following a 4-week vapor inhalation exposure. Dow
5	Chemical Company, Toxicology & Environmental Research and Consulting, Midland,
6	MI, Study ID 031020, July 5, 2006, unpublished.
7
6/8/2009
F-40
-------