vvEPA
United States
Environmental Protection
Agency
Office of Health and
Environmental Assessment
Washington DC 20460
EPA/600/8-87/029A
June 1987
External Review Draft
Research and Development
Technical
Analysis of
New Methods and
Data Regarding
Dichloromethane
Hazard
Assessments
Review
Draft
(Do Not
Cite or Quote)
NOTICE
This document is a preliminary draft. It has not been formally
released by EPA and should not at this stage be construed to
represent Agency policy. It is being circulated for comment on
its technical accuracy and policy implications.
-------
-------
DRAFT
DO NOT QUOTE OR CITE
EPA/600/8-87/029A
June 1987
Review Draft
TECHNICAL ANALYSIS OF NEW METHODS AND DATA
REGARDING DICHLOROMETHANE HAZARD ASSESSMENTS
Prepared by:
The Interagency Hazard/Risk
Assessment Committee of the Integrated
Chlorinated Solvents Project
NOTICE
THIS DOCUMENT IS A PRELIMINARY DRAFT. It has not been formally
released and should not at this stage be construed to represent
policy. It is being circulated for comment on its technical
accuracy and policy implications.
Published by:
Office of Health and Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, D.C.
-------
DISCLAIMER
'(
This document is an external draft for review purposes only
and does not constitute policy. Mention of trade names or
commercial products does not constitute endorsement or
recommendation for use.
ii
-------
CONTENTS
Tables ..... ...... ....... .......... ,vi
Figures .......... ................. vii
Preface ........ .............. ..... viii
Abstract .......... .................. ix
Authors, Contributors, and Reviewers .......... . . . .xi
1. INTRODUCTION. ..... ..... ....... ...... 1
2. PHARMACOKINETICS AND PHARMACOKINETIC MODELS ........ 5
2.1. INTRODUCTION. . . .......... . ....... 5
2.2. PHARMACOKINETIC MODELS AND METABOLISM . ..... . . 6
2.2.1. Model Used by Andersen and Reitz. ...... 6
2.2.2. Model Used by Angelo et al .......... 17
2.2.3. Summary . ........... ...... .23
3. METABOLISM OF DICHLOROMETHANE ....... . ....... 29
3.1. IN VIVO METABOLISM .......... „ ....... 30
3.2. REACTIVE INTERMEDIATES. . / ............. 42
3.3. USE OF IN VITRO DATA TO PREDICT IN VIVO METABOLISM. .46
3.4. IN VITRO METABOLISM ......... ........ 49
3.5. SUMMARY ........ .............. ^.55
4. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODEL: METABOLIC
KINETIC CONSTANTS ... ........ .......... 58
5. MUTAGENICITY ASSESSMENT OF DICHLOROMETHANE: REVIEW OF
STUDIES PERFORMED BY CEFIC ......... „ ...... .66
5.1. PHASE I TESTS . .. .............. ... ,.66
5.1.1. In vivo Unscheduled DNA Synthesis ...... 66
5.1.2.^ In vitro Unscheduled DNA Synthesis ...... 68
5.1.3. Covalent Binding to DNA in vivo ....... 70
5.1.4. Summary of Phase I Tests . ......... 74
5.2. PHASE II TESTS. ........ ........... 75
5.2.1. Mouse Micronucleus Test . ..... ..... 75
5.2.2. Induction of S-*Phase Hepatocytes
5.2.3. Summary of Phase II Tests
6. EPIDEMIOLOGY: RECENT KODAK STUDY ......... ... .79
6.1. INTRODUCTION. . . .................. 79
6.2. HRAC REVIEW OF THE STUDY BY HEARNE ET AL. (1987). . .81
iii
-------
6.3. APPLICATION OF EPIDEMIOLOGY TO QUANTITATIVE CANCER
RISK ASSESSMENT AND COMPARISON WITH ANIMAL-BASED
ESTIMATES 85
6.3.1. Previous Quantitative Cancer Risk Assess-
ments for DCM 85
6.3.2. Quantitative Risk Estimation Based on the
Study of Kodak Employees 89
6.3.2.1. Excess or Additive Risk Model . . .89
6.3.2.2. Multiplicative or Relative Risk
Model .91
6.3.2.3. Data .93
6.3.2.4. Results 96
6.4. DISCUSSION 97
6.5. SUMMARY ........ 101
7. SCALING RISK ACROSS SPECIES USING DELIVERED DOSE .... 103
7.1. INTRODUCTION. 103
7.2. SCALING APPLIED DOSE TO EXTRAPOLATE RISK ACROSS
SPECIES 105
7.3. PHARMACODYNAMICS 108
7.4. PHARMACOKINETICS, PHARMACODYNAMICS, AND THE
SURFACE AREA CORRECTION Ill
7.4.1. Assuming that the Surface Area Correction
Accounts for Pharmacokinetics . 112
7.4.2. Assuming that the Surface Area Correction
Accounts for Pharmacodynamics 114
7.4.3. Other Possible Assumptions for PD and PK. . 116
7.4.4. High- to Low-Dose Extrapolation 117
7.5. THE PHARMACOKINETIC MODEL USED BY ANDERSEN ET AL.
FOR DCM 120
7.5.1. Breathing Rates 121
7.5.2. Using the Surface Area Correction ..... 123
7.5.3. Developing a Unit Risk Based on Internal
Dose: Incorporation of High- to Low-Dose
Differences and Species-to-Species
Differences » 126
7.5.4. Developing a Unit Risk Based on Internal
Dose: Incorporation of Only High- to Low-
Dose Differences 133
7.5.4.1. Review of Metabolism Data .... 134
7.5.4.2. Robustness of Model Output. ... 138
7.5.4.3. Using Pharmacokinetics for High-
to Low-Dose Extrapolation .... 142
iv
-------
7.6 CONCLUSIONS
References. . . . . ,
144
147
-------
TABLES
1. Metabolism of DCM to CO and CC>2 in mice 38
2. The mutagenic effect of dihalomethanes 44
3. The mutagenic effect of DCM on S_. typhimurium TA100
in the presence of various rat liver fractions 45
4. Kinetic constants for the metabolism of DCM to carbon
monoxide by liver microsomes 46
5. Kinetic constants for the metabolism of DCM to
formaldehyde by liver cytosol 47
6. Kinetic constants for the metabolism of dihalomethanes
to carbon monoxide by rat liver microsomes 48
7. The metabolism of 2,4-dinitrochlorobenzene
by rat liver and lung cytosol 55
8. Comparison of previous incremental cancer risk
estimates for methylene chloride based on pancreatic,
lung, and total cancer response in,the Kodak
employees' study and the Dow study. 87
9. Person-years of observation and observed and expected
deaths from lurig cancer for Kodak employees exposed to
DCM with follow-up through 1984 94
r
10. Person-years of observation and observed and expected
deaths from pancreatic cancer for .Kodak employees
exposed to DCM with follow-up through 1984 95
11. Comparison of incremental cancer risk estimates for
DCM based on the NTP female mouse lung response with
estimates based on lung and pancreas cancer death
response in 1013 Kodak employees 98
12. Sensitivity analysis for kj- 140
13. Sensitivity analysis for Vmax » 14d
14. Sensitivity analysis for % . 141
15. Sensitivity analysis for Vmax/Kjj 142
vi
-------
FIGURES
1. End of exposure HbCO levels in naive, 2,3-EP, and
pyrazole pretreated rats following 4-hour exposures to
DBM 31
2 End of exposure HbCO levels in naive, 2,3-EP, and
pyrazole pretreated rats following 4-hour exposures to
DCM .32
3. End of exposure HbCO levels in naive, 2,3-EP, and
pyrazole pretreated rats following 4-hour exposures to
bromochl or ome thane. 33
4. Dependence of plasma inorganic bromide levels on the
ambient concentrations of bromochloromethane following
4-hour exposures using naive, 2,3-EP, and pyrazole
pretreated rats 35
5. Dependence of plasma inorganic bromide levels on the
ambient concentrations of dibromomethane following
4-hour exposures using naive, 2,3-EP, and pyrazole
pretreated rats . . . . . . . .. . . ... .. . .36
6. Proposed pathways for dihalomethane metabolism .50
vii
-------
PREFACE
This document was prepared by the Health/Risk Assessment
Committee (HRAC) of the Integrated Chlorinated Solvents Project,
a committee comprised of representatives from four federal
regulatory agencies. This interagency committee was established
to evaluate the health effects caused by dichloromethane (DCM)
and five other halogenated solvents. As part of its work on
halogenated solvent compounds, the HRAC reviewed and evaluated
the information recently submitted to EPA and other federal
agencies on DCM's potential to cause cancer and other toxic
effects.
This document provides an extensive analysis of the new data
that have become available since the publication of EPA's Health
Assessment Document (HAD) for Dichloromethane (DCM) and Addendum
in 1985, and also discusses alternative methodologies for
determining risk for DCM. Thus, this document is not intended to
replace the 1985 reports, but to provide an evaluation of the
recent data and risk assessment methodologies. Its purpose is to
provide the agencies with the latest background information on
DCM that can be used by each agency in developing its own risk
assessment.
Vlll
-------
ABSTRACT
New information on cytotoxicity, genotoxicity, and
/
epidemiology has raised some questions about the federal
regulatory agencies' cancer risk assessments for dichloromethane
(DCM, methylene chloride). In addition, physiologically based
pharmacokinetic models have been developed/ showing that tissue-
level delivery of metabolically activated DCM may be
disproportionately reduced at low exposure levels. These studies
suggest to some that the clear carcinogenic response seen in mice
under chronic high exposures does not imply substantive human
risk at low doses..
The Health/Risk Assessment Committee (HRAC), comprising
representatives of four federal regulatory agencies, was convened
to conduct joint analyses of these new data. This document
reports on the HRAC's consideration of the data and the questions
they raise about human cancer risk from DCM. It serves as a
source of up-to-date analyses that may be drawn upon 'by each
agency as it considers modifying its cancer risk assessment.
The HRAC finds that, despite new data, the mechanism of
carcinogenic action of DCM remains problematical; there is no
basis at present to conclude that carcinogenic response is unique
tG mide or confined to high exposure levels. Uncertainties in
current pharmacokinetic models for DCM are examined, and their
application to extrapolating animal-based risks to humans is
discussed extensively. Negative epidemiologic studies do not
ix
-------
contradict the human risk estimates extrapolated from
experimental animals.
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
This document represents the joint efforts of scientists
from several federal regulatory agencies who are members of the
Hazard/Risk Assessment Committee of the Integrated Chlorinated
Solvents Project.
EPA's Office of Health and Environmental Assessment had
overall responsibility for coordination and direction of the'
document preparation and production effort (Jerry N. Blancato,
Project Manager).
AUTHORS
Steven Bayard
Carcinogen Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
David L. Bayliss
Carcinogen Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Jerry No Blancato
Exposure Assessment Group
Office of Health and Environmental Assessment
U.S« Environmental Protection Agency
Miriam Bloom
U.S. Consumer Product Safety Commission
Murray Cohn
U.S. Consumer Product Safety Commission
William H. Farland
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Chapter 6
Chapter 6
Chapters 1 & 2
Chapter 5
Chapters 4 & 7
Chapter 5
xi
-------
David Jacobson-Kram
Reproductive Effects Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Lorenz Rhomberg
Carcinogen Assessment Group
Office of Health and Environmental Assessment
(formerly with the Office of Toxic Substances)
U.S. Environmental Protection Agency
Hugh L. Spitzer
Office of Regulatory Analysis
U.S. Environmental Protection Agency
Chapter 5
Chapter 7
Chapter 3
CONTRIBUTORS
Robert Brown
Food and Drug Administration
Jane Hopkins
Office of Solid Waste and Emergency Response
(formerly with the Office of Toxic Substances)
U.S. Environmental Protection Agency
Ronald Lorentzen
Pood and Drug Administration
REVIEWERS
The following individuals reviewed earlier drafts of this
document and provided valuable comments.
Karl P. Baetcke
Office of Toxic Substances
U.S. Environmental Protection Agency
Diane D. Beal
Office of Toxic Substances
U.S. Environmental Protection Agency
Karen Blanchard
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
xii
-------
Chao W. Chen
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Margaret M.L. Chu
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Fred DiCarlo
Office of Toxic Substances
U.S. Environmental Protection Agency
Amanda Edens
Occupational Safety and Health Administration
U.S. Department of Labor
Ernest Falke
Office of Toxic Substances
U.S. Environmental Protection Agency
Richard N. Hill
Office of Pesticides and Toxic Substances
U.S. Environmental Protection Agency
Charalingayya B. Hiremath
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Carl Mazza
Office of Toxic Substances
U.S. Environmental Protection Agency
Paul White
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
i
Harold Zenick
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
xiii
-------
-------
1. INTRODUCTION
This document is the product of the Hazard/Risk Assessment
Committee (HRAC) of the Integrated Chlorinated Solvents Project.
The following federal regulatory agencies participated in this
effort: the U.S. Consumer Product Safety Commission (CPSC), the
U.S. Environmental Protection Agency (EPA), the Food and Drug
Administration (FDA) of the U.S. Department of Health and Human
Services, and the Occupational Safety and Health Administration
(OSHA) of the U.S. Department of Labor. Scientists from these
agencies cooperated in an intense effort to review the numerous
technical papers on dichloromethane (DCM, methylene chloride)
that had been submitted to the agencies or published since the
publication of EPA's Health Assessment Document (HAD) for
Dichloromethane and Addendum in 1985. The chapters that follow
report on the extensive analyses of the new data and alternative
methodologies for determining risk for DCM.
This document does not replace previously published
documents, nor is it a risk assessment per se. It is the H&AC's
intention that this document be used as background when each
agency develops its most up-to-date risk assessment for DCM for
its own mandated purpose.
Chapters 2 through 4 review the physiologically based
pharmacokinetic models used by Andersen et al. (1986, 1987) and
Angelo et al. (1984) to describe and attempt to predict the
disposition of DCM and its metabolites in the body. An ultimate
-------
goal of such models oould bo to quantitatively account for
interspecies differences in metabolism and pharmacokinetics.
Also, the models can be employed to account for pharmacokinetic
nonlinearities that arise when dose-to-dose extrapolations are
performed in the risk assessment process.
Although the models are capable of predicting some facets of
DCM disposition, several concerns result regarding the structure
and parameters of the models. The most crucial uncertainty
appears to revolve around the estimates chosen for the key rate
constants in the metabolio pathways that transform DCM into other
products, including putative carcinogenic species, other
studies, submitted by the European Council of Chemical
Manufacturer's Federation (CEFIC, 1986a, e, f), were also
extensively reviewed in an attempt to reduce the uncertainty
associated with the metabolic parameters.
Chapter 5 reviews several submitted studies (CEFIC I986b, c,
d, g) regarding DCM's potential mechanism of carcinogenicity.
Although not in themselves conclusive, the results of these
studies could be consistent with considering DCM to be a weak
genotoxic agent.
A recent update of the epidemiology of Kodak workers with
known exposure to DCM (Hearne et al., 1987) was reviewed in
Chapter 6. The quality of the study was assessed and the study
results were compared for consistency with risks calculated from
animal-based experiments. The HRAC concluded that the risks
calculated from animal-based experiments can be considered to be
-------
consistent with the results of the epidemiology studies.
Chapter 7 illustrates two possible applications of how
\ '
pharmacokinetic information and data may be incorporated into the
quantitative risk assessment. One method incorporates
interspecies differences in pharmacokinetics while the other only
incorporates differences resulting from high- to low-dose
extrapolation. The major uncertainties associated with each
approach are discussed. For the present, when using the
pharmacokinetics for high to low dose only, the estimated upper
bound on the risk would be reduced from the applied dose
estimates, at a minimum, slightly more than twofold. On the
other hand, using pharmacokinetic data and models for
interspecies extrapolation would result in a reduction of risk
from the applied dose estimate by almost ninefold. The two
methods differ in the assumptions that are made and are not
equally sensitive to one of the key metabolic rate constants.
For example, the ninefold reduction resulting after incorporating
pharmacokinetics into the interspecies extrapolation could be
substantially altered with alternative estimates for the
metabolic rate constant for the conversion of DCM by a
glutathione-S-transferase (GST) mediated pathway.
In formulating methodology to incorporate pharmacokinetic
information into risk assessments, a number of generic questions
are revealed. The relative importance of interspecies
differences in pharmacokinetics and pharmacodynamics will have to
be discerned to further reduce uncertainty in the risk assessment
-------
process. Pharmacokinetic models may not necessarily settle all
the questions regarding equivalency of doses between different
species; however, valuable insight regarding the magnitude and
consequence of metabolic and pharmacokinetic differences between
species can be gained from their use. Exposure-related
questions/ such as comparing the effect of sporadic high level
with low level sustained exposure can be more readily and
accurately answered using pharmacokinetic models.
Pharmacokinetic data and pharmacokinetic models do not answer all
of the "old" questions that have faced previous risk assessors
but hopefully provide new insights into understanding how major
uncertainties can be reduced.
-------
2. PHAHMACOKINETICS AND PHARMACOKINETIC MODELS
2.1. INTRODUCTION
One of the most challenging issues concerning the HRAC's
risk assessment of DCM has been with respect to pharmacokinetics.
It is very important to ascertain whether or not the estimate of
risk for this compound is appreciably altered after considering
pharmacokinetic data. Basically two physiologically based
pharmacokinetic (PBPK) models have been formulated to predict the
disposition of DCM and its metabolites in tha body. Th© first
model is based on an earlier model formulated to describe the
disposition of styrene (Ramsey and Andersen, 1984). This model
has been modified by Andersen et al. (1986, 1987) and Reitz et
al. (1986) to account for DCM exposure. Earlier drafts of these
studies were supplied to the various federal regulatory agencies
for review. It is postulated by Andersen et ale (1986, 1987) and
Reitz et al. (1986) that if the EPA considers the results of this
model, the risk number would be greatly reduced from what is
presently in the HAD for DCM.
A second PBPK model formulated by Angelo et al. (1984) has
also been used to describe the disposition of DCM and its
metabolites in the body. A series of accompanying papers have
been published (Angelo and Pritchard, 1984; Angelo et al., 1986a,
b) that provide data on several important pharmacokinetic
considerations. After reviewing both the Andersen and Angelo
papers, it becomes quite obvious that the two models have
• 5
-------
significantly different structures. This chapter will discuss
the use of pharmacokinetics in risk assessment and will examine
i
both models and elucidate differences between the two. It will
also assess some of the assumptions that have gone into the
formulation of the models.
2.2. PHARMACOKINETIC MODELS AND METABOLISM
2.2.1. Model Used by Andersen and Reitz
This PBPK model for DCM is based on a similar model
published by Ramsey and Andersen (1984) to describe the
disposition of styrene. The modal has been modified for the case
of DCM (Andersen et al., 1986, 1987) to account for metabolism by
two pathways. Both metabolic pathways are assumed to occur in
the liver as well as in the lung. One path is mediated by the
P-450 system and is considered to exhibit saturation kinetics at
the given exposures. The second pathway is mediated by the
glutathione-S-transferase (GST) system. This pathway is
considered to be first order at the given exposure conditions.
Most importantly, it is assumed by Andersen et al. (1986, 1987)
and Reitz et al. (1986) that this pathway is the only source of
carcinogenicity. Further, it is postulated that the activity of
this pathway in humans is less than that of mice and might only
become significant when the P-450 pathway has been saturated, a
process occurring only at concentrations above those expected in
conditions of human exposure. Reports presented by CEFIC (1986e)
further imply that the activity of this pathway in humans is at
least two orders of magnitude lower than in mice, and in fact,
-------
may be totally nonexistent. Thus, it is inferred that human risk
is practically zero.
For the inhalation case the model assumes that the inhaled
air in the lung and pulmonary blood quickly achieves and
maintains steady-state conditions throughout the course of
exposure. None of the other organs are assumed to be in steady
state. This steady-state assumption is reasonable for this type
of compound under the conditions of exposure outlined by Andersen
©t al. (1986, 1987). The model is structured 00 that the lung is
divided into two subcompartmentss a gas exchange compartment and
a metabolism compartment. The lung was modeled with this
subcompartmentalization because the DCM rapidly equilibrates
between air and lung blood. Thus, it was assumed that this
equilibration was completed before the DCM entered the lung
tissue (Andersen et al., 1987). Given, the possibly high
*
metabolic activity of some of the lining cells (CEFIC, 1986a),
this may not be a totally accurate picture of the lung. However,
it is probable that error, if introduced, would be small.
Some key physical parameters that go into the model are
/
partition coefficients. These partition coefficients actually
represent the ratio of distribution of DCM between a tissue
compartment and th© blood at conditions where the blood and the
tissue are at equilibrium or between blood and air. Although
they may be determined in a variety of ways, often partition
coefficients are determined in an experimental situation with
dosing performed so that blood concentrations are at steady
-------
state, such as may be achieved with a constant infusion.
Alternately, they are determined mathematically from data
obtained at various times after exposure. Usually these data
should be the result of exposure at more than one dose. The
method used by Andersen et al. (1986, 1987) for DCM, however, is
significantly different. Tissues are homogenized and then
partition coefficients are determined with a vial equilibration
technique. This essentially means that air/tissue partition
coefficients are determined for various tissues including the
blood. A ratio of the air/blood to air/tissue coefficient is
then the tissue blood/partition coefficient. Considerable
concern arises over whether or not this results in an accurate
determination of this parameter.
Essentially, distribution of a compound among different
compartments is governed by its reversible binding with proteins
*
and/or other constituents and permeability of the various
membranes across which it may pass (Terasaki et al., 1984). The
homogenization of tissues clearly alters their normal
architecture and thus leads to several experimental problems that
could effect the accuracy of the resultant partition
coefficients. For example, it is not clear how, after the
homogenization process, artifacts such as disruption of normal
membranes, degeneration of normal binding components, and altered
metabolism should be taken into account.
Angelo at al. (1984) used a different method for determining
partition coefficients which will be discussed more fully in
8
-------
subsequent sections. The model used by Angelo et al. uses
coefficients for two subcompartments in each organ, a vascular
compartment and' an extravascular compartment. The extravascular
compartment is thought to be perhaps lipid containing.
Comparisons of the coefficients reported by Andersen et al.
(1986, 1987) paper with those reported for the vascular
compartment in th® Angelo et al. (1984) paper reveal a close
correlation. Thus, in this case it is probable that the
partition coefficients used by Andersen et al. (1986, 1987) are
within acceptable ranges for describing partitioning across the
vascular membrane. It is possible that some inaccuracy could
result if an organ has a significant lipid fraction and
partitioning occurs from the vascular region into the lipoid
region. If tissue and blood values for DCM were determined and N
reported over time, the distribution values could be estimated by
other means (King et al., 1983), and thus a mor® direct
comparison of in vivo and in vitro values could be made. In
addition, in a memo sent to the U.S. EPA dated March 9, 1987
(TSCA docket no. OPTS-62045), Harris reports that work done at
the Central Toxicology Laboratory of the Imperial Chemical
Sildtiiiries-United Kingdom (ICI-UK) shows that rats and mice h&vfe
the same air/blood partition coefficient, in contrast with the
differences reported by Andersen et al. (1986, 1987). The actual
quantitative impact that this discrepancy could have on the model
has not yet been determined, but it does raise some questions
about the methods used by Andersen et al. (1986, 1987). In fact,
-------
in earlier discussions ICI-UK scientists stated that in order for
the model to coincide with data generated in their laboratories,
the fat/blood partition coefficient had to be raised almost an
order of magnitude. This might make the value for this parameter
in excess of expected and other reported values. Also, a
combination of errors in the metabolic rate constants and the
partition coefficients may cause the model used by Andersen et
al. (1986, 1987) to be in error. While few definitive
conclusions can be drawn at this time, the sensitivity of the
model to these parameters serves to demonstrate that PBPK models
depend on a complex set of parameters. Because of the
interactions of those parameters within the model, estimating
error for any one of the parameters can become a very difficult
task.
A short discussion regarding the breathing rate parameters
that have been used by Andersen et al. (1986, 1987) and Reitz et
al. (1986) is warranted. Andersen et al. (1986, 1987) determined
breathing rates for humans and mice by direct observation. The
value for mice is higher than EPA's standard assumption, while
that for humans is markedly lower. The Andersen value for humans
is for a person at rest, but the EPA, FDA, and CPSC use a Value
considered typical of average activity (almost twice as high as
Andersen's value) or occupational activity (nearly three times
higher than Andersen's value). The values chosen by Andersen et
al. (1986, 1987) and Reitz et al. (1986) are not necessarily
incorrect but apply to the specific exposure conditions that
10
-------
those experimenters observed for a given set of experiments. To
compare any estimate of body burden or risk estimation with that
typically calculated by the federal regulatory agencies, the
breathing parameters should be the same for both methods.
Andersen et al. (1986, 1987) compared the federal regulatory
agencies' prediction of risk with predictions based on the PBPK
model using the original breathing rates; i.e., risks based on
the higher human breathing rate without pharmacokinetics have
been compared to risks based on the lower breathing rate with
pharmacpkinetics. This issue is discussed further in Chapter 7
which contains an evaluation of the effects of the PBPK model on
the risk assessment.
Another key set of parameters is the metabolic constants,
which are discussed in more detail in subsequent chapters.
Briefly, Andersen et al. (1986, 1987) and Reitz et al. (1986)
applied their model to data on the disappearance of DCM from a
closed inhalation chamber, as the compound is taken up and
metabolized by mice, rats, and hamsters. They developed values
on the rates of metabolism by each pathway for the model, not
from direct experimentation but by mathematical optimization, to
provide the best fit of the model to the data on the
disappearance of DCM from the chamber. Because human data were
judged to be inappropriate for the optimization routine, the
value for the human GST metabolism constant (kF) was determined
by scaling based on body weight to a power. Andersen ©t al.
(1986, 1987) observed that clearance (mL/hr) for the GST pathway
11
-------
in the liver appeared to have an allometric relationship in the
three rodent species, as demonstrated by the fact that intrinsic
clearance (rate constant times the liver volume divided by body
weight raised to the 0.7 power) was nearly equal in all three
species. The highest experimentally observed intrinsic clearance
(60 mL/hr/kg) was then used to estimate the clearance in 70-kg
humans. From this clearance, the human kF was determined. As is
discussed in Chapters 3 and 4, the HRAC has some reservations
with the method and the results of such an extrapolation. The
potential for error in this method is illustrated by the fact
that the findings of CEFIC (1986e) contradict the extrapolation
prediction of substantial GST metabolism in human liver.
Although the HRAC feels that the sensitivity of the method used
by CEFIC (1986e) is very limited, results of those in vitro
experiments indicate very littl© GST activity towards DCM in
human liver tissues. However, it is also well to note that no
agreed-upon methods exist to extrapolate metabolic constants from
in vitro to in vivo either.
As far as the model is concerned, the metabolic parameters
are very important. The premise with this model is that the risk
is associated with the formation of metabolites in the liver and
the lung. Thus, the actual desired and important output of the
model is the metabolite production.
It appears from some investigations (Chapter 4) that the
varying of the metabolic parameters results in a different
prediction of metabolite production while still fitting the data
12
-------
regarding the disappearance of DCM from the inhalation chamber.
The result is that an inaccurate metabolic parameter in the model
would not be noticed if model output was compared simply to
chamber data. Thus, it is very important to have actual
metabolite production data to "test" the model. This is
particularly important if metabolite production is key to risk
estimation. Although the gathering of preliminary data by Reitz
et al. (1986) is a first step in this process, more such data
will be needed to properly determine the level of certainty of
this model.
Additional key metabolic parameters in the model are those
that specify the relative activities of the two pathways between
metabolically active organs, in this case the liver and lung.
Andersen et al. (1986, 1987) partitioned activity of the MFO and
GST pathways between liver and lung using data from Lorenz et. al.
(1984) on the relative activities in each tissue toward surrogate
substrates (2,4-dinitrochlorobenzene for GST and 7-ethoxycoumarin
for MFO activity). On the basis of the Lorenz data, Andersen et
al. (1986, 1987) set the proportion of MFO metabolism occurring
in the human lung at a very low level compared to the mouse Xttfif»
Lorenz and coworkers noted, however, that their human lung
preparation contained endogenous inhibitors of the MFO pathway.
Concerning GST "surrogates," it is not clear whether.or not
isoenzymes exist, and if so, whether they would all show the same
level of activity. (Specific reviews of this procedure are
discussed in Chapter 3.) The model would be very sensitive to
13
-------
errors in the partitioning of metabolism. In particular, an
error in the partitioning of GST activity between liver and lung
would result in essentially a proportional error in the model's
prediction of metabolite, and thus of the estimated risk.
CEFIC (1986f) have gathered data in vivo in both rats and
mice in an attempt to validate existing PBPK models of DCM
disposition. Briefly those findings indicate that the
disposition in mice is different than that in rats. These
studies do substantiate the fact that mice convert DCM more
efficiently than do rats. In addition, both species show a two-
phase elimination profile after exposure ceases. The slower
second phase, much more pronounced in rats, is probably due to
release of DCM which had sequestered in lipid-rich compartments
of the tissues during exposure. One probable explanation for the
longer second phase elimination in rats is that, due to lower
metabolic rates, rats tend to sequester the parent compound in
lipid-rich compartments during exposure and slowly release it
/•
after the exposure period has terminated. These differences
between rats and mice are observed at levels above which
mechanisms for production of carbon monoxide formation from DCM
are saturated.
These data further indicate that after saturation of the
carbon monoxide pathway in rats, there is a nearly linear
iildfease in the area under the curves (AUC) for DCM in the bleoti
with increasing exposure concentration. On the other hand, in ,
the mouse, the AUC for DCM in the blood does not show a linear
14
-------
relationship with exposure concentration but rather an increase
in a sublinear fashion. Thus, it appears that the ratio of the
Vmax for the P-450 pathway and the kp (first-order rate constant)
for the GST-mediated pathway is different in the mouse as
compared to the rat. Therefore, the finding of linearity of the
blood concentration of the parent compound with increasing
exposure, concentration in the rat, but not in the mouse, need not
necessarily lead to the conclusion that no GST activity occurs in
the rat. Rather, it could simply mean that the ratio of Vmax to
kp is greater in the rat than in the mouse.
Careful examination of the profiles of DCM in the blood of
both rats and mice reveals additional interesting findings. It
appears that at the lowest exposure concentration (500 ppm) rats
are more efficient at removing DCM from the blood than are mice.
Yet, at; this 500-ppm exposure both species show the same level of
saturation of hemoglobin with carbon monoxide. Thus, it might be
deduced that rats, at low exposure conditions, have some
mechanism in addition to P-450 metabolism to remove DCM from the
blood. There are two possible explanations for this more
efficient blood removal of the parent compound at low exposure
concentrations by the rat. First, the rat is able to sequester
more of the parent compound in fat-containing regions of the
body. Angelo et al. (1984) found that DCM elimination could only
be explained (after intravenous dosing in a vehicle containing
polyethylene glycol and water) if some sequestering into lipid-
rich compartments of the organs were considered. A second
15
-------
possibility is that rats show some other metabolic process at
very low exposure concentrations. However, it is clear from
examining carbon dioxide elimination patterns that carbon
dioxide-producing pathways, such as the GST pathway, although
present, are not as active even at low doses in rats.
An apparent explanation for at least part of this more
efficient elimination of DCM from the blood by rats is that due
to a larger fat compartment, they are able to sequester the
compound more efficiently at these lower concentrations than are
mice. Also, the "filling" of the fat compartment can be thought
of as a saturation of the another process of eliminating DCM from
the blood. It appears that both the fat compartment and the
P-450 metabolism mechanism are saturated by 1000 ppm. In
conclusion, these data strongly suggest that the metabolic
capabilities of mice are different than those of rats regarding
DCM. Mice are able to continue metabolizing the compound even
after the oxidative carbon monoxide-producing pathway has been
saturated, as opposed to rats which appear to have a lower level
of activity of the GST pathway at the tested exposure conditions.
CEFIC (1986f) has not reported on any comparison of their in
vivo data with predictions of the model of Reitz et al. (1986).
However, in discussions with the HRAC they have stated that the
fat/blood partition coefficient had to be increased by an order
of magnitude in order for the model to fit. This increase would
be beyond any expected range for the value for this parameter*
Thus, it is apparent, according to those workers, that some other
16
-------
parameter in the model must be in error, or that the model's
structure is in error. Green and coworkers are under the
assumption that the metabolic parameters selected by Reitz et al.
(1986) are in error, and have embarked on a series of experiments
that may greatly aid in aatablishingf more accurate values for
fenbie parameters. In general, the HRAC agrees that any error in
the model is likely to be with the metabolic parameters; however,
a question remains regarding partition coefficients (section
2.2.1.) and the model's structure. This question arises when
comparing the model used by Andersen et al. (1986, 1987) with one
developed by Angelo et al. (1984) for DCM after intravenous and
per os exposure (discussed more fully in the next section).
2.2.2. Model Used by Angelo et al.
The PBPK model used by Angelo et al. (1984) is structured
somewhat differently from the model used by Anderson et al.
(1986, 1987) in that it was formulated to describe the
disposition of DCM after an intravenous dose. The model also
takes into account metabolism to carbon dioxide and carbon
mdfto^ide by two pathways. However, this model considers most 6£
the body organs to have two subcompartments. One, termed the
vascular region, is blood flow limited, and the other, termed the
extravascular regi'on, is membrane diffusion limited. This model
is somewhat more complex than the model used by Andersen et al.
(1986, 1987), and more parameters are required as input. The
structure of the model was found to best describe the data
gathered after an intravenous exposure.
. 17
-------
In the model used by Angelo et al., the partition
coefficients are determined by a much different process than
those in the model used by Andersen et al. (1986, 1987), that is,
they are determined mathematically by optimizing from several
time points at two doses. This method is used for a variety of
reasons. The most important reason is that for each organ the
partition coefficient for the lipoid region would be very
difficult to determine experimentally. This method is similar to
that used by King et al. (1983) and it appears to be a reasonable
way to determine the partition coefficients. It would be
interesting to have had an experiment in which the whole organ
partition coefficients were determined from a steady-state
infusion technique. The organ partition coefficients could then
have been compared to the vascular region coefficients in the
model used by Angelo et al. as well as those derived from the
equilibration technique in the model used by Andersen et al.
(1986, 1987).
Pulmonary clearance rates, metabolism constants, and
permeability-area products are determined from in vivo data.
Pulmonary clearance of DCM i« calculated by dividing the total
amount excreted from the lung following a 60-minute time period
after dosing by the integral of arterial blood concentration
(Angelo et al., 1984). The amount of DCM that is excreted is
determined experimentally. The metabolic clearance is determined
in a similar fashion using the total amount of each metabolite
produced (carbon dioxide and carbon monoxide) divided by the
18
-------
integral of the venous concentration of DCM in the venous blood,.
The apparent in vivo metabolism constants (KM and Vmax) are then
mathematically derived from these data. The permeability-area
products are determined from experimental data. The data needed
to determine the metabolic constants are the venous blood
concentration of DCM, the total organ concentration of DCM, and
the integral of the venous concentration of DCM. These are
considered initial estimates for all of the parameters. Then,
given the data from two doses, an optimization routine results in
a "final fit" for all.
Given intravenously, DCM appears to sequester in certain
tissues while simultaneously disappearing from the blood. In
subsequent work (Angalo ©t al., 1986a, b) this phenomenon was
investigated further. After reviewing these papers, it does
indeed appear that DCM remains in some of the tissues for quite
some time after its disappearance in the blood. Angelo et al.
(1986a, b) concluded that this, is due, for some reason, to the
vehicle of administration. They reported that, when dosing by
the oral route with an aqueous vehicle, the disposition profile
is not the same as when dosing by an intravenous route with a
polyethylene glycol vehicle. ,
A strong point of the model used by Angelo et al. is its
faVbrable comparison to the data. The model was particularly
successful in predicting the disposition of DCM, carbon dioxide,
and carbon monoxide for repeated oral dosing. It also predicted
some components well after an intravenous dose. The predicted
19
-------
values for DCM elimination from the lung were 10% to 15% higher
than the actual data. The prediction for carbon dioxide and
carbon monoxide elimination were excessive at short times but
closer at predicting the ultimate amounts of the formation of
these substances.
Predictions using this model indicate that DCM,
when given repeatedly for 14 days to Fisher 344 rats by daily
oral gavage in water at a dose level of 200 mg/kg, did not
sequester in the liver from day to day (Angelo and Pritchard,
1984). That is, by the time of the next daily gavage the
previous day's material had been eliminated from the liver. The
prediction, therefore, appeared to agree with the data. Further,
the model showed that a corn oil vehicle would greatly increase
th« tinm r«quir«d for «limination to occur in th« llv«r compared
to an aqueous vehicle. The model's prediction showed the corn
oil effect to be less pronounced for venous blood. Thus, blood
profiles are not necessarily accurate reflections of tissue
disposition profiles. This point needs to be considered when
comparing model output to data. Agreement of model output with
blood data is not always indicative of agreement with tissue
data.
Angelo et al. (1986a) found, upon comparing profiles after
intravenous dosing in water with oral dosing in water, that the
elimination phase of DCM in the blood is similar for both routes
of administration. However, the tissue concentrations were fbttftd
to significantly differ between^intravenous and oral routes.
20
-------
Intravenous administration resulted in elevated tissue levels
over time in the lung and kidney, but not in the carcass. Also
of note is that following an intravenous administration, blood
profiles do not resemble tissue profiles as they do following
oral administration. Thus, it could be misleading to infer
tissue disposition merely from blood profiles in the case of
intravenous administration. It is not clear at this time whether
or not this lack of correlation between blood and tissue occurs
after inhalation exposure. This again illustrates some problems
when measuring the accuracy of a model against data; that is, a
model predicting blood data may not be accurately predicting
tissue data.
The effect of route on metabolic profiles is also of obvious
consequence when comparing intravenous versus oral
administration. Because the liver has greater metabolic activity
than the lung, a greater fraction of DCM is metabolized upon
"first pass" after oral administration. Thus, less unchanged DCM
is available for pulmonary excretion when compared to intravenous
administration. In this latter case the lung is the "first pass
offgith," and thus a greater fraction of DCM can be eliminated
unchanged.
The model used by Angelo et al. (1984) was modified to
account for inhalation exposure (Angelo et al., 1986c), and the
authors report that the model described the distribution of DCM
in the blood of rats during a 6-hour inhalation exposure using
the data of McKenna et al, (1982). However, the results of that
21
-------
simulation are not provided (Angelo et al., I986c). Assuming the
simulations accurately describe the data, one must question
whether a model, such as the one used by Andersen et al. (1986,
1987) , which does not account for sequestering into lipid-rich
compartments of some organs, is accurate.
The model was also used to determine "equivalency" between
oral and inhalation doses. The correlation varies depending on
the target one is observing. The nature of the relationship
between routes depends on which tissue, end product, parent, or
metabolite is being examined.
The model was able to predict the amount of DCM retained
during steady-state inhalation conditions. The predictions were
that less than 15% of DCM would be retained in the body at
exposure concentrations between 50 and 1500 ppm. Whether or not
this indicates error in any previous extrapolation based on
applied dose (which assumes 100% absorption) would depend upon
whether or not the fraction absorbed is different in experimental
animals versus humans and whether or not the absorbed fraction is
different at the exposure concentrations of the bioassay from low
exposure concentrations. When properly applied, PBPK models are
able to account for such differences.
To summarize, three important points can be gained from
examining the data and model developed by Angelo et al. (1984;
1986a, b, c) and Angelo and Pritchard (1984). First, tissue
concentrations of DCM are not accurately reflected by blood
concentrations; thus, to determine how well a model predicts
22
-------
tissue disposition, tissue data are necessary. Second, it is
possible that the disposition of DCM after an intravenous dosing
regimen is most likely different from the disposition after an .
oral dosing. The implication that this difference may have an
effect on inhalation dosing is not clear. Third, during steady-
state inhalation exposure, much less than 100% absorption would
be predicted. PBPK models are able to account for this factor
and can account for changes in this factor with different
exposure concentrations.
2.2.3. Summary
The model being proposed by Andersen et al. (1986, 1987) and
Reitz et al. (1986) needs to be further tested against
concentrations of DCM in the tissues. More data regarding the
production of carbon monoxide and carbon dioxide are required to
further substantiate the model's ability to accurately predict
metabolism. Also, these additional data regarding tissue
disposition will elucidate whether or not the complexity of the
model used by Angelo et al. is necessary.
Further, according to discussions with the ICI-UK
scientists, it appears that some reevaluation of certain
parameters is necessary. Those investigators feel that the
Metabolism parameters are in need of readjustment in order for
the model to be a valid predictor of tissue concentrations of DCM
and its metabolites. The HRAC agrees with this opinion but also
feels that there is some possibility that the model may have to
be structurally altered to account for sequestering of the parent
23
-------
compound into lipid-containing compartments. This could be
significant when using the model for humans who may have repeated
exposure over a long period of their lifetimes.
In addition, several key questions remain, but ICI-UK
scientists are conducting studies that may greatly reduce the
uncertainties associated with these. First, there remains some
uncertainty as to whether or not the GST pathway is the sole
path to carcinogenicity. As previously discussed, the evidence
to support this assumption is quite strong, but one cannot be
100% certain. Without such assurance some level of uncertainty
will always be associated with risk assessments for DCM. When
some of the uncertainties with other questions are reduced, it
may be possible to estimate the level of certainty with this
first assumption which is truly related to the exact mechanism of
action. The remaining questions address the activity of the GST
pathway. It is obvious at this time that there are species
differences in the activity of metabolism by this path. Due to
the lack of sufficient sensitivity of the recently completed
analyses by ICI-UK, it is not clear whether or not the pathway is
totally nonexistent (regarding DCM in rats, hamsters, and
humans). If it is proven to be nonexistent and one assumes that
this path is the major or only route to carcinogenicity, then the
risk estimate to humans could be substantially lowered. It is
hoped that ongoing studies at ICI-UK will add a great deal of
knowledge in this area. Presently ICI-UK scientists are plattflin^f
and/or conducting studies using 36C1-DCM to determine, in vitro,
24
-------
the lavela of GST activity towards DCM in liv«r, and possibly
lung, of various species, including humans. Properly performed,
these analyses could answer the question as to whether or not
this pathway is active in species, such as the hamster, which
have not shown the same tumor response.
Questions regarding the salivary and mammary gland adenomas
and possible human pancreas tumors will remain, however, unless
the same or comparable pertinent metabolic studies are performed
on those tissues as well. The model used by Andersen et al.
(1986, 1987) does not, at present, account for "toxic" metabolite
production in any tissues except the lung and liver. If other
tissues are deemed important in the scheme of potential
carcinogenicity to humans, then appropriate parameters would have
to be determined and the model modified to describe and predict
metabolite disposition in those tissues as well»
The epidemiologic evidence that pancreatic tumors develop
after exposure €o DCM is equivocal at present. Yet, some studies
(Black and Howerton, 1984; Mukhtar et al., 1981) suggest
significant GST activity in animal and human pancreatic tissue.
Also, at this tim® the significance of the rat salivary
gland tumors is not considered remarkable, and reports have been
published (Russo et al., 1982) which indicate that benign
adenomas of the mammary gland are histogenically different from
malignant adenocarcinomas and are not believed to have a high
malignant potential. A note of caution should be raised here as
well. Tumors that are shown to exhibit altered genomes or other
••••••'' ' ' 25 ' . •'"'•' '
-------
chromosomal aberrations within their cells should be considered
to be more significant than those that do not exhibit such
changes. These changes, regardless of the tumor type, should be
considered relevant when assessing the cancer risk for humans.
That is, a substance that may cause a mutagenic change in one
mammalian cell should be considered potentially capable of
causing such a change in another mammalian cell, including human
cells. There are numerous measures to determine whether or not
DCM or an active metabolite is mutagenic in specific tissues.
Studies to determine the rate of unscheduled DNA synthesis and
the amount of covalent binding to DNA have not been performed in
mammary cells; thus, it would be impossible to state with
absolute certainty that DCM or its metabolites do not have
mutagenic potential. However, the mammary tumors in the NTP
bioassays were fibroadenomas, and these do not show
characteristics indicative of malignant potential. Thus, the
significance of these tumors, although uncertain, is probably not
great. In fact, Ackerman and Rosai (1974) stated "... that the
malignant transformation of a fibroadenoma is exceptional and for
practical purposes can be disregarded in the management of this
lesion. However we have seen a few cases in which part of the
epithelial component of a fibroadenoma had the microscopic
appearance of carcinoma. The prognosis of tumors limited to the
fibroadenomas was excellent."
Additional studies being planned by ICI-UK scientists using
the stable isotope effect will provide information necessary to
26
-------
answer the question regarding the source of carbon dioxide formed
after exposure to DCM. CEFIC (1986e, f) and Reitz et al. (1986)
assumed that the GST pathway in mice operates only at
concentrations that are high enough to saturate the P-450
pathway, and CEFIC (1986e) further assumed that the GST pathway
has very little activity in humans. The question then arises as
to where the carbon dioxide observed in mice at low doses and
rats and humans arise from? CEFIC (1986f) and Reitz et al.
(1986) assumed that some carbon dioxide results from metabolism
from the P-450 system; thus, the carbon dioxide observed is not
•inconsistent with the lack of GST activity in rats and humans and
lack of GST activity in mice at low doses. Previously it had
been assumed that carbon dioxide could only arise from the GST
pathway. If correct, this assumption would imply that some GST
activity must be occurring in mice at low doses as well as in
rats and humans. Determining which of these assumptions is
correct could obviously have significant implications on the risk
estimate. The planned studies using deuterium replacement for
hydrogen, and thus taking advantage of the stronger bond energy
that results from substitution, will help to discern the source
of the carbon dioxide. Carbon dioxide resulting from P-450
metabolism would be subject to the isotope effect, and such a
pathway would exhibit reduced carbon dioxide generation rates.
On the other hand, carbon dioxide arising from the GST pathway,
due to no carbon deuterium bonds being broken, is not affected in
this case, and its production should not be altered by the
27
-------
isotope effect. Thus, in in vivo studies with deuterated DCM,
reduction of carbon dioxide production would be indicative of the
amount of carbon dioxide produced by the P-450 pathway. If no
reduction of carbon dioxide production is observed, then it can
be concluded that the GST pathway is the sole source of carbon
dioxide. This would imply that the GST pathway is active at low
exposure concentrations in mice and rats. If this is the case,
then the structure of the model, its parameters, and its
underlying hypothesis must be reevaluated.
To summarize, it is hoped that these studies will serve to
remove some doubt as to whether or not the existing evidence can
be interpreted to mean that the predominant pathway at low doses
is the MFO pathway and at high doses is the GST pathway.
In conclusion, the model used by Andersen et al., (1986,
1987), despite some questions regarding structures and parameter,
appears to be a good approach for high- to low-dose extrapolation
within a species. The questions and possible errors would not be
judged to have a great impact (Chapter 7). However, these errors
would have a much greater impact if the model was used for
species-to-species extrapolation.
28
-------
3. METABOLISM OF DICHLOROMETHANE
In a 1985 analysis of the relevant literature, EPA staff
scientists concluded that the results of both in vitro and in
vivo studies indicated that DCM is metabolized via two pathways
i
(U.S. EPA, 1985a, b). One pathway yields carbon monoxide (CO) as
an end product, and the other pathway yields carbon dioxide (C02)
as an end product. Each pathway involves formation of a
metabolically active intermediate that is theoretically capable
of irreversibly binding to cellular macr©molecules (Ahmed et al.,
1980). In vivo data suggest that when rats or mice are exposed
to high concentrations of DCM (50 mg/kg or 500 ppm or more), they
exhale more CO2 than CO (Yesair et al., 1977; McKenna et al.,
1982) . At exposure to low concentrations of DCM (1 to 10 mg/kg
or 50 ppm) both pathways are utilized about equally (Yesair et
al., 1977; McKenna et al., 1982). These observations are
consistent with the data showing that the oxidative pathway
(yielding CO) is saturated at relatively low exposures (< 500
ppm) while the pathway yielding CO2 appears to be first order
even at exposures of 1500 ppm.
Recent reports (Gargas et al., 1986; CEFIC, I986e, f; Reitz
et al., 1986) have suggested that: (1) at low doses DCM is
almost exclusively metabolized by the oxidative pathway; (2) the
oxidative pathway is capable of yielding significant amounts of
CO2 in addition to CO; (3) there is a significant difference in
the enzymatic activity of the two pathways from species to
29 ,
-------
species; and (4) there is a significant difference in toxicity
between the reactive intermediates generated by the two pathways.
The purpose of this chapter is to review the data used to support
these conclusions and to alter the EPA HAD for DCM if needed.
3.1. IN VIVO METABOLISM
The conclusions that DCM is almost exclusively metabolized
by the oxidative pathway at low exposure levels and that the
oxidativ© pathway is capable of yielding significant amounts of
CO2 are based on a series of in vivo experiments in rats and mice
(Gargas et al., 1986; Reitz et al., 1986).
Gargas et al. (1986) studied the in vivo metabolism of a
series of dihalomethanes in rats. These investigators monitored
blood bromide during inhalation of dibromomethane (DBM) and
bromochloromethane, and blood levels of carboxyhemoglobin (HbCO)
during inhalation of DBM, bromochloromethane, and DCM. Gargas et
al. (1986) compared control values to those obtained from animals
pretreated with pyrazole (inhibits microsomal P-450 oxidation) or
2,3-epoxypropanol (2,3-EP) (which depletes glutathione) in order
to assess the relative contribution of each pathway to the total
metabolism of dihalomethanes. Animals pretreated with pyrazole
had a very significant decrease in blood HbCO during exposure to
DBM, DCM, and bromochloromethane (Figures 1, 2, and 3). Animals
pretreated with 2,3-EP had a significant increase in blood HbCO
during exposure to DCM and bromochloromethane but not DBM.
Pyrazole pretreatment only slightly reduced the blood bromide
30
-------
NAIVE
AND
2,3-€P
PYRAZOLE
2400
Figure 1. The end of exposure HbCO levels in naive, 2,3-EP, and
p'yr'a'lgQle pretreated rats following 4-hour exposures to DBM.
Three to six animals were used for each exposure and the smooth
Curves were predicted by the computer model using the kinetic
constants for DBM listed in Table 2. The stoichiometric yield of
CO from the oxidative pathway was 100% for both naive and 2,3-EP
pretreated rats.
SOURCE: Gargas et al., 1986.
31
-------
CH2CI2
15.0
PYRAZOLE
0.00
240O
Figure 2. The end of exposure HbCO levels in naive, 2,3-EP, and
pyrazole pretreated animals following 4-hour exposures to DCM.
Three to six animals were used for each exposure and the smooth
curves were predicted by the computer model using the kinetic
constants for DCM listed in Table 2. The stoichiometric yield of
CO frbm" the oxidative pathway was 100% for 2,3-EP pretreated rats
and was lowered to 70% for naive rats.
SOtfRCEi Gargas et al., 1986.
32
-------
CH2BrC!
PYRAZOLE
0.0©
0
240O
Figure 3. The end of exposure HbCO levels in naive, 2,3-EP, and
pyrazdlfe pretreated animals following 4-hour exposures to
broinbchloromethane. Three to nine animals were used for each
exposure and the smooth curves wore predicted by the computer
model using the kinetic constants for bromochloromethane. The
stoichiometric yield of CO from the oxidative pathway was 100%
for 2,3-EP pretreated rats and was lowered to 70% for naive rats,
SOURCE: Gargas et al., 1986.
33
-------
concentration in rats during exposure to DBM .or concentration in
rats during exposure to DBM or bromochloromethane (Figures 4 and
5) compared to the reduction in .blood HbCO. Animals pretreated
with 2,3-EP had only a minimal decrease in blood bromide compared
to controls.
Gargas et al. (1986) concluded from their study that the
effect of pyrazole pretreatment, which nearly abolished CO
production, provides support for the involvement of cytochrome
P-450 in the oxidation of DCM to CO. They also suggested that
the metabolic intermediate, formyl chloride, produced during the
oxidative metabolism of DCM, has a longer half-life in vivo than
the formyl bromide intermediate produced during the oxidative
metabolism of DBM. Therefore, "this would probably allow attack
by a cellular nucleophile such as GSH [GTS], on the formyl
chloride and would result in a reduced stoichiometric yield of CO
in rats with normal GSH concentrations." Based on this line of
reasoning, Gargas et al. (1986) suggested that a significant
portion of the formyl chloride (20% to 30%) may react with other
nucleophiles probably yielding CO2.
Gargas et al. (1986) also reported that the experimental
data did not show a stoichiometric relationship between the yield
of CO and blood bromide in animals pretreated with 2,3-EP. They
concluded from this observation that any alteration in the
metabolic pathways (by inhibitors) must not compromise total
metabolism (i.e., halide release). Thus, the kinetic constants
calculated by the authors to assess in vivo metabolism were
34
-------
^ s.oo
4.00
3.00
2.0O
1.0O
0.00
Q
I
O
NAIVE
2,3-EP
PYRAZOLE
12OO
PPM
24OO
4. The dependence of plasma inorganic bromide levels on
the ambient concentrations of bromochloromethane following
4-hour exposures using naive, 2,3-EP (O), and pyrazole (A)
pretreated rats. Three to nine animals were used for each
exposure and data are X + SE (error indicated by the vertical
lines). Individual data points with no apparent error bars are
those points where the SE is less than the size of the symbol.
The smooth curves were generated by using the kinetic constants
for bromochloromethane.
SOURCE: Gargas et al., 1986.
35
-------
CH2Br2
5.00
NAIVE
2,3-EP
PYRA2OLE
0.00
120O
PPM
24OO
Figure 5. The dependence of plasma inorganic bromide levels on
the ambient concentrations of dibromomethane following 4-hour
exposures using naive, 2,3-EP, and pyrazole pretreated rats.
Three to six animals were used for each exposure and the smooth
curves were generated by the computer model usin§ the kinetic
constants for dibromomethane.
SOURCE: Gargas et al., 1986.
36
-------
based on gas uptake and plasma bromide concentration. The
calculated kinetic constants derived from the data obtained in
these studies have led these investigators to label the oxidative
pathway as "high affinity, low capacity" and the GST pathway as
"low affinity, high capacity."
However, the lack of stoichiometry suggests an alternative
explanation of the data obtained by Gargas et al. (1986). In
animals pretreated with 2,3-EP, there is a significant increase
in blood HbCO which could suggest an underutilisation of the
oxidativa pathway. Thus, th® l&ok of stoiahiowuatry. between CO
production and blood bromide could be explained by a significant
underestimation of the metabolism of DBM via the GST pathway. If
this alternative explanation is correct, in control animals the
CO pathway would appear to be saturated at very low concentra-
tions and is consistent with the observed increase in blood HbCO
in animals pretreated with 2,3-EP. Since Gargas et al. (1986)
did not measure the metabolism of dihalomethanes to C02, it is
not possible to directly estimate the metabolic activity
attributed to the GST pathway nor the effect of 2,3-EP on
metabolic activity attributed to that pathway.
Reitz et al. (1986) exposed groups of male B6C3F1 mice to
either 50 or 1500 ppm 14C-DCM in order to study the in vivo
conversion of DCM to 14CO and 14CO2. Groups of four mice were
exposed to 14C-DCM for 6 hours and 14CO, 14CO2/ and urine wer^s
collected during the exposure period and for 24 hours post-
exposure. Reitz et al. (1986) compared control values to those
37
-------
obtained from animals pretreated with pyrazole or pretreated
with diethylmaleate and buthionine sulfoxime (BSO/DEM) (which
depletes glutathione) in order to assess the relative
contribution each pathway makes to the total metabolism of DCM.
In animals pretreated with pyrazole, there was an apparent
decreased amount of DCM metabolized to CO and C02 (Table 1). In
animals pretreated with BSO/DEM, there was also an apparent
decrease in the amount of DCM metabolism to CO and C02 (Table 1).
However, BSO/DEM pretreatment resulted in a more significant
decrease in DCM metabolism to CO2 than CO. The authors
interpreted the effect of pyrazole as showing that the microsomal
oxidation of DCM yields both CO and C02. The authors did not
offer an explanation for the decreased conversion of DCM to CO in
animals pretreated with BSO/DEM, but did intrepret the reduced
conversion to CO2 as supporting previous observations that DCM is
metabolized by a system requiring glutathione.
TABLE 1. METABOLISM OF DCM TO CO AND CO2 IN MICE
C0a
Exposure 50 ppm
control
BSO/DEM
Pyrazole
Exposure 1500 pom
ControL
BSO/DEM
Pyrazole
0.118
0.070
0.040
0.908
0.685
0.125
0.130
0.102
0.062
1.88
0.905
0.565
aValues are mM/kg.
SOURCE: Reitz et al., 1986.
38
-------
A review of available data on the use of pyrazole as an in
j
vivo inhibitor indicates that this compound affects a variety of
metabolic systems including thyroid function and alcohol
dehydrogenase activity (Szabo et al., 1978; Cornell et al.,
1983)» The observation that pyrazole inhibits alcohol
dehydrogenase suggests that it might also inhibit the activity of
formaldehyde dehydrogenase. At present, the available data do
not exclude the possibility that pyrazole also inhibits the
cytosolic metabolism of DCM to C02« Thus, it would seem
premature, at this time, to attribute the observations of Reitz
et al. (1986) to pyrazole inhibition of microsomal enzyme
activity. The caution is especially applicable to the use of
pyrazole as an inhibitor in the elucidation of in vivo DCM
metabolism, since the available data suggest that a variety of
physiologic functions are affacted by pyrazol®.
Reitz et al. (1986) also observed that animals pretreated
with DEM (see Table 1), a known depletor of cellular glutathione,
reduced the metabolism of DCM to both CO and CO2° Reitz et al.
(1986) did not offer an explanation for the reduced exhalation of
CO in animals pretreated with DEMo
Stevens et al. (1980) demonstrated that rats pretreated with
DEM show a significant decrease in the conversion'of DBM to blood
CO compared to controls. They also showed that the conversion of
DBM to CO was significantly reduced in isolated hepatocytes and
by isolated microsomes. These observations are consistent with
the suggestion of Anders et al. (1978) that DEM may diredtly
39
-------
inhibit certain cytochrome P-450 catalyzed oxidations. These
data suggest that the attempt by Reitz et al. (1986) to
characterize the in vivo metabolism of DCM by pretreating animals
with various inhibitors will require revision.
Lastly, a search of the literature did yield data that would
support the presence of significant amounts of GST activity
associated with the microsomal fraction of the cell. Morgenstern
et al. (1979) and Boyer et al. (1982) demonstrated that GST
activities are associated with rat liver microsomes. Using
2,4-dinitrochlorobenzene as a substrate, the activity of the
microsomal GST was 81 nmol/min/mg protein (Boyer et al., 1982).
Boyer et al. (1982) also showed that the GST associated with the
microsomal fraction of the liver could be activated increasing
the specific activity to 1220 nmol/min/mg protein. This latter
observation has suggested to Boyer et al. (1982) that the GST
associated with the microsomal fraction of the liver could be an
important detoxification mechanism. A search of the literature,
however, did not yield information that demonstrated the use of
this proposed detoxification mechanism. Thus, the available data
supporting the hypothesis that microsomal oxidation of DCM yields
significant amounts of CO2 should be considered very limited vat
this time. Research will be required to reduce the uncertainties
concerning the significances of the branch in the microsomal
pathway as well as the characterization of the metabolic activity
attributed to the proposed branch before it can be quantitatively
factored into an analysis of DCM metabolism.
40
-------
It is not clear from the hypothesis proposed by Gargas et
al. (1986) why the glutathione intermediate produced by the
microsomal pathway is significantly less reactive, and therefore
less genotoxic,.than the proposed glutathione intermediate in the
cytosolic pathway. The intermediate proposed by Gargas et al.
(1986) is similar in structure to the one proposed by Ahmed and
Anders (1976) for the GST pathway and, therefore, could be
expected to be equally toxic. Lastly, it should be noted that
there are some data to indicate that the metabolic
intermediate(s) produced by the microsomal pathway is mutagenic
(Jongen et al., 1982) and, therefore, probably genotoxic.
CEFIC (1986f) studied the in vivo metabolism of 14C-DCM in
rats and mice. Groups of three animals were exposed to 500,
1000, 2000, or 4000 ppm DCM for 6 hours. These investigators
monitored blood levels of the parent compound and HbCO during and
after the exposure period. In addition, they monitored
exhalation of 14CO and •L4CO2 postexposure. In rats, the blood
level of DCM is proportional to the dose at exposures above 500
ppm. In mice, the blood level of the parent compound is not
proportional to dose until exposures reached 2000 ppm. The
authors interpreted this observation as a difference in the
metabolic capacity of the species, i.e., mice can metabolize
significantly more DCM via the GST pathway than rats. The data
on blood levels of DCM in rats supports the conclusion that there
is a dose-response relationship between blood levels and exposure
above 500 ppm. However, the data from the mouse studies are
41
-------
highly variable which suggests that conclusions based on average
values used by CEFIC (1986f) must be viewed with caution. For
example, the blood level of DCM in mice exposed to 2000 ppm
varied from 40 ug/mL at 1.5 hours to 20 ug/mL at 6 hours.
Following exposure to 4000 ppm DCM, mice exhale more C02
than rats. Based on this observation, CEFIC (1986f) concluded
that mice metabolize more DCM via the GST pathway than rats. The
authors used differences in blood levels of the parent compound
as a surrogate to suggest that there is a species difference in
DCM metabolized to CO2 during exposure. The data obtained by
CEFIC (1986f) are significantly different from those found in the
published literature. McKenna et al. (1982) exposed groups of
rats to 50, 500, or 1500 ppm and found little differences in the
percent conversion of DCM to CO and C02 at each dose over the
exposure range tested. Angelo et al. (1984) exposed mice or rats
to either 10 or 50 mg/kg DCM and clearly demonstrated that the
percent conversion to CO and C02 were similar at both dose
levels and in both species. The McKenna et al. (1982) and Angelo
et al. (1984) studies question correctness of the assumption made
by CEFIC (1986f) that the blood level of the parent compound is
an appropriate surrogate for metabolism of DCM to C02.
3.2. REACTIVE INTERMEDIATES
In the elucidation of DCM metabolism, Ahmed and Anders
(1976) and Kubic and Anders (1978) predicted formation of a
reactive intermediate by both the microsomal and cytosolic
pathways. Andersen et al. (1986, 1987), however, suggested that
42
-------
the metabolism of DCM via the microsomal pathway does not result
in the formation of a reactive intermediate. These investigators
cite two observations to support this hypothesis. First, they
suggest that the lack of tumor development in the drinking water
study (National Coffee Association, 1982a, b; 1983) can be
explained by assuming that at the low doses all DCM is
metabolized via the microsomal pathway. It should be noted,
however, that the Carcinogen Assessment Group (U.S. EPA, 1985a,
b) predicted, based on the tumor response in the National
Toxicology Program (NTP) study, that there would not be a
significant tumor response in the drinking water study without
invoking differences in the utilization of metabolic pathways.
The second observation cited by Andersen et al. (1986,
1987) , that the metabolism of DCM via the microsomal pathway does
not result in the formation of a reactive intermediate, comes
from the work of Green (1983). Green (1983) investigated the
mutagenic potential of DCM and chlorofluoromethane (CFM) using
Salmonella typhimurium strain TA100 in an Amas type assay. s_.
tvphimurium exposed to either DCM or CFM gave a mutagenic
response without the addition of rat liver fractions. The
addition of rat liver post-mitochondrial supernatant (S9) did H6%
significantly increase the mutagenic response of TA100 exposed to
DCM, but did increase the mutagenic response of TA100 exposed to
CFM. Green (1983) also observed that the addition of microsomes
did not increase the mutagenic response of TA100 exposed to DCM,
but the addition of cytosol did increase the mutagenic response
43
-------
(Table 2). He concluded from these observations that the lack of
mutagenic response by the intermediate generated by microsomal
metabolism could be explained by the fact that the reactive
intermediate was highly unstable (exists at -60°C in inert
solvents).
TABLE 2. THE MUTAGENIC EFFECT OF DIHALOMETHANES
Addition
DCMa
CFMa
None
89
Cytosolb
Microsomesk
Boiled S9
Air control
386 + 24
458 + 24
490 ± 19(698)c
375 ± 31
376 + 28
80 + 1
521 ± 27
793 ± 17
618 + 14(812)c
623 + 27
^Values are revertants/plate ± SD after 3-day exposure.
"Cytosol and microsomal concentrations (mg protein) are
equivalent to those in the S9/cofactor mixture.
°Cytosol concentration increased threefold.
SOURCE: Adapted from Green, 1983,
However, the inability of Green (1983) to demonstrate a
mutagenic response by the reactive intermediate generated during
microsomal oxidation of DCM may be the result of experimental
design, specifically the long incubation period. Green (1983)
carried out incubations for 3 days while other investigators,
performing similar types of experiments, have used much shorter
incubation periods. Van Bladeren et al. (1980) studied the
mutagenic response of £3. typhimurium TA100 exposed to dibromo- or
diiodomethane. These investigators, using a 15-minute incubation
44
-------
period, clearly showed that the addition of microsomes or cytosol
markedly increased the number of revertants/plate using either
dibromo- or diiodomethane as a substrate. Similarly, Jongen et
al. (1982), using a 6-hour incubation period, also demonstrated
that the addition of microsomes or cytosol increased the number
of revertants/plate of S. typhimurium exposed to DCM (Table 3).
TABLE 3. THE MUTAGENIC EFFECT OF DCM ON S. TYPHIMURIUM TA100
IN THE PRESENCE OF VARIOUS RAT LIVER FRACTIONS
Addition
No. of Revertantsa
Aroclor
Controlb
Microsomes
Cytosol
Cytosol+DEM
Phenobarbital
Control
Microsomes
Cytosol
Cytosol+DEM
939
1201
1407
889
540
624
806
567
number of revertants of three plates.
"Control equals microsomes minus cofactors.
SOURCE? Adapted from Jongen et al., 1982.
Using an Ames bioassay, the available data on the formation
of reactive intermediates are limited to a few studies. All
studies reported to date have shown that the metabolism of DCM
using a liver cytosolic fraction leads to formation of a reactive
intermediate. The various studies, taken together, suggest, but
45
-------
do not prove, that the metabolism of dihalomethanes via
microsomal oxidation lead to formation of reactive intermediates.
The data also suggest that the reactivity of the intermediates
formed using different dihalomethanes as substrates may not be
equal. Regarding the metabolism of DCM, the data from some
studies suggest that the metabolism of DCM by the microsomal
cytochrome P-450 pathway leads to formation of a reactive
intermediate. Further research is required to resolve the lack
of agreement among investigators on formation of a reactive
intermediate during microsomal metabolism of DCM.
3.3. USE OF IN VITRO DATA TO PREDICT IN VIVO METABOLISM
Gargas et al. (1986) characterized the microsomal metabolism
of DCM as "high affinity - low capacity" and the cytosolic
metabolism as "low affinity - high capacity." The data used to
support this characterization comes from the observation that the
microsomal enzyme has a much lower %(Table 4) than the cytosolic
enzyme (Table 5).
TABLE 4. KINETIC CONSTANTS FOR THE METABOLISM OF DCM
TO CARBON MONOXIDE BY LIVER MICROSOMES
Species
KM
(mM)
'max
(nmol/mzn/mg
protein)
Mouse
Rat
Hamster
0.79
0.86
2.83
1.94
0.58
1.85
SOURCE: Adapted from CEFIC, 1986e.
46
-------
TABLE 5. KINETIC CONSTANTS FOR THE METABOLISM OF DCM
TO FORMALDEHYDE BY LIVER CYTOSOL
Species
(HIM)
vmax
(nmoles/min/mg
protein)
Mouse
Rat
Hamster
Man
86
21
No detectable rate
No detectable rate
36.4
2.9
SOURCE: Adapted from CEFIC, 1986e.
However, the KM and Vmax values reported by CEFIC (1986e)
(1986a) are significantly different from those published by Ahmed
and Anders (1976) and Anders et al. (1978). These investigators
obtained KM and Vmax values of 50.1 mM and 5.4 nmol CO/min/mg
protein for the metabolism of 14C-DCM to CO using a rat liver
microsomal preparation (Table 6). They also reported % and Vmax
values of 48 mM and 16 nmol formaldehyde/min/mg protein for the
metabolism of DBM by a rat liver cytosolic preparation.
Thus, the KM and VTOax values obtained by Ahmed and Anders
(1376) and Anders et al. (1978) do not support the conclusion by
Gargas et al. (1986) that the microsomal pathway should by
considered "high affinity - low capacity" and the cytosolic
pathway should by considered "low affinity - high capacity." The
difference in the KM value of the microsomal/cytosolic pathway is
1/23 according to the data reported by CEFIC (1986e) but only
1/2.4 based on the data from Ahmed and Anders (1976) and Anders
et al. (1978). Furthermore, the KM and Vmax values obtained by
47
-------
TABLE 6. KINETIC CONSTANTS FOR THE METABOLISM OF
DIHALOMETHANES TO CARBON MONOXIDE BY RAT LIVER MICROSOMES
Author
(mM)
vmax.
( nmol es/min/mg )
protein)
Substrate
CEFIC (1986e)
Anders et al. (1978)
Ahmed and Anders (1976)
0.86
50.1 + 5.2
48.1 ± 6.2
0.58
5.4 + 1.7
15.5 + 2.5
DCM
DCM
DBM
Ahmed and Anders (1976) are consistent with the in vivo
observations made by McKenna et al. (1982) and Angelo et al.
(1984) that over a wide range of doses there is little difference
t
in the percentage of DCM metabolized to CO and CO2.
In addition, neither Gargas et al. (1986) nor CEFIC (1986e)
included in their in vitro to in vivo extrapolation a
consideration of differences in total enzyme in the tissue nor
did they account for the effect of cellular architecture on the
distribution of DCM within the cell. The available data suggest
that DCM is taken-up by cells via passive diffusion. Thus, given
our current understanding of cellular architecture, it can be
assumed that both the endoplasmic reticulum (ER) (microsomes) ttild
the cytosolic fraction of the cell will be exposed equally to
DCM. Since the cytosolic fraction of the cell occupies
significantly more volume than the ER, one could reasonably
predict that a significant amount of DCM would be available for
metabolism by the cytosolic enzyme. In addition, on a mass
basis, the microsomal fraction of the cell is about 2% to 5%
48
-------
while the cytosolic fraction of the cell is about 10%.
Consistent with this analysis is the observation that animals
exposed to low concentrations of DCM exhale significant amounts
of CO2 (McKenna et al., 1982; Angelo et al., 1984; Reitz et al.,
1986). Furthermore, this analysis is also consistent with the
observation that animals exposed to high concentrations of DCM
exhale significantly more C02 than CO.
3.4. IN VITRO METABOLISM
Gargas et al. (1986) predicted that microsomal metabolism of
DCM might yield both CO and C02. The Gargas et al. (1986)
hypothesis suggests that microsomal metabolism of DCM proceeds
via the mechanism outlined in Figure 6. The mechanism predicts
that following the initial oxidation of DCM to formyl chloride,
this intermediate combines with glutathione and is then further
metabolized by a GST to CO2. Gargas et al. (1986) suggested that
the formation of the formyl glutathione intermediate is possible
assuming that formyl chloride is a relatively stable
intermediate. The assumption that the formyl chloride is
relatively stable, thus allowing for formation of the formyl
glutathione, is in marked contrast to the observation made by
Green (1983) that formyl chloride was nonmutagenic because it was
such an unstable intermediate. The only data supporting this
prediction is the observation by Reitz et al. (1986) that mice
pretreated with pyrazole exhale less CO and C02.
CEFIC (1986e) recently submitted to the HRAC information on
the in vitro metabolism of DCM in tissues from the rat, mouse,
49
-------
W»l
X~GS
CYTOSOL f
i
GS-CH2-X
A
_»_i i
x***^ 2
^""•^M^"
i
®/"»e— r»u ..fiu
(ao wrl2 ""Win
[NAD*
T 0
^»
GS-CN
r i . r-«»ow pi v^y
Os OL ' Ol
NADPH ' ° ^ X
X
H
¥ ^ V™
•\
*^ t
^X~
1
CO2
o
\ H/J
HV .
,
i
x-NUCLEOPHILE
C ( i.e. GSH)
Srx-
o
+ II
G-S-CV
H ^ @
i
V^H*
i
\?
GS-C
C«f%
•0
?
t
H COOH * GSH
CO2
GSH
CO.
Figure 6. The proposed pathways for dihalomethane metabolism.
Pathways 1 and 2 were taken from Anders et al. (1977) and pathway
3 is proposed by Gargas et al. (1986).
SOURCE: Gargas et al., 1986.
50
-------
hamster, and human. These investigators measured the conversion
of DCM to CO using microsomal preparations from lung and liver
tissue and the conversion of DCM to formaldehyde by the cytosolic
fraction from lung and liver tissue. In addition, they measured
the cytochrome P-450 content of lung and liver microsomes and the
GST activity in the cytoisolic fraction of lung and liver tissue
using 2,4-dinitrochlorobenzene as a substrate.
CEFIC (1986e) reported that the highest microsomal-specific
activities (DCM converted to nmol CO/mg min/protein) were
measured in hamster liver followed by mouse lung and liver. The
specific activity of rat liver microsomes was about one-third
that of mouse liver. The single measurement for human liver
microsomal metabolism of DCM to CO was 150% of the value obtained
for rat liver. The authors state, based on this single
measurement, that the value for human liver microsomal metabolism
of DCM to CO was similar to that for rat liver. Mouse, rat, and
\
hamster lung microsomal preparations were also assayed for DCM
metabolism to CO. Mouse lung microsomes had the highest
IBIiiv'ity, hamster next, and rat,the lowest, having about
on^-tenth the activity of mouse lung microsomes. Human lung
microsomes were not assayed for DCM metabolic activity.
The in vitro data presented on DCM metabolism to CO suggests
that there are significant metabolic differences between species
(CEFIC, 1986e). Unfortunately, these investigators did not carry
out experiments that allowed for determining the stoichiometric
relationship between the amount of DCM metabolized to the amount
51
-------
of CO produced. The lack of stoichiometric measurements does not
allow one to determine if DCM metabolism to other end products
such as CO2 could explain the apparent species differences.
Indeed, if the hypothesis put forth by Gargas et al. (1986) and
the data from Andersen et al. (1986, 1987) and Reitz et al.
(1986) are correct, then one could expect formation of
significant amounts of other metabolites during DCM metabolism.
Lastly, the cytochrome P-450 content of lung and liver microsomes
reported by CEFIC (l?86e) differ from those previously reported
in the literature (Souhaili-El Amri et al., 1986). The
differences are small, less than 20% for rat liver and 168% for
mouse liver, but large for human liver, 368%. The significance
of these differences and the correlation between cytochrome P-450
content and DCM metabolism will require additional research.
CEFIC (1986e) assayed mouse, rat, hamster, and human liver
cytosolic fractions for the metabolism of DCM to formaldehyde by
the GST pathway. Mouse and rat liver, but not hamster or human
liver, had detectable levels of DCM metabolized to formaldehyde.
In addition, mouse, but not rat or hamster, lung cytosolic
fractions had detectable levels of DCM metabolized to
formaldehyde. The investigators concluded from these observa-
tions that the presence of DCM metabolism to formaldehyde is
consistent with the tumorigenic- response in the mouse lung and
liver and the lack of tumorigenic response in the rat and
hamster. However, a review of the methodologic approach used to
assay for DCM metabolism in the cytosolic fraction raises a
52
-------
number of concerns. First, it is not clear why these
investigators decided to measure only formaldehyde. The approach
used presents a number of variables for which control data have
not been reported. These include: species differences in the
metabolism of formaldehyde, and stoichiometry of the reaction.
Indeed, the lack of detectable levels of formaldehyde might
readily be explained by the large excess of formaldehyde
dehydrogenase found in both lung and liver tissues (Uotila and
Kiovusalo, 1981). Also, it is not clear why these investigators
selected the Nash (1953) method for assaying for formaldehyde
formation. The method is not specific for formaldehyde and gives
a high background level because of its reactivity with a variety
of substrates.
Ahmed and Anders (1976) also used the Nash method to
demonstrate the metabolism of dihalomethane (DBM) to formaldehyde
by rat liver cytosol. These investigators, however, reported
values significantly greater than those reported by CEFIC
(1986e). Ahmed and Anders (1976), unlike CEFIC (1986e), dialyzed
the cytosolic preparation before conducting the assay and found
that the rate of formation of formaldehyde was three times
greater for dialyzed cytosol compared to undialyzed. These
investigators also could not demonstrate a stoichiometric
relationship between bromide release and formaldehyde formation
With the observed concentrations of free bromide being about
twice the predicted based on the observed concentration of
formaldehyde. The values reported by Ahmed and Anders (1976)
53
-------
suggest that formaldehyde formation may have been underestimated.
CEFIC (1986e) also assayed liver and lung cytosolic
fractions for GST activity using 2,4-dinitro-chlorobenzene as a
substrate. The values reported by CEFIC (I986e) for 2,4-
dinitrochlorobenzene metabolism in all tissued assayed were
significantly smaller than those reported by other investigators
(Lorenz et al., 1984). The difference for all tissues was about
40% of the value reported by Lorenz et al. (1984) but similar to
the value reported by Moron et al. (1979) for rat liver cytosol.
However, both Lorenz et al. (1984) and Moron et al. (1979)
reported similar and significantly larger values than CEFIC
(1986e) for rat lung GST using 2,4-dinitrochlorobenzene as a
substrate (Table 7). Since stability of enzymes can vary, it may
be necessary to determine if there are changes in the ability of
different tissues to metabolize DCM when stored for various
periods of time.
The methodology used by CEFIC (1986e) indicates that the
amount of protein used to assay for enzymatic activity varied
significantly from tissue to tissue and species to species. *he
difference in protein concentration may significantly affect the
amount of metabolic end product measured since the reactive
intermediates formed during the metabolism of DCM bind to protein
and lipid (Anders et al., 1977). In addition, because of the
differences in protein concentration, it is not clear that the
glutathione concentrations were optimal in all experiments.
Thus, in the absence of stoichiometric data and control
54
-------
TABLE 7. THE METABOLISM OF 2,4-DINITROCHLOROBENZENE BY RAT
LIVER AND LUNG CYTOSOL
Author
Livera
Lunga
CEFIC (19866)
Lorenz et al. (1984)
Moron et al. (1979)
504.9
1380 + 110
567 + 184
44.2
77.5
99.8 + 11.4
aValues are nmol/min/mg protein.
experiments, it is not possible to determine if the experiments
were carried out under optimal conditions nor if the values
reported represent total metabolism. Lastly, it should be noted
that the values reported by CEFIC (1986e) represent a single
experiment using pooled tissue samples. If one is to conclude
that there are significant species differences in the metabolism
of DCM, then it will be necessary to expand the data base to
demonstrate statistical significance.
The in vitro data reported by CEFIC (1986e) may allow for
alternative approaches to interspecies extrapolation in assessing
the effects of exposure'to DCM. However, before such
extrapolations are useful, the uncertainties raised about the
quality of the data reported need to be resolved. In addition,
it would be useful to expand the human tissue data base beyond a
single data point.
3*5. SUMMARY
The recent reports by Gargas et al. (1986), CEFIC (1986e,
f), and Reitz et al. (1986) have raised a number of potential
55
-------
issues regarding species differences in the metabolism of DCM
which have suggested to these investigators alternative
approaches for assessing human risk from exposure to DCM. The
hypothesis put forth by these investigators is supported by
very preliminary data. An analysis of the data suggests that
there is a need to conduct critical experiments to support the
hypothesis as well as to rule out alternative explanations. In
addition, there is a need to conduct experiments to help explain
why some of the observations made by these investigators are
significantly different from those found in the published
literature.
The hypothesis that microsomal oxidation of DCM leads to
formation of significant amounts of C02 is based almost entirely
on a series of assumptions and some indirect measurements.
Therefore, one must be very cautious in accepting the conclusion
that microsomal oxidation of DCM leads to CO2 formation given the
lack of data to support similar types of metabolism by
microsomes.
Gargas et al. (1986) and CEFIC (1986e), using in vitro
enzyme kinetic values, have made the assumption that at low doses
DCM is almost exclusively metabolized in vivo by the microsomal
pathway. The appropriateness of this extrapolation is based
almost exclusively on indirect measurements and a series of
unvalidated assumptions. At present, the available data do not
allow for distinguishing differences in the utilization of the
microsomal and cytosolic pathways at low versus high doses of DCM
56
-------
in in vivo metabolism. First, there is a significant body of
data in the published literature which strongly suggests that
both the microsomal and cytosolic studies by Gargas et al. (1986)
can be interpreted in a way which would strongly support the
assumption that at low doses significant metabolism of DGM via
the cytosolic pathway occurs in vivo. Thus, at present, data are
lacking to support the assumption that, at low doses, DCM is
almost exclusively metabolized via the microsomal pathway.
The data supporting possible differences in the
characterization of the reactive metabolic intermediates formed
by microsomal and cytosolic metabolism of DCM are limited. The
available data suggests that further characterization of the
reactive intermediates may be useful to better delineate species
differences in the metabolism of DCM.
The observation by CEPIC (1986e) that human tissue does not
metabolize DCM by the cytosolic pathway must be viewed with a
great deal of reservation. The observation is based on a single
sample of tissue, and control experiments indicate metabolic
activity that is significantly less than reported by others
(Lorenz et al., 1984). The CEFIC (1986e) study, however, does
suggest that development of a data base that includes metabolic
activity in human tissues could substantially reduce the
uncertainties in our present risk assessment.
57
-------
4. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODEL:
METABOLIC KINETIC CONSTANTS
Some of the critical determinations for the use of the
physiologically based pharmacokinetic (PBPK) model used by
Andersen et al. (1986, 1987) (Chapter 2) are the values of the
kinetic constants (kp, Vmax, KJJ) for the metabolism of DCM and
how these are related among the various species. The following
discussion makes the assumption that the model is valid (which is
not necessarily true, as discussed in Chapter 2). This
assumption includes an important provision relating to the
kinetic constants, that is, that kp is the constant relating to a
pathway that involves GST, and that this pathway is nonsaturable
(therefore needing to be. defined by only one constant). Vmax and
KM, on the other hand, are standard kinetic constants which apply
to a second, saturable mixed-function oxidase (MFO) pathway.
Classically, the GST pathway leads to carbon dioxide and the MFO
pathway leads to carbon monoxide as end products, using DCM as a
substrate.
Originally, these kinetic constants were based on the
specific activities of GST and monooxygenase measured in samples
of lung and liver tissue from four species by Lorenz et al.
(1984) using l-chloro-2,4-dinitrobenzene (for GST) and
7-ethoxycoumarin (for monooxygenase) as substrates. Since these
critical constants were not based on experimentation using DCM
itself as a substrate, the assumption that the Lorenz et al. data
58
-------
served as an appropriate substltuta was tenuous.
In a subsequent version of the model used by Andersen et al.
(1986, 1987) used a curve-fitting exercise to obtain the "whole
tissue" values of the three kinetic constants in animals. With
all other physiological and biological parameters kept constant,
the values of Kj^, Vmax, and kp were varied in an intricate
computer optimization procedure. The pharmacokinetic model was
used to arrive at predictions that best approximate the actual
experimental chamber data in which disappearance of DCM over time
was monitored. •
For humans, values of KJJ (0.58 mg/L) and Vmax (119 mg/hr)
for the oxidative pathway were estimated (Andersen et al., 1987)
from unpublished human experimental data in which DCM
concentrations in expired air were measured following exposure by
inhalation. For the critical GST pathway, however, no human data
were available upon which an estimate of kp can be based. The
authors noted that allometric scaling, based on the concept of
clearance, served to adequately relate the mouse, hamster, and
rat data, using a factor of body weight to the 0.7 power. Using
this scaling procedure, the authors obtained a human value of
0.53 hr""1 for kp. This value, by the authors' methodology, leads
to a reduction in risk of somewhat less than an order of
magnitude, as compared to risks calculated without incorporating
species-to-species and high- to low-dose extrapolative
pharmacokinetic information (all other risk assessment
assumptions being unchanged) . Also, the value results from
59
-------
assuming that (1) the GST pathway is the only toxic pathway, and
(2) carcinogen risk is directly related to concentration at a
"target" site.
If other combinations of KM, Vmax, and kF could be found
which, when put into the pharmacokinetic model, approximate the
experimental concentration versus time curves to an extent
similar to that found by the authors' optimization scheme, then
the scaling approach would be suspect. This is the case, in
fact; values of kF, which vary as much as fourfold from the
optimized values, were used in the model, leading to alternative
combinations that fit the experimental data nearly as well as the
optimized values.
It is therefore concluded that, although the optimization
procedure may lead to kinetic constants giving an optimal fit
(using the pharmacokinetic model) to the experimental chamber
data, there are alternative combinations of the three kinetic
constants that can virtually do the same thing. This may be due,
perhaps, to the fact that the "optimized" solution may have a
goodness-of-fit that only trivially exceeds those from other fits
having substantial differences in the underlying constants. In
other words, there are multiple adequate solutions to the problem
(redundancy). On this basis, the scaling approach for the
determination of a human kp value, a critical value with regard
to DCM carcinogenicity given that the model used by Andersen et
al. (1986, 1987) is valid, is subject to uncertainties which
render the approach suspect. Thus, the use of the scaling
60
-------
approach is still in the preliminary stages of development, and
more research and validation is required before it can be
accepted as a reasonable basis for pharmacokinetic modeling in
the case of DCM.
New data have become available subsequent to the proposal of
a pharmacokinetic model for DCM by Andersen et al. (1986, 1987).
These experiments have the potential to measure the human GST
pathway; thus, other approaches for obtaining human GST data,
such as the scaling approach described above, would not be
necessary. The data released by ICI-UK (CEFIC, 1986e, f) are
interpreted by those investigators as indicating that there is
little or no risk of cancer to humans as a result of exposure to
DCM. The cornerstone of this argument is the same as proposed by
Andersen et al. (1986, 1987)—that the toxic intermediate is
produced only by the GST pathway. The results of this study
(CEFIC, 1986e, f) are interpreted as showing that there is no-GST
activity in humans with regard to DCM, as opposed to mice, where
such activity is easily detected.
CEFIC's (1986e) critical experiment used human liver derived
from accident victims. Fractions of the liver, dosed with DCM,
were assayed for formation of formaldehyde. The amount of
formaldehyde produced was considered to be a measure of the
fraction's GST activity towards DCM. No excess (over background)
of formaldehyde was found at any dose of DCM. In livers derived
from mice, a dose-related formation of formaldehyde was observed.
If humans indeed have no GST pathway for DCM, kp (as
61
-------
described above) would be zero, and there would be no risk if
this were the only toxic pathway. There are, however, many
problems associated with the CEFIC (1986e) experimentation; some
of the more important ones are briefly discussed here. First,
CEFIC (1986e) did not present information regarding whether or
not the critical control experiment was done. That is, adding a
known amount of formaldehyde (for example, the amount^at the
limit of detection) to the reaction mixture and seeing if this
amount could be recovered by CEFIC's (1986e) assay procedure.
This information is needed because formaldehyde is very reactive;
it can bind to many subcellular fractions or enter other
metabolic pathways. Small amounts actually generated by human
liver could go undetected. [In April 1987, Harris of ICI-UK
informed the HRAC by telegram that this experiment had been
performed and that "less than 10% of the formaldehyde was lost
over the incubation period." Details of the procedure, including
the dose of formaldehyde, and results have not be provided.]
Second, it is difficult to calculate the maximum velocity at
which human liver could metabolize DCM by the GST pathway. The
calculation by CEFIC (1986e) suggests that the metabolic
capability of the human liver GST could be anywhere from zero to
one-sixtieth the mouse value (on a nmol/min/mg protein basis)
based on the limit of detection. [Other preliminary calculations
suggest that the mouse value may be even closer to the human
value based on the data reported by CEFIC (1986e).] One-eighth
of the mouse value is the relationship suggested by Andersen et
62
-------
al. (1986, 1987) based on the scaling approach discussed above.
Obviously, the uncertainty encompasses orders of magnitude.
Third, there is some question regarding the nature of the
human liver samples used. That is, whether the events that
occurred between the time that the liver was operating normally
in the human and the time of the GST assay somehow affect whether
or not GST activity would be detected (for example, what happens
to the liver between the time of an accident and the time it is
removed from the body). Such factors include introduction of
drugs, and how the liver was maintained.
Fourth, there is difficulty in using the CEFIC (1986e) data
in the model used by Andersen et al. (1986, 1987). Additional
information is needed on the relationship between activity per
amount of protein in the CEFIC (1986e) in vitro assays and how
much of this activity would be in the entire liver. Assumptions
can be made in this area, but additional uncertainty is the
result.
Fifth, a number of incidental questions have not been
answered, such as where the carbon dioxide comes from in vivo
experiments using rats and hamsters treated with DCM if the GST
pathway is negligible or absent using DCM as a substrate, as
indicated from the CEFIC (1986f) analyses in these species. One
hypothesis is that the MFO pathway can also produce carbon
dioxide; the CEFIC (1986f) data show that all species tested,
including humans, have MFO activity. However, this is only a
hypothesis that may be questioned, as discussed previously.
63
-------
Another question is whether or not other fractions of the liver
homogenate, not just the cytosol, were assayed for DCM GST
activity. Although this activity might be expected to be in the
cytosol, other possibilities should not be ruled out on the basis
of what is expected.
Sixth, CEPIC (1986e) only assayed human liver GST at a
single pH value (pH 7.4) using DCM as a substrate. Since
individual isozymes may have sharp pH optima in the acidic,
neutral, or alkaline range, the CEFIC (1986e) assay may have
selected against the detection of a human DCM-specific GST
isozyme. Furthermore, it has been reported by Seidegard and Pero
(1985) and Seidegard et al. (1985, 1986) that some humans (as
many as 50% of the population) lack a specific principal GST
isozyme that is present in the rest of the population. Thus, if
one or more of the particular subjects studied by CEFIC (1986e)
lacked a potential DCM specific GST isozyme, there would be a
decreased sensitivity, or an inability to detect activity towards
this substrate.
Finally, and of great importance, only human liver was
assayed by CEFIC (1986e). Human lung, or other potential target
organs such as the pancreas, were not assayed. If the premise of
no human risk by virtue of the lack of metabolic capability by
the GST pathway is to be assumed, then information from the liver
only is not adequate to rule this metabolic capability out for
the entire body.
In conclusion, there are many ways to calculate the kinetic
64
-------
input into a pharmacokinetic model for DCM in order to estimate
"target site" dose. The ultimate effect on estimates of human
risk, using Andersen et al.'s methodologies and assumptions
(which include that the GST pathway is the only toxic pathway,
and that risks are related to concentration at a "target site"),
ranges from a finding of little difference to a finding of no
human risk at all from the traditional "applied dose" procedure.
Additional investigations* (some of which are currently ongoing
by ICI-UK and NTP as outlined in Chapter 2) and more data will be
needed to narrow this large range of uncertainty.
*0n May 26, 1987, HRAC members received a letter from Dr. R.
Reitz of Dow Chemical Company and Dr. T. Green of ICI-UK
indicating that both labs were able to find human GST activity
using 36C1-DCM. Further data will be forthcoming on these
findings. The ICI-UK lab, directed by Dr. Green, is the same
lab that conducted the previously cited studies by CEFIC (1986a,
b, c, d, e, f).
65
-------
5. MUTAGENICITY ASSESSMENT OF DICHLOROMETHANE:
REVIEW OF STUDIES PERFORMED BY CEFIC
5.1. PHASE I TESTS
In Phase I of its investigation into the carcinogenic
mechanism of DCM, CEFIC (1986b, c) tested the effect of the
compound on rats and mice in two genotoxic assays: unscheduled
DNA synthesis (UDS) in vivo and in vitro, and covalent binding to
DNA in vivo.
5.1.1. In vivo Unscheduled DNA Synthesis
DCM was tested for its ability to induce DNA repair
synthesis in liver cells of Fischer 344 rats and B6C3F1 mice
treated in vivo (CEFIC, 1986b). Animals were exposed by
inhalation to DCM at 2000 and 4000 ppm for either 2 or 6 hours.
At the end of the exposure period, animals were perfused to
generate hepatocyte cell suspensions which were then placed into
culture. After allowing up to 2.5 hours for cell attachment,
tritiated thymidine was added to the culture medium. After an
additional 4 hours of incubation with fresh medium, the cultures
were washed and incubated overnight with medium containing no
label. The following day, the ceils were fixed, dried, and
prepared for autoradiography. Those cells that were undergoing
"long patch" DNA repair during the 4-hour incubation period would
be expected to show higher than background levels of exposed
silver grains over hepatic nuclei.
There are several serious weaknesses in this study which
66
-------
cast into doubt the usefulness of the results. First, only male
animals were tested and no reason was given for not testing
females. Second, the assay of in vivo UDS has inherent
theoretical difficulties. Its use, rather than an in vitro
method, is only justified in situations where extrahepatic
metabolism may be a factor. Even then, this assay should be used
only when metabolites generated in other tissues are sufficiently
stable to travel to the liver and generate DMA damage in
hepatocytes. A serious weakness of this assay is the relatively
long period of time between agent administration and the assay of
DNA repair. Excision repair presumably begins shortly after
exposure to the test chemical. Before UDS can be measured,
however, the animal first must be perfused, and then the
hepatocytes isolated, placed into culture, and allowed to attach
(the final step alone requiring approximately 2 hours). It has
not been shown that UDS continues for a sufficient period of time
to be reliably detected in vitro. These uncertainties are
exacerbated further by the nature of the positive controls. A
convention has developed that the positive control animals in
/
this assay do not have to be similarly exposed in vivo (Mirsalis
et al., 1980). Instead, hepatocytes isolated from untreated
animals are exposed in vitro to diethylnitrosamine. In the
opinion of the HRAC, this does not constitute an adequate
positive control.
The third weakness of this study relates to dose selection.
The authors selected the dosing regimens in order to deliver
67
-------
doses equivalent to those used in the NTP bioassay. In effect,
the investigators chose a dose normally used in a chronic study
for this acute assay. This, of course, nearly guarantees a
negative result for this study. Exposure levels for a lifetime
bioassay are selected so that animals can be chronically dosed
with sufficient numbers of animals surviving to the end of the
study. Indeed, in the NTP study, the animals were given 2000 and
4000 ppm for 6 hours per day, 5 days per week, for 102 weeks.
For an acute study such as the in vivo UDS, much higher doses
should have been selected.
5.1.2. In vitro Unscheduled DNA Synthesis
For the in vitro UDS studies, hepatocyte cultures derived
from untreated rats and mice were exposed to 500, 1000, 2000, and
4000 ppm DCM for 8 hours (CEFIC 1986b). The methods used in this
study are similar to those described for the in vivo study except
of course that dosing occurred in vitro. Dose selection again
presents a major problem for assessing the value of results
obtained in this assay. The criteria used to select the exposure
levels were not presented. Generally, in such a study,
concentrations are selected by first determining a dose which can
cause profound cytotoxic effects as judged by trypan blue
exclusion, LDH release, cell detachment, or altered morphology.
Normally, test doses that are significant fractions of the
cytotoxic dose would not be selected for testing. However, there
is no indication that any cytotoxicity measurements were made in
.»*
these studies.
68
-------
A second criticism of the study results from the method used
to score exposed silver grains over hepatocytes. In the standard
protocol the authors state: "For each cell the number of grains
over the nucleus and the highest number of grains in an
equivalent adjacent area of cytoplasm were counted. The
cytoplasmic count was subtracted from the nuclear count to give a
net grain value for each cell." This method of selecting the
highest number of grains in an equivalent adjacent area has been
commonly used in the past. New guidelines from the American
Society for Testing and Materials, agreed upon by experts in the
field, state that an average number of grains from several
equivalent adjacent areas of cytoplasm should be used to quantify
the background. This change could significantly alter the data
and conclusions in a study such as this one in which the results
at the high dose are equivocal.
Finally, data interpretation presents an additional problem
for evaluating the merits of this study. Both the rat and mouse
in vitro studies were judged to be negative; however, the HRAC
judges them as equivpcal in result. For example, in the mouse
study, 1.3% of the cells were in repair for the control, while
13.3% were in repair for the 500 ppm cultures, 10.0% for the 1000
ppm cultures, 14.0% for the 2000 ppm cultures, and 19.3% for the
4000 ppm cultures. Thus, there would appear to be an obvious
dose response. The authors provide no statistical basis for
calling these results negative. Clearly, this difference is much
larger than that seen between the negative and positive controls
69
-------
in the rat study: 8.5% and 10.6%, respectively. Therefore, if
the difference for the mouse assay is not significant, then the
controls for the rat assay are not valid.
Finally, in the discussion section the investigators state
"It is however clear that methylene chloride is exerting a low
level non-specific effect on both nuclear and cytoplasmic grain
counts." The authors' explanation of a low level non-specific
effect was not clear. Futhermore, the authors speculated that
this effect may represent "an early indication of elevated levels
of replicative (scheduled) DNA synthesis." It is not clear how
the authors may have arrived at this conclusion, since DNA
replication still takes place in the nucleus.
5.1.3. Covalent Binding to DNA in vivo
This study (CEFIC, 1986c) attempts to determine whether the
carcinogenic action of DCM may be mediated through its ability to
bind covalently to DNA and thereby lead to mutations. Fischer
344 rats and B6C3F1 mice were exposed to 4000 ppm 14C-DCM for 3
hours. DNA was isolated from lung and liver tissues 6, 12, and
24 hours after the start of exposure, then hydrolyzed and
analyzed by high performance liquid chromatography. As a
control, an additional group of rats and mice were injected with
14C-formate to determine if radioactivity associated with DNA
originated from incorporation of radiolabeled carbon from the
free carbon pool. The rationale for this study is based on the
observations that DCM, which can be metabolized in several ways,
is believed to yield reactive intermediates capable of alkylating
70
-------
DNA and thus leading to mutations, but DCM can also generate
formate as an intermediate which can enter the free carbon pool.
Since this pool is used in the normal biosynthesis of nucleotide
bases, radioactivity associated with DNA after exposure to
labeled DCM may be a normal component of DNA and not an alkylated
product. Based on the results of this study the authors indeed
conclude that all the radioactivity associated with DNA after
exposure to 14C-DCM results from incorporation of formate into
nucleotide bases and not from DNA covalent binding. The authors
further conclude that based on these results, DCM is not
genotoxic and is another example of an "epigenetic carcinogen."
This study can be criticized on several grounds. First, it
is important to note that chemicals can cause mutations in DNA
without alkylation. For example, intercalating agents can cause
frameshift mutations without covalent binding, and some metals
have been shown to cause mutations in DNA by interfering with the
fidelity of DNA polymerase. Therefore, the assertion, that
because DCM does not bind to DNA it is not genotoxic, is invalid.
Second, the authors again performed the studies only in male
animals without explanation.
The major weaknesses of this study, however, appear to be in
design and interpretation. First, it is not clear why the
investigators chose to dilute 14C-DCM with unlabeled compound to
produce a lower specific activity exposure environment. Clearly,
the higher the specific activity, the greater the probability of
detecting a minor DNA adduct. Second, the authors present an
71
-------
elaborate argument to support their conviction that the
radioactivity in the second fraction of the mouse liver
chromatogram is really a protein contaminant containing
radiolabeled amino acids which were also synthesized from the
free carbon pool. To support this argument the authors point out
that a similar peak (although relatively smaller) is also present
in the chromatogram from livers of mice exposed to 14C-formate
and in the chromatogram of lungs from mice exposed to 14C~DCM.
However, the relative size' of this peak in the lung chromatogram
is greatly diminished. In order to reconcile this apparent
discrepancy, the authors present the following highly elaborate
but specious argument:
The fact that similar contamination by radioactive
protein is not seen in mouse lung DNA may be a reflec-
tion on the known specificity of methylene chloride
for a single cell type in the lung, the Clara cell
(CTL/P/1432). In the liver, even though there are
small differences in the distribution of metabolizing
enzymes across the hepatic parenchyma most hepatocytes
will metabolize methylene chloride to a similar extent.
Consequently protein and DNA in each hepatocyte will
have been exposed to a similar concentration of methy-
lene chloride metabolites and the macromolecules
isolated from the whole organ will be reasonably
representative of each single cell within the liver.
In contrast to the liver, the specific damage to mouse
lung Clara cell suggests that metabolism occurs largely
in a single cell type which is known to contain the
highest concentration of cytochrome P-450 enzymes in
the lung (Boyd, 1977). Because the Clara cell consti-
tutes approximately 5% of the total cell types in mouse
lung, specificity for this cell would result in a 20-
fold dilution in the specific activity of protein iso-
lated from the whole organ. In contrast, l4C-formate
incorporation into DNA will presumably occur in all
cell types of the lung dependent only on their rates of
DNA synthesis. The effect therefore of protein contami-
nation of DNA from the lung will be markedly less than
for hepatic DNA.
72
-------
Such reasoning fails to consider that if only the Clara cell
metabolized DCM to formate, permitting 14C into the free carbon
pool to be used in protein synthesis, the radioactivity in the
DNA will also only be derived from Clara cells. Thus, the ratio
of radiolabeled DNA to protein should remain the same, although
the specific activity of both macromolecules should be lower in
lung. The specific activity of DNA from the two tissues is
essentially the same, while the specific'activity for lung
proteins is not provided in the report. In any case, this issue
could have been resolved easily and directly by simply injecting
one of the mice with a tritiated amino acid and then analyzing
for both labels in the chromat©graphic fractions. If the
authors' hypothesis is correct, tritium should only be detected
in fraction 2.
Finally, the authors have ignored all of the labeled peaks
in fractions 22 through 27 in the chromatogram of mouse liver
DNA. Yet, these peaks could have coincided with alkylated bases.
Indeed, the use of a higher specific activity DCM would probably
have enhanced these peaks. The sensitivity of the assay poses a
further problem. In examining the chromatograms of the DNA
hydrolyzates, it is clear that there is radioactivity associated
with all nucleotide bases except cytosine. This is as expected
eili66 the free carbon pool is used by the cell in the synthesis
of purine rings but not in the synthesis of pyrimidine ring
structures. However, radiolabel is expected to be associated
with thymidine since the 5-methyl group is derived from the free
73
-------
carbon pool. Concern is raised because there is no peak
associated with 5-methyl cytosine. This modified base is created
after cytosine is incorporated into DNA and should comprise
approximately 5% of total cytosine; the source of this methyl
group is the free carbon pool. One can conclude, therefore, that
this assay is incapable of detecting a single species of
alkylated nucleotide even when that species comprises 5% of the
total amount of that base.
5.1.4. Summary of Phase I Tests
The results of mutagenicity testing of DCM are mixed. DCM
clearly is a bacterial mutagen giving positive responses in the
Ames assay (Barber et al., 1980; Green, 1983; Jongen et al.,
1982). Both positive and negative responses have been reported
in the yeast Saccharomyces cerevisiae (Callen et al., 1980;
Simmon et al., 1977) in Drosophila (Abrahamson and Valencia,
1980; Gocke et al., 1981), as well as for the induction of sister
chromatid exchange in cultured hamster cells (Jongen et al.,
1982; Thilagar and Kumaroo, 1983). Highly significant levels of
chromosomal aberrations were induced in CHO cells in vitro
(Thilagar and Kumaroo, 1983) but were not detected in vivo (Dow
Chemical Company, 1980); the doses tested in this latter study
may not have been sufficiently high. The weight of evidence
suggests that DCM is mutagenic, although it is perhaps a weak
mutagen. The results suggest that DCM, like other mutagenic
carcinogens, initiates cancer through genetic alterations. In
addition, other health effects associated with exposure to
74
-------
mutagens may also be of concern, e.g., heritable (germ line)
mutations, teratogenicity, and reproductive effects. The data
presented by the ICI-UK provide nothing to alter this conclusion.
5.2. PEASE II TESTS
In Phase II of its investigation into the carcinogenic
mechanism of DCM, CEFIC (1986d, g) tested the effects of the
compound on rats and mice in two additional assays: the mouse
micronucleus test and induction of S-phase hepatocytes.
5.2.1. Mouse Micronucleus Test
DCM was tested (CEFIG, 1986d) in the mouse micronucleus test
using both male and female C57BL/6J/Alpk mice. Preliminary
studies were performed to determine the approximate MLD/7 (mean
lethal dose after 7 days). The micronucleus test was performed
using doses of 4000 mg/kg, 2500 mg/kg, and 1250 mg/kg, which are
equivalent to 80%, 50%, and 25% of the MLD/7, respectively. Ten
mice were included in each dose group, along with positive
(cyclophosphamide) and solvent (corn oil) controls. Animals were
sacrificed and bone marrow smears prepared at 24, 36, 48, and 72
hours after dosing. The frequencies of micronuclei per
polychromatic erythrocyte (PCE) were within the control range for
all dose levels and all time points. The percentage of PCEs was
decreased for the 24-hour time points at the two highest doses,
although statistical significance was achieved only at the 2500
mg/kg dose.
The authors concluded that DCM is not clastogenic in the
mouse micronucleus test. The HRAC believes a more accurate
75
-------
conclusion would be that DCM was not positive in the mouse bone
marrow micronucleus test (under the conditions of the assay).
This is because the micronucleus test does not detect all
clastogenic activity, but only acentric fragments and some
spindle effects. Also, although bone marrow is generally used
for micronucleus assays, bone marrow is not a target for DCM, and
reactive metabolites may not reach it in sufficient quantity to
produce a detectable effect.
5.2.2. Induction of S-phase Hepatocvtes
DCM was tested (CEFIC, 1986g) for its ability to induce DNA
synthesis in liver cells of B6C3F1 mice. The induction of
S-phase hepatocytes was measured after either one or two
inhalation exposures to DCM for 2 hours. After exposure the
animals were injected with [methyl-3H]' thymidine using one of
three radiolabeling regimens, and DNA synthesis was quantified
using autoradiography at 24 and 48 hours after exposure. Sodium
phenobarbitone was used as the positive control. The use of two
of the three experimental protocols revealed small but
statistically significant increases in the numbers of S-phase
cells.
The results of these studies are difficult to interpret for
a number of reasons. The induction of S-phase hepatocytes is not
a dommonly used assay, and the results, either positive,
negative, or as in this case equivocal, are of unknown
significance. The authors state:
The carcinogens [carbon tetrachloride, trichlorethylene,
and polybrominated biphenyls] do, however, induce liver
76
-------
cell turnover in vivo, as monitored by increased DNA
scheduled synthesis. It has therefore been suggested
that such as increased cell turnover in the liver may be
mechanistically involved in the hepatocarcinogenicity to
the mouse of compounds found to be non-genotoxic in vivo.
Mirsalis et al. (1985) have suggested that the apparent
correlation between induced cell proliferation in the
liver and hepatocarcinogenicity indicates that any com-
pound that induces increased cell turnover in the B6C3F1
mouse liver may produce liver tumors in a two year bio-
assay.
Based on the explanation in the section above and the use of
phenobarbital as the, positive control, the S-phase induction
assay appears to be a test for potent promoters in a system that
is genetically "initiated." It is clear, however, that simply
inducing liver cell turnover alone cannot be the entire action of
the hepatocarcinogens described above/ since a partial
hepatectomy, a very efficient inducer of hepatocyte DNA synthesis
is not, in and of itself, carcinogenic (European Chemical
Industry Ecology and Toxicology Center, 1982). Therefore, such
"non-genotoxic hepatocarcinogens" must have other actions in
addition to the induction of cell replication. Futhermore, since
the authors claim that the small increases seen in this study may
have ho biological significance, one may conclude that DCM does
not fall into this class of carcinogens, and therefore the low
levels of genotoxicity seen in in vitro assays may be responsible
for its carcinogenic activity.
The equivocal results seen in this study may be based on the
doses of DCM that were selected for study: 4000 ppm for 2 hours
for a single or two inhalation exposures. This was the same dose
used in the NTP bioassay, but, in the latter, mice were exposed
77
-------
to this same level for 6 hours a day, 5 days a week, for a total
of 2 years. The exposures used in this acute study may have been
too low or of insufficient duration to maximally induce DNA
synthesis.
Finally, the liver was not the only target organ of DCM-
induced tumors. Lung tumors were induced as well.
5.2.3. SajTmnary of Phase II Tests
The mouse bone marrow micronucleus test was conducted at
adequate exposure levels (CEFIC, 1986d). The negative results
may have been due to the fact that bone marrow is not a target
organ for DCM. The induction of S-phase, even if it were clear-
cut, could not be interpreted at this time in a way that would
help to elucidate the mechanism of DCM carcinogenesis.
There is, at present, no way to discriminate between
nongenotoxic and weakly genotoxic activity for DCM.
78
-------
6. EPIDEMIOLOGYS RECENT KODAK STUDY
6.1. INTRODUCTION
Since the Health Assessment Document (HAD) for DCM was
published, Friedlander et al. have updated their study of
Eastman-Kodak employees, increasing by some 262 men the size of
the original cohort and presenting some detailed dose-response
analyses. An early report dated July 2, 1985, had been submitted
to the EPA; subsequently, the authors presented an updated
analysis at the Winter Toxicology Forum in Washington, D.C., on
February 18, 1986, and later at a meeting with scientists from
the EPA, FDA, and CPSC. In June 1986, this analysis was
submitted to EPA under the authorship of Hearne et al. (1986) as
a prepublication copy. This paper has been published recently
with only minor changes (Hearne et al., 1987). Two main themes
are emphasized in these recent updates; first, that DCM is safe
for humans at the occupational exposure levels; and second, that
EPA's upper-limit incremental cancer risk estimates based on
animal studies, significantly overpredict the cancer experience
of the Kodak employees, and therefore should be lowered.
Other actions are also of note. The series of Kodak studies
through July 2, 1985, were sent to Professor Genevieve Matanowski
for review by EPA's Office of Toxic Substances; she submitted her
report on October 15, 1985. This review was in turn responded to
by Friedlander et al. on April 15, 1986. Matanowski responded to
the Friedlander et al. response, but was, at the same time, part
79
-------
of a team reviewing the Hearne et al. (1986) paper (Batelle,
1986). Matanowski's review dealt with qualitative rather than
quantitative issues. Briefly, she noted some methodological
problems which preclude her from agreeing with the authors'
conclusions regarding the safety of DCM. One of her major
criticisms was that the cohort actually was a survivorship
cohort, with the associated bias being one of selection of
persons remaining because of tolerance to the exposure
conditions. She suggested that this bias must be removed before
a meaningful analysis could be done. Matanowski also found
problems with the control populations. Comparison of the Kodak
s
death rates with the New York State death rates raised the
"healthy worker effect" issue. While comparisons with the total
Kodak hourly workers adjusted for the healthy worker effect,
shortcomings of such a comparison were described. Matanowski did
not, however, suggest a quantitative estimate of the bias
involved in these SMR comparisons; as a result, no adjustment
could be made.
The following sections present the HRAC review and analysis
of the most recent Kodak study (Hearne et al., 1987). Section
6.1. presents a critical review of the methodology and results
of the study; section 6.2. uses the data to calculate a
quantitative cancer risk estimate, and compares this estimate
with others that have been calculated both from human and animal
studies; section 6.3. discusses the significance of the
pancreatic cancer response; and section 6.4. summarizes the
80
-------
results.
6.2. HRAC REVIEW OF THE STUDY BY HEARNE ET AL. (1987)
Hearne et al. (1987) completed a historic prospective study
on all 1013 male employees of the film casting division of
Eastman-Kodak in Rochester, NY (Kodak-Rochester), who were
employed at least one year, of which some portion of that
employment had to overlap into the period from January 1, 1964
through December 31, 1970. The cohort was followed through
December 31, 1984, and vital status ascertainment was 99.1%
complete. Causes of death in this cohort were contrasted«with
those expected based on rates of two referent comparison groups:
(1) the general male population of upstate New York (excluding
New York City) from 1965 to 1980, and (2) an occupational
population of greater than 40,000 male employees of Kodak-
Rochester (excluding this plant) from 1964 to 1985.
DCM has been used since 1944 in this plant as a solvent in
the manufacture of cellulose triacetate film base. This process
is accomplished in large hooded "casting machines" within which
heated air is circulated and drawn off into a solvent recovery
facility. Employees must enter the casting machines from time to
time (up to six entries per day) to make manual adjustments arid
do maintenance. Samples from personnel industrial hygiene
monitors and surrounding air indicated that concentrations of DCM
range from 5000 to 10,000 ppm within the casting machines and
between 30 to 100 ppm outside the casting machines but within the
workroom. Air-supplied respirators were provided for use
\ -
81
-------
whenever entry into the casting machines became necessary, but
they were not always used.
The measurement of exposure appears to have been quite
well-done. Since 1945 more than 1200 samples of area air and
from personnel monitors had been collected at several locations
in the vicinity of the casting machines, usually in the
breathing zone of the employees. Additionally, to assist in
operations, thousands of samples were taken by engineering
personnel at critical points within the machines.
The characterization of DCM exposure also appears to have
been done thoroughly. Regression analysis by the authors failed
to reveal any significant trend in seasonally adjusted DCM
concentrations within the production room between 1953 and 1985.
An estimate of lifetime exposure for each employee was then
derived by summing the products of the duration in months spent
in each of these jobs by the average exposure level in ppm
(normalized to an 8-hour time-weighted average) in each of those
jobs. Three career exposure categories were defined: under 350
ppm-years, 350 to 749 ppm-years, and 750 ppm-years or greater*
The median latency in each of these categories were 17, 31, and
37 years, respectively.
A priori hypotheses, based on the results of an NTP bioassay
(1986) and in vivo studies of the metabolism of DCM, led the
authors to consider malignancy of the lung and liver as well as
ischemia heart disease as possibly high risk causes of death.
The authors found a statistically significant deficit of
82
-------
deaths from all causes combined compared to the death rates of
males living in upstate New York or employed at Kodak-Rochester
[176 observed versus 253.2 expected (upstate New York) or 205.8
expected (Kodak-Rochester)]. With respect to total malignant
neoplasms, a nonsignificant deficit of deaths was noted [41
observed versus 59.3 expected (upstate New York) or 52.7 expected
(Kodak-Rochester)]. Deficits, albeit nonsignificant, were also
seen for respiratory cancer [14 observed versus 21.0 expected
(upstate New York) or 16.6 expected (Kodak-Rochester)] as well as
liver cancer [0 observed versus 0.8 expected (upstate New York)
or 0.5 expected (Kodak-Rochester)]. A statistically significant
deficit of colon-rectal cancer is apparent [2 observed versus 9.8
expected (upstate New York)], as well as a nonsignificant excess
risk of pancreatic cancer [8 observed versus 3.2 expected
(upstate New York) or 3.1 expected (Kodak-Rochester)] attributed
by the authors to posssible misdiagnoses or incorrect coding of
underlying cause of death.
This study appears to be well-conducted in all respects
except one. There is a distinct possibility that the cohort
Selection process may have been faulty. Approximately 75% ot the
cohort began work prior to January 1, 1964, but after 1944. A
total of 48% of the cohort actually began employment prior to
1954. Anyone who terminated his employment for whatever reason
prior to January 1, 1964, was not included in the cohort. This
cohort selection criterion has the potential for providing a
study cohort that consists mainly of "survivors" (non-
83
-------
susceptibles, exceptionally careful health-conscious persons, or
persons who just did not succumb to a work-related illness).
This phenomenon is different from the healthy worker effect which
was present when observed deaths were contrasted with upstate New
York death rates but not when contrasted with deaths rates
generated from Kodak-Rochester employees. However, even in the
latter case severe deficits still persist with some exceptions,
i.e., lung cancer. These results are suspicious unless the
survivorship question is considered as a contributing factor.
That this survivorship effect is present is supported by the
analysis of respiratory cancer by latency and exposure. There
appears to be a nonsignificant elevated risk of respiratory
cancer in the lowest exposure and latent categories where one
would expect that the experience of the more recent hires would
predominate. This is followed by a decrease in risk as exposure
or period of follow-up increases. A larger proportion of the
experience of the pre-1964 hires tends to predominate in the
higher dose and latent categories. This phenomenon has also been
seen in other studies but most notably in the Fox and Collier
(1978) studies of vinyl chloride workers.
One possible solution to this problem would be to either
include those employees terminated prior to 1964 in the cohort
and blend their mortality experience in with the present cohort.
Alternatively, the analysis could exclude all employees who began
prior to January 1, 1964, and examine only the mortality
experience of those employed on or after January 1, 1964. The
84
-------
main deficiency with the latter solution is that very few
person-years would accumulate in the higher latency and dose
categories thus leading to a tremendous drop in the power to
detect an increased risk if one were present. If these
suggestions are incorporated into the study design and the
results of the reanalysis are nonsignificant, then it seems
likely that this study could be used to determine an upper-bound
risk estimate for cancer. In the latter solution, however, where
the power is dramatically reduced, the results could hot be
assumed to imply that exposure to DCM does not increase one's
risk of cancer.
6.3. APPLICATION OF EPIDEMIOLOGY TO QUANTITATIVE CANCER RISK
ASSESSMENT AND COMPARISON WITH ANIMAL-BASED ESTIMATES
This, section discusses several previous quantitative cancer
risk assessments for DCM. In the next section a comparison will
be made of the upper-limit incremental unit cancer risk estimate
for DCM predicted from the NTP mouse inhalation lifetime bioassay
(1985, 1986) with that predicted from the Hearne et al. (1987)
study of Kodak employees.
6*3.1, Previous Quantitative Cancer Risk Assessments for DCM
the Carcinogen Assessment Group (GAG) has previously
reviewed both the animal studies (NTP, 1985) and epidemiologic
studies (up to and including Hearne and Friedlander, 1981) in the
HAD for DCM and Addendum (U.S. EPA, 1985a, b). Based on the
studies, GAG calculated an upper-limit incremental unit risk of
= 1.4 x 10"
(ppm)""1 for inhalation based on lung and liver
85
-------
tumors in mice. The GAG has also calculated 95% upper-limits of
expected deaths based on the early epidemiology study and
concluded that the "Friedlander et al. study does not have the
power to rule out an overall cancer risk, or . . .a lung cancer
risk, that is predicted using the upper-bound slope derived from
the NTP study." Because the Hearne and Friedlander (1981) update
was considered a negative response for cancer, if the upper-limit
risk estimate based on the epidemiology study had been less than
that based on the animal data, a human data estimate would have
been used.
Allen et al. (1986) prepared a risk assessment of the Hearne
and Friedlander 1981 update of their original 1978 study as part
of an overall effort to correlate risk assessments based on
animal and human data for the CA6. Allen et al. based their
assessment on a relative risk model and 24 total malignant
neoplasms observed in the cohort update of the 1981 paper versus
28.64 expected. Only average exposure data for the entire cohort
were available. The results of this analysis are presented in
Table 8. Also included in Table 8 are estimates which Allen et
al. calculated based on a study.of Dow Chemical Company '
feraplbyees. This upper-limit estimate, 1.6 x 10~2 (ppm)"1, i^
also consistent with that based on total cancer deaths in the,,.
Kodak study, but Allen et al. used a methodology for uncertainty
which tends to inflate the upper-limit estimates versus the EPA
methodology. . ;
With respect to the Hearne et al. (1987) update, the U.S.
•• • • _• ••• 86 '' ' • - ;"';" V-- .•••..••;••••.•• •:' • ••• '.
-------
co
Q 63
W M
^ \~.
3y iTi
co
H §
p, «
o p
§
CO
63 iX
EH /* u}
••H CD
rH
M -P
i-J i — I
O Q
C ^i
5^
rH
•"'§1
£ e 3
• • f
4J
CO
oo
o
r-H
X
0
oo
OO
o
rH
X
r~-
•
V.0
O
m
w
n3
2
o
DJ
0)
•H
JJ
-H
3
1
a
00 OO
-------
Consumer Product Safety Commission (CPSC) has examined the
authors' claim that upper-limit estimates based on animal data
are too high (Cohn and Rock, 1986). The reanalysis by CPSC found
that extra cases predicted for the Kodak employees from animal
data were between 2.2 and 8.7, estimates which bordered the 4.8
excess of pancreatic cancers observed in Kodak employees in the
most recent update. Another agency, the California Department of
Health Services, prepared a report (1986) with an analysis of
more extensive dose-response data obtained personally from the
authors of the Kodak study. In that analysis upper-limit unit
risk estimates were devloped based on both lung and pancreatic
cancers in the Kodak workers. The upper-limit estimates based on
pancreatic cancers are within a factor of plus or minus 2 of the
EPA's upper-limit estimates based on the NTP female mouse
bioassay, which is qj = 1.4 x 10~2 (ppm)"1. These estimates are
presented in Table 8.
One other analysis of the NTP mouse data, an internal Food
and Drug Administration (PDA) memorandum dated May 4, 1985,
estimated a unit risk value roughly 27.6 times less than EPA's.
There are also other analyses using epidemiologic data, including
the Dow study on DCM workers (U.S. EPA, 1985a). The bulk of
these analyses, however, discuss the epidemiologic studies in
terms of statistical power concepts, i.e., the probability of
detecting an increase in cancers if one is actually associated
with DCM exposure. While EPA has dealt with this concept in a
previous document on DCM (U.S. EPA, 1985a) and other chemicals,
88
-------
the focus here will be on the analysis of actual risk estimation
based on the Kodak employee cohort, which is presented in the
following section.
6.3.2. Quantitative Risk Estimation Based on the Study of Kodak
Employees
The following analysis provides maximum likelihood and 95%
upper-limit estimates of incremental cancer risk based on the
cancer death response in the lung and pancreas in the most recent
update of the Kodak study (Hearne et al., 1987). Both additive
and relative risk models are used. This type of analysis has
been used previously with epidemiologic studies in several EPA
documents; the HADs for nickel and cadmium are two recent
examples. It also follows closely the analysis used by the State
of California (1986), but the results differ—in some cases only
slightly, in others quite a bit—because of an adjustment made
for latency. A description of the models follows.
6.3.2.1. Excess or Additive Risk Model—This model follows the
assumption that the excess cause-age-specific death rate due to
DGM exposure, hx(t), is increased in an additive way by an amount
proportional to the cumulative exposure up to that age. In
mathematical terms, this is
BXt
where X^ is the cumulative exposure up to age t, and B is the
proportional increase. The total cause-age-specific rate h(t) is
89
-------
then additive to the background cause-specific rate h0(t) as
follows:
h(t) = h0(t) + hx(t)
Under the assumptions of this model, the parameter B can be
estimated by summing the expected rates to yield
> Ej = E0j + BXjWj
where Ej is the total number of expected cancer deaths in the
observation period from the group exposed to cumulative exposure
Xj. E0j is the expected number of cancer deaths due to
background causes; for the Kodak study, expected deaths based on
both total Kodak employees and New York State cancer death rates
are available. The total Kodak employee death rates were used
since they eliminated the "healthy worker effect," although, as
discussed in Section. 6.2l, this control group may not be entirely
suitable. Wj is the number of person-years of observation for
the jth exposure group, and the parameter B represents the slops
of the dose-response model. To estimate B, the observed number
of cause-specific deaths, oj, is substituted for Ej. Oj is
assumed to be distributed as a Poisson random variable. The
parameter estimate, b, can be tested for being significantly
greater than zero. A statistically significant result is
evidence of an additional cancer effect due to cumulative DCM
90
-------
exposure.
Under the above assumptions, the solution by maximum
likelihood proceeds as follows: the likelihood equation is
L = 2 [exp-(Eoj + BXjWj)] [Eoj + BXjWj
where N = the number of separate exposure groups. The maximum
likelihood estimate (MLE) of the parameter B is obtained by
solving the equation
N
d In L = 2 [-X-SW-; +
dB j=l J J
/ (E0i +
J
= 0
for B.
The asymptotic variance for the parameter estimate b is
2 2
t
(Eoj
i-l
Where b is the MLE. This variance can then be used to obtain
SpgrSximate 95% upper and lower bounds for b. Lifetime
incremental cancer risk estimates for 1 ppm continuous exposure
are estimated by multiplying b by 70 if X is in units of lifetime
continuous exposure.
6.3.2.2. Multiplicative or Relative Risk Model—This model ,
follows the assumption that the background cause-age-specific
rate at any age t is increased in a multiplicative way by an
91
-------
amount proportional to the cumulative dose up to that age. In
mathematical terms this is
h(t) = h0(t)(l + BXt)
As above, summing over the observed and expected experience
yields, for each exposure group,
E
J/EOJ
= 1 + BX
j
Again, to estimate B, the observed number of cause-specific
deaths, Oj, assumed to be a Poisson random variable, is
substituted for Ej. Following the same procedure as above, the
MLE, b, is the solution to
N
d In L = £ [-E0iX-< + (0-4X^)7(1 + BX-O] == 0
dB j=l J J JJ J
with asymptotic variance
N
C 2 (Eojx2j)/(i +
Lifetime incremental risk estimates per ppm under this model are
obtained by multiplying b by the background lifetime cause-
specific risk of death, P0. The values P0 are derived using life
table methods and 1980 U.S. death rates. For lung and pancreas
cancers these are 0.037 and 0.008, respectively.
92
-------
6.3.2.3. Data—The data for the above models, presented in
Tables 9 and 10, are derived from information supplied to the
State of California by Dr. Friedlander of Kodak and are
reproduced from the State of California's criteria document.
Table 9 presents the observed and expected deaths from lung
cancer and Table 10 presents the pancreatic cancer death
experience by separate age and exposure categories. The three
career cumulative exposure categories (<350, 350-749, and 750+
ppm-years) are the same as those presented in the Hearne et al.
(1987) update and are comprised of 350, 353, and 310 employees,
with mean exposure levels of 16, 22, and 42 ppm (8-hour time-
weighted average) respectively. Mean duration of exposure (13,
26, and 29 years) and median latency (17, 31, and 37 years) also
increased for the higher level exposure categories. The mean
cumulative exposures were 153, 531, and 1212 ppm-years,
respectively. Lifetime continuous exposure (LCE), as presented
in Tables 9 and 10, were calculated as follows:
LCE
= Xi = 16 ppm X 8/24 X 240/365 X 13/70 =0.65 ppm
For the two higher exposure categories, substitution of mean
' i
exposure levels 22 and 42 ppm, and mean duration of 26 and 29
years, respectively, leads to LCEs of-1.79 ppm and 3.81 ppm.
The State of California analysis used all age and exposure
grdupS for every analysis. Here ail experience prior to age 45
is eliminated, on the assumption of a 20-year latency and the
93
-------
CO
r-l
32 tC
§H CD
^j »5
§ g
CO
SB
uJ J
x. o
uu £j
Jj
2 uu
1^*
Q EH
s §
J5S
53 E
§?5
Q S
Cr3
O O
M gj
«rf uj
^
1^ r«
§£j
Q
pf_i f\.
O S
O* vj
*3
a\
CM O> CO
VD rH o ro
O CO
ro vo
CM
in
co r- o
CM ro O **3*
•H ro
<£> ro
rH
CTl
•^f in in o
-H r-H ro CM CM
*3*
y?
rH
ro ro
rH ro u?
r-^OO tDOrH CNOO
^* 00
^D O^
rH rH
CTl
VD
ro CN rH
-H
rH
rH
**D i — I
rH CO O
cooo c~-oo moo
o ro
co r~-
CM ' — 1
0
O
l^ 0
*-? o o* m
co ro
OO rH
ro
O
CM
CTl O
ID
rH
0 0
0 0
o o
• , •
O O OOO
0 O
0 0
0 O
CTlOO OOO OOO
O
rH
rH
to to
2 <0 — - -r-l tO & (0
•~* (S -r-iO- ^1 — 0)
ID -a o H ro . -o
rn •• y) ^-^ ^^* d) •* to
in— ~^Ji-i _ >— ->-I>H
nj in ro CD TJ 'Q L en (o cu
0 ^D 0 u 0 o) S r*^ CD o
loirowo arHiro
EIICCJ00 II CO
Q. *r— O tO Q« CT» •!— tO
O W (1) 3 | W m 3
inocMJ ocjCii-5
ro i-3 in i3
v — * ro ^—
-j§
— . "-TV |s ro "• T-I
•r—j o • — 0 TO O
O W ^ -O O H
~^ >H — SH SH
'O'D roi — iro0roro
00 d3co0u00
5 o t1 ro" i* § 15 o
00) SIICU00
U3 Qj O TOO 10 Qi
8& ttX2B1SS
4- w 0 3
0 CJ PU rJ
in 13
a
a
in
>£>
•
o
II
0
ro «
CO
X o>
rH
m
ro (5
o c
^* ^i
<* (0
CN C_)
CO t+H
X
0
g 4J
Ui <0
O, 4J
VD
rH
II W
o s
ij O
(0 O)
94
-------
CO
rH
cc
CO D
53 O
H Q5
^ ra
fr( £_(
Q
ft
PS
ft tj
as
Q &•
«2J p
^L H-|
J2
Q
pa s
la
Ed
§g
^*^
Q rd
s i
O LtJ
M
rtj W
t> DQ
Pn ^
H Q
CO >3
° §
fe
CD ^d
CO Cj
K O
^ fci
W
i O
S 2.1
O
CO p^
w u
ft ^
O
0 O
rH M
EH
a
oj in 03 o 'O *o i
0 LO 0 -.H 0 0 g
> . >,jj p> jj a
I o i 03 d o a
S 11 C CO 0 Q)
G -r-iO !u W Qi ? ^
JL ^ C O W r~
. O W 0 03 I
in cj ft ft o
ro ij LO
v »-" ro
ro
rH
O)
OJ rH
^
00
o . o
OJ O O*
ro
rH
ro
00 OJ
ro rH o
CO
ro
ov
o ^
•* o o'
ro
ro
rH
O
rH
cri ro
VO rH O
CO
en
( — I
ro
CO
rH O
r^ o o
ro
r-
rH
rH
O
O
in o o
ro
rH
O
0
,o
0 00*
*•
CO
JJ
03
-8
— u
(LCE=Xj=1.79):
Person-years (Wj
Pancreatic cance
Ctoserved (Oj )
Expected (EQJ)
750+ ppm-years
in
OJ
•
^* rH
IJD
rH
00 rH
•* o o'
r-
rH
rH
in
oj ro
in ro o
o
o
rH
r-
o
U"> 1J3
ro rH o
^*
vo •
rH '
J*
ro rH
OJ O O
rH
rH
rH
00
O
00 O
m o o
rH
O
o
0
0 00
o
o
o
0 O 0
• *
co
5
03
•3
— i-t •—
(LCE=Xj=3.18):
Person-years (W^
Pancreatic cance
Ctoserved (Oj)
Expected (EQJ
•
LO
CO
rH
03
• rJ
1
• H
rH
S
IJ-I
0
1
JJ
CO
95
-------
observation that only 1110 person-years of observation occurred
before the age of 25. This implies that very few men started
work before the age of 25, so that all experience up to age 45
would be within a 20-year latency period. Overall, there were
29,364.4 person-years of experience among the 1013 men for an
average observation (or latency) of 29 years per man. Removing
the experience before age 45 takes away 9337 or 31.8% of the
person-years. However, there were no observed deaths and less
than one expected death for lung or pancreatic cancer in these
age groups. By comparing Tables 9 and 10, one can note that both
the person-years and the LCEs are the same in corresponding
groups, and that for any age group (column), the ratio of
expected deaths from pancreatic cancer to expected deaths from
lung cancer is constant.
6.3.2.4. Results—The results of the analyses, presented in
Table 11, compare MLE and 95% lower- and upper-limit estimates
based on animal (both administered and metabolized dose) and
human cancer experience. As -can be seen, estimates based on the
human pancreatic cancer deaths are very close to those of the
female mouse based on administered dose. The MLE estimates, 1.2
x 10~2 (ppm)"1 and 5.0 x 10~3 (ppm)""1, based on the additive and
multiplicative models, respectively, are slightly higher than thi
MLE of 3.2 x 10~3 (ppm)"1 based on combined lung and liver tumors
in the mouse. The 95% upper-limit estimates based on the Kodak
experience bound the upper-limit estimate of 1.4 x 10~2 (ppm)"1
based on lung and liver tumors in the mouse.
96
-------
Table 11 also compares the results of the Kodak employees'
analyses with estimates based on the metabolized dose in animals.
It can readily be seen that the upper-limit risks predicted from
human lung cancers by either the additive or multiplicative model
are very close to those predicted using metabolised dose from the
animal data. Further, it can be seen that the upper-limit risk
predicted from human pancreatic cancer deaths is greater than any
of the risks predicted from animal data using metabolized dose.
Also included in Table 11 are tests of significance of the
estimates based on the Kodak employees' experience. As can be
seen, the p-values based on pancreatic cancer are borderline
significant (p = 0.05) by the asymptotic normal test. The
p-values based on the likelihood ratio test are p = 0.02 for the
relative risk model and p = 0.04 for the additive risk model.
For lung cancer, none of the p-values are statistically
significant, and in fact, the MLE dose-response estimates are
actually negative as might be expected from the data.
6.4. DISCUSSION
The above analysis shows that the estimates of incremental
unit cancer risk based on the NTP female mouse inhalation study
are about the same as those calculated from an analysis based on
pancreas cancer deaths in the Kodak employees. The larger
question remaining, however, is whether the pancreas cancer
Sfeijpbnse is a true response. Two factors that appear signifidaht
are the concurrent decrease in colon-rectum cancers and the
appearance of most of these pancreatic cancers between the last
97
-------
CO
fcj &]
CO pt3
T~> ^^
is
PJ
iv^ g
12
3 §i
Ccj Q
O
A| jy*
2* CO
MO
15-
62
3 CO
Q
Q Du
ta en
«2 S
tn
5J |H
CJ »rf
p^ Cu
O
DH
O OS
£i Q3
H 3
jrf
§ IS
EH M
CO 04
2 p
e£ 3
03
J 2
*^ O
is
SB
, ^i
fei gj
g£
co Cxj
p^ EC
«£ t-i
& B
^5
u w
* o
<— t &
*~* s
&3 Q5
1
H
j^
03
•H CU
)4 S-l !J
CD o ail
O u O S
ex -a
(0 CD dP r-l
O in
03 CTi
•38
•^3 Tl
C C 00
flj •(-{ C Q rj
S 4-J 3 jC^ (ti
35 C £ *|~' E
MO -3 r-l -M
O CJ X 4 n >-j •-!
(0
CO
1— 1
(0
a
•iH o; w
[ 1 ^ (jj
S ro *
GJ*'™' ^
>1 (C 4J
W > 03
«a; Q)
s-< w
CD CD
4-1 4J
§(rt
5 ,O
S-l 4J
(0 03
PJ CD
S-l
0) CD
o a
i-H
o> -n
TO
0 0
O* 0
oo oo
I i
o o
r-H i— 1
X X
r- o
.-I I-H
•3' •*
I i
0 0
i— 1 r-H
X X
** 00
• •
i— 1 i-H
1
03
(C
B S1
05 D
M-l I—I
CD
jj
•H
TD
1o3
rH
X X
^f O\
• in
|H 4
rH
co
|
0
rH 1
X 1
CN
CO*
o o
0) Q)
03 03
0 0
, , Tj T3
C
O T3 T3
•8§.S
SCO O
•H £)
•5 4J
s d)
'O IS
rtl
"S 'g
<0 C
M -H
3-H Q
rH 0
T3 a,
CD Dl
N fg
•r-l 4J
S-l 03
(0 -r-l
(1) 4->
C rH
'/j 3
to
0)
03 (Q
o g
a y-i
f
Q)
1
iS
fO
Cti
4-J
o
4J
03
Q)
r~\
O
•r-(
S
a
1
05 •
d)
T? 4-J
rH S
fC 4-1
4-1 • 03
1 r- a)
o
S S-l (D
4-1 flj >
c • a 4-1
O O n3 (0
T3 CJ 4)
0) X C
03 03
<0 £> 03 II
(S CO
(T3 O O 2
98
-------
observation date, 1977, and the current one, 1984. The Batelle
.report (1986) suggests that in the earlier Friedlander et al.
(1978) report there was an excess of stomach and gastrointestinal
cancer deaths and that this excess might represent some
misdiagnosed pancreatic cancer deaths. Regardless, Batelle
suggests that the current standardized mortality ratio (SMR) of
250 for pancreatic cancer in a healthy population, an apparent
rapid rise in the last 8 years of follow-up, plus the fact that
no other disease shows such an effect, is, at least, cause for
further investigation.
Hearne et al. (1987) argue the opposite. They suggest that
when "a large number of non-hypothesized causes [are] evaluated,
it is likely that [a significant result might happen] by chance
alone." They point to "deficits of the same magnitude . . . for
such sites as colon-rectum (2 vs. 9.8) and prostate and bladder
(3 vs. 8.0)." They also suggest five other issues which, they
claim, demonstrate the pancreatic cancer observation as a chance
occurrence. i These are: (1) no dose-related effect; (2) no
evidence that DCM or its metabolites concentrate in the pancreas
or produce any other toxic pancreatic response; (3) the histology
for these cancers were adenocarcinomas, the most common
pathologic diagnosis for this site; (4) potential concurrent
exposure to suspected animal carcinogens 1,2-dichloroethane and
1,2-dichloropropane; and (5) each of the eight cases had one or
more non-occupational risk factors [smoking (7), diabetes (2),
and alcohol abuse (1)].
99
-------
While both sides have merit, the HRAC believes that the
above analysis, showing a statistically significant
exposure-response trend for human pancreatic cancer of
0.012 among motion picture
projectionists in Washington State between 1950 and 1971.
However, the results were based on a highly unreliable
proportionate mortality analysis, and the study design allowed
for no additional analyses. Furthermore, there were no exposure
measurements.
Finally, the argument of whether or not the pancreatic
cancer response is a bona fide response has a bearing on the
quantitative risk estimation. If the evidence for pancreatic
100
-------
cancer in humans is considered limited and DCM is classified as a
Bl carcinogen according to EPA's classification scheme, then the
cancer risk estimate would be based on the MLE, 1.2 x 10~2
(ppm)"1 or 5.0 x 10~3 (ppm)""1 for the additive or multiplicative
model, respectively. Both of these estimates are slightly lower
than the upper-limit estimate based on the administered dose in
the NTP study. However, they are both slightly higher than the
upper-limit estimates based on the metabolized dose in the NTP
study. If the pancreatic cancer response is considered possibly
a chance occurrence and DCM receives a B2 classification, then
the study would be used only to get the 95% upper-limit estimate
on a negative-response study, and the risk estimates would be
even higher. Thus, whether DCM is classified as a Bl or a B2
affects the quantitative risk estimates but in an inverted
manner. This all, of course, assumes that the control population
of all Kodak hourly employees is a proper comparison group.
6.5. SUMMARY
A quantitative risk extrapolation based on pancreatic cancer
deaths in Eastman-Kodak employees exposed to an 8-hour time-
weighted average of 30 to 125 ppm DCM yields 95% upper-limit
estimates in the range of 9.9 x 10~3 to 2.5 x 10~2 (ppm)™1. This
range (Table 11) exceeds the proposed EPA upper-limit estimate
derived from lung and liver tumors in the NTP female mouse
inhalation study (based on metabolized dose). Furthermore, a
test for exposure-response trend yields borderline statistical
significance, additional evidence that the pancreatic cancer
101
-------
response in the Kodak employees is exposure-related and not a
chance occurrence. At this time, however, the HRAC believes that
the evidence is not strong enough to determine if this response
is bonafide and can merely conclude that the estimates based on
animal cancers do not appear to overestimate the risk. Even if
pancreatic cancer deaths are discounted, the 95% upper-limit
estimates based on lung tumors from the epidemiologic analysis
(Table 11) would be between 1.1 x 10~3 and 2.7 x 10"~3 (ppm)"1,
which are comparable to the upper-limit estimates derived from
the mouse metabolized dose data.
102
-------
7. SCALING RISK ACROSS SPECIES USING DELIVERED DOSE
7.1. INTRODUCTION
Using experimental animal exposures to putative carcinogens
as a means for estimating risk to humans relies on the general
similarity of mammalian species in anatomy, physiology, and
biochemistry. The biological processes that underlie
carcinogenicity are not well understood, but it is reasonable to
suppose that they will operate in a more or less similar manner
in rodents and humans, at least in most cases. Although these
processes may be supposed to be qualitatively similar, the rates
at which they proceed will vary among species, leading to
differences in the carcinogenic potency of a substance.
Much of the variation in the rates of the underlying
processes will be the result of the pronounced difference in
scale between humans and experimental animals. Humans are some
2000 times heavier than mice, they live about 35 times longer,
and their physiological processes operate at a generally slower
rate. The amount of a carcinogenic compound that engenders equal
lifetime cancer risks in mice and humans clearly should depend on
such differences. If one can properly take into account
differences in scale, one can quantitatively predict human risks
from observations on the doses that produce certain risks in
animals.
Much controversy exists over the scaling factor that ought
to be applied; different factors are preferred by different
103
-------
federal regulatory agencies. FDA generally scales doses by body
weight; doses are considered to be of equivalent lifetime
carcinogenicity when the daily rate of dosing is equal in units
of mg/kg. EPA and CPSC scale doses by body surface area. That
is, doses are equivalent in units of mg/kg2/3/day, since surface
area is proportional to (body weight)2/3. (Both of these factors
adjust for life span in, that equivalent lifetime risks are based
on daily exposure rates, even though humans are exposed for 35
times more days in a lifetime.)
The extrapolation of potencies of carcinogens using such
scaling factors has some empirical support. For some chemicals,
direct estimates of risk from defined exposures are available
from epidemiologic data. Such estimates agree (at least
generally) with projections based on scaling the potencies
estimated from animal studies (National Academy of Sciences,
1975). Crump et al. (1985) and Allen et al. (1986) have shown
for 23 "chemicals" (including industrial chemicals, drugs, and
cigarette smoke) that the directly observed human cancer risks
are well predicted from animal studies if the applied doses are
scaled by some measure of body size and life span. Their data
have insufficient resolution to clearly favor body weight or
surface area as a scaling factor, however. Two points of caution
should be noted. First, the chemicals in the comparison are ones
producing rather high human risk, high enough to be estimated in
epidemiologic studies. Thus, the extrapolation is from
experimental animals at high bioassay doses to humans usually at
104
-------
similarly high doses. Second, there is a good deal of variation
among chemicals which reflects on the accuracy of the
extrapolation; there is considerable room for improvement in
prediction, especially in the area of "explaining" the outliers
to the general pattern. Nonetheless, these studies are
reassuring in their .suggestion that current risk extrapolation
processes are not completely ill-founded.
The question addressed in this chapter is how information on
the internal target-organ doses of a putative carcinogen should
affect the process of extrapolating risks observed in animals to
humans. Extrapolation on the basis of applied or external dose
can be criticized for failing to take into account species
differences in metabolism and disposition, as well as for
ignoring the changes in these factors from high doses to the low
levels of exposure for which human risk estimates are usually
/
desired. If information is available on internal doses in the
tissues subject to carcinogenic induction, a dose measure much
closer (in terms of the underlying causal processes) to the
biological end point of concern can be examined. A discussion
about the relationship of such data to the scaling process
requires an examination of the scaling process as it is currently
used on applied doses.
7.2. SCALING APPLIED DOSE TO EXTRAPOLATE RISK ACROSS SPECIES
The applied dose basis for risk extrapolation makes no
explicit attempt to specify the underlying biological
processes/factors that mediate the relationship between the dose
•' ' - 105 '•••'":•.
-------
and the ultimate response. The same set of underlying processes
is presumed to operate in experimental animals and in humans; the
scaling factor is employed to account for the alteration in
potency that results from the different values for factors
affecting the underlying processes in larger, longer-lived humans
compared to rodents. The scaling factor has to account for the
overall scaling of the whole relationship of applied dose to
response, even though that relationship comprises a large number
of components. These components include the pharmacokinetic
processes of absorption, distribution among the tissues,
metabolism, and excretion, all of which will affect the degree to
which the internal dose at the site of action remains
proportional to the applied dose. Other components are those
that affect the tissue's sensitivity to this internal dose, and
include such factors as the cells' ability to scavenge free
\
radicals, rates of DNA repair and cell turnover, the number of
cells at risk, immunosurveillance, and others. The general
empirical success of the scaling factors usually used by the
federal regulatory agencies, discussed previously, argues that
the combined effect of differences in all of the underlying
factors is well estimated in many cases. However, no single
factor can be identified as the key to explaining the
interspecies differences in a carcinogen's potency, nor is the
specific contribution of any one element to the overall scaling
effect identifiable.
If pharmacokinetic data on humans and animals were available
106
-------
on a number of the chemicals considered by Allen et al. (1986),
one could compare the internal dose differences between species
to the observed differences in the potency of carcinogens,
yielding an estimate of the contribution of differences in
pharmacokinetics to the scaling of potency across species. Such
data are not currently available, however. Information on this
question is being compiled by EPA with the hope of providing
insights that will aid in interspecies extrapolation.
In order for pharmacokinetic models or data on internal
doses to alter the risk extrapolation based on applied dose, one
must replace,the contribution of pharmacokinetics that has been
assumed as part of the general scaling across species with the
particular data for the compound at hand. In order to do so, one
must make explicit the assumptions about the contribution of
pharmacokinetics to interspecies scaling that one assumes in the
applied dose extrapolation procedure. (Lacking data on the other
elements contributing to sensitivity differences among species,
one must assume that their contribution is the same as in the
applied dose extrapolation.) If other pharmacodynamic factors
are known or assumed based on appropriate evidence, the degree to
which the data show that pharmacokinetic differences between
humans and experimental animals do not follow the assumed pattern
is the factor by which the risk extrapolation should be altered.
Siiftgiy showing that species differences in pharmacokinetics exllfe
does not help in deciding how to change human risk predictions,
since the observed differences may be in accord with the general
107
-------
scaling pattern between species that the applied close scaling
factor already encompasses.
The process of apportioning expected species differences in
a carcinogen's potency between pharmacokinetic factors and other
pharmacodynamic factors is difficult, and is discussed in the
following section.
7.3. PHARMACODYNAMICS
Pharmacodynamics is "what the dose does to the body."
Although pharmacodynamics can be defined to include
pharmacokinetic considerations, it is defined for purposes of
this discussion as including the biological processes that govern
the target tissue's degree or probability of response to a given
delivered dose, which may or may not be proportional to that
delivered dose.
Our ignorance of pharmacodynamics is, if anything, greater
than our ignorance of pharmacokinetics, especially when cancer is
the end point. As in the case of pharmacokinetics, we ought to
consider how pharmacodynamic differences from high to low doses
and from species to species affect our ability to extrapolate
risk from animal bioassays to human experience. Generally, if a
carcinogen acts by enhancing or accelerating an ongoing
biological process involved in the production of "background"
tumors, we expect the carcinogenic potency of a delivered dose
(at least at low levels) to be directly proportional to (i.e., to
vary linearly with) that dose. However, a number of biological
processes can lead to marked nonlinearity in the curve of
108
-------
magnitude of effect against delivered dose. For example, the
swamping of DNA repair at higher doses of an agent could cause
the dose-effect curve to be convex (leading to lower than
expected risks when extrapolated to low delivered doses). A
carcinogen that must induce cytotoxicity to cause cancer could
have a very convex dose-effect curve. On the other hand, if low
doses of an agent fail to induce DNA repair enzymes, a somewhat
concave dose-effect curve would result, and low dose risks may be
underestimated. Although many possible nonlinear effects can be
listed, their actual importance in real situations is currently
unknown.
Clearly, different hypotheses about the mechanism of
carcinogenic action profoundly affect the process of high- to
low-dose extrapolation, even after pharmacokinetic non-
linearities are accounted for. One can make the assumption (if
data are lacking) that within-species pharmacodynamics is linear;
that is, the effect is directly proportional to the delivered
dose. It is well to realize, however, that this assumption
introduces uncertainty, and could be somewhat anti-conservative
in some cases. On the other hand, risks from an epigenetic
carcinogen that has little effect below a certain threshold will
be overestimated.
The effect of pharmacodynamics on interspecies extrapolation
of risk is a good deal more problematic.- There are many factors
that differ between rodents and humans that can be expected to
influence the degree of toxic reaction to a given tissue-level
109
-------
exposure. In the case of carcinogenicity, these include the
slower rate of cell turnover in humans, the greater cellular
defenses against free radicals and other highly reactive
compounds, and possible differences in the efficiency of DNA
repair. One major factor is that humans have many more cells in
any given organ than do rodents, each of which will be at equally
increased risk of carcinogenic transformation when exposed to a
given tissue-level concentration of carcinogen. Since only one
cell need be transformed to initiate a tumor, all else being
equal, humans ought to be many times more sensitive to a given
tissue concentration of a carcinogen than are rodents.
Clearly/ all else is not equal. Interspecific factors that
tend to increase human susceptibility to carcinogens are to a
large degree balanced by other factors that tend to decrease
susceptibility. The key question for interspecies extrapolation
of carcinogenicity, which remains unanswered at the present time,
is exactly where this balance is struck in humans vis-a-vis
rodents. In other words, we do not yet know what to use for an
"interspecies pharmacodynamic correction factor" to account for
interspecific differences in the carcinogenic response of tissues
to a given delivered dose.
The following section discusses the "surface area
correction" as a means of scaling doses for interspecies
extrapolation, in light of the uncertain contributions of
pharmacodynamics and pharmacokinetics to the overall scaling of
carcinogenic potency across species.
110
-------
7.4. PHARMACOKINETICS, PHARMACODYNAMICS, AND THE SURFACE AREA
CORRECTION
The surface area correction is the interspecies correction
factor by which human applied doses (in mg/kg/day) are modified
in order to be equivalent to animal doses (in the same units)
under the EPA and CPSC assumption that doses are of equal risk
when expressed in terms of mg/kg2/3/day. Humans have a smaller
surface area to volume ratio than do rodents, since surface area
is approximately proportional to the 2/3 power of body weight,
while volume is approximately proportional to weight. Hence, an
equal dose per kg2/3 will be less per kg in humans. In
extrapolating from mice weighing 0.0345 kg [which may be used for
the weight of female mice in the NTP (1985) inhalation bioassay
t
of DCM] to 70 kg humans, the human doses in mg/kg are divided by
the cube root of the ratio of body weights, or a factor of 12.7.
In general terms, the overall interspecies correction factor
for applied dose (F, which is equal to 12.7 in this case) must
include the effects due both to interspecies pharmacokinetic
differences (PK) and to interspecies pharmacodynamic differences
(PD)t both of which are usually unknown. One can represent this
as a simple equation,
F = PK x PD
where F is the multiplicative factor that relates human risk to
mouse risk at the same applied dose in mg/kg/day, PK is the
111
-------
factor that relates human delivered dose to mouse delivered dose
at a given applied dose, and PD is the factor that relates human
response at a given delivered dose to the mouse response at that
same delivered dose. (To the extent that PK and PD display
within-species nonlinearities, their values may change at
different levels of applied dose; this phenomenon is ignored for
the time being.)
When pharmacokinetic data are available, one must consider
replacing the assumed value of PK with an empirical determination
of this factor, PK* (where the "*" distinguishes an observed
value from the a priori assumption). The new interspecies
extrapolation factor F* (and hence the new estimate of potency in
humans) would thereby be altered by a factor PK*/PK. The
question is, what is the assumed value of PK?
7.4.1. Assuming that the Surface Area Correction Accounts for
Pharmacokinetics
One approach is to take the case of the surface area
correction factor of 12.7 between mice and humans (and hence the
interspecies differences in a carcinogen's potency) as being
entirely due to differences in pharmacokinetics (i.e., assume
that PK «• 12.7). A given tissue-concentration is taken to be
equally toxic in all species (i.e., PD - 1). The rationale for
this approach comes from the common pharmacokinetic observation
that the volume of distribution of a compound tends to scale
across species in proportion to body weight, while the rate at
which the compound is cleared from that volume (by metabolism
112
-------
and/or excretion) tends to scale as surface area.. As a result,
the concentration in the blood attenuates more slowly in larger
animals, resulting in a larger area under the concentration-time
curve, and presumably a larger biological effect (e.g.,, Dedrick
et al., 1970). This approach is commonly used in scaling up
doses of experimental drugs from mice to humans. Under this
view, the reason that humans are assumed to be 12.7 times more
sensitive than mice to a dose in mg/kg is that they experience an
internal dose (area under the curve) that is 12.7 times higher.
If actual pharmacokinetic data for a certain compound showed
that, contrary to this expectation, human internal doses were,
say, one half as large as in mice following equal applied doses
(i.e., PK* is observed to be 0.5), owing perhaps to some
idiosyncracies of the compound's metabolism in humans, then this
knowledge should prompt the lowering of the estimated human risk
by a factor of 12.7/0.5 =25.4.
There are reasons to question whether the basis cited for
this view of the surface area correction (the volume of
distribution versus clearance argument) applies to carcinogens,
however. Many carcinogens seem to require metabolic activation
from a comparatively innocuous parent compound to a highly
reactive metabolite. Thus, the concentration of parent compound
and the rate of its clearance from the body may not be directly
relevant to carcinogenic potency. For example, if the two means
6f clearance of a parent compound from the body, excretion and
metabolism, both scale between species in proportion to surface
113
-------
area, their ratio would be fairly similar in rodents and humans.
Although human metabolism may be slower, the loss of parent
compound by excretion is also slower to about the same degree.
In other words, the same proportion of the dose may be
metabolized for the compound in mice and humans, albeit over a
longer time scale in humans. In this case, rather than 12.7, PK
would be 1.0 between mice and humans, that is, the applied dose
and internal dose are in the same proportion in different
species. Of course, if metabolism and non-metabolic clearance do
not remain in the same proportion, PK would assume some other
value (for example, CEFIC studies are interpreted by the authors
as showing that mice metabolize DCM by the GST pathway at least
60, rather than 12.7, times as fast as humans).
7.4.2. Assuming that the SurfaceArea Correction Accounts for
Pharmacodynamics
If internal doses are assumed, as in the above example, to
be generally proportional to applied doses, even across species,
and if the overall extrapolation factor F is still assumed to be
given by the surface area correction of 12.7, then the correction
is being assumed to be due solely to species differences in the
sensitivity of the tissues to carcinogenic transformation by a
given internal dose, i.e., PD is assumed to be 12.7. A possible
explanation for sensitivity scaling as surface area could be that
some of the components -of variation in tissue sensitivity, such
as cell division and turnover rates, DNA repair rates, scavenging
of free radicals, and so on, are related to tissue aging rates
114 /
-------
and to life span. Boxenbaum (1983, 1984) relates life span and
aging to different scales of "physiological time" among species,
which tends to vary in proportion to body surface area. On the
other hand, factors such as number of cells at, risk would not
scale" in this fashion. A further complication is how to
interpret the effect of a number of components each possibly
scaling by surface area. Taken together, all of the components
could lead to a factor of 12.7; alternatively, a multiplicative
solution might be needed. For example, if each of the three
components just listed caused a 12.7-fold effect, the net effect
would be 12.7*12.7*12.7, but if number of cells caused an
opposite effect of 12.7*12.7, the result would be a 12.7-fold
effect. Obviously, the value of PD for any compound is subject
to many considerations.
If, however, it is assumed that the surface area correction
is applied in this way (PD = 12.7), it is a correction not on
dose, but on risk from a given internal dose. The hypothetical
compound mentioned above, for which pharmacokinetic data showed
PK* as 0.5, would, under this method, have only half of the
pharmacokinetic difference expected by this view of the surface
area correction to applied dose. Thus, the risk estimate would
be lowered by a factor of 1/0.5 = 2 (instead of by a factor of
24.5, as in the original example). When the surface area
correction is viewed as a correction for tissue sensitivity
differences across species, internal doses must be adjusted by
the surface area correction just as applied doses are, because
115
-------
the difference in sensitivity of the tissues between species is
not addressed by pharmacokinetics.
7.4.3. Other Possible Assumptions for PD and PK
If there is an unusual pharmacokinetic pattern for a
compound (i.e., PK* is shown to be different from its expected
value, but PD retains the appropriate expected value), correction
for species-to-species extrapolation should be made. A study of
many compounds for which there are animal and human risk data,
the purpose of which is to obtain estimates of PK*, may lead to
clues regarding the values of PK and PD for this spectrum of
compounds. If pharmacokinetics, for example, seems to always
lower risks using 12.7 for PD as in the above example, the
empirical success of applied dose scaling (which places some
constraint on PK x PD) suggests that human sensitivity, PD, is
perhaps being underestimated by the surface area correction.
One could start from a different set of assumptions about PK
and PD. One cannot regard the use of pharmacokinetic information
as replacing the need for assumptions about the "usual" scaling
across species, since the scaling can always be determined
empirically. This can only be done if a value of PD, the
sensitivity differences between species, can be settled on. It
is important to note that if one makes the decision to use
pharmacokinetic data for species-to-species extrapolation, one
cannot extrapolate across species using internal dose without
making an assumption as to the value of PD. The approach
described above is one attempt to show a derivation for PD.
116
-------
Obviously, there is a large uncertainty in its value for a given
chemical, and PD may vary considerably for other chemicals. When
comparing risks extrapolated to humans on the basis of applied
dose to risks extrapolated using internal or delivered dose, one
must use the same value of PD (unless independent information on
the biological basis of PD is also adduced to suggest that it too
should be changed).
7.4.4. High- to Low-Dose Extrapolation
The preceding discussion focussed on the interspecies
extrapolation, temporarily ignoring the fact that both PK and PD
can change as a function of dose level. In the case of the
pharmacodynamic factor, PD, the changes of response with internal
dose are estimated by fitting a dose-response curve. The methods
for doing this, and the assumptions that must be made, are
essentially the same as when the dose measure is applied dose
(the principal difference is that any influence of
pharmacokinetic nonlinearities on the applied dose versus
response curve is not falsely ascribed to pharmacodynamic
differences with dose). These basic risk assessment methods are
well discussed in federal regulatory agency risk assessment
documents.
The relationship between the externally applied dose and the
subsequent internal dose at the site of toxic action can vary
with dose level because of saturation of metabolism, action of
different biochemical pathways at different substrate
concentrations, nonlinear patterns of binding of the compound,
117
-------
and other factors. The assumption made by the applied dose
procedure that similar biological processes underlie the action
at low doses of a compound in animals and humans may be violated
because the experimental animals are exposed to much higher dose
levels. The equation describing the component assumptions of the
applied dose extrapolation can be modified to
P - (PK x HL) x PD,
where the new factor, HL (for "high-to-low"), is the change in
the ratio of human applied dose to internal dose between high
doses (as high as rodents in the bioassay) and the low doses
usually characteristic of the human exposures for which risk
estimates are desired. In the applied dose procedure, HL is
assumed to be 1; i.e., the possibility of such changes is
ignored.
As a general principle, even very nonlinear pharmacokinetic
systems usually approach linearity at low doses, below the level
at which saturation effects are important. [The model used by
Andersen et- al. (1986, 1987) for instance, predicts GST
metabolism that varies almost linearly with applied dose below
about 10 ppm.] This is important because it means that the high-
to low-dose extrapolation is the same for most exposure levels of
concern (since they are all in the low-dose region), allowing a
unit risk to be used in risk calculation. In other words, it is
possible to calculate an observed value for the high- to low-
118
-------
dose correction, HL*, which can replace the assumed value of
HL = 1. Whether or not species-to-species extrapolation is
employed, the HRAC agrees that nonlinearities in the dose-
response curve due to pharmacokinetic differences must be
accounted for in the risk assessment process, if the pertinent
data are available.
HL is separable from PK and PD, and unlike the latter,, has
an assumed value for high- to low-dose extrapolation in the
applied dose procedure—it is clearly assumed to be equal to one.
Therefore, it is possible to incorporate pharmacokinetic
information relating only to dose level into quantitative risk
assessment. This method holds that the current lack of
understanding of interspecies differences in tissue sensitivity
(PD) precludes using internal doses to extrapolate across
species, but assumes that the results of pharmacokinetic
investigations can be used for high- to low-dose extrapolation;
the pharmacodynamic factor is held constant within humans.
The extrapolation between species is done using the applied
dose method. PK and PD need not be individually specified, only
their product is fixed by whatever assumption the risk assessor
has traditionally used (e.g., mg/kg/day or surface area
correction). This method relies on the empirical success of
applied dose extrapolation (Crump et al., 1985; Allen et al.,
1986) discussed previously. These papers examine extrapolation
only to high human doses, and in such cases the applied dose
scaling factors work fairly well. The pharmacokinetic
119
-------
information on humans is then used to define an observed non-
proportionality (HL*) of applied and internal doses when high
doses are compared to low doses. The extrapolation' factor F is
then adjusted to equal PK x PD x HL*. This method accounts for
nonlinearities in internal dose across exposure levels that arise
from dose-dependent changes in absorption, disposition,
excretion, and saturation or other changes in metabolism.
However, it forgoes modifying the interspecies component of
extrapolation on the basis of pharmacokinetics. New data
changing the estimate of interspecies pharmacokinetic
differences, PK*, will not result in different risks under this
procedure.
The preceding pages have laid an extensive groundwork on the
various factors that must be borne in mind when using
pharmacokinetic information in the extrapolations inherent in
quantitative risk assessment. The practical value of drawing the
distinctions outlined above will be demonstrated in the next
section, in which an analysis is presented of the interpretation
by Andersen et al. (1986) of the results of their pharmacokinetic
model for DCM vis-a-vis the EPA (1985b) and CPSC (1985) risk
assessments.
7.5. THE PHARMACOKINETIC MODEL USED BY ANDERSEN ET AL. FOR DCM
Andersen et al. (1986) submitted to EPA and CPSC a report
(which was subsequently published in 1987) on a physiologically
based pharmacokinetic model along with interpretations of its
results that, according to these authors, show the two agencies'
120
-------
risk assessments to overestimate human risks from inhalation by
167-fold for liver tissue and 144-fold for lung tissue. The
model used by Andersen et al. (1986, 1987) is a modification of
an earlier model developed to describe the disposition of styrene
(Ramsey and Andersen, 1984). The model, its merits, and its
shortcomings have been discussed in Chapter 2. Here the intent
is to analyze the contention of great overestimation of risks
when extrapolation from mice to humans is done on the basis of
applied dose instead of on internal doses as provided by the
model. The model is used as presented by Andersen et al. (1986).
The HRAC finds that the differences between risk estimates
derived from EPA/CPSC's applied dose method and from using
internal doses from the model may be interpreted as being much
smaller, only a few fold. The disagreement stems from two
factors: (a) breathing rates, and (b) the appropriate
application of the surface area correction to dose, as discussed
previously.
7.5.1. Breathing Rates
In EPA's Addendum to the Health Assessment Document fotf SGM
(1985b), and CPSC's risk assessment (1985), applied doses are
calculated as the amount of DCM breathed in per kg of body weight
per day, estimated using empirically-based breathing rates
(m3/day) and assuming 100% absorption. Because larger animals
breathe less air per unit of body weight, humans receive a
smaller applied dose from a given exposure to a certain vapor
concentration than do mice. The model used by Andersen et al.
121
-------
(1986, 1987) incorporates the assumption that humans have a
smaller input of DCM per kg than mice, but the model uses a set
of breathing rates different from EPA's and CPSC's. The model's
>:
value for human breathing rate (12.5 m3/day) was measured for a
man at rest, and is consequently much lower than EPA/CPSC's
assumption (20 m3/day) based on average daily activity. When the
model is used in assessing risks from actual human exposures, its
parameters should reflect normal human activity levels.
Furthermore, the model's breathing rate value for mice
(0.084 m3/day) is much higher than EPA/CPSC's assumption (0.043
m3/day). Andersen et al. (1986) compared the results of their
model to EPA's procedure without accounting for the fact that the
two methods use different breathing rates, both for mice and for
humans. One may debate about which set of rates is most
appropriate, but the same set of values ought to be used when
comparing the two methods. Since the breathing rate values
appear both in the model and in the applied dose calculations,
using the model's rates throughout implies a smaller human
applied dose (and a bigger mouse applied dose) than the EPA/CPSC
procedure does, but these applied doses are the ones that are
truly associated with the model's estimates of internal dose.
When the comparison is made in this way, using the Andersen et
al. breathing rates in both procedures, the non-proportionality
between applied dose and internal dose reported by Andersen et
al. (1986) is reduced by a factor of 3.1, reflecting the
different assumptions in the model and in the EPA/CPSC procedure
122
-------
as to the ratio of mouse to human breathing rates. (The change
in breathing rates in the EPA/CPSC procedure lowers the applied
dose estimate, which is then less different from the model's
results. If the EPA/CPSC breathing rates are used in both the
model and the applied dose calculations, however, the model's
results are not changed by 3.1-fold, but by a smaller amount,
especially in mice, owing to nonlinear effects within the model.
In a following section on developing a unit risk based on
internal dose, the EPA/CPSC rates are in fact used.)
When the same breathing rates are used, the difference
between the model and the applied dose procedure are not 167-fold
and 144-fold for liver and lung, respectively, but rather 54-fold
and 46-fold.
7.5.2. Using the Surface Area Correction
The second difference that the HRAC has with the analysis of
Andersen et al. (1986) is in the use of the surface area
correction on applied dose. Andersen et al. follow the first of
the two interpretations of the surface area correction, as
disetissed in the earlier section. That is, they attribute t®
EPA/CPSC the assumption that, since the overall scaling factor
for applied dose extrapolation across species is F = 12.7, the
expected pharmacokinetic difference between mice and humans is
PK.» 12.7 (with PD = 1).
The model used by Andersen et al. (1986, 1987), using their
"optimized" values for DCM kinetic constants (see previous
sections for a discussion of the sensitivity of this procedure),
123
-------
shows that the internal dose (metabolism by the GST pathway per
liter of tissue) is not 12.7-fold greater in humans than in mice
at the same applied dose, but is in fact somewhat lower in
humans. In terms of the notation developed previously, the.
observed difference in the ratio of applied to internal dose when
comparing mice at high doses to humans at low doses is PK* x HL*
=4.3 for liver tissue, while in lung tissue the value is 3.6.
[In liver, the interspecies extrapolation actually shows the
applied dose method to underestimate the internal dose somewhat
at 4000 ppm (PK* - 0.60), but HL* is 7.2, leading to a combined
interspecies and high- to low-dose comparison of 4.3-fold; in
lung, PK* is about 1.6 and HL* is 2.2.]
If the observed deficit of internal dose in humans vis-a-vis
mice of 4.3-fold for liver is compared to the expectation of an
excess of 12.7-fold, then the applied dose procedure
overestimates internal dose (and therefore risk) by 12.7 x 4.3 =
54-fold. In lung the calculation is 12.7 x 3.6 = 46-fold. [If
one adds back in the 3.1-fold inflation due to comparing
different breathing rates, which is really extraneous to the
ddfrigarison of applied to internal dose, then the factors arfe I&71
and 144, as Andersen et al. (1986) originally reported them.]
If, on the other hand, one takes a second interpretation of
the surface area correction—that the correction is to account
for interspecies differences in sensitivity of the tissues to the
internal dose as explained in the example above—then the assumed
values of PK and PD are not 12.7 and 1, but rather 1 and 12.7,
124
-------
respectively. The 4.3-fold deficit of internal dose in human
liver vis-a-vis high-dose mice is thus compared to an expectation
of 1. The change in the risk calculations is thus not 54-fold,
but 4.3-fold, when internal doses are used in place of applied
doses. The factor for lung is 3„6-fold, by the same argument.
Since PD is assumed to be 12.7, the surface area correction is
still applied to doses, even though they are internal doses, to
correct for the anticipated greater sensitivity to those doses in
humans compared to mice. If another value (other than 1.0 or
12.7) is assumed for PD, results would vary accordingly—one can
see how important an assumption for PD is.
If one does not assume PD, but uses the high- to low-dose
information only for incorporation of pharmacokinetic data into
the risk assessment process, the difference between the Andersen
et al. proposal and EPA/CPSC applied dose estimates is 7.2 or 2.2
for liver and lung, respectively (i.e.,-HL*).
The comparison by Andersen et al. assumes that the sole
reason for applying the surface area correction factor of 12.7 to
^
applied doses when extrapolating across species is to adjust for
an'tidipated differences in pharmacokinetics between species«
HRAC feels that it is strongly arguable that the surface area
correction is not a correction on dose to allow for
pharmacokinetics, but rather a correction on risk to allow for
many factors, including pharmacodynamics. The HRAC's
interpretation of the results of the model used by Andersen et
al. (1986, 1987) is that, although it indicates some
1 125
-------
overestimation of internal dose of GST metabolism in humans
compared to the level in mice in the NTP bioassay, this effect
can easily be only a few fold.
Nonetheless, it is important to judge the impact of such
results on the DCM risk assessment process. The following
section develops a unit risk for DCM based on extrapolating risks
from mice to humans on the basis of GST metabolism in liver and
lung tissue.
7.5.3. Developing a Unit: Risk Based on Internal Dose;
Incorporation of High- to Low-Dose Differences and
Species-to-Species Differences
This section outlines the development of a unit risk based
on the amount of metabolism by the GST pathway, as estimated by
the model used by Andersen et al. (1986, 1987). It was assumed
that PD = 12.7, as outlined in the above example, i.e., that the
carcinogenic potency of DCM in corresponding organs of mice and
humans is assumed to be equal when the daily amount of metabolism
by the GST pathway per liter of target organ tissue (the
"internal dose") is scaled by body weight to the 2/3 power, which
is proportional to the relative surface area to volume ratio of
mice and humans. This is accomplished by dividing the output of
the model for humans by a factor of 12.7 before risks are
calculated to account for a presumed difference between mice and
humans in tissue sensitivity to a given tissue-level dose.
(Surface area scaling could be based on organ weight rather than
on body weight, but as organ weights are nearly a constant
126
-------
proportion of body weight, this would make little difference in
the eventual unit risk, a difference of only 3% greater for lung
t
and 8% greater for liver.)
Lung and liver do not generate equal amounts of metabolites
during exposure, so it is necessary to extrapolate risks for each
organ separately. (This exercise does not imply that the HRAC's
position is that humans and animals must respond at similar
sites. The HRAC provides this example for methodological
illustration; this problem is discussed further below.) Separate
incidence rates of lung and liver tumors in female mice (the more
sensitive sex) were drawn from the NTP (1985) bioassay. For
hepatocellular adenomas and/or carcinomas the incidences of
tumor-bearing animals were: 3/45 in the control group, 16/46 in
the 2000 ppm group, and 40/46 in the 4000 ppm group; for
alveolar-bronchiolar adenomas and/or carcinomas the incidences
were 3/45 among controls, 30/46 at 2000 ppm, and 41/46 at 4000
ppm. Animals dying before the appearance of the first tumor
(which occurred in week 68 for both tumor types) have been
eliminated from the denominators, since the capacity of these
individuals to develop tumors is not fully tested.
<;
Internal doses were estimated using the model used by
Andersen et al. as presented in Andersen et al. (1986), with th@
following exception: the HRAC feels that the federal regulatory
agencies' long-standing assumptions about breathing rates better
reflect the general activity levels of both mice and humans than
do the rates used in the model as presented. The mouse and human
127
-------
breathing rates used in the model were adjusted accordingly.
Cardiac output was adjusted by the same proportion as the
breathing rates. These changes represent less a modification of
the model than a change in the assumption about the activity
level at which the model is to estimate internal doses.
The modified mouse pharmacokinetic model was used to
estimate internal doses under the exposure regime of the NTP
bioassay, that is, 6 hours a day at 2000 or 4000 ppm. The
results showed that virtually no DCM remained in the body after
the 18 hours following one exposure and preceding the onset of
the next day's exposure, so accumulation of compound in the body
over time is not an issue in this case. The internal doses
resulting from a day's exposure were multiplied by 5/7 to provide
an average daily internal dose (since dosing was for 5 days per
week). The resulting values are 727.8 and 1670 mg/L for liver
and 111.4 and 243.7 mg/L for lung in the low- and high-exposure
groups, respectively.
These internal doses, along with the corresponding tumor
incidences, were then used to construct dose-response curves
based on the multistage model procedure, using the computer
program GLOBAL86 (Howe et al., 1986). The liver data were fittt&
by a two-stage model, while a one-stage model was fitted to the
lung tumor data. (The number of stages was chosen as that number
from 1 to 6 leading to a model with the minimum q*, the 95% upper
bound on the estimated linear term. Since qj is used as the
basis for low-dose extrapolation, all alternative possibilities
128
-------
for the number of stages in the model would result in higher risk
estimates.)
The extrapolation of the observed risks at high doses to
those estimated at low doses was accomplished using the 95% upper
bound on the fitted curve. This corresponds to choosing the
curve with the largest linear component (which will dominate the
low-dose shape of the curve) subject to the constraint that the
curve fits the data reasonably well in the observed range. This
process reflects the difficulty in determining the shape of the
dose-response curve at low doses. Most conceptions of the
carcinogenic process lead to the expectation that, at worst, risk
should decrease in direct proportion to decreasing dose at
sufficiently low doses. If the true dose-response curve is
convex at low doses, actual low dose risks will be less, perhaps
much less, than those estimated by the upper-bound curve.
Next, the human pharmacokinetic model was used to estimate
the internal doses from a continuous exposure to a low dose
[1 ppm, as used by Andersen et al. (1986)]. The rate of
metabolism by the GST pathway under such an exposure is 0.07011
mgVVday in liver and 0.008386 mg/L/day in lung. These internal
doses were modified by the surface area scaling described
previously, and the risks that result from such tissue-level
exposures were estimated from the curve of internal dose versus
response developed from the female mouse tumor data. These risks
give the lifetime probability of developing cancer from
continuous exposure to 1 ppm, under the stated set of
129
-------
assumptions. Because internal dose varies directly with vapor
concentration at such low exposure levels, and because (by
assumption) risk is taken to vary directly with internal dose,
the estimated risks at 1 ppm can be used as unit (or incremental)
risks expressed in units of vapor concentration. When these unit
risks are converted to units of (ug/m3)"*1, they are 1.34 x 10~7
for liver and 3.33 x 10~7 for lung.
The risk based on mouse lung tumors is a bit over twice that
based on mouse liver tumors. In fact, however, the risk in lung
tissue per unit of internal dose is about 20-fold higher than for
liver tissue, but for any given exposure to DCM vapor, the liver
has a much higher internal dose, owing to its greater metabolic
activity, so that overall risks in the two tissues are nearly
comparable. Small errors in the model's allocation of GST
metabolism between liver and lung could have large consequences
on risk estimates, since metabolism in the lung evidently
engenders much more risk.
The above phenomenon illustrates a difficulty that arises in
ihterspecies extrapolation when tissue-specific doses and risks
are calculated. If one expects strict site concordance across
species, then the liver and lung unit risks estimate the organ-
specific risks in humans. But the tissue-specific extrapolations
run into difficulty when they are used to predict an overall
level of human cancer risk, which may be manifested in other
organs beside liver or lung. For example, it is problematic to
arrive at a prediction of possible risk to human pancreas, given
130
-------
that there is no internal dose estimate for this organ, nor any
information on tissue-specific risk from a given internal dose
(which can vary considerably, as shown by the mice).
In mice, the occurrence of liver and lung tumors is
independent; developing a tumor at one site does not affect the
probability of developing a tumor at the other site. If this is
also true in humans, then overall estimates of total cancer risk
from DCM exposure are given by the sum of the liver- and lung-
based risk extrapolations. Thus, one can arrive at an overall
unit risk by simply adding the tissue-specific unit risks.
Adding the individual unit risks yields a human unit risk for
continuous inhalation exposure to 1 mg/m-* of 4.7 x 10*~7. This
unit risk is 8.8-fold lower than the EPA's (1985b) published unit
risk based on applied dose (4.1 x 10™6 per ug/m3), which is based
on the same bioassay data.
If it were decided that the surface area correction ought
not be applied to internal doses before risk calculation, the
procedure would be the same as outlined above, with the sole
exception that the division of human internal dose by 12.66
before risk calculation would not be done. Omitting this step
eSrifesponds to the assumption that corresponding human and mouse
tissues are equally sensitive to carcinogenic transformation by a
given internal dose. The unit risk calculated in this way would
be lowered ,by 12.66-fold from that calculated above. It would
thus be 111-fold less than the unit risk in EPA's applied dose
analysis (U.S. EPA, 1985b). On the other hand, if PD turns out
131
-------
to be 100, there would be no difference in the applied versus
internal dose-based estimated risks.
Lifetime extra risks over background from continuous and
constant low-level exposure to DCM may be estimated by
multiplying the vapor concentration by the internal unit risk
value. The HRAC's analyses of the model used by Andersen et al.
(1986, 1987) indicate that, so long as vapor concentrations
remain low (below 100 ppm), single exposures, intermittent
exposures, and exposures to varying vapor concentrations, are all
nearly equivalent in the internal doses they produce to a
continuous exposure to DCM at the time-weighted average level.
In other words, noncontinuous and other-than-lifetime exposures
can be converted to lifetime average daily equivalent exposures
before risks are calculated.
If vapor concentrations exceed 100 ppm or so for any part of
an exposure, substantial nonlinearities begin to appear that tend
to invalidate the assumptions allowing the unit risk to be used.
Under such conditions the MFO pathway begins to show saturation,
resulting in disproportionally more DCM being available to GST
metabolism, which results in disproportional increases in
internal dose. Exposures involving high vapor concentrations can
have estimated risks that are several-fold above the levels
implied by the "equivalent" time-weighted average exposure. The
reader is also reminded that the unit risk assumes a breathing
rate of 20 m3/day. Occupational exposures, or other exposures
occurring during more-strenuous-than-average activity, will
132
-------
consequently have risks somewhat underestimated.
7.5.4. Developing a Unit Risk Based on Internal Dose;
Incorporation of Only High- to Low-Dose Differences
Because of the current lack of data bearing on the question
of interspecies differences in tissue sensitivity to carcinogens,
it is exceedingly difficult to construct a sound argument as to
why a particular value of PD should be settled upon. Not only is
it difficult to argue for a PD value of 12.7 versus 1, as
outlined above, but other more widely ranging values are possible
as well, constrained only by the limited data showing that the
product F = PK x PD is in the neighborhood of between surface
area scaling of dose and body weight scaling of dose (Allen et
al., 1986).
Furthermore, there is a great deal of uncertainty in the
estimates of internal dose generated by the model used by
Andersen et al. (1986, 1987), as discussed earlier in this
document. In view of these uncertainties, it is wise to take
careful account of the metabolism data from all sources, and to
define a set of conclusions that seem the most robust despite the
unSVc-idable uncertainties. Then, the extrapolation of risk from
hig^i-dosed mice to humans can be examined in the light of What
can be concluded from these findings, from the viewpoint of using
those pharmacokinetic data which, due to a strong weight-of-
evidence indication, the HRAC feels should at a minimum be
incorporated into assessments of risk for DCM. The following
section reviews the data and weight of evidence, and then
133
-------
develops an extrapolation of risk to humans that uses the most
robust conclusions about pharmacokinetics to extrapolate from
high human doses to low doses, accounting for the effect of
saturation of the MFO pathway at high exposure levels.
7.5.4.1. Review of Metabolism Data—As discussed previously, DCM
is metabolized by two known pathways s the MFO pathway leading to
carbon monoxide as an end product, and hypothesized by some to
lead to carbon dioxide as well, and the GST pathway, leading to
carbon dioxide as the end product. The available data, as
thoroughly discussed in previous chapters, indicate that the MFO
pathway is saturable in animals at levels below the tested
inhalation levels in .the NTP bioassay. Based on the exhaustive
data base in animals and what little human data are now
available, it is likely that this pathway saturates in humans as
well. The available data, however, indicate that the GST pathway
is not saturated, even at the highest levels tested in the NTP
bioassay (the term "non-saturating" applies up to this exposure
level; obviously, the pathway will saturate at much higher
levels). Based on new in vitro data just submitted to HRAC, this
would seem to be the case in humans as well. At higher levels,
once the MFO pathway saturates, whatever output occurs from iihe1
GST pathway for any dose increment is likely to be linearly dose
dependent, due to the projected first order kinetics of the GST
pathway even at high dose levels; this approach would not be
inconsistent with the applied dose method at high doses (both are
linear at high doses in the case of DCM). However, at lower
134
-------
doses, where the MFO pathway is not saturated, the amount of
parent compound available to the GST pathway, relative to applied
dose, would be less than that at MFO—saturating levels—for the
obvious reason that for any dose increment, the MFO pathway
removes some of the parent compound that would otherwise be
available to the GST pathway. Thus, given that the MFO pathway
saturates, a nonlinear output from the GST pathway is expected at
low (non-saturating) doses relative to higher (saturating) doses.
It is thus important to elucidate the role of these two
pathways with regard to the carcinogenic potential of DCM. In
doing so several possibilities must be considered:
(1) The GST pathway is responsible for the carcinogenic response
(2) The MFO pathway results in the production of carbon monoxide
and possibly (hypothesized) in the production of carbon
dioxide. One or the other of these routes (or both) is
responsible for the carcinogenic response.
(3) The parent compound is responsible for the carcinogenic
response.
(4) Some combination of the above is responsible for the
response.
The saturation' of the MFO pathway is based on results
looking for carbon monoxide generation. According to the
recently available DCM data for model input, and PBPK models
developed for DCM, this portion of the MFO pathway is expected to
be saturated not only in the NTP inhalation bioassay at all dose
levels in mice, but also in the National Coffee Association (NCA)
135
-------
drinking water study at the highest dose level in mice.
[According to the model used by Andersen et al. (1986, 1987), MFO
pathway output is an average of 2550 and 3200 mg DCM
metabolized/liter of tissue/day for the NTP low dose and NCA high
dose, respectively.] If the carbon monoxide portion of the MFO
pathway were of primary importance, a response similar in
magnitude to that observed in the NTP bioassay would have been
expected at the highest dose of the NCA study, since the output
of this saturable pathway is estimated to be approximately the
same in these two cases. However, this was not observed; the
response in the NCA study was nonsignificantly elevated, much
smaller than the NTP response, but of a magnitude consistent with
a linear extrapolation from the higher NTP doses. Also, a dose-
response relationship for lung and liver would not be expected to
be as readily apparent as seen in the NTP bioassay, if the MFO
pathway was of primary importance, since again this pathway is
expected to be saturated, with a similar output, at both doses.
[The model used by Andersen et al. (1986, 1987) predicts MFO
output as an average of 2550 and 2650 mg DCM metabolized/liter of
tissue/day for the NTP low and high doses, respectively.,] This
argument does not eliminate the MFO pathway from consideration
with regard to a role in the carcinogenic response of DCM, but it
does indicate that some other pathway or chemical species is
likely to be of greater importance.
This leads to the likelihood that the GST pathway is of
primary importance for the carcinogenic response. The GST
136
-------
pathway intermediates are easily envisioned as being able to
interact with the genetic material, supporting an important role
for this pathway. Short-term tests indicate that DCM itself must
be metabolized before genotoxic effects are observed; although
this does not eliminate the parent compound from having a role,
it emphasizes a potential role for GST metabolites. In this
case, estimated risks at lower doses, especially those below the
point at which the MFO pathway saturates, would reflect the
nonlinear output of the GST pathway as described in the beginning
of this section, since they are based on observed risks at higher
doses.
However, even if there is a nonsaturable portion to the MFO
pathway (e.g., leading to carbon dioxide) which has a role in
carcinogenesis, or if the parent compound has a role, the effect
on these would be similar in concept as described above for the
GST pathway (i.e*,' the effects of a saturable system regarding
high- to low-dose differences on other nonsaturable systems).
Again, it should be remembered that some or all of the
intermediates of the various pathways, or the parent compound•,
may contribute to the carcinogenic process.
Thus, the HRAC believes that, based on the weight of
evidence/ some adjustment using pharmacokinetic data should be
made when extrapolating from high to low dose. As explained
previously in this section, however, the HRAC at this time is
unsure about how to use pharmacokinetic data that indicate
metabolic differences between species. This is because
137
-------
extrapolation between species involves many factors, one of which
is metabolism/pharmacokinetics. The ability to elucidate one
component of a species difference does not necessarily indicate
what, if any, adjustments should be made; it does not provide
more certainty than the empirical process currently used. Thus,
in the following section the HRAC performs an analysis of how
high- to low-dose effects on metabolism may affect ultimate risk.
7.5.4.2. Robustness of Model Output—In the case of DCM, the
HRAC has examined the effects on GST output in the lung and
liver, using the model used by Andersen et al. (1986, 1987),
relative to dose. As explained elsewhere in this document, the
model is generally able to estimate levels of DCM in various
situations, such as chamber disappearance, blood levels, etc.
Thus, the model is viewed to be reasonably able to predict levels
of DCM in tissues such as blood, lung, and liver. As explained
elsewhere as well, the model seems to be sensitive to changes in
the metabolic constants kF, KM, and Vmax.
At the values that Andersen et al. (1986, 1987) estimated
for these three constants in humans (kp, 0.53; KJJ, 0.58; Vmax,
118.3), the nonlinear contribution to the projected human db^e**
response curve due to consideration of DCM pharmacokinetic data
is 2.2 for the lung and 7.2 for the liver. These numbers are
derived by dividing the outputs of the model used by Andersen et
al. [RISK2L or RISK2P: measures of the output of the GST
pathway; the 2 stands for the GST pathway, and the L or P stands
for liver or lung (pulmonary), respectively] at 4000 ppm by
138
-------
RISK2L or RISK2P at 1 ppm, and then dividing that by 4000 to get
the difference relative to dose.
Having established that some correction, i.e., HL*, to
"target" dose levels due to high- to low-dose extrapolation (the
factors 2.2 and 7.2 are examples of such a correction) is
necessary, it is important to determine the sensitivity of these
ratios as the underlying metabolic constants input into the model
are varied. In other words, even though the absolute output of
the model is sensitive to metabolic constant variation, the
relative output (e.g., factors such as 2.2 and 7.2) may not be.
This is important in the determination of the magnitude of the
correction the HRAC recommends to be applied on the basis of
pharmacokinetic data.
Table 12 displays the results of a sensitivity analysis for
kF based on varying the metabolic constants input into the model
used by Andersen et al. (1986, 1987) for DCM. The ratio is,
again, RISK2L or RISK2P at 4000 ppm divided by RISK2L or RISK2P
at 1 ppm, and then divided by 4000 to get a ratio relative to
dose. The ratio is relatively insensitive to the value of kF,
the GST pathway metabolic constant. Thus, even if kp is off by
more than an order of magnitude, no real change in the ratio is
expedited. (The value of 4-fold was chosen based on the
sensitivity analysis performed on kF elsewhere in this document;
the value of 10-fold to look at order of magnitude differences,
and the value of 53-fold lower to approximate the possible
reduction in kF based on the CEFIC data; see previous sections
139
-------
for a review of these data.)
TABLE 12. SENSITIVITY ANALYSIS FOR kF
kp
V,
max
Comments
Ratio
Lung Liver
0.01
0.053
0.13
0.53
2.12
5.3
0
0
0
0
0
0
.58
.58
.58
.58
.58
.58
118
118
118
118
118
118
.9
.9
.9
.9
.9
.9
kp
kp
kF
decreased
decreased
decreased
53-fold
10-fold
4-fold
Model values
kF
kp
increased
increased
4-fold
10-fold
2.
2.
2.
2.
2.
1.
3
3
3
2
1
9
7
7
7
7
6
5
.6
.4
.4
.2
.5
.5
Tables 13 and 14 look at similar analyses for Vmax and KM/
the MFO pathway metabolic constants. Again, the lung ratios are
not too dissimilar from the ratio using the Andersen et al.
values when varied by an order of magnitude (only at a 10-fold
decrease does the ratio start to approach no difference as
opposed to an approximate 2-fold difference); the liver values
TABLE 13. SENSITIVITY ANALYSIS FOR Vmax
kp
'max
Comments
Ratio
Lung Liver
0.
0.
0.
0.
0.
53
53
53
53
53
0.
0.
0.
0.
0.
58
58
58
58
58
11.89
29.70
118.9
475.6
1189
vmax
Vmax
Model
Vmax
Vmax
decreased
decreased
values
increased
increased
10-fold
4-fold
4-fold
10-fold
1
1
2
2
1
.4
.7
.2
.2
.8
1.
2.
7.
20
32
7
7
2
140
-------
TABLE 14. SENSITIVITY ANALYSIS FOR %
Ratio
kF
5.3
5.3
5.3
5.3
5.3
%
0.
0.
0.
2.
5.
058
145
58
32
8
Vmax
118
118
118
118
118
.9
.9
.9
.9
.9
Comments
KM
decreased
decreased
10-fold
4-fold
Model values
KM
KM
increased
increased
4-fold
10-fold
Lung
2.
2.
2.
1.
1.
7
6
2
6
3
Liver
63
26
7.
2.
1.
2
5
6
vary more considerably, but interestingly, the ratio does not
fall much below two over this 100-fold range.
A final analysis (Table 15) looks at changes in both KM and
Vmax, but altering them simultaneously by the same magnitude
(looking at the sensitivity of the ratio of Vmax/Kj^) • The ratio
is much less sensitive to simultaneous changes in Vmax/KM than to
either variable alone. Consideration of this type of analysis is
important in cases where KJJ is large enough that the rate of
response for the MFO pathway is essentially proportional to
vmax/KM' in effect, a first-order rate constant somewhat like kp
[§iriee for a saturating system, rate is related to Vmax*C/ (Kjj+C) ,
Where C is concentration; at high KJJ relative to C, the equation
«*
can be approximated as Vmax*C/Kj,j, and thus the rate at various
concentrations is essentially related to Vmax/KM]•
The HRAC interprets the above analysis to indicate that even
if the Andersen et al. constants are in error by an order of
magnitude in either direction, the ratio of metabolism by the GST
pathway at low versus high dose, relative to dose, will always be
141
-------
TABLE 15. SENSITIVITY ANALYSIS FOR Vmax/%
Ratio
5
5
5
5
5
kp
.3
.3
.3
.3
.3
0
0
0
2
5
KM
.058
.145
.58
.32
.8
vmax
11.89
29.70
118.9
475.6
1189
Comments
vmax/KM
Vmax/%
decreased
decreased
10-fold
4-fold
Model values
vmax/KM
vmax/KM
increased
increased
4-fold
10-fold
Lung
2
2
2
1
1
.4
.3
.2
.9
.6
Liver
7
7
7
5
4
.8
.8
.2
.7
.1
decreased by a factor of, at least, approximately two for either
lung or liver. Furthermore, the analysis may be able to tolerate
even more error. The ratio is least sensitive to kp; variation
of this constant by nearly two orders of magnitude results in
little variation of the ratio. The value, kp, according to
Andersen et al. (1986), is the only constant that was truly
optimized and scaled, for humans; Vmax and KJJ were based on human
data (which are currently being analyzed by the HRAC), which
ostensibly are subject to less error.
7.5.4.3. Using Pharmacokinetics for High- to Low-Dose
Extrapolation—Given the above sensitivity analysis, the HRAC has
developed the following procedure for the calculation of human"
risk due to exposure to DCM.
(1) Calculate human doses to be input into a risk assessment
model from animal dose data by whatever species-to-species
conversion factor has been conventionally used (e.g.,
mg/kg/day, surface area correction).
(2) Modify these human doses by dividing the applied dose by the
appropriate factor derived from human lung GST metabolism
142
-------
model from animal dose data by whatever species-to-species
conversion factor has been conventionally used (e.g.,
mg/kg/day, surface area correction).
(2) Modify these human doses by dividing the applied dose by the
appropriate factor derived from human lung GST metabolism
according to the model used by Andersen et al. (1986, 1987).
The equation to derive the factor for a given dose X is:
Factor = RISK2P4000 * X / (RISK2PX * 4000)
where RISK2P4000 is RISK2P at 4000 ppm and RISK2PX is RISK2P
at X ppm.
(3) Use these doses, and the animal responses, for input into
the mathematical extrapolation model. When using the output
of the mathematical extrapolation model to calculate risk
for a specific applied environmental dose, apply the above
factor to the environmental dose before using the model to
predict a risk for that dose. Realize that the ,
pharmacokinetic model should be run to account for'duration
of exposure.
(4) All other calculations, such as proportion of lifetime
exposed, remain unchanged.
This procedure allows the incorporation of a factor, due to
consideration of high- to low-dose pharmacokinetic differences in
the case of DCM, of between 1.0-fold (at high doses) and about
2.2-fold (at low doses). The factor may indeed be greater, but
143
-------
the HRAC does not recommend that this procedure, for a minimum
pharmacokinetic-based adjustment, go beyond the limits of the
data as currently available. However, the weight of evidence
does justify, in the opinion of the HRAC, reducing the upper
bound or any other DCM risk estimate based on applied dose by the
aforementioned factor, at minimum. Furthermore, the HRAC
realizes that if some other tissue, such as the pancreas, is a
target, the above rationale would still hold since saturation of
the MFO pathway is still a very likely hypothesis for any tissue,
and the GST pathway is found in many tissues, including the
pancreas (Mukhtar et al., 1981). Finally, as in the approach
assuming PD = 12.7, HL*'s can be calculated if no exposure is
above 100 ppm, and thus lifetime average daily equivalent
exposures can, in such situations, be calculated before
mathematical risk assessment models are employed.
7.6. CONCLUSIONS
It is clear that, once estimates or measurements of internal
dose at the sites of toxic action are obtained, there are still
many difficult issues to be faced in deciding how to use such
data in the extrapolation of risk from experimental animal^ t6
humans. The problem is not confined to DCM, nor does it result
from faults in the information on pharmacokinetics for this
compound. It is a general problem, reflecting the lack of
understanding of the pharmacodynamics of carcinogenesis.
The use of pharmacokinetics only for high- to low-dose
extrapolation, outlined above, has the advantage that it is quite
144
-------
insensitive to the major uncertainties in the pharmacokinetic
data. This method accounts for the nonlinearities in internal
dose across exposure levels that arise from dose-dependent
changes in absorption, distribution, excretion, and saturation of
metabolism. The principal effect is the nonlinearity of GST
metabolism between high and low doses as a result of saturation
of the competing MFO pathway, which is relatively well
characterized. The method forgoes using pharmacokinetics for
interspecies extrapolation, due to uncertainty in assessing the
impact of a given internal dose difference in view of the lack of
knowledge of sensitivity differences. It thereby avoids the
question of the relative contribution of pharmacokinetics and
pharmacodynamics to interspecies scaling of carcinogenic potency,
but it assumes that the combined effect of these (the product PK
x PD) is more or less reliably given by cross species
extrapolation based on applied dose. This method results in a
minimum lowering of the implied risk to humans that seems
necessary in view of the data. If data became available
, indicating that human GST metabolism is much lower than
previously estimated, the risk calculation by this method would
not change, since the interspecies difference in potency is not
informed by metabolic differences between species (although such
differences clearly have an effect on potency, the method assumes
that the effect cannot be estimated).
The use of pharmacokinetics for interspecies extrapolation
as well as high- to low-dose extrapolation, discussed earlier in
•'' • 145
-------
this chapter, does incorporate interspecies differences in
metabolism/ and changes in this comparison will change the
estimates of human risk accordingly. The evident importance of
differences in metabolism among rats, mice, and hamsters to DCM's
carcinogenic potency in these species makes the use of metabolic
differences desirable, at least provisionally, in the estimation
of human risk. To make the interspecies extrapolation, however,
this method must make a further assumption about the relative
differences in tissue sensitivity to internal doses between '
experimental animals and humans. That is, one must make an
assumption about the value of PD, and not just about the product
PK x PD. Uncertainty in the proper value for species differences
in pharmacodynamics leads to widely divergent risk estimates.
This method is also more sensitive to errors in the
pharmacokinetic model revolving around the determination of kF
(see previous sections).
Both methods must assume that the relationship of internal
dose to risk within species is adequately characterized by the
process of fitting dose-response curves and extrapolating them to
low doses. The shape of the dose-response curve at low doses
depeiids heavily on the mechanism of action involved in DCM'S
carcinogenicity. This mechanism is very poorly understood, and
the uncertainties 'about low-dose extrapolation of pharmaco-
dynamics probably greatly exceed those about interspecies
extrapolation of potency.
146
-------
REFERENCES
Abrahamson, S.; Valencia, R. (1980) Evaluation of substances of
interest for genetic damage using Drosophila melanogaster.
Prepared for the Food and Drug Administration. Contract No.
233-77-219. Unpublished.
Ackerman, L.V.; Rosai, J. (1974) Surgical Pathology.
St. Louis, MO; The C.V. Mosby Co.
Ahmed, A.B.; Anders, M.W. (1976). Metabolism of dihalomethanes
to formaldehyde and inorganic halide. Drug Metab. Dispos.
4:357-361.
Ahmed, A.E.; Kubic, V.L.; Stevens, J.L.; Anders, M.W. (1980)
Halogenated methanes: metabolism and toxicity. Fed. Proc.
39(13):3150-3155.
Allen, B.C.; Shipp, A.M.; Crump, K.S.; Kilian, B.; Hogg, M.;
Tudor, J.; Keller, B. (1986) Investigation of cancer risk
assessment methods. Final Report: Summary. Prepared for '
the U.S. Environmental Protection Agency, Department of
Defense, Electric Power Research Institute, Risk Science
Institute.
Anders, M.W.; Kubic, V.L.; Ahmed, A.W. (1977) Metabolism of
halogenated methanes and macromolecular binding. J.
Environ. Pathol. Toxicol. 1:117-124.
Anders, M.W.; Kubic, V.L.; Ahmed, A.E. (1978) Bio-organic
mechanisms of the metabolism of dihalomethanes to carbon
monoxide, formaldehyde, formic acid and inorganic halide.
International Congress Series Excerpta Medica 440:22-24.
Andersen, M.E.; Clewell, H.J., III; Gargas, M.L.; Smith, F.A.;
Reitz, R.H. (1986) Physiologically based pharmacokinetics
and the risk assessment process for methylene chloride.
Submitted for publication.
Andersen, M.E.; Clewell, H.J., III? Gargas, M.L.; Smith, F.A.;
Reitz, R.H. (1987) Physiologically based pharmacokinetics
and the risk assessment process for methylene chloride.
Toxicol. Appl. Pharmacol. 87:185-205.
Angelo, M.J.; Pritchard, A.B. (1984) Simulations of methylene
chloride pharmacokinetics using a physiologically based
model. Regul. Toxicol. Pharmacol. 4:320-339.
147
-------
Angelo, M.J.; Bischoff, K.B.; Pritchard, A.B.; Presser, M.A.
(1984) A physiological model for the pharmacokinetics of
methylene chloride in B6C3F1 mice following i.v.
administrations. J. Pharmacokinet. Biopharm. 12:(4)413-436.
Angelo, M.J.; Pritchard, A.B.; Hawkins, D.R.; Waller, A.R;
Roberts, A. (1986a) The pharmacokinetics of
dichlormethane. I. Disposition in B6C3Fi mice following
intravenous and oral administrations. Food Chem. Toxicol.
24(9):965-974.
Angelo, M. J.; Pritchard, A.B.,° Hawkins, D.R.; Waller, A.R.;
Roberts, A. (1986b) The pharmacokinetics of
dichloromethane. II. Disposition in Fischer 344 rats
following intravenous and oral administrations. Food Chem.
Toxicol. 24(9):975-980.
Angelo, M.J.; Pritchard, A.B.? Nelson, D.A. (1986c) Dosimetry
comparisons between inhalation and oral exposures of
dichloromethane in rats. Presented at the 25th annual
meeting of the Society of Toxicology, New Orleans, LA.
March 3-7.
Barber, E.D.; Donish, W.H.; Mueller, K.R. (1980) Quantitative
measurement of the mutagenicity of volatile liquids in the
Ames Salmonella/microsome assay. Proceedings of the llth
annual meeting of the Environmental Mutagen Society.
Abstract, p. 39.
Battelle. (1986) Independent review of Kodak rebuttal and
methylene chloride mortality study update. Task 1-19. EPA
Contract No. 68-02-4246. Subcontract No. K-8958 (0730)-801.
E. Margosches, Project Officer.
Black, O.; Howerton, B.K.; (1984) Glutathione S-transferase
activity in rat pancreas. J. Natl. Cancer Inst. 72(1):121-
123.
Boyd, M.R. (1977) Evidence for the Clara cell as a site of
cytochrome P450-dependent mixed function oxidase activity in
lung. Nature 269:713-715.
Boyer, T.D.; Zakim, D.; Vessey, D.A. (1982) Studies of
endogenous inhibitors of microsomal glutathione S-
transferase. Biochem. J. 207:57-64.
Boxenbaum, H. (1983) Evolutionary biology, animal behavior,
forth-dimensional space, and the raison d'etre of drug
metabolism and pharmacokinetics. Drug Metab. Rev.
14(5):1057-1097.
148
-------
Boxenbaum, H. (1984) Interspecies pharmacokinetic scaling and
the evolutionary-comparative paradigm. Drug Metab.
Rev. 15(5 & 6):1071-1121. ^
Callen, D.F.; Wolf, C.R.; Philpot, R.M. (1980) Cytochrome
P450-mediated genetic activity and cytotoxicity of seven
halogenated aliphatic hydrocarbons in Saccharomyces
cerevisiae. Mutat. Res. 77:55-63.
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986a) Hext, P.M.; Foster, J.; Millward, S.W. Methylene
chloride: 10-day inhalation toxicity study to investigate
the effects on rat and mouse liver and lungs. ICI Central
Toxicology Laboratory, Report No. CTL/R/1432. January 10,
1986.
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986b) Trueman, R.W.; Ashby, J. Methylene chloride: In
vivo and in vitro unscheduled DNA synthesis studies in the,
mouse and the rat. ICI Central Toxicology Laboratory,
Report No. CTL/P.1444. January 21, 1986.
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986c) Green, T.; Provan, W.M.; Collinge, D.C.; Guest,
A.E.. Methylene chloride: interaction with the rat and mouse
liver and lung DNA in vivo. ICI Central Toxicology
Laboratory, Report No. CTL/R/851. January 22, 1986.
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986d) Sheldon, T.; Richardson, C.R.; Hamilton, K.;
Randell, V.; Hart, D.; Hollis, K. Methylene chloride; an
evaluation in the mouse micronucleus test. ICI Central
Toxicology Laboratory, Report No. CTL/P/1603. September
19, 1986.
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986e) Green, T.; Nash, J.A.; Mainwaring, G. Methylene
chloride: in vitro metabolism in rat, mouse and hamster
liver and lung fractions and in human liver fractions.
ICI Central Toxicology Laboratory, Report No. CTL/R/879.
September 22, 1986.
CEFlC (European Council of Chemical Manufacturer's Federation)
(1986f) Green, T.; Provan, W,M.; Nash, J.A.; Gowans, N.
Methylene chloride: in vivo inhalation pharmacokinetics and
titetabolism in F344 rats and B6C3F1 mice. ICI Central
Toxicology Laboratory, Report no. TL/R/880. September 22,
1986.
149
-------
CEFIC (European Council of Chemical Manufacturer's Federation)
(1986g) Lefevre, P.A.; Ashby, J. Methylene chloride:
induction of S-phase hepatocytes in the mouse after in vivo
exposure. Id Central Toxicology Laboratory, Report No.
CTL/R/885. September 22, 1986.
Conn, M.S.; Rock A.R. 1986. Review of epidemiology study on
methylene chloride by Friedlander. U.S. Consumer-Product
Safety Commission. April 25, 1986.
Consumer Product Safety Commission (CPSC). (1985, June 19)
Briefing package on methylene chloride from Sandra Eberle,
program manager, Chemical Hazards Program to the Commission.
131 pp.
Cornell, N.W.; Hansch, C.; Kim, K.H.; Henegar, K. (1983) The
inhibition of alcohol dehydrogenase in vitro and in isolated
hepatocytes by 4-substituted pyrazoles. Arch. Biochem.
Biophys. 227(1):81»90.
Crump, K.S.; Silvers, A.; Ricci, P.F.; Wyzga, R. (1985)
Interspecies comparison for carcinogenic potency to humans.
In: Ricci, P.F., ed. Principles of health risk assessment.
Englewood Cliffs, NJ: Prentice-Hall.
Dedrick, R.L.; Bischoff, K.B.; Zaharko, D.S. (1970)
Interspecies correlation of plasma concentration history of
methotrexate (NSC-740). Cancer Chemotherap. Rep. Pt 1,
54:95-101.
Dow Chemical Company. (1980) Methylene chloride: a two-year
inhalation toxicity and oncogenicity study in rats and
hamsters. Follow-up response A. U. S. Environmental
Protection Agency, Office of Toxic Substances, Information
Management Division. TSCA Docket No. FYI-OTS-0281-0097.
European Chemical Industry Ecology and Toxicology Center. (1982)
Hepatocarcinogens in laboratory rodents: relevance for man.
Monograph No. 4. Brussels, Belgium.
Fox, A.J.; Collier, P.F. (1976) Low mortality rates in
industrial cohort studies due to selection for work and
survival in the industry. Br. J. Prev. Soc. Med.
30:225-230.
Friedlander, B.R.; Hearne, T.; Hall, S. (1978) Epidemiologic
investigation of employees chronically exposed to methylene
chloride. Mortality analysis. J. Occup. Med. 20:657-666.
150
-------
Friedlander, B.R.; Pifer, J.W.; Hearne, F.T. (1985) 1964
methylene chloride cohort mortality study. Update through
1984. Eastman Kodak Company. Rochester, N.Y. July 2,
1985. U.S. Environmental Protection Agency, Office of Toxic
Substances, Information Management Division. TSCA Docket
No. OPTS-48503.
Friedlander, B.R.; Hearne, F.T.? Pifer, J.W.; Grose, F.H. (1986)
Epidemiologic evidence regarding methylene chloride.
Presented at the Winter Toxicology Forum, Washington, D.C.,
February 18.
Friedlander, B.R.; Hearne, F.T.; Pifer, J.W.; Grose, F.H. (1986)
Response to "Review of Kodak Studies of Workers Exposed to
Methylene Chloride at Eastman Kodak Company." Rochester,
New York. April 15, 1986. U.S. Environmental Protection
Agecy, Office of Toxic Substances, Information Management
Division. TSCA Docket No. OPTS-62045.
Gargas, M.L.; Clewell, H.J.; Anderson, M.E. (1986) Metabolism
of inhaled dihalomethanes in vivo: differentation of kinetic
constants for two independent pathways. Toxicol. Appl.
Pharmacol. 82:211-223.
Gocke, E.; King, M.T.; Eckhardt, K.; Wild, D. (1981)
Mutagenicity of cosmetics ingredients licensed by the
European communities. Mutat. Res. 90:91-109.
Green, T. (1983) The metabolic activation of dichloromethane
and chlorofluoromethane in a bacterial mutation assay using
Salmonella typhimurium. Mutat. Res. 118(4)5277-288.
Hearne, F.T.; Friedlander, B.R. (1981) Follow-up of methylene
chloride study. J. Occup. Med. 23:660.
Hearne, F.T.; Grose, F.; Pifer, J.W.; Friedlander, B.R. (1986)
Methylene chloride mortality study update. Eastman Kodak
Company. Rochester, N.Y. June 16, 1986.
Hearne, F.T.; Grose, F.; Pifer, J.W.; Friedlander, B.R.; Raleigh,
R.L. (1987) Methylene chloride mortality study: dose-
response characterization and animal model comparison. J.
Occup. Med. 29(3):217-226.
Howe> R.B.; Crump, K.S.; Van Landingham, C. (1986) GLOBAL86* A
computer program to extrapolate quantal animal toxicity data
to low doses. Prepared for the U.S. Environmental
Protection Agency. Contract no.68-01-682.
151
-------
Jongen, W.M.F.; Harmson, E.G.M.; Alink, G.M.; Koeman, J.H.
(1982) The effect of glutathione conjugation and microsomal
oxidation on the mutagenicity of dichloromethane in S.
tvphimurium. Mutat. Res. 95:183-189.
King, F.G.; Dedrick, R.L.; Collins, J.M.; Matthews, B.B.?
Birnbaum, L.S. (1983) Physiological model for the
pharmacokinetics of 2,3,7,8-tetrachlorodibenzofuran in
several species. Toxicol. Appl. Pharmacol. 67:390-400.
Kubic, V.L.; Anders, M.W. (1978) Metabolism of dihalomethanes
to carbon monoxide. III. Studies on the mechanism of the
reaction. Biochem. Pharmacol. 27(19):2349-2355.
Lorenz, J.; Glatt, H.R.; Fleischmann, R.; Ferling, R.; Oesch, F.
(1984) Drug metabolism in man and its relationship to that
in three rodent species: monooxygenase, epoxide hydroylase,
and glutathione S-transferase activities in subcellular
fractions of lung and liver. Biochem. Med. 32:43-56.
Matanowski, G.M. (1985, October 15) Review of Kodak studies of
workers exposed to methylene chloride. Prepared for the
U.S. Environmental Protection Agency, Office of Toxic
Substances. Contract No. 68-02-4246.
McKenna, M.J.; Zempel, J.A.; Braun, W.H. (1982) The
pharmacokinetics of inhaled methylene chloride in rats.
Toxicol. Appl. Pharmacol. 65:1-10.
Milham, S., Jr. (1976) Occupational mortality in Washington
State 1950-1971. Washington, DC: U.S. Government Printing
Office. NIOSH Pub. No. 76-175A.
Mirsalis, J.C.; Tyson, C.K.; Butterworth, B.E. (1980)
Detection of unscheduled DNA synthesis in hepatocytes
isolated from rats treated with genotoxic agents: an in
vivo and in vitro assay for potential carcinogens and
mutagens. Carcinogenesis 1:621-625.
Mirsalis, J.c.; Tyson, C.K.; Loh, E.N.; steinmetz, K.L.; Bakke,
J.P.; Hamilton, C.M.; Spak, O.K.; Spaulding, J.W. (1985)
Induction of hepatic cell proliferation and unscheduled DNA
synthesis in mouse hepatocytes following in vivo treatment.
Carcinogenesis 6:1521-1524.
Morgenstern, R.; Depierre, J.W.; Ernster, L. (1979) Activation
of microsomal glutathione S-transferase activity by
sulfhydryl reagents. Biochem. Biophys. Res. Commun. 87:657-
663.
152
-------
Moron, M.S.; Depierre, J.W.; Mannervik, B. (1979) Levels of
glutathione, glutathione reductase and glutathione S-
transferase activities in fat lung and liver. Biochim.
Biophys. Acta 582:67-78.
Mukhtar, H.; Zoetemelk, C.E.; Baars, A.J.; Wijnen, J.T.;
Blankenstein-Wijnen, L.M.; Meera Khan, P.; Breimer,, D.D.
(1981) Glutathione S-transferase activity in human fetal
and adult tissues. Pharmacology 22(5):322-329.
Nash, T. (1953) The colorimetric estimation of formaldehyde by
means of the Hantzsch reaction. Biochem. J. 55:416-421.
National Academy of Sciences (NAS) (1975) Contemporary pest
control practices and prospects: the report of the executive
committee. Vol 1. Pest control: an assessment of present
and alternative technologies. Washington, DC: National
Academy Press.
National Coffee Association (NCA) (1982a) Twenty-four month
chronic toxicity and oncogenicity study of methylene
chloride in rats. Final report. Prepared by Hazleton
Laboratories, America, Inc., Vienna, VA. Unpublished.
National Coffee Association (NCA) (1982b) Twenty-four month
chronic toxicity and oncogenicity study of methylene
chloride in rats. Addition to the final report. Prepared
by Hazleton Laboratories America, Inc., Vienna, VA.
National Coffee Association (NCA) (1983) Twenty-four month
oncogenicity study of methylene chloride in mice. Prepared
by Hazleton Laboratories, America, Inc., Vienna, VA.
National Toxicology Program (NTP) (1985) NTP technical report
on the toxicology and carcinogenesis studies of
dichloromethane in F-344/N rats and B6C3F1 mice (Inhalation
studies). NTP-TR-306.
National Toxicology Program (NTP). (1986, January) NTP
technical report on the toxicology and carcinogenesis
studies of dichloromethane in F344/N rats and B6C3F1 mice
(inhalation studies). NIH publication no. 86-2562. NTP TR
306. U.S. Department of Health and Human Services, Public
Health Service, National Institutes of Health.
Ramsey, J.C.; Andersen, M.E. (1984) A physiologically based
description of styrene in rats and humans. Toxicol. Appl.
Pharmacol. 73:159-175.
153
-------
Reitz, R.H.; Smith, F.A.; Andersen, M.E. (1986) In-vivo
metabolism of 14C-methylene~ chloride (MEC). Presented at
the annual meeting of the Society of Toxicology. New
Orleans, LA.
Russo, J.; Tay, L.K.; Russo, I.E. (1982) Differentiation of the
mammary gland and susceptibility to carcinogenesis. Breast
Cancer Res. Treat. 2(l):5-73.
Seidegard, J.; Pero, R.w. (1985) The hereditary transmission
of high glutathione transferase activity towards trans-
stilbene oxide in human mononuclear leukocytes. Hum. Genet.
69:66-68.
Seidegard, J.; DePierre, J.W.,; Pero, R.W. (1985) Hereditary
interindividual differences in the glutathione transferase
activity towards trans-stilbene oxide in resting human
mononuclear leukocytes are due to a particular isozyme(s)*
Carcinogenesis 6(8):1211-1216.
Seidegard, J.; Pero, R.W.; Miller, D.G.; Beattie, E.J. (1986) A
glutathione transferase in human leukocytes as a marker for
the susceptibility to lung cancer. Carcinogenesis
7(5):751-753.
Simmon, V-F-/' Kauhanen, V.; Tardiff, R.G. (1977) Mutagenic
activity of chemicals identified in drinking water. Dev.
Toxicol. Environ. Sci. 2s249-258.
Souhaili-El Amri, H.; Batt, A.M.; Siest, G. (1986) Comparison
of cytochrome P-450 content and activities in liver
microsomes of seven animal species, including man.
Xenobiotica 16(4):351-358.
State of California (1986) Health assessment for
darcinogenicity criteria document: methylene chloride.
State of California, Department of Health Services.
Berkeley, CA.
Stevens, J.L.; Ratnayake, J.H.; Anders, M.w. (1980) Metabolism
of dihalomethanes to carbon monoxide. IV. Studies in
isolated rat hepatocytes. Toxicol. Appl. Pharmacol.
55:484-489.
Szabo, S.; Horvath, E.; Kovacs, K.; Larsen, P.R. (1978)
Pyrazole-induced thyroid necrosis: a distinct organ lesion.
Science 199:1209-1210.
154
-------
Terasaki, T.; Sugiyama, Y. ? Hanano, M. (1984) Pharmacokinetic
study on the mechanism of tissue distribution of
doxorubicin: interorgan and interspecies variation of
tissue-to-plasma partition coefficients in rats/ rabbits,
and guinea pigs. J. Pharm. Sci. 73(10):1359-1363.
Thilagar, A.K.; Kumaroo, V. (1983) Induction of chromosome
damage by methylene chloride in CHO cells. Mutat. Res. 116:
361-367.
Uotila, L.; Koivusalo, M. (1981) Formaldehyde dehydrogenase.
In: Jakoby, W., ed. Methods in enzymology, Vol. 77.
Detoxication and drug metabolism: conjugation and related
systems. New York, NY: Academic Press.
U.S. Environmental Protection Agency (EPA). (1985a, February)
Health assessment document for dichloromethane (methylene
chloride). Final report. EPA/600/8-82/004F.
U.S. Environmental Protection Agency (EPA). (1985b, September)
Addendum to the health assessment document for
dichloromethane (methylene chloride): updated
carcinogenicity assessment of dichlormethane (methylene
chloride). EPA/600/8-82/004FF.
Van Bladeren, P.J.; Breimer, D.D.; Rotteveel-Smijs, G.M.T.; Mohn,
G.R. (1980) Mutagenic activation of dibromomethane and
diidomethane by mammalian microsomes and glutathione S<-
transferases. Mutat. Res. 74:341-346.
Yesair, D.W.; Jaques, P.; Shepis, P.; Liss, R.H. (1977) Dose-
related pharmacokinetics of 14C-methylene chloride in mice.
Fed. Proc. Am. Soc. Exp. Biol. 36:998 (abstract).
, 155
-------
-------
-------
JO
"^
Q) O
c
CO
EPA/600/8-
00
•vl
o
to
>
If you do not wis
detach, or copy
upper left-hand
o :r 3*
° Z' o
3 r» .1
5 o a
CD CD^
OJ CD
§i
2 3
D -O
o 9
^5
a
o. ^
-t O
2g
C o>
3 •?
HI
o to
!1 w
3 S
3 m
3-cJ
re tb
c _.
11
> m i
CQ 3 :
(D <
'
(C
3
T)
^
O
il
»?!
o;
U1
to
en
oo
------- |