SEPA
United States
Environmental Protection
Agency
Office of Health and
Environmental Assessment
Washington DC 20460
EPA/600/8-87/030A
July 1987
External Review Draft
Research and Development
Update to the
Health Assessment
Document and
Addendum for
Dichloromethane
(Methylene
Chloride):
Pharmacokinetics,
Mechanism of
Action, and
Epidemiology
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(Do Not
Cite or Quote)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 EPA/600/8-87/030A
DO NOT QUOTE OR CITE July 1987
Review Draft
UPDATE TO THE HEALTH ASSESSMENT DOCUMENT
AND ADDENDUM FOR DICHLOROMETHANE (METHYLENE
CHLORIDE): PHARMACOKINETICS, MECHANISM OF ACTION,
AND EPIDEMIOLOGY
NOTICE
THIS DOCUMENT IS A PRELIMINARY DRAFT. It has not been formally
released by the U.S. Environmental Protection Agency 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. J
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 Agency policy. Mention of trade names or
commercial products does not constitute endorsement or
recommendation for use.
ii
-------
CONTENTS
Foreword v
Preface ix
Abstract xi
Authors, Contributors, and Reviewers xii
1. INTRODUCTION 1
2. OVERVIEW OF DICHLOROMETHANE CANCER HAZARD/RISK ISSUES ... 5
3. APPLIED-DOSE RISK ASSESSMENT 10
4. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS 12
4.1. MODEL STRUCTURE 15
4.2. MODEL INPUT 16
4.2.1. Partition Coefficients 16
4.2.2. Breathing Rates 17
4.2.3. Metabolic Parameters 18
4.3. VALIDATION OF THE MODEL 26
5. ASSUMPTIONS CONCERNING THE CARCINOGENIC PATHWAY 30
6. ASSUMPTIONS CONCERNING THE MECHANISM OF ACTION 35
6.1. GENOTOXICITY 35
6.1.1. Bacteria, Yeast, and Drosophila 35
6.1.2. Mammalian Cells in vitro 36
6.1.3. Mammalian Cells in vivo 37
6.1.4. Summary 37
6.2. ALTERNATIVE MECHANISMS OF CARCINOGENIC ACTION ... .38
7. IMPACT OF THE PHYSIOLOGICALLY BASED PHARMACOKINETIC
MODEL USED BY ANDERSEN AND REITZ ON HUMAN RISK
ESTIMATES 41
8. USING INTERNAL DOSES AS A BASIS FOR HUMAN RISK
ESTIMATION 47
8.1. METHOD 1: COMPARISON OF INTERNAL DOSES TO
ALLOMETRIC EXPECTATION 50
8.2. METHOD 2: USE OF PHARMACOKINETICS ONLY FOR HIGH-
TO LOW-DOSE EXTRAPOLATION 62
8.3. COMPARISON OF METHODS 1 AND 2 71
iii
-------
9. IMPACT OF THE CEFIC EXPERIMENTAL DATA ON RISK ESTIMATES . .88
10. IMPACT OF EPIDEMIOLOGIC EVIDENCE ON HUMAN RISK
ESTIMATION 92
11. EPA'S CONCLUSIONS CONCERNING THE RISKS TO HUMANS FROM
EXPOSURE TO DICHLOROMETHANE 100
References
113
IV
-------
FOREWORD
This document is the result of a yearlong intensive review
and evaluation of the latest data that have a bearing on risk
assessment for dichloromethane (DCM; methylene chloride). These
data, and the most current methods for their analysis, have been
reviewed in an effort to improve the quantitative risk estimates
for DCM. In addition, I hope that this review and the attempt at
applying state-of-the-art methodologies will help to advance the
science of risk assessment.
Several key issues have arisen as a result of the new data
that have been provided to the federal regulatory agencies. The
most notable is the impact of pharmacokinetic analysis on risk
extrapolation. Physiologically based pharmacokinetic models are
used to estimate organ-level concentrations of DCM and its
metabolites. Such organ-level concentrations can serve as
measures of the delivered dose. This type of pharmacokinetic
analysis raises two important questions. First, can sufficiently
reliable estimates of delivered dose be made so that they
constitute an improvement over the use of applied dose as a basis
for risk extrapolation? Second, how are such delivered-dose
estimates to be used in the quantitative extrapolation of risk
from experimental animals to humans exposed to much lower doses?
Two possible applications of how pharmacokinetic information
and data may be incorporated into the quantitative risk assessment
are developed. The first (Method 1) incorporates interspecies
-------
differences in pharmacokinetics while the second (Method 2) only
incorporates differences resulting from high- to low-dose
extrapolation. The major uncertainties associated with each
approach are discussed. For the present, using pharmacokinetic
data and models for interspecies extrapolation results in a
reduction of the estimated risk from the applied-dose estimate by
almost ninefold. Similarly, using the pharmacokinetic models for
high- to low-dose extrapolation only, the risk would be reduced
from the applied-dose estimates slightly more than twofold. The
two methods differ in the assumptions that are made and are not
equally sensitive to one of the key metabolic rate constants.
For the present, EPA scientists have used Method 1, which
results in an ninefold reduction from previous unit risk estimates
based on applied dose. CPSC scientists, on the other hand, have
used Method 2 which results in an approximate twofold reduction.
OHSA and FDA scientists are in the process of deciding what, if
any, modifications to make to present risk numbers.
It would be unwise to read too much importance or
significance into changes in the unit risk of a few fold when
pharmacokinetic data are used in either Method 1 or Method 2.
This document outlines uncertainties in the structure and
parameter values of the pharmacokinetic models. Although it is
difficult to define these uncertainties in quantitative terms, it
is clear that model projections of internal doses could vary,
perhaps up to several fold, without contradicting currently
available model validation data. Moreover, there are many
VI
-------
uncertainties as to the biological effects of those internal doses
that overshadow any error in their estimation. Species
differences in responsiveness—and within-species differences in
susceptibility of various tissues—are unclear. Perhaps the
largest uncertainty lies in the question of the relative
carcinogenicity of high and low doses, owing to the lack of
knowledge about the mechanism of DCM's carcinogenic action.
Rather than focusing on exactly how much the risk
extrapolation has been changed by the use of pharmacokinetic
information, it is instructive to examine how little it has been
changed. Perhaps the most important result of the these analyses
is that, in the case of DCM, pharmacokinetic considerations have
not revealed a great error inherent in using applied dose as a
surrogate for internal or delivered dose.
Thus, Method 1 has been chosen not because it is felt to be
the only valid approach (it is not yet clear which approach, if
any, can be used with confidence), but rather because it
represents the best use of the available information in a manner
closely consistent with what EPA has done with other chemicals,
such as tetrachloroethylene and trichloroethylene. In view of the
uncertainties involved, the changes in DCM's carcinogenic potency
that result from different uses of the available pharmacokinetic
information are not, in practical terms, very different.
Discussion of the issues has been worthwhile because of their
theoretical importance rather than their practical significance in
the present case. For other compounds (or for DCM itself, upon
vii
-------
itself, upon the introduction of new data), the distinction among
extrapolation methods may have much greater practical consequences
and the question of choice among methods will have to be re-
examined. EPA staff will then be required to look for data and
ways with which to reduce the uncertainties around a given
estimate. I hope that the peer review process that this document
will undergo will also help in the quest for this knowledge.
Peter W. Preuss
Director
Office of Health and Environmental
Assessment
viii
-------
PREFACE
This document was prepared by staff of the Office of Health
and Environmental Assessment (OHEA) and the Office of Toxic
Substances (OTS) of the U.S. Environmental Protection Agency (EPA)
as part of the Integrated Chlorinated Solvents Project. It
represents the latest technical review and assessment of
dichloromethane (DCM). The document is not intended to replace
the Health Assessment Document (HAD) and Addendum for
Dichloromethane but to update them by providing an evaluation of
data and risk assessment methodologies that have become available
since their publication. The update comprises two documents: one
prepared by EPA and a second prepared by the Health/Risk
Assessment Committee (HRAC) of the Integrated Chlorinated Solvents
Project.
This EPA document represents EPA's analysis of the weight of
evidence regarding DCM's carcinogenic potential for humans; it
includes revised cancer risk estimates which take into account the
new information submitted to date on pharmacokinetics, mechanism
of action, and epidemiology, and discusses in some detail methods
for incorporating pharmacokinetic information into the cancer risk
assessment. The EPA document draws on the body of work developed
by the HRAC, an interagency committee that was established to
evaluate the health effects caused by DCM and five other
halogenated solvents. As part of its work on halogenated solvent
compounds, the HRAC reviewed and evaluated all of the information
ix
-------
recently submitted to EPA and other federal agencies on DCM's
potential to cause cancer and other toxic effects.
The HRAC analyses, which comprise a second document titled
"Technical Analysis of New Methods and Data Regarding
Dichloromethane Hazard Assessments," deal with several aspects of
the risk assessment.
-------
ABSTRACT
This document represents EPA's analysis of the weight of
evidence regarding the carcinogenic potential of dichloromethane
(DCM, methylene chloride) for humans. It includes revised cancer
risk estimates that take into account the newest information on
pharmacokinetics, mechanism of action, and epidemiology, and it
discusses, in some detail, methods for incorporating
pharmacokinetic information into the cancer risk assessment. This
document draws on the body of work developed by the Hazard/Risk
Assessment Committee (HRAC) of the Integrated Chlorinated Solvents
Project, an interagency workgroup. For the present, EPA
scientists have used methods that result in an approximate
ninefold reduction from previous unit risk estimates which were
based on applied dose.
xi
-------
AUTHORS, CONTRIBUTORS, AND REVIEWERS
This document represents the joint efforts of the staff of
the Office of Health and Environmental Assessment (OHEA) and the
Office of Toxic Substances (OTS) of the U.S. Environmental
Protection Agency (EPA).
OHEA had overall responsibility for coordination and
direction of the document preparation and production effort (Jerry
N. Blancato, Project Manager).
AUTHORS
Jerry N. Blancato
Exposure Assessment Group
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
Jane Hopkins
Office of Solid Waste and Emergency Response
(formerly with the Office of Toxic Substances)
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
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
xii
-------
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
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
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
Harold Zenick
Office of Health and Environmental Assessment
U.S. Environmental Protection Agency
xiii
-------
-------
1. INTRODUCTION
Dichloromethane (DCM, methylene chloride) is a multipurpose
solvent with application in paint stripping, metal cleaning, foam
blowing, electronics, and chemical processing. It is also a
component in certain aerosol propellant mixtures. Because of its
many applications, a large number of people may be exposed to DCM
in the workplace, through use of consumer products, or from
emissions to ambient air.
Animal bioassays published in the 1980s raised concern that
DCM could cause cancer in humans. Several bioassays reported an
elevation in tumor incidences in mice and rats exposed to this
chlorinated solvent. In response to these bioassay data and
concerns about other potential chronic effects, EPA's Office of
Air and Radiation (OAR) requested that the Office of Health and
Environmental Assessment (OHEA) prepare a Health Assessment
Document (HAD) on DCM. The HAD, which reviewed data relevant to
acute and chronic effects, was published in February 1985.
New bioassays published in 1985 by the National Toxicology
Program (NTP) intensified concern for DCM's potential to cause
cancer. EPA's Office of Toxic Substances (OTS) began a
regulatory review to determine whether the risks posed by any or
all uses of DCM were sufficient to warrant priority consideration
for regulatory action under section 4(f) of the Toxic Substances
Control Act (TSCA). OAR, already involved in investigating
-------
whether ambient sources of DCM should be regulated under the
National Emissions Standards for Hazardous Air Pollutants
(section 112) provision of the Clean Air Act (CAA), requested
that OHEA prepare an addendum to the HAD to evaluate the new data
from the NTP.
In May 1985, OTS published (U.S. EPA, 1985c) a finding under
TSCA section 4(f) that DCM may present a significant risk to
humans of serious and widespread harm from cancer. The new NTP
bioassays provided sufficient evidence to upgrade DCM from a
possible to a probable human carcinogen and led to a positive
4(f) finding. Following the 4(f) action, OTS and OAR combined
efforts and a single coordinated assessment was published by OHEA
in July 1985 as an Addendum to the HAD for DCM. In the Addendum,
EPA classified DCM as a probable human carcinogen (Group B2) as
defined by the EPA Guidelines for Carcinogen Risk Assessment
(U.S. EPA, 1986). The assessment led to publication of an EPA
Advance Notice of Proposed Rulemaking in October 1985 (U.S. EPA,
1985d).
EPA was not alone in responding to the NTP's new bioassay
data. Following release of the NTP data, several other federal
regulatory agencies began to investigate the health effects posed
by DCM. The Consumer Product Safety Commission (CPSC) released a
briefing package in June 1985 announcing the staff's intent to
pursue voluntary labeling actions. Subsequently, CPSC began a
rulemaking procedure to determine whether or not DCM should be
designated a hazardous substance, and published a Federal
-------
Register notice (U.S. CPSC, 1986) requesting comment on the
proposed rule. In November 1985, the Food and Drug
Administration (FDA, 1985) proposed a ban on the use of DCM in
hairsprays. The Occupational Safety and Health Administration
(OSHA) began an investigation into the risks associated with
occupational exposures to DCM, and published an Advance Notice of
Proposed Rulemaking in November 1986 (OSHA, 1986).
The multiple assessments called for a coordinated effort on
the part of the various agencies to ensure consistency in
regulatory decision-making. To that end, EPA proposed developing
a common integrated assessment covering not only DCM, but other
chlorinated solvents that could serve as substitute chemicals as
well. The agencies developed an integrated strategy to assess
the major occupational, ambient, and consumer sources of exposure
and risk from DCM, tetrachloroethylene, trichloroethylene, carbon
tetrachloride, methyl chloroform, and CFC-113.
The integrated strategy led to the establishment of a
committee composed of representatives from EPA, CPSC, OSHA, and
FDA to assess data relevant to the health risks that might be
associated with exposure to the chlorinated solvents. The
committee, called the Health/Risk Assessment Committee (HRAC) of
the Integrated Chlorinated Solvent Project, was charged with
reviewing and updating, if necessary, the existing hazard/risk
assessments on the six solvents. Separate committees were set up
to resolve issues concerning ambient air, occupational and
consumer exposure, and to address economic issues through the
-------
development of a risk/benefit analysis.
In response to the NTP bioassay and the federal regulatory
agencies' investigations, industry and other external groups
developed new data on DCM, focusing primarily on species
differences in metabolism and the mechanism of tumor initiation.
The HRAC has made it a top priority to evaluate these data,
acknowledging the need for review of EPA's final risk assessment
by the Science Advisory Board, rulemaking procedures underway by
CPSC (1986), OSHA (1986), and FDA (1985), and the responsibility
of addressing public comments. The HRAC's evaluations of the new
data serve as the basis for EPA's updated risk assessment of DCM.
-------
2. OVERVIEW OF DICHIOROMETHANE CANCER HAZARD/RISK ISSUES
In May of 1985, EPA's Office of Toxic Substances found that
DCM met the criteria for priority review under section 4(f) of
the Toxic Substances Control Act (TSCA). Underpinning the 4(f)
decision was the conclusion that DCM should be considered a
probable human carcinogen, as defined by EPA's Guidelines for
Carcinogen Risk Assessment.
In assessing the cancer hazard posed by exposure to DCM,
primary consideration was given to the evidence of
carcinogenicity from the NTP's animal studies (1985). The NTP
carcinogenesis bioassays clearly demonstrate that DCM is
oncogenic in two species of laboratory animals, rats and mice,
exposed at different dose levels via the primary route of human
exposure to DCM, inhalation.
In the mouse bioassay, DCM induced a dose-dependent,
statistically significant increase in liver and lung adenomas and
carcinomas in male and female mice exposed through inhalation for
a lifetime at concentrations of 2000 or 4000 ppm. Tumor
incidences were as follows: at 2000 ppm, 30/48 female mice and
27/50 male mice developed lung tumors; 16/48 female mice and
24/49 male mice developed liver tumors. At 4000 ppm, 41/48
female mice and 40/50 male mice developed lung tumors; 40/48
female mice and 33/49 male mice developed liver tumors.
In the rat bioassay, DCM induced a statistically significant
increase in benign mammary gland tumors, of a type not expected
-------
to progress to malignant tumors (McConnell et al., 1986), at the
two highest doses in female rats exposed at 1000, 2000, or
4000 ppm. Male rats developed mammary gland fibroadenomas at
4000 ppm, but only at a marginally significant rate. The NTP
interpreted their study as showing clear evidence of animal
carcinogenicity, and data from the NTP bioassay on mice are the
basis of the regulatory agencies' estimates of human risks at
expected human exposures.
A study of Syrian golden hamsters exposed to DCM at inhaled
doses of 500 to 3500 ppm was negative, but several chronic
studies of mice and rats, including inhalation studies by Dow
Chemical Company (1980, 1982) and a drinking water study by the
National Coffee Association (NCA) (1982 a, b; 1983), reported an
increase in tumors in rats and mice at sites corresponding to the
sites observed in the NTP bioassay. One of the Dow studies
(1980) (inhalation at 1500 to 3500 ppm) reported an increase in
salivary gland sarcomas in male rats. These tumors have not been
repeated in other studies. Results of the Dow and NCA studies,
conducted at doses below those used in the NTP bioassays, were
not statistically significant, with the exception of the salivary
gland tumors in male rats.
Based on an estimated risk comparison with the NTP bioassay
data, EPA concluded that despite the lack of statistical
significance, the results of the Dow and NCA studies were not
clearly inconsistent with those of the NTP bioassays. For
example, comparing the NCA and NTP unit risk numbers (estimated
-------
using the multistage model) for liver tumors in male mice, the
95% upper confidence limit (UCL) for the NCA study was estimated
to be 0.78 x 10~3; the UCL derived from data on male mice in the
NTP study was 0.195 x 10~3 (U.S. EPA, 1985b).
At the time of the 4(f) decision, data on humans exposed to
DCM in the workplace were considered to be inadequate for judging
carcinogenic potential. Data from two epidemiologic studies did
not show evidence of a significant increase in deaths from lung
or liver cancer in exposed workers, but these studies had
insufficient statistical power to detect increased risks as
predicted using the upper-bound estimate derived from the NTP
bioassay on mice.
Based on the evidence, EPA concluded that DCM should be
classified as a probable human carcinogen, group B2. This
classification signifies that evidence of animal carcinogenicity
as provided by the NTP bioassays is sufficient, but data from
human studies are inadequate. CPSC, FDA, and OSHA, after
reviewing the DCM database, came to similar conclusions.
In response to EPA's 4(f) announcement in 1985 and the
initiation of investigations by CPSC, OSHA, and FDA, a number of
comments and studies were submitted to the federal agencies
advancing reasons why the results of the NTP bioassay on DCM in
rats and mice should not lead to the conclusion that DCM presents
a high risk to humans. The major criticisms of the preliminary
assessments suggest that (1) current DCM risk estimates
overestimate risks to humans because they ignore species
-------
differences in metabolism and pharmacokinetics; or (2) the
carcinogenic response shown by mice is unique to that species,
i.e., the mechanism by which DCM causes cancer in mice is not
expected in humans.
Addressing these criticisms calls for a brief review of DCM
metabolism. DCM is metabolized in mice, the species which showed
a clear carcinogenic response, by two routes; one mediated by the
cytochrome P-450 oxidative system [often referred to as the mixed
function oxidase (MFC) pathway], and the other by the
glutathione-S-transferase system (also known as the GST pathway).
Both pathways may be active in mice at low doses, but at higher
doses the MFO pathway becomes saturated and the metabolic load is
increasingly shifted to the alternative GST pathway. Recent
studies (CEFIC, 1986e) indicate that the GST pathway is less
active in rats, hamsters, and humans than in mice.
Arguments against the conclusion that DCM presents a risk to
humans take the position, in general, that the carcinogenicity of
DCM is due to reactive metabolites produced by the GST metabolic
pathway, and that this pathway is significantly active only
following saturation of the MFO pathway, i.e., only at high
doses. Further, the GST pathway is assumed to be the sole
carcinogenic pathway and to be far less active in humans than in
mice, the test species in which malignant tumors have been
observed. Finally, some hypothesize that the metabolites of the
GST pathway are not reactive with DNA, but initiate cancer in
mice through some alternative mechanism such as specialized cell
-------
toxicity or increased cell turnover, events unlikely to occur at
low doses and possibly irrelevant to humans. One might conclude
from these assumptions that the human risk for developing tumors
from exposure to DCM is very low, that it may not exist below
some threshold level, or that there may be no risk to humans
whatsoever.
-------
3. APPLIED-DOSE RISK ASSESSMENT
The position that DCM presents a low risk to humans is in
contrast to the federal regulatory agencies' earlier risk
assessments, which indicate that risks to humans at expected
human exposure levels may be high. The agencies' initial
assessments are based on the applied-dose multistage model
procedure for estimating risk (excepting FDA's assessment which
is based on exposure concentration and a linear extrapolation
procedure similar in principle to the underlying principle of the
multistage model), and range from 2.3 x 10"6 per ug/m3 (CPSC) to
4.1 x ID"6 per ug/m3 (EPA). This range reflects relatively small
differences in assumptions concerning such factors as scaling
between species and the use of upper confidence limits (UCL)
versus maximum likelihood estimates (MLE). The major assumptions
of these assessments, including FDA's, are the same, however.
The applied-dose procedure as used by the federal agencies
assumes that risks are a function of the dose taken into the
body. It treats the body as a whole unit in which cancer may
potentially arise sometime after exposure to DCM. Further, it
assumes that the general similarity of mammalian anatomy,
physiology, and biochemistry justifies extrapolation from rodents
to humans. It incorporates the concept, known as low-dose
linearity, that the carcinogenic response seen at low doses would
be directly proportional to low-dose exposure.
The applied-dose procedure has several drawbacks: it does
10
-------
not account for certain metabolic differences between species; it
does not account for differences in metabolic reactions at high
and low doses; and it does not differentiate between the parent
compound and its metabolite(s) or combinations of these, as the
carcinogenic species. These shortcomings mean that the procedure
does not consider any nonlinearities that may exist between the
dose taken into the body (applied dose) versus the dose that
actually reaches the target tissues (internal dose), and it does
not estimate risk based on the level of a specific carcinogenic
species, i.e., the level of a particular metabolite, for example.
Risks estimated by this method may, therefore, be over- or under-
estimates.
Nevertheless, when definitive knowledge of species-specific
biological traits that might affect carcinogenicity, of high- to
low-dose differences in metabolism, or of the carcinogenic
species are lacking, extrapolating risks to humans using applied
dose and the multistage model provides a reasonable approach for
dealing with these kinds of uncertainties. This procedure may
have some empirical support. For about two dozen chemicals,
carcinogenic potency can be estimated directly from available
epidemiology and these estimates agree, in general, with
projections based on applied-dose extrapolation from animal data
[Allen et al. (1986) analyzed 23 substances including industrial
chemicals, drugs, food additives, and tobacco smoke].
11
-------
4. PHYSIOLOGICALLY BASED PHARMACOKINETIC MODELS
Some of the uncertainties associated with the applied-dose
method may be reduced through information provided by
physiologically-based pharmacokinetic models. Such models
describe the organ level disposition of a chemical and its
metabolites as they vary over time and dose. Given the proper
parameters describing blood flows, metabolic rates, and exchanges
of compound between tissues, blood, and air, well-validated
pharmacokinetic models should be able to account for any non-
linearities in the level of target-organ dose of the parent
compound or any of its metabolites that may occur with increasing
or decreasing applied dose. Further, by adjusting these
parameters to reflect species differences, pharmacokinetic models
may answer some questions concerning interspecies differences
in chemical disposition.
Before applying pharmacokinetic models to risk assessment,
however, one should have a clear idea of how such models can
reduce the uncertainties inherent in risk extrapolation using the
applied-dose procedure. A pharmacokinetic model can account for
differences in the disposition and metabolism of a chemical in
the body as they vary from dose to dose. Thus, it may be quite
useful in reducing the uncertainties associated with
extrapolating from high- to low-dose exposures. Questions, such
as differences in sensitivities of target-tissues (both within
12
-------
species and between species) to concentrations of chemicals,
whether the site of carcinogenic response would be the same from
one species to the next (site concordance), and which chemical is
actually the carcinogenic species, cannot be answered by
pharmacokinetics. The extrapolation of carcinogenic potency
across species depends on many factors in addition to
pharmacokinetic differences. Even when pharmacokinetic models
are well validated as to their ability to determine internal
doses, questions remain as to how this information should be used
to alter the extrapolation of risk from rodents to humans. In
the sections that follow methods of developing and using internal
dose information to assess risks associated with exposure to DCM
are discussed.
Two physiologically based pharmacokinetic models have been
formulated to predict the disposition of DCM and its metabolites
in the body. The model developed by Andersen et al. (1986, 1987)
is based on inhalation exposure to DCM and is essentially a
modified version of an earlier model developed to describe the
disposition of styrene (Ramsey and Andersen, 1984). Andersen and
coworkers have stated that if the results of their model were
taken into account in estimating risk, risk estimations would be
greatly reduced from those published in the EPA Addendum to the
HAD for DCM (U.S. EPA, 1985b) and documents from CPSC, OSHA, and
FDA.
Angelo et al. (1984) developed a physiologically based
pharmacokinetic model describing the disposition of DCM following
13
-------
exposure by intravenous dosing and by gavage. The HRAC's review
(1987) of this model and a series of papers (Angelo et al.,
1986a/ b; Angelo and Pritchard, 1984) revealed the model used by
Andersen and Reitz and the model used by Angelo et al. to be
quite different in structure. The existence of two
pharmacokinetic models for DCM, differing markedly in structure,
demonstrates that the appropriate way to mathematically model
such a complex biological process is not self-evident. The model
used by Andersen and Reitz, however, is based on inhalation
exposure, the route of exposure most important to human risk, and
its results are more directly applicable to the federal
regulatory agencies' estimates of risk than are those from the
model used by Angelo et al. The HRAC (1987) studied the model
used by Angelo et al. chiefly for the insights it provides
concerning the model used by Andersen and Reitz.
The model used by Andersen and Reitz is designed to estimate
the tissue-level doses of DCM (or a metabolite) that result from
any given animal or human exposure. Confidence in the model's
ability to improve risk estimation depends, however, upon the
degree to which the model can be shown to be valid. The HRAC
(1987) thoroughly evaluated the model used by Andersen and Reitz
and found it to provide a plausible description of the uptake and
distribution of DCM. Development of such a model offers great
promise for improvement in the accuracy of estimating
target-tissue concentrations. At present, however, certain
components of the model are not well validated, a matter which
14
-------
must be considered when determining the degree of confidence that
currently can be placed in the model's results.
4.1. MODEL STRUCTURE
Three points about the structure of the model should be
noted but are considered to be of minor importance. First,
Andersen et al. (1986, 1987) chose to model the lung as a two-
compartment region. One compartment acts as the gas exchange
region between blood and air and the other as the metabolic
region for the metabolism of DCM. Although such a model of the
lung is less than a true physiologic representation, it
introduces no obvious source of error.
Second, the GST pathway is represented in the model by a
single rate constant; Andersen et al. assumed that the rate of
metabolism by this pathway will be directly proportional to DCM
concentration, even at high exposures. In contrast, data from
the European Council of Chemical Manufacturers' Federation
(CEFIC, 1986f) indicate the rate of GST metabolism to be
saturable at very high substrate concentrations. The CEFIC rate
is described by Michaelis-Menton kinetics, which requires the
estimation of two metabolic parameters. The two representations
of GST metabolism are essentially equivalent in the range of
substrate concentrations that the model is concerned with,
however.
A third structural element may have more significance. The
model does not account for compartmentalization of DCM within the
tissues of the individual organs. Data from Angelo et al. (1984)
15
-------
indicate that under some conditions of exposure, DCM may
sequester into the lipid rich regions of the various organs.
Compartmentalization may affect the rate of disposition of DCM or
its metabolites in the body's tissues; thus, failure to take
Compartmentalization into account may lead to errors in the
estimation of DCM metabolism over time and exposure.
4.2. MODEL INPUT
4-2.1. Partition Coefficients
Andersen et al. (1986, 1987) determined tissue/air partition
coefficients in vitro by observing the equilibrium distribution
of DCM between homogenized tissue and air. This method does not
determine these coefficients for intact tissue; it may,
therefore, introduce error into the values of the partition
coefficients which, in the model, represent intact tissues. As
might be considered a priori, other studies have shown that using
homogenized tissues does not account for all the physical and
physiologic processes going on in vivo that may affect the
partitioning.
Further, the blood/air partition coefficients, which
Andersen et al. measured, reflect unexpectedly large differences
between the coefficients for mice and rats. CEFIC (1987b)
indicated that their measurements of preliminary blood/fat
partition coefficients do not repeat Andersen et al.«s findings.
The final results of the CEFIC work on blood/fat partition
coefficients are expected to be available in the summer of 1987
and may improve the model's representation of these parameters.
16
-------
4.2.2. Breathing Rates
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 value, while; that for humans is
markedly lower. The Andersen et al. value for humans is for a
person at rest, but the federal regulatory agencies use a value
considered typical of average human activity (almost twice as
high as Andersen et al.'s value) or occupational activity (nearly
three times higher than Andersen et al.'s value).
For the model used by Andersen and Reitz to apply to
expected human exposures, the breathing rates would have to be
increased to reflect the higher breathing rates typical of active
humans. This is an important issue when comparing the federal
agencies' estimates of risk based on linear extrapolation from
applied dose to the results of extrapolations based on the
pharmacokinetic model. In assessing the impact of the
pharmacokinetic model on risk estimations, the human breathing
rate factor should be kept constant and only the use of
pharmacokinetic information should vary. Andersen et al. (1986,
1987) have compared the federal agencies' risk estimates (based
on the higher human breathing rate and the applied dose method)
to risks estimated using the pharmacokinetic model and the lower
breathing rate. This issue is discussed further in the section
evaluating the effect of the pharmacokinetic model on risk
assessment.
17
-------
4.3.3. Metabolic Parameters
The remaining parameters of the model are those that
characterize how much of each metabolic pathway's activity occurs
in lung versus liver tissue and how the rates of metabolism along
each pathway vary with tissue concentration of DCM.
It is difficult to estimate from experiments how the
relative rates of metabolism by the MFO and GST pathways vary
with DCM concentration in the tissues, because the end products
of DCM metabolism are not pathway-specific. In the past (Anders
et al., 1978; Ahmed and Anders, 1978), DCM was thought to produce
carbon monoxide via the MFO pathway and carbon dioxide via the
GST pathway. If this were so, the relative production of carbon
monoxide and carbon dioxide could be monitored as indicators of
the relative metabolic rates of the MFO and GST pathways,
respectively. Gargas et al. (1986) presented data, however,
which suggest that the MFO pathway produces a substantial (but
unknown) amount of carbon dioxide as well as carbon monoxide.
If, in fact, carbon dioxide is produced by both pathways, its
production cannot be used as an indicator of GST metabolism
alone.
The question of whether or not carbon dioxide is produced by
the MFO pathway has other implications as well. Crucial to the
conclusion that humans, rats, and hamsters have little GST
activity toward DCM is the assumption that the carbon dioxide
produced by DCM metabolism in these species comes primarily (or
18
-------
totally) from the MFO pathway. If the MFO pathway does not, in
fact, produce carbon dioxide, the observation of carbon dioxide
production in these species following DCM exposure would be
inconsistent with little or no GST activity. CEFIC has underway
new studies using deuterated 14C-DCM, which, by exploiting the
stable isotope effect, should be able to distinguish carbon
dioxide produced by the MFO versus the GST pathway. These
studies can be conducted in vivo at low doses and should resolve
questions concerning the extent of GST metabolism at exposures
well below the MFO saturation point.
Because of the difficulty in estimating metabolic rates from
experimentation, Andersen et al. (1986, 1987) estimated these
parameters for the model indirectly by a mathematical
optimization procedure. Metabolic rates were set at the rates
that optimized the model's ability to predict the loss of DCM
from a closed inhalation chamber, as the compound was taken up
and metabolized by mice, rats, and hamsters; i.e., values were
selected that gave the best fit of the model's predictions of the
rate of DCM disappearance to actual data on the disappearance of
DCM from the inhalation chamber.
Although this "curve fitting" procedure may result in a good
characterization of total metabolism, it is an inexact means of
allocating metabolism between the MFO and GST pathways. The HRAC
(1987) has shown that if the rate of metabolism along the GST
pathway is reduced fourfold, for instance, from the optimum
values used by Andersen and Reitz (and the MFO pathway parameters
19
-------
are then readjusted), the fit of the model to the chamber data is
virtually as good as before. Fairly large deviations from the
"best" metabolic rate parameter values are required before this
fit deteriorates appreciably. The HRAC's analysis shows that
many alternative sets of metabolic parameters, implying different
amounts of metabolism by the GST and MFC pathways, are
supportable by the data. As a consequence, the model's crucial
output with respect to risk assessment, that is, estimates of
tissue level concentrations of DCM's metabolites, has an
uncertainty factor of several fold.
Because it was inappropriate to obtain human data on the
disappearance of DCM from an inhalation chamber, Andersen et al.
estimated the metabolic rate parameters for humans by
extrapolating from the values estimated for rats and mice.
Andersen et al.'s extrapolation procedure may compound the
uncertainty already inherent in the rodent values discussed
previously (HRAC, 1987). Data from CEFIC (1986c) point to a
possible large error in the model's estimate of human GST
metabolism in the liver, for example.
Once the parameters describing the overall amount of MFO and
GST metabolism in the body are estimated, it is necessary to
specify the relative activities of the two pathways between
metabolically active organs, e.g., liver and lung. Andersen and
coworkers partitioned the 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
20
-------
substrates (2,4-dinitrochlorobenzene for GST activity and
7-ethoxycoumarin for MFO activity).
The surrogate substrates used may not accurately reflect the
activities in each tissue toward DCM, however. On the basis of
the Lorenz et al. data, Andersen et al. set the proportion of MFO
metabolism occurring in the human lung at a very low level
compared to the mouse lung. Lorenz and coworkers noted, however,
that their human lung preparation contained endogenous inhibitors
of the MFO pathway. EPA believes that a higher level of MFO
activity in the lung than Andersen et al. chose is probably more
realistic (HRAC, 1987). Errors in the partitioning of metabolism
are considered important if only one metabolic pathway is
responsible for DCM's carcinogenicity. An error in the
proportion of MFO versus GST metabolic activity occurring in a
particular tissue could lead to over- or underestimates of
tissue-specific risks.
CEFIC (1986e) provided direct in vitro measurements of GST
and MFO metabolic rate parameters in liver tissue from rats,
mice, and humans. The most important of these experiments are
those reporting no detectable metabolism of DCM by cytosolic
preparations of hamster and human liver tissue, implying that the
GST pathway operates on DCM in these tissues at a level below the
limit of detection of the test system, at most. This finding is
in conflict with the predictions by Andersen et al. (1986, 1987)
of substantial GST metabolism in human liver.
Direct comparison of the in vitro measurements with the
21
-------
Andersen et al. in vivo estimates is difficult because of the
problem of extrapolating from metabolic parameters estimated on
isolated cell fractions (which reflect CEFIC's work) to the rates
expected in intact organs. The motivation for attempting such a
comparison, despite the difficulties, is to provide a check on
the metabolic parameters used by Andersen and Reitz. Preliminary
work by the HRAC (1987) indicates that some metabolic parameters
extrapolated from the in vitro data may disagree with the values
in the model used by Andersen and Reitz by up to an order of
magnitude or more, depending on the set of assumptions used in
the extrapolation. Looking at GST metabolism in the human liver
specifically, the GST metabolism rate may be approximately
7.5-fold below the value used in the by Andersen and Reitz, if
the level of GST metabolism is assumed to be at the experimental
limit of detection reported by CEFIC (1986e).
The CEFIC work and its implications for the metabolic
parameters in the model used by Andersen and Reitz must be
interpreted cautiously, however. The in vitro (1986e) experiment
tested tissue from four human livers. CEFIC (1987c) has given
the following details concerning the treatment of the human
tissue samples. The samples obtained were from renal transplant
donors killed in accidents. The donors, all males in the age
group ranging from 13 to 34 years, were previously healthy and
were not receiving any form of medication. Prior to the
transplantation procedure, the donors were maintained on life
support machines and given a mixture of dopamine, rogitine,
22
-------
heparin, dibenzylamine, largactil, and lasix to maintain the
viability of the kidneys. The livers were perfused in situ with
ice-cold medium within 7-8 minutes of cardiac arrest occurring.
Following removal of the kidneys, samples of liver were taken and
kept on ice. Subcellular fractionation of these samples was
carried out within one hour.
Viability of the livers was tested with a series of markers,
namely: ECOD, EROD, ALE, LAH, and cytochrome P-450 content.
Specific GST activity was not assayed at this time, but was
analyzed during the DCM studies using the broad-spectrum
substrate l-chloro-2,4-dinitrobenzene. The levels measured,
relative to those in animal livers, were comparable to those
reported in the literature by other workers using biopsy
techniques (Lorenz et al., 1984).
The DCM-specific GST in mouse and rat liver has been shown
to be stable for at least one year under conditions equaling
those under which the human livers were stored (-70° C). CEFIC
concluded, therefore, that there is no reason to suspect that
there was any loss of GST activity during the removal of the
livers or the storage fractions.
It does not appear from the methodology described by CEFIC
that tissue viability is an issue in considering the results of
the in vitro experiments. A sample of four livers, however,
provides an uncertain basis from which to generalize to all
humans in view of the known polymorphism among humans in activity
of GST toward certain substrates (Seidegard and Pero, 1985). The
23
-------
GST enzyme is actually a complex of similar isozymes, each of
which may have different activity (or no activity) toward any
given substrate. The mix of isozymes varies among individuals,
raising the possibility that some humans may have substantial GST
activity toward DCM while others do not. At present, there is no
direct evidence of variability among humans in GST activity
toward DCM, however.
Also important is the insensitivity of the test system used
to detect GST activity. CEFIC measured GST metabolism in vitro
by monitoring the formation of formaldehyde, the final in vitro
metabolite. The assay used is not strictly formaldehyde-specific
and appears to have a high background level of formaldehyde
formation as well, resulting in a test system capable of
detecting only relatively high rates of GST metabolism. The
results obtained in the CEFIC experiments should not be
interpreted, therefore, as indicating total absence of GST
activity. The in vitro experiments do point to lower GST
activity in human liver than predicted by the model used by
Andersen and Reitz, but how much lower is uncertain, at present.
CEFIC has underway new studies using an improved method of
detection of the GST pathway in human and animal tissues. The
new studies will determine in vitro GST metabolism using 36C1-DCM
which may improve the level of detection of GST activity by up to
two orders of magnitude. In a May 1987 letter to EPA, Green and
Reitz stated that these studies, and similar studies conducted by
the Dow Chemical Company, have in fact detected low levels of GST
24
-------
activity toward DCM in human tissues. No quantitative data are
yet available, however. CEFIC plans to use these data in
conjunction with new in vivo experiments to arrive at estimates
of human metabolic parameters that, while still somewhat
indirect, are based on the only experimental data obtainable from
humans. Such estimates could improve the model used by Andersen
and Reitz by substituting experimentally-based values of
metabolic parameters for those developed indirectly by scaling or
by using surrogate substrates. CEFIC expects to make the results
of their studies available in the summer of 1987.
Even if the level of GST activity is assumed to be at the
level of detection of the current formaldehyde assay, however, it
is difficult to know exactly how to use the in vitro data to
estimate in vivo parameters, as noted above. The method of
extrapolation used is to a certain degree arbitrary; it was
developed in the absence of experimental data indicating the true
relationship between the rates of reaction in isolated cell
fractions and intact organs (HRAC, 1987). EPA considers the
estimates based on this method to be fairly uncertain.
Turning next to the parameters for the human lung, CEFIC was
not able to determine in vitro GST metabolism rates for this
tissue, owing to a lack of availability of human tissues. The
results on liver tissue in vitro are not necessarily indicative
of a similar lack of activity in lung tissue. As noted
previously, the GST enzyme is actually a complex of similar
isozymes, each of which may have different activity (or no
25
-------
activity) toward a given substrate. Although the isozymes in
human liver apparently have low activity toward DCM, they are
highly active when assayed with 2,4-dinitrochlorobenzene, a
substrate that is metabolized by all GST isozymes. Similarly,
the differences among species in the ability of liver GST to
metabolize DCM are not due to differences in total GST, but to a
different array of isozymes. This array of isozymic forms varies
not only across species, but also among tissues; thus, the low
level of GST activity (or lack of it) toward DCM in human liver
does not necessarily preclude such activity in human lung tissue,
or any other tissue.
Andersen et al. (1986, 1987) estimated the level of GST
metabolism in human lung by assuming that the relative activity
of lung and liver toward the general substrate
2,4-dinitrochlorobenzene will reflect differences in GST
metabolism of DCM as well. The substrate 2,4-dinitro-
chlorobenzene measures the activity of all GST isozymes, and
therefore this method provides a poor estimate of the ability of
GST in the lung to metabolize DCM, since the lung may have a
subset of isozymes different from the liver.
4.3. VALIDATION OF THE MODEL
Of more major concern than the certainty of individual
parameters is whether the pharmacokinetic model as a whole can be
shown empirically to be a good representation of the processes
governing the body's disposition and metabolism of DCM over a
range of exposures and times. For the pharmacokinetic model to
26
-------
be fully validated, its output, that is, its predictions of the
amount of metabolism via the MFO and GST pathways, should be
measured against actual data.
Andersen et al. (1986, 1987) present several examples of
their model's ability to predict concentrations of DCM in the
blood at various times for mice, rats, and humans. The modelfs
predictions fit the data fairly well, indicating that the model
gives a reasonably good picture of the uptake and total body
clearance of DCM. The issue of the individual rates of
metabolism by the two pathways is not clarified by such
comparisons, however.
Validation of the kind desired is not possible, at present,
owing to a lack of data on the actual production of pathway-
specific metabolites in the tissues for both pathways. Reitz et
al. (1986) arrived at a rough experimental determination of the
rates of the individual pathways by comparing the modelfs
predictions of carbon monoxide and carbon dioxide production from
the MFO pathway and carbon dioxide production from the GST
pathway with experimentally observed levels of carbon monoxide
and carbon dioxide production in mice. This comparison assumed
that the ratio of carbon monoxide to carbon dioxide resulting
from MFO metabolism remains fairly constant at all doses. Reitz
et al. attributed all carbon dioxide production, above the level
they expected from MFO metabolism, to the GST pathway. The model
predicted the production of carbon monoxide and carbon dioxide
within a factor of 2 for two different doses of DCM. EPA
27
-------
believes, however, that the assumption concerning the proportion
of carbon dioxide that comes from the MFO versus the GST pathways
lacks verification.
To summarize, EPA finds that uncertainties in the model used
by Andersen and Reitz arise chiefly from the input data rather
than from the model's structure. Full validation of the model
used by Andersen and Reitz is not possible, at present, because
DCM uptake and blood level data do not distinguish between
disappearance of DCM via the GST pathway and the MFO pathway.
Without data on the rates of metabolism by each pathway in the
tissues, a wide range of relative activities of the two metabolic
pathways can be shown to fit the data on the parent compound.
In a presentation to the HRAC on February 19, 1987, Dr. T.
Green outlined work underway (sponsored by CEFIC) on validating
the model used by Andersen and Reitz. Dr. Green stated that the
model, although apparently correct in overall structure, fails to
predict much of CEFIC's data on blood levels of DCM and
carboxyhemoglobin (which results from carbon monoxide
production). He noted that in order for the model to more
accurately predict experimental data, the fat/blood partition
ratio had to be perturbed beyond the expected range of this
parameter and concluded that some other parameter, in probability
the metabolic rate constants, must be in error in the model.
CEFIC is currently engaged in developing new experimental values
for partition coefficients, tissue volumes, breathing rates, and
metabolic parameters. This work is expected to be available in
28
-------
the summer of 1987.
From the discussions above, it is clear that the model
used by Andersen and Reitz may be improved by additional data and
validation. Nevertheless, EPA believes that the structure of the
model is sufficiently well developed at present to provide a
means of considering the available knowledge of DCM metabolism
and pharmacokinetics, as it relates to risk assessment, in a way
that is not possible through the applied-dose method. While
confidence in the results of the model are expected to increase
upon further model validation, the development of preliminary
estimates using the results of the model as currently developed
provides insight into the effect on risk estimates of metabolism
and pharmacokinetic information.
29
-------
5. ASSUMPTIONS CONCERNING THE CARCINOGENIC PATHWAY
Fundamental to the position that knowledge gained from the
pharmacokinetic model, despite its uncertainties, leads to
lowered risk estimates for DCM, are the assumptions that the
product(s) of only one metabolic pathway, the GST pathway, is
responsible for tumorigenesis and that this pathway is less
active in humans than it is in mice. CEFIC (1987a) and Andersen
et al. (1986, 1987) assume this to be the case.
CEFIC conducted a series of experiments examining MFO and
GST metabolic pathway activity in animals and in human tissue to
determine whether species differences in metabolism could explain
observed differences in carcinogenic response seen in animal
bioassays. An in vivo experiment (CEFIC, 1986f) compared the
exhalation rates of carbon dioxide and blood levels of DCM in
rats and mice exposed to DCM at high concentrations. In this
experiment, the MFO pathway was presumed to be saturated; thus,
the further increases in levels of expired carbon dioxide above
the levels expected from the saturated pathway were taken as an
indicator of increasing GST activity at high doses. Based on the
fact that rats had much higher levels of parent DCM in the blood
but much lower levels of carbon dioxide exhalation than did mice,
CEFIC concluded that mice have the greater capacity, by far, to
metabolize DCM via the GST pathway. Results of in vitro
experiments were consistent with the in vivo findings. The rate
of GST metabolism by the rat liver was low, about 12 times lower
30
-------
than the rate found in the mouse liver. These preliminary assays
failed to find evidence of GST activity in hamster or human liver
tissue. The insensitivity of the assays used prevents the
conclusion that there is no GST activity toward DCM in these
tissues, but the assays are sufficient to suggest that the level
of activity is low.
The results of the CEFIC studies track well with the
hypothesis that it is GST metabolic activity which produces the
carcinogenic response observed in mice. Mice apparently use the
GST pathway to a greater extent than other species tested (rats,
hamsters, and humans), and, in bioassays, mice developed tumors
at sites shown to metabolize DCM by the GST pathway. Rats, which
developed neither liver nor lung tumors, showed only low levels
of GST metabolism in the liver and no detectable rate in the
lung. Hamsters, which did not develop tumors, showed no evidence
of the GST pathway in either the lung or the liver. Human lung
tissue has not been tested, but CEFIC reports that the
preliminary studies of human liver tissue showed no measurable
activity of the GST pathway toward DCM. Thus, it appears that
tumor incidence correlates with GST activity.
The MFO pathway has been considered to be theoretically
capable of producing reactive intermediates and, thus,
potentially carcinogenic. But the available evidence does not
show correlation between the rate of metabolism via the MFO
pathway and the susceptibility of species to the development of
cancer following exposure to DCM. The CEFIC in vivo and in vitro
31
-------
experiments indicate that the MFC pathway is saturated at about
the same level across the species tested; yet, in bioassays
conducted at doses above the MFO saturation level, there is
considerable variability in the tumor pattern across species.
Moreover, in the species that did develop tumors, tumor incidence
increased with increasing dose despite the fact that MFO activity
was at a constant maximum level. Thus, it is very unlikely that
the MFO pathway is the source of these tumors. Although,
conceivably, a very low level of carcinogenic activity on the
part of the MFO pathway is a possibility, there is virtually no
experimental evidence to support it.
The Andersen et al. (1986, 1987) comparison of the results
from the National Coffee Association (NCA) study (1983) and the
NTP (1985) mouse bioassay also implicates the GST pathway.
Andersen et al. (1986, 1987) determined that mice in the NCA
bioassay, which were exposed to relatively low levels of DCM, had
levels of MFO metabolic activity (both in the liver and lung)
similar to the levels found in mice exposed to high doses of DCM
in the NTP bioassay. The fact that tumors were observed in the
NTP mice but not in the NCA mice (at a statistically significant
increase), indicates that the MFO metabolites could not have been
responsible for the excess tumors in the NTP mice. Andersen et
al. concluded that these must then have been produced either by
the parent compound or by the GST metabolites.
The parent compound was not considered a likely candidate
for several reasons. DCM is relatively inert and does not have
32
-------
direct alkylating activity. Its mutagenicity in the Ames test
appears to be due to bacterial metabolism, and is enhanced after
activation by cytosolic GST (Green, 1983). Moreover, the
appearance of liver tumors in animal studies does not correlate
well with parent DCM concentration; in the NTP bioassay, the rat
liver showed no tumor response despite the presence of DCM in
concentrations higher than those in mice (in which a liver tumor
response was observed).
CEFIC (1986a) also looked for evidence from the mouse lung
where cell toxicity appears to be most intense in certain
specialized cells known as Clara cells. Clara cells are thought
to have a high potential for metabolizing DCM by the GST pathway.
If, in fact, DCM itself were the toxic species (with metabolism
acting to detoxify it), one would expect to find a significant
response in all of the lung's cells rather than in the Clara
cell, where metabolism appears to occur more quickly.
Andersen et al. (1986, 1987) and CEFIC (1987a) are
reasonably certain, based on the evidence, that the GST pathway
is the sole route producing a carcinogenic response. A few
points of caution should be noted, however. The mechanism by
which DCM causes carcinogenesis may not involve DNA alkylation;
the lack of alkylating activity does not, by itself, eliminate
parent DCM as a suspect. The Clara cell response could be due to
some membrane phenomenon allowing easier penetration of DCM
itself into this cell type. Further, the argument concerning the
Clara cell assumes that the cytotoxic response given by the Clara
33
-------
cells is a precursor to tumor formation, which has not been
established. Exposures to other chlorinated solvents
(tetrachloroethylene for example) lead to similar Clara cell
lesions without subsequent tumorigenesis.
But although it is possible to argue with some of the
individual lines of evidence, collectively the evidence makes a
strong case implicating the GST pathway. EPA concludes from the
weight of evidence that DCM's carcinogenicity is most likely
produced as a result of metabolism via the GST pathway. Parent
DCM and the MFO pathway are considered to be possible rather than
probable sources of tumorigenic potential. Neither one
correlates with tumor incidence, either across species or across
doses. Further, the parent compound is thought to be chemically
unreactive.
34
-------
6. ASSUMPTIONS CONCERNING THE MECHANISM OF ACTION
Before proceeding to a discussion of the implications of the
pharmacokinetic model and theories concerning the carcinogenic
pathway for quantitative risk estimation, the relevancy of the
response seen in mice to humans must be addressed. It has been
proposed (CEFIC, 1986a, g) that the response of mice to DCM may
be limited to that species based on hypotheses about the
mechanism by which DCM causes cancer. If DCM can be shown to
cause cancer in laboratory animals by a mechanism that is lacking
in humans, there is no evidence of a cancer risk to humans. The
genotoxicity and cytotoxicity of DCM have both been investigated
as possible mechanisms of carcinogenic action.
6.1. GENOTOXICITY
6.1.1. Bacteria. Yeast, and Drosophila
Reviews of the evidence as to the genotoxicity of DCM (U.S.
EPA, 1985a; United Kingdom Health and Safety Executive, 1985;
CEFIC, 1987a) lead to the conclusion that DCM is mutagenic, with
and without metabolic activation, in most bacterial assays for
point mutation (Salmonella tvphimurium. Escherichia coli). The
United Kingdom Health and Safety Executive (UKHSE, 1985) noted
that, as S. tvphimurium is capable of converting DCM to a
mutagenic chemical species, the positive results obtained without
an exogenous metabolizing system do not indicate that DCM is a
direct-acting mutagen. Rather, while DCM does not appear to
require activation by mammalian liver microsomes, the chemical
35
-------
must still be metabolically transformed in order to obtain
mutagenic activity. Results of point mutation and mitotic
recombination tests in yeast (Saccharomvces cerevisiae^ are
mixed; negative and positive responses were given up to toxic
doses (360,000 ppm), at which one strain showed a clear positive
response. Sex-linked recessive lethal mutation assays in
Drosophila also gave mixed results; only one out of four studies
of recessive lethal mutations in Drosophila reported a positive
result (Gocke et al., 1981) and the reliability of this study has
been questioned (CEFIC, 1986b) because of problems in the
methodology and the weakness of the response observed.
6.1.2. Mamma-Han Cells in vitro
Studies on chromosomal damage have shown DCM to be
clastogenic, with and without the addition of a metabolic system,
in mammalian cells in culture. When tested for its ability to
induce transformation in a variety of cell systems, DCM gave
negative results in tests of mouse cells and in most tests of rat
cells (U.S. EPA, 1985a; UKHSE, 1985; CEFIC, 1987a). It tested
positive in one assay of Syrian hamster cells (Hatch et al.,
1983). In vitro studies of unscheduled DNA synthesis (UDS),
indicative of DNA repair, provided negative results in most
assays of rat hepatocytes, human fibroblasts, and human
lymphocytes (U.S. EPA, 1985a; UKHSE, 1985; CEFIC, 1986b). CEFIC
reports (I987a) that a "marginal" positive result was observed in
a rat primary hepatocyte UDS study (Thilagar et al., 1984), but
notes that details of the study are not available. Chromosomal
36
-------
mutation assays gave mixed results; only one study (Thilagar and
Kumaroo, 1983) of Chinese hamster ovary cells can be judged
positive.
6.1.3. Mammalian Cells in vivo
In vivo mutagenicity assays testing DCM in both rats and
mice have given uniformly negative results, including a recently
completed mouse micronucleus test at gavage doses up to 4000
mg/fcg (CEFIC, 1986b) . Results in other test systems have been
also, for the most part, negative. Recent work from CEFIC
(1986c) on UDS in rat hepatocytes from rats exposed in vivo gave
negative results, as did studies of DNA binding of DCM in rat and
mouse liver and lung (CEFIC, 1986d). CEFIC's UDS experiments
with B6C3F1 mice gave limited but statistically significant
evidence of the induction of mitosis in the liver. CEFIC notes
(1987a) that further experiments are in progress to establish the
biological significance of this effect.
The CEFIC in vivo studies have been criticized (HRAC, 1987)
on the grounds that, given the short exposure periods, doses
tested (2000 or 4000 ppm of DCM by inhalation for periods up to 6
hours) were below the level which might have given a positive
response. EPA does not, therefore, consider the results of these
in vivo studies to be definitive.
6.1.4. Summary
Reviews of the full range of genotoxicity studies of DCM
(U.S. EPA, 1985a; UKHSE, 1985; CEFIC, 1987a) have concluded that
there is clear evidence of mutagenicity in bacteria, but no clear
37
-------
and consistent evidence of mutagenicity in mammalian systems.
DCM has been shown to be mutagenic in bacteria, gave mixed
results in tests of yeast, Drosophila, and mammalian cells in
culture, and gave largely negative results in mammalian cells in
vivo. CEFIC (1987a) hypothesizes that the positive mutagenic
response of bacteria are caused by the metabolism of DCM to
short-lived proximate mutagens, e.g., formyl chloride and
S-chloromethyl glutathione. They consider these species to be so
chemically unstable as to be unlikely to survive long enough to
affect DNA protected by a nuclear membrane in the nuclei of
higher organisms. Scientific evidence to support this hypothesis
is lacking, however.
Given the evidence of in vitro clastogenicity and the
insensitivity of the in vivo UDS and DNA binding studies, EPA
concludes that DCM may be a weak mutagen in mammalian systems,
weak enough to be below the level of detection of the in vivo
studies. A weak genotoxic mechanism is considered to be a
possibility due in part to the failure, at present, to identify
an alternative mechanism of action for DCM, as discussed in the
following section.
6.2 ALTERNATIVE MECHANISMS OF CARCINOGENIC ACTION
CEFIC (1986a, b, c, d) has addressed the question of whether
DCM acts through an epigenetic mechanism (e.g., cell toxicity
leading to tumor promotion) rather than through genotoxicity. A
10-day inhalation study to investigate the effects of DCM on the
rat and mouse liver and lung showed no significant toxic effects,
38
-------
in mouse or rat livers or in the rat lung. Some transient
cytotoxic effects, including cytoplasmic vacuolation, were
observed in the Clara cell of the mouse lung, apparently brought
on by MFO metabolites, but the significance of this response to
potential carcinogenicity is unclear (CEFIC, 1987a; HRAC, 1987).
Similar responses to other solvents (e.g., tetrachloroethylene)
are not associated with tumor production. CEFIC has hypothesized
that the destruction of smooth endoplasmic reticulum in the Clara
cells results in reduction of MFO activity, which in turn leads
to greater GST metabolism than would otherwise be expected. If
the GST products are not genotoxic, it is unclear how such
metabolism would enhance tumorigenicity since cytotoxicity alone
appears to be insufficient.
It is also unclear whether carcinogenesis induced through
cytotoxicity would be specific to the mouse lung. The
histogenesis of Clara cells, which are more prevalent in mice
than other species, and type II pneumocytes, which may be more
prevalent in other species (including humans) than in mice, are
quite similar. Clara cells and type II pneumocytes are both
active secretory cells and may be similar biochemically. Thus,
any process that produces a response in the Clara cells in mice
must be suspected of producing a similar response in the type II
pneumocytes in other species as well.
Seeking information on the mechanism of action in the liver,
CEFIC (1986g) examined the possibility that DCM induces a
carcinogenic response in the mouse liver through increased cell
39
-------
turnover. In an S-phase hepatocyte study, small, variable
increases in the incidence of S-phase cells after DCM exposure
were observed. CEFIC concluded that although these increases
were statistically significant, the biological significance of
such small changes is unclear.
It is widely agreed (CEFIC, 1987a; HRAC, 1987) that much
uncertainty remains about the mechanism of action of DCM. There
is little evidence at this point to support either a genotoxic
mechanism or some epigenetic effect such as cytotoxicity, or to
indicate that the mechanism is limited to mice. Additional data
on mechanism may be furnished by the National Institute of
Environmental Health Sciences (NIEHS) which plans to investigate
further the role of cell replication in DCM tumorigenesis and the
pattern of oncogene activation in spontaneous and dose-related
tumors, and by CEFIC which is conducting studies to evaluate the
effects of DCM in the Clara cell in relation to tumor
development.
40
-------
7. IMPACT OF THE PHYSIOLOGICALLY BASED PHARMACOKINETIC
MODEL USED BY ANDERSON AND REITZ ON HUMAN RISK ESTIMATES
In comparing risk estimates based on the physiologically
based pharmacokinetic model to EPA's risk estimates derived from
the applied-dose procedure, Andersen et al. (1986, 1987)
concluded that EPA's method overestimates internal dose (and
consequently risk) by a factor of 167 for the human liver and 144
for the human lung. (The internal dose is taken in this
comparison to be the amount of metabolite production per liter of
tissue by the GST pathway.) These factors actually represent the
combined effect of several components, only some of which EPA
feels are appropriate to the comparison. The following section
separates the components and examines them individually.
The factors, as Andersen et al. derive them, comprise the
following: (1) factors of 13.5 and 11.3 for liver and lung,
respectively, representing the non-proportionality between
applied dose (mg/kg/day) and internal dose (mg-equivalents of GST
pathway metabolism/L/day) in these tissues; and (2) an additional
factor of 12.7, which Andersen et al. attribute to an arbitrary
"interspecies correction factor" applied by EPA to account for
anticipated species differences in pharmacokinetics for DCM (13.5
x 12.7 = 167; 11.3 x 12.7 = 144). EPA finds the differences
between risk estimates derived by the two procedures to be far
less. The disagreement stems from two factors: (a) breathing
rates, and (b) the appropriate application of the "surface area
41
-------
correction" to dose.
In EPA's Addendum to the Health Assessment Document for
Dichloromethane (1985b), 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 Anderson and Reitz incorporates the
fact that humans have a smaller input of DCM per kg than mice,
but the model uses a set of breathing rates different from those
used by EPA. The model's value for human breathing rate (12.5
mVday) was measured for a man at rest, and is consequently much
lower than EPA's estimate (20 m3/day) which is 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.
Further, the model's breathing rate value for mice (0.084
m3/day) is much higher than EPA's estimate (0.043 m3/day).
Andersen et al. (1986, 1987) 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 the mice and for
humans. A more accurate comparison is achieved when both methods
use the same breathing rates. If the Andersen et al. values
(12.5 m3/day for humans and 0.084 m3/day for mice) are used in
both EPA's applied dose and the calculations used by Anderson and
42
-------
Reitz, the non-proportionality between applied dose and internal
dose is only a factor of 4.3 in liver and 3.6 in lung, rather
than the factors of 13.5 and 11.3, respectively, implied by
Andersen et al.'s comparison. (The decrease of 3.1-fold from the
factors in the Andersen et al. comparison reflects the difference
in the ratio of mouse to human breathing rates between the model
and EPA's assumptions. If the EPA breathing rates are adopted in
both the model and EPA's calculations, the non-proportionality of
applied and internal dose is somewhat larger, 7.7-fold for liver
and 9.4-fold for lung. Substitution of the EPA breathing rates
for the Andersen and Reitz breathing rates does not lead to a
proportional change in internal dose estimates, hence the
difference in the two sets of estimates. For the sake of clarity
in analyzing the Andersen et al. comparison of the EPA applied-
dose risk estimates and the pharmacokinetic model-based internal
dose risk estimates, further discussion of the comparison
incorporates the Andersen and Reitz set of breathing rates. But
when the pharmacokinetic model is used to actually generate risk
estimates, EPA uses the EPA set of breathing rates. These
calculations are discussed further in Chapter 8„)
Andersen and coworkers note that EPA assumed the human dose
(in mg/kg/day) producing a given risk level to be lower than the
mouse dose producing that same risk by a factor of 12.7. This
assumption is known as the "surface area correction," and
corresponds to the assumption that it is the amount of applied
dose per unit of surface area, rather than per unit of body
43
-------
weight, that is equivalent in risk across species. Andersen et
al. interpret this to mean that EPA expects internal doses to be
12.7-fold higher in humans for a given applied dose, with risk
being directly proportional to tissue concentration, independent
of species. The higher internal dose in humans is presumably
expected owing to pharmacokinetic differences between species
that arise as a result of the scaling of key pharmacokinetic
parameters in proportion to body surface area rather than to body
weight.
In other words, Andersen et al. have interpreted the surface
area correction as a correction to applied dose to produce an
estimate of anticipated internal dose. They show that, when
internal doses are actually estimated with their model, they are
not in fact 12.7-fold higher in humans, as anticipated, but
rather an estimated 4.3-fold* lower in liver and 3.6-fold* lower-
in lung. To Andersen et al. (1986, 1987) this implies that use
of the surface area correction is incorrect for DCM, and that
EPA's risk estimates are too high by 12.7 x 4.3 = 54-fold* and
12.7 x 3.6 = 46-fold for liver and lung, respectively. (Their
paper cites the figures as 167-fold and 144-fold since the
3.1-fold inflation due to breathing rates differences had not
been corrected for).
*These numbers represent estimates based on the model and
assumption for breathing routes used by Andersen and Reitz.
When EPA's breathing rate assumptions are used in the
pharmacokinetic model, these factors become 7.7 for liver and
9.4 for lung without the surface area correction factor; 97 8
(7.7 x 12.7) for liver and 119.4 (9.4 x 12.7) for lung with the
surface area correction factor.
44
-------
In summary, EPA finds that Andersen et al.'s estimate of
167-fold difference between liver cancer risks (for example)
estimated from the applied-dose method and the pharmacokinetic
model can be attributed to the following factors:
Pharmacokinetic nonlinearity 4.3-fold
(as predicted using the model)
Use of different breathing parameters 3.1-fold
(by Andersen et al. versus EPA)
Surface area correction 12.7-fold
Composite 167-fold
It should be clear that only the factor of 4.3 comes from
the model used by Andersen and Reitz. That is, when comparing
mice at high bioassay doses to humans at low exposure levels, the
non-proportionality of applied dose to internal dose amounts to
only a few fold. Although the model reveals some differences
between species and from high to low doses in the proportion of
an applied dose of DCM that is metabolized by the GST pathway,
these differences are apparently not large.
The other two factors are features of the applied-dose
procedure, against which the modelfs results are being compared.
That is, they are part of the question about how much impact the
finding of a 4.3-fold non-proportionality between applied and
internal dose has on risk estimation. The validity of their
inclusion in the calculation of the impact of the model used by
45
-------
Andersen and Reitz on risk extrapolation rests on whether they in
fact correctly represent the use of applied doses by EPA. As
noted previously, EPA believes that the 3.1-fold factor that
arises from the use of different breathing rates in the applied
dose calculations and in the model is an extraneous factor, and
should be eliminated from the comparison by using a single set of
breathing rates.
This leaves for consideration the contentious interspecies
correction factor which, in the EPA procedure, is equal to the
relative surface-to-volume ratios of mice and humans, giving a
factor of 12.7 by which human applied doses in mg/kg/day are
divided to be of equal lifetime risk to mouse doses. The
question revolves around what the use of this factor implies
about the expectations for the underlying pharmacokinetics, the
relation of applied to internal dose, and the reasons for
interspecies differences in carcinogenic potency. EPA feels that
the surface area correction is not simply a correction on applied
dose to account for an expected higher internal dose in humans,
as presumed by the analysis of Andersen et al. (1986, 1987).
That treatment oversimplifies a complex issue, which is discussed
in the following chapter.
46
-------
8. USING INTERNAL DOSES AS A BASIS FOR HUMAN RISK ESTIMATION
The interspecies correction factor is applied to account for
the difference in expected potency of a given applied dose in
experimental animals and humans. Many biological differences
exifit among species that can be expected to enter in to such
differences in potency. Compared to mice, humans have a slower
excretion rate of compounds from the body, their rates of
metabolism are generally slower, and they have a lower rate of
cell division. They may have more efficient DNA repair and may
be more effective in scavenging free radicals and other reactive
compounds. On the other hand, they have many more cells at risk,
only one of which need undergo carcinogenic transformation to
produce a tumor, and they may be exposed to carcinogens for a
much longer lifetime. Many other potential factors could be
named. Some, but not all, of the above factors are
pharmacokinetic; that is, they can be expected to influence the
relationship of internal doses in the tissues to the amount of
dose applied. Other factors, however, can be expected to be
manifested in species differences in the degree of carcinogenic
responsiveness of the tissues to a given internal dose. The
potential contribution of these other factors continues to be
problematic even after interspecies pharmacokinetic differences
are accounted for. It is clear that, even when internal dose
information is available, no matter how reliably known, there
remain important questions about the degree to which different
47
-------
species will respond to those internal doses.
The interspecies extrapolation factor must take into account
the combined effect on carcinogenic potency of all underlying
biological differences between species. The few data that exist
(Crump et al., 1985; Allen et al., 1986) suggest that fairly good
estimates of human risk are achieved (on average) when applied
doses are scaled according to some measure of body size (surface
area or weight) and/or some measure of life span. No one factor
in the mechanism of carcinogenesis can be identified as the key
to interspecies differences in potency, however. Since
pharmacokinetic data address only part of the interspecies
extrapolation question, such data alone cannot solve the
difficulties of extrapolating risks across species. Some
elements of the former assumptions must be retained in the
extrapolation procedure, even when internal dose estimates are
available. The question is, how much of species differences in
potency (if any) can be ascribed to species differences in
pharmacokinetics, and how much (if any) can be ascribed to the
remaining factors that lead to differences in "sensitivity" of
the tissues?
The disagreement about how and when to apply the surface
area correction factors to applied dose illustrates that there is
some question about the basis for specifying those
pharmacokinetic differences between species implicit in the
surface area correction procedure (i.e., the differences expected
as a consequence of the differences in scale rather than to
48
-------
chemical-specific metabolic factors, saturation of metabolism,
and so on). Using internal dose estimates in place of applied
doses in risk extrapolation should change the risk estimates by
the degree that the original assumptions about pharmacokinetics,
implicit in the applied-dose procedure, are shown by the new data
to be in error. The impact of pharmacokinetic data, then,
depends on one's view of what can be said about the prior
presumptions about pharmacokinetic differences between species,
and how those presumptions relate to the use of the surface area
correction on applied dose. To discuss this, the rationale for
the use of surface area scaling must be examined.
One way to develop an interspecies extrapolation factor is
to examine the problem conceptually, from first principles.
Although rodents and humans share a basically similar mammalian
anatomy, physiology, and biochemistry, they differ considerably
in size. The field of allometry studies the way in which the
magnitude of different features varies with animal size (e.g.,
Schmidt-Nielsen, 1984; Mordenti, 1986). While many features tend
to scale across species in proportion to body volume (or weight),
others, including many rates of physiological processes, tend to
scale in proportion to an animal's body surface area. The aim of
the allometric approach to interspecies extrapolation is to
discover what pharmacokinetic differences are expected between
rodents and humans solely as a result of their differences in
size.
A second way to approach the interspecies extrapolation
49
-------
factor is as an empirical correction. That is, justification for
the use of a particular factor (such as the surface area
correction) comes through the extent to which it can be shown to
be successful in predicting human cancer potencies from those
determined in animal experiments. Such comparisons require
epidemiologically based direct estimates of human risks against
which to test the predictions from animals. It is not necessary
to specify or understand the underlying biological processes that
produce interspecies differences in cancer potency—only the
overall combined effect, which comprises unknown contributions of
pharmacokinetics and species differences in responsiveness, is
assumed to be given by the surface area correction.
These two approaches lead to somewhat different methods for
incorporating pharmacokinetic data into risk extrapolation, which
are outlined below.
8.1. METHOD 1: COMPARISON OF INTERNAL DOSES TO ALLOMETRIC
EXPECTATION
Surface area scaling of doses is routinely used in the area
of predicting appropriate human dosage levels of experimental
drugs from data on their effect and toxicity in experimental
animals. The allometric rationale, which is supported by
experimental evidence (e.g., Dedrick et al., 1970; Dedrick,
1973), is that the volume of distribution of a drug tends to
scale between species in proportion to body weight, while the
rate at which the drug can be cleared from the body (i.e.,
removed by excretion and/or metabolism) tends to scale up in
50
-------
proportion to body surface area. Larger animals (such as humans)
have a smaller surface area to volume ratio than smaller ones.
Compared to a 35-g mouse, for instance, a human must clear drug
from a volume some 2000 times larger, but the rate of clearance
(in mL of blood per minute from which drug is removed) is only
159 times faster. The result is that a given initial blood
concentration attenuates much more slowly in the human, and the
area under the blood concentration-time curve (which is the
integrated or cumulative amount of exposure of the tissues) is
larger. Scaling the applied dose by surface area results in a
smaller initial blood concentration in the human, but a similar
total tissue exposure (area under the curve) as is experienced by
the mouse. It is this rationale for surface area scaling of
applied dose that is attributed to EPA by Andersen et al. (1986,
1987) .
The above argument concerns exposure to the parent compound,
but carcinogens are often thought to require metabolic activation
to a more reactive chemical form; it is exposure of the tissues
to such metabolically activated species that is at issue. A
somewhat different rationale for surface area scaling, also
arising from allometric observations, is based on the idea that
metabolic rates can be expected to scale across species in
proportion to surface area (Davidson et al., 1986). It has been
argued (e.g., Ramsey and Gehring, 1980) that humans are thereby
expected to have a lower rate of activation of procarcinogens,
and hence should be less—rather than more—responsive to a given
51
-------
concentration of parent compound, as the usual use of surface
area scaling of applied doses suggests.
It should be pointed out, however, that both the slower
clearance of parent compound by humans, as well as its slower
metabolism to activated carcinogen, are occurring simultaneously.
Humans may have a lower metabolic rate constant than mice, but
the rate of metabolism depends both on this constant and on the
concentration of parent compound. In fact, if it is assumed that
both metabolism and non-metabolic clearance scale as surface area
(as is presumed by their use in providing a basis for the surface
area correction), their individual "surface area corrections"
cancel out. That is, humans have a greater area under the blood
concentration-time curve for a given applied dose, but this
exposure to parent compound is not the relevant one. The rate of
carcinogenic activation of this parent compound by metabolism is
lower (per kg) than in mice for a given parent compound
concentration in the tissues, but that concentration stays high
longer, owing to the slow clearance, with the result that the
same fraction of a dose is metabolized in each species.
Another way to view the same phenomenon is to realize that
metabolic and non-metabolic clearance (i.e., excretion) are both
acting to remove parent compound from the body; in fact, all of
the compound introduced into the body must eventually find its
way out through one or the other process. They both act more
slowly in humans, since they are both allometrically related to
surface area. But, so long as they both scale the same way, the
52
-------
ratio of metabolic to non-metabolic clearance is the same in mice
as in humans. The same proportion of an administered dose will
end up being metabolized in each species, although the process
will take longer to complete in humans.
The case of inhalation exposure is slightly more complex,
but the same principle applies. In inhalation exposures of
sufficient duration, a steady-state blood concentration is
reached in which input of new parent compound is offset by loss
through metabolism and excretion (which, as is the case with DCM,
is usually through the lungs). Cardiac output, minute volume of
breathing, and rate of metabolism (for a given parent compound
concentration) all tend to scale allometrically in proportion to
a species1 surface area. (Actually, the argument requires only
that all the factors scale in the same way, be it surface area or
otherwise, but scaling close to surface area is empirically
observed.) Under these circumstances it can be shown that, while
breathing a given vapor concentration, steady-state blood
concentrations of parent compound are expected to be equal in all
species. The amount of parent compound metabolized into the
active carcinogen per kg of body weight per unit time is less in
humans (by the amount of the surface area correction), owing to
their smaller metabolic rate constant. However, the applied dose
experienced by a human breathing a given air concentration for a
given time is also smaller than that of a mouse, and by the same
amount, owing to the lower breathing rate per kg in humans. (As
noted earlier, applied dose is figured by multiplying breathing
53
-------
rate by air concentration and duration of exposure, giving the
amount, in mg/kg, that is inhaled during the course of exposure.)
In other words, under a variety of circumstances, from bolus
dosing to continuous inhalation, the balance of metabolic and
non-metabolic clearance leads to an equal proportion of an
applied dose being metabolized in rodents and humans. Focusing
on the scaling of metabolic rate constants while ignoring the
parent compound concentration (or vice versa) looks at only one
element of the two interacting processes.
Of course, it is not necessarily true that the important
rates all scale across species in just the same way, i.e., in
proportion to surface area. For example, mice breathe somewhat
more air per minute than predicted by strict surface area
scaling, and many physiological measures scale closer to the 0.75
power or the 0.59 power of body weight than to the 0.67 power.
Furthermore, peculiarities in species with regard to the
properties of their metabolic enzymes, saturation of metabolism
at higher doses, shifts among alternative metabolic pathways, and
the like, will all cause some deviations from the pattern
outlined above. Rate constants of specific metabolic pathways
may not scale in the same way as the overall metabolic process.
But the question here is about the a priori expectation for the
relation of applied to internal dose across species. The
relevance of surface area to the question of choosing an
interspecies scaling factor to be used on applied doses comes
from the idea that crucial elements of the physiological
54
-------
processing of an applied dose may have this allometric
relationship to body size. The above argument suggests that,
considering the way this scaling is usually conceived, applied
dose and internal dose ought to be directly proportional across
species. In other words, differences in body size do not, in
themselves, result in any inherent difference among species in
the proportion of an applied dose that is metabolized, all else
being equal. The extent to which actual pharmacokinetic and
physiological data show all else not to be equal is the extent to
which the risks estimated under the applied dose basis for
extrapolation ought to be changed.
It is interesting to note that the model used by Andersen
and Reitz predicts internal doses that very closely adhere to the
argument developed above. According to the model, humans at 4000
ppm have a ratio of applied to internal dose (using the model's
breathing rates) that is 0.60 times that of mice in liver tissue,
and 1.6 times that of mice in lung tissue. That is, the a priori
assumption that applied and delivered dose are in the same ratio
across species appears to differ by less than a factor of 2. At
low doses the model's predicted difference reaches about a factor
of 5 in the liver, but the ratio for the lung is virtually equal
to the assumption of 1.0. Thus, despite the lack of adherence to
strict surface area scaling assumptions, saturation of
metabolism, multiple metabolic pathways, and so on, the a priori
expectation appears to be incorrect by only a small amount.
In view of the expectation of internal doses in mice and
55
-------
humans outlined above, how is the surface area correction on
applied dose to be regarded? It is not a correction to account
for presumed differences between species in internal dose, since
internal dose is expected to be a constant proportion of applied
dose. Rather, it can be viewed as a correction to the expected
species differences in risk from a given internal dose. That is,
the assumption that humans have an equal cancer risk from a dose
that is 12.7-fold smaller per kg of body weight can be ascribed,
to a greater "responsiveness" to that internal dose. (Actually,
one must keep in mind that risks are presumed equal for lifetime
exposures at a daily rate that is 12.7-fold lower in humans.
Since humans live some 35 times longer than mice, their total
cumulative exposure leading to the same presumed risk is actually
35/12.7 = 2.8 times greater than for mice. Thus, the assumption
is that they are slightly less sensitive than mice to a given
lifetime tissue-level exposure.)
Why should one assume a 12.7-fold greater "responsiveness"
to internal dose in humans than in mice? The assumption that
applied doses ought to be scaled by surface area to be of
equivalent risk has been used by EPA for over a decade on many
chemicals, including most of those to which DCM is being compared
when possible substitution of a chemical with less risk is
considered. If the above argument that the general scaling
principles that have been invoked in fact lead to an a priori
expectation of internal doses being proportional to applied
doses, even across species, then the last decade's use of the
56
-------
surface area correction to predict expected human risks has been,
in practice, a correction for the relative tumorigenicity across
species of a given internal dose. This is true even if the
stated reason for the surface area correction has been that it
corrects for "metabolic differences" and some other factor.
Since pharmacokinetic data on DCM provide no new information on
species differences in responsiveness, the original de facto
assumption ought to remain unchanged. In other words, the factor
of 12.7 should continue to be applied to internal doses from the
pharmacokinetic model because this corresponds to the assumption
that has in effect been used all along. There is no information
to justify changing this assumption at this time.
One may, of course, question whether the assumption is
correct. There are some reasons to think that it may be
reasonable. As noted above, it nearly corresponds to the
assumption that humans and mice are equally sensitive to a total
cumulative lifetime dose in mg/kg (it is off by a factor of only
2.8 from this). One may also note that some of the components of
tissue sensitivity to a carcinogen, such as cell division and
turnover rates, DNA repair rates, scavenging of free radicals,
and so on are related to tissue aging rates and to life span.
Boxenbaum (1983, 1984) relates life span and aging to a different
scale of "physiological time" for each species, which tends to
vary across species in approximate proportion to body surface
area. These observations are not the reason for adopting a
surface area factor for species differences in responsiveness—
57
-------
that reason is the historical precedence cited above. But these
observations do indicate the sorts of factors that might be
examined to reach further understanding of tissue sensitivity.
If new data become available to alter this assumption, then risk
calculations using internal doses may be changed accordingly, but
so must risks calculated using the applied-dose procedure.
Returning to the particular case at hand, how should the
results of the model used by Andersen and Reitz alter EPA's
estimation of human risk from the levels calculated using applied
doses? Under the allometric view of interspecies extrapolation,
and assuming that metabolism by the GST pathway at the target
tissue is the relevant internal dose measure, one should simply
replace the applied dose with the internal dose in the usual
extrapolation procedure. To the extent that the model predicts
internal doses that are not in fact proportional to applied dose,
the resulting risk estimates will change. That is, both species
differences in the proportion of the applied dose that is
metabolized and high- to low-dose differences in this proportion
affect the low-dose human risk estimate. The steps are: (a) use
the model to estimate internal doses received by mice in the NTP
bioassay; (b) fit a dose-response curve by the usual methods,
using the NTP mouse responses and these internal doses; (c) use
the model to estimate internal doses in a human subjected to a
low-level continuous exposure, say 1 ppro; (d) scale that internal
dose by the surface area correction; and (e) calculate the risk
from that scaled dose, according to the dose-response curve.
58
-------
The scaling in step (d) is applied not to alter the estimate of
human internal dose, but to account for the presumed greater risk
that the dose engenders in a human vis-a-vis a mouse. That is,
the correction is applied in the same way as for the applied dose
procedure, as a correction for species differences in
responsiveness to a given dose (internal or applied).
The actual calculations of this method are shown in the HRAC
(1987, Chapter 7) report. The model used by Andersen and Reitz,
as presented by Andersen et al. (1986, 1987), was followed with
one exception: EPA believes that its 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 breathing rates in the model were
adjusted accordingly. Cardiac output was then adjusted in the
same proportion as the breathing rates.
The lung and liver do not generate equal amounts of
metabolites during an exposure, so it is necessary to extrapolate
risks for each organ separately. The risk based on mouse lung
tumors is slightly over twice that based on liver tumors, despite
the fact that the liver has much higher internal doses, owing to
its higher metabolic activity. In fact, the risk in lung tissue
per unit of internal dose is about 20-fold higher than for liver
tissue. In mice used in the NTP bioassay, the occurrence of
liver and lung tumors was independent; developing a tumor at one
site did not affect the statistical probability of developing a
tumor in the other tissue. If this is also true in humans, then
59
-------
the overall unit risk can be derived by simply adding the tissue-
specific unit risks. This process yields a human incremental
risk for continuous inhalation of 1 ug/m3 of 4.7 x 10~7. This
unit risk is 8.8-fold lower than EPA's published unit risk based
on applied dose (4.1 x 10~6 per ug/m3), which is based on the
same NTP female mouse bioassay data.
The above analysis extrapolates both across species and
across doses using the internal dose estimates from the model
used by Andersen and Reitz. The difference in estimated
carcinogenic potency of DCM between the NTP bioassay doses to
mice and the lower doses to which humans are exposed is
influenced by both species-to-species differences in metabolism
(which are minor in the model as currently constituted—the
results nearly correspond to the a priori assumption based on
allometry) and high- to low-dose differences in metabolism within
humans, resulting from the saturation of the competing MFO
pathway.
One of the more uncertain parameters in the model used by
Andersen and Reitz is the first-order rate constant kF,
describing GST metabolism. The human value of kF cannot be
estimated directly, and so Andersen et al. (1986, 1987) derived
an estimate based on allometric scaling from observed values in
rodents. (These authors used body weight to the 0.7 power,
rather than the 0.67 power corresponding to a strict geometric
surrogate measure of surface area for similarly shaped objects.)
Thus, this element of the human model is essentially filled with
60
-------
the a priori assumption of the allometric approach, which results
in the model's prediction that approximately equal proportions of
an applied dose are metabolized by the GST pathway in mice and
humans. (Most of the 8.8-fold lowering of estimated risk that
emerges from method 1 is attributable to high- to low-dose non-
linearities in metabolism by the human GST pathway, rather than
to species differences in metabolism). The CEFIC data on in
vitro GST metabolism in human liver fractions suggest that in the
model used by Andersen and Reitz, the scaled value of kp for
humans may be a good deal too high, however. If this proves to
be the case as more reliable data become available, it would
serve as an example of chemical- and species-specific deviations
from that pattern of metabolic variation that is expected as a
result of allometry. The human risks calculated by extrapolating
across species on internal dose would change in approximate
proportion to the change in the model's value of kp. That is,
human risks would be lowered to the degree that humans metabolize
a smaller proportion of an applied dose by the GST pathway than
do mice.
As discussed previously, gauging the quantitative impact of
the currently available CEFIC data is difficult, since they are
in vitro data that do not directly yield an estimate of the in
vivo value of kp. Furthermore, there is some question about the
limit of detection of the assay. These issues prevent the use of
the current CEFIC data as a basis for estimating the human kp
directly, rather than by the scaling procedure used by Andersen
61
-------
et al. (1986, 1987). Nonetheless, a preliminary estimate
suggests that a lowering of the human kF by at least sevenfold or
so may be indicated, which, if correct, would lower human risk
estimates by about the same degree. If human GST metabolism is
in fact well below the limit of detection of these studies,
rather than at the limit, then further proportional lowering of
human risk may be indicated. New studies from CEFIC, expected
this summer, will examine human GST metabolic activity toward DCM
with a much more sensitive in vitro assay in both liver tissue
and (for the first time) in lung tissue. CEFIC intends to use
these studies in conjunction with new in vivo work to develop
more direct estimates of human GST metabolism in vivo, which can
then be used (along with other new parameter estimates) to more
reliably formulate the model's representation of interspecies
differences in metabolism of DCM. Extrapolation of risk to
humans, when calculated by the extrapolation method described
above, will vary from current estimates in approximate proportion
to the degree that mice and humans are shown to metabolize a
different proportion of a delivered dose.
8.2. METHOD 2: USE OF PHARMACOKINETICS ONLY FOR HIGH- TO LOW-
DOSE EXTRAPOLATION
Interspecies extrapolation is considered to be a problematic
conversion. It is difficult to specify how a given metabolic
difference should be combined with a host of non-pharmacokinetic
factors in determining carcinogenic potency differences between
mice and humans. if uncertainty exists about the assumptions as
62
-------
to how this should be done, one can turn to a second method of
incorporating pharmacokinetic considerations into risk
assessment. This method (Method 2) presumes that the current
lack of understanding of interspecies differences in tissue
sensitivity precludes using internal doses to extrapolate across
species, but assumes that the results of the pharmacokinetic
model can be used for high- to low-dose extrapolation.
All extrapolations between species must make some
assumptions regarding the dosimetry between species. Some factor
(F) is taken as a measure of interspecies correspondence. In the
case of extrapolation performed on an applied dose basis, F is
based on the fact that differences between species result from
differences in both pharmacokinetics (PK) and tissue response or
pharmacodynamics (PD). The EPA and CPSC have traditionally based
the value of F on the differences between the surface areas of
different animal species (HRAC, 1987, Chapter 7). The FDA, on
the other hand, bases the value of F on the differences between
the weights of different animal species. Other values of F could
be selected, for example, the differences in organ weight or
total protein mass of a tissue. Regardless of the particular
basis for F, once pharmacokinetic differences are accounted for
by models, some adjustment of F may still be needed. Obviously,
if F is due to differences in PK and PD and if either PK, PD, or
both are adjusted by accurate quantitative means, then F must
also be adjusted. The fundamental issue to be resolved involves
how to adjust F when only PK is known.
63
-------
Method 2 is based on the assumption that by merely knowing
the species differences in PK, without any knowledge of species
differences in PD, it is not possible to make any adjustment in F
from the value used in the extrapolation based on the applied
dose basis. Put in other terms, knowing about pharmacokinetic
differences in species still does not necessarily mean that
equivalent doses in different species would yield the same
response in different species. For example, does 1 ug of toxin
per liter of mouse liver yield the same number of detrimental
events per organ as would 1 ug of toxin per liter of human liver?
Method 2 takes the position that information is not yet available
to answer that question (pharmacodynamics); thus, some value for
F must still be applied. Further, it is not possible to discern
how much of F is due to pharmacokinetic differences and how much
is due to pharmacodynamic differences. Physiologically based
pharmacokinetic models are not able to account for
pharmacodynamic differences.
Figure 1 is a diagrammatic simplification of extrapolation
between species. The horizontal arrow between applied doses
represents the extrapolation factor F, which has been
traditionally applied prior to incorporating pharmacokinetic
data. The three horizontal dotted arrows represent interspecies
correlations for the pharmacokinetic and some of the
pharmacodynamic factors. Assuming that the pharmacokinetic
factors are understood for a particular case, one is left with
the problem of assigning factors to account for interspecies
64
-------
SPEC IES 1
SPEC IES 2
i ed Dose
APR Ii ed Dose
Phys i o I Q
-------
differences in the pharmacodynamics. Little information
currently exists from which accurate estimates of these factors
could be calculated.
At present there is some preliminary empirical evidence to
suggest that applied dose factors (F) used by the various federal
regulatory agencies are not in great error. However, it is not
at all clear that those factors can be applied to account for
pharmacodynamic events without some quantitative and perhaps
qualitative modification. Method 2 implicitly assumes that the
present data do not allow for determination of that adjustment.
How then can pharmacokinetic information be used if not for
interspecies correlations? After reviewing the data and models
with regard to DCM, it became apparent that there were
pharmacokinetic differences related to differences in physiology
and metabolism of the different species. It was also apparent
that there were differences in the pharmacokinetics of high-dose
exposure as compared with those of low-dose exposure. Method 2
accounts for differences in the high- to low-dose extrapolation
within a species. However, for reasons described in previous
paragraphs, interspecies differences in pharmacokinetics are not
directly taken into account. The method is more thoroughly
described by the HRAC (1987, Chapter 7), but briefly is
accomplished as follows: Because it is not deemed possible to
discern the true value of PD once PK is accounted for, the
interspecies extrapolation is done by using F, that is, on an
applied dose basis at the NTP dose. The physiologically based
66
-------
pharmacokinetic model is then used to extrapolate from the high
dose in the human to the low dose at which the unit risk is
calculated. When this is performed, the effect of accounting for
the nonlinearity in the pharmacokinetics from high to low dose
results in a reduction of risk from the applied-dose method. The
reduction is 2.1-fold for the lung and 4.4-fold for the liver.
This reduction is the same regardless of what basis of F was used
(Chapter 7 of the HRAC report provides a detailed derivation of
these numbers). The sensitivity analyses (HRAC, 1987, Chapter 7)
indicate that the 2.1-fold reduction is a minimum reduction that
appears to be necessary, and so it is used as a basis for
altering the low-dose human risk estimate. The resulting
estimated human risk from continuous lifetime exposure to 1 ug/m3
is 1.8 x 10~6. This is 2.1-fold below the EPA estimate based on
applied dose of 4.1 x 10"6.
When extrapolations are performed, several assumptions are
made. The examination of these assumptions and their
implications on the risk assessment process is as critical as the
estimation procedures themselves. In Method 2, some assumptions
are generic in nature and others apply to the present case of DCM
exposure.
First, this method assumes that at present it is not
possible to discern if a unit of toxin per unit of mouse tissue
would result in the same response as an equal unit of toxin per
same unit of human tissue. As previously stated, this method
assumes that the interspecies correlation factor used, results
67
-------
from interspecies differences in both pharmacokinetics and
pharmacodynamics. At present for DCM, this method assumes that
the data are insufficient to determine how much of the
correlation factor is due to differences in pharmacodynamics.
Thus, it does not apply the interspecies correlation factor only
to pharmacodynamics but to applied dose. A disadvantage of this
assumption is that interspecies differences in pharmacokinetics,
which can be accounted for quantitatively by modeling, are not
weighed in the extrapolation procedure.
Second, this method actually quantifies pharmacokinetic
differences that result in nonlinearities between high and low
doses. As a result, for this case, the method appears to be
somewhat insensitive to variations in the linear metabolic term,
the first-order rate constant for metabolism by the GST pathway.
This is both advantageous and disadvantageous. As described in
subsequent chapters, the rate constant associated with the GST
pathway is one of the most uncertain parameters in the
physiologically based pharmacokinetic model. Because of the
approach's insensitivity to this parameter, any errors do not
greatly affect (within certain limits) the risk estimation.
However, this insensitivity means that the approach will not
reflect the impact of any new data that may be generated which
show that GST activity in humans is significantly different than
the present estimates.
Third, because of the nature of the pharmacokinetics of DCM
as described by the pharmacokinetic model (Andersen et al., 1986,
68
-------
1987), the reduction of 2.1-fold for lung and 4.4-fold for liver
would also apply if the parent compound contributes to
carcinogenicity.
Fourth, the method depends on several assumptions regarding
the relative roles of the two metabolic pathways to
carcinogenicity. The modification of the risk number calculated
assumes the products of the MFO pathway to be of less importance
than the products of the GST pathway in the possible mechanism of
carcinogenicity. Although the evidence for this assumption, as
discussed in the HRAC (1987) report, is strong, some doubt does
linger.
Fifth, the model used by Andersen et al. (1986, 1987)
partitions metabolism between lung and liver for both pathways.
The methods used to determine the relative apportioning of
metabolism between the two organs and the sensitivity of the
model to those parameters are discussed in the HRAC (1987)
report. The parameters used in the model for the estimation in
this method are those used by Andersen et al. (1986, 1987) and
are subject to uncertainty. The impact of any error in those
parameters on the risk estimation could be significant.
Sixth, the method assumes that the structure of the model is
correct. This implies that carbon dioxide observed at low doses
arises from the MFO pathway. If true, this would be consistent
with the theory that the GST pathway is not operational at low
doses. The HRAC (1987) report discusses in great detail the
uncertainties with this assumption. Obviously, if this
69
-------
assumption were incorrect, one interpretation could be that the
GST pathway is operational at low doses, and thus the model's
structure would have to be significantly modified.
In summary, this method assumes that most of the parameters
used in the model of Andersen et al. (1986, 1987) are within
reasonable limits (with some adjustment for breathing rates). It
assumes that the GST pathway is the predominant path to
carcinogenicity. However, even if there is a nonsaturable
portion to the MFO pathway (e.g., leading to carbon dioxide)
which has a role, 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. However, if the
estimates of partitioning of metabolism between liver and lung
are incorrect, then the method could be in error. Further, this
method makes no assumption regarding equivalency of delivered
doses between species. It applies interspecies correlation
factors in a manner that do not require knowledge regarding how
much of the difference between species response is due to
pharmacokinetics and how much is due to pharmacodynamics.
70
-------
8.3. COMPARISON OF METHODS 1 AND 2
As with any risk estimation that involves extrapolation,
many assumptions affect the actual risk numbers, and Methods 1
and 2 are certainly no exception. Both use pharmacokinetic
information in an attempt to reduce uncertainties inherent in
risk assessments. It is commonly, although erroneously,
conceived that by incorporating pharmacokinetic information into
a risk assessment, magical reductions of uncertainity are
achieved. Actually, the examination and application of
pharmacokinetic data and models for this compound have revealed
something quite different. As one becomes more familiar with
developing and using this type of information, new and often
more complex questions arise. The utility of pharmacokinetics
is, in fact, this very point. It allows for a systematic
analysis of a chemical's disposition in the body, an important
component of the risk assessment. In applied-dose approaches,
assumptions are frequently made which, although sometimes based
on empirical evidence, are often inflexible and thus in error at
some conditions of the human exposure. For example, absorption
fraction is frequently set at some arbitrary value determined
from some empirical evidence or from assuming the "worst case" of
100% absorption. Pharmacokinetic modeling, when properly
performed, is able to account more logically and realistically
for amounts absorped on a time basis. Pharmacokinetic models
seek to account for instantaneous concentrations and changes in
those concentrations that are due not only to changes in exposure
71
-------
conditions, but changes in the physiologic responses as well.
As observed in earlier discussions and in the HRAC (1987)
report, while many uncertainties are reduced, several of the "old
problems" remain, and in fact, new challenges arise. Only with
continued work and trial applications will the science continue
to mature. Only two possible methods have been applied here, but
given the body of evidence and the development of the science, at
this time these two possibilities are considered the most
reasonable.
Methods 1 and 2 employ many of the same assumptions, and yet
vary in some very significant ways. Although the actual
calculated numbers are almost identical (within a factor of 4 for
liver and much less for lung), the methodologies are quite
different. When in error even some of the common assumptions
have different implications depending on the method chosen.
However, the differences are mainly in the "last step" of the
risk assessment process, that is, how to actually use delivered
dose to calculate a risk number.
A major and fundamental assumption that EPA has made for
both methods is that the physiologically based pharmacokinetic
model used by Andersen et al. (1986, 1987) is a reasonable method
for describing and predicting the disposition of DCM and its
metabolites in human tissues. This would include acceptance of
the model's structure and input parameters. The HRAC (1987)
report raises several important questions that are deemed
important, and future elucidation for purposes of methodology
72
-------
development will be necessary. However, for the present, EPA has
applied the model with some minor changes. It was felt that the
uncertainity raised by questions regarding model structure were
no greater than those raised by a conventional applied-dose risk
assessment. In fact, because the model is able to quantitatively
describe numerous physiologic and biochemical processes, it is
highly probable that model structure questions pose less
uncertainty than the traditional approach. The HRAC is less
certain about some of the input parameters, such as the metabolic
rate constants. The consequences of errors in these could be
great, and the impact may be somewhat different depending on
whether Method 1 or Method 2 is employed.
There are three major sources of uncertainity with the
metabolic scheme and parameters in the model. First, the model
assumes that any carbon dioxide observed at low doses is being
produced from the MFO pathway. The implication is that the
carbon dioxide observed by several investigators at low doses is
still compatible with the assumption built into the model that
the GST pathway is virtually inactive at low doses. If incorrect
it would mean that the GST pathway is active at low doses, where
the model is predicting that it is not. Both methods would be in
significant error in predicting risk at low doses
(underprediction).
A second uncertainity common to both Methods 1 and 2 are the
values of the input parameters which apportion metabolism by both
pathways between the liver and lung. As discussed in the HRAC
73
-------
(1987) report, there is great concern over the values estimated
for these parameters. The pharmacokinetic model is quite
sensitive to these parameters, and thus any error would be
reflected in model predictions. Such error would be significant
in both methods (over- or underprediction).
The third uncertainty regarding metabolism is with the
values of the metabolic rate constants. As discussed in the HRAC
(1987) report, most questions remain with the value determined
for the first-order rate constant for the GST-mediated pathway.
Determined by allometeric scaling and "curve fitting" of the
model to exposure chamber data, the value of this parameter is
uncertain. In fact, data from CEFIC (1987b) indicate that the
value selected by Andersen et al. (1986, 1987) is in error.
Significant questions also remain regarding the methodology and
results of the CEFIC experiments. It is reported that
experiments are being conducted which may reduce some of the
uncertainty with regard to this rate constant. Method 1 is more
sensitive to this parameter than is Method 2; thus, any error
would result in greater error in the risk number calculated using
Method 1. However, because of this sensitivity, if the value of
this parameter is established more accurately, Method 1 would
better reflect the impact of such findings.
As described previously, numerous uncertainties are
associated with the assumptions that have been made to develop a
risk assessment using pharmacokinetic information for DCM. Apart
from the major generic difference between Methods 1 and 2, there
74
-------
is also the uncertainty associated with the metabolic rate
constant of the GST-mediated metabolic pathway. The risk
calculation resulting from Method 1 would be greatly lowered by
significant reductions in the estimate of this parameter.
Also, one might want to allow for some possible minor role
of metabolism by the MFO-mediated pathway in the carcinogenic
process. No current evidence suggests a contribution by this
pathway, and a large role is ruled out by the very low tumor
response in the National Coffee Association drinking water
bioassay (NCA, 1982a, b; 1983), in which the MFO metabolism was
saturated at a level similar to those associated with the highly
tumorogenic exposures in the NTP inhalation bioassay (NTP, 1985,
1986). However, even a small contibution to DCM's
carcinogenicity by MFO metabolism at high doses might have a
meaningful impact on low-dose risks, since the proportion of the
dose metabolized by this pathway increases at low exposure
levels. One might, for example, hypothesize that 5% of the
carcinogenic risk to mice at the bioassay doses resulted from MFO
metabolism. (Any higher contribution begins to conflict with the
observed lack of correlation of MFO metabolism—and clear
correlation of GST metabolism—with tumor incidences.) Such a
small contribution to tumor production by MFO metabolites would
not have a major impact on the human risk estimates as they have
currently been calculated using Method 1. If, however, the human
GST metabolic rate constant were greatly reduced from the present
estimate, resulting in much lower predicted human risk from this
75
-------
pathway, that same hypothetical 5% contribution from MFO
metabolism would have a far greater impact on the total human
risk estimate at low doses. Neglecting a contribution of as
little as 5% by the MFO-mediated pathway towards carcinogenesis
would, under those circumstances, greatly underestimate the risk.
If new data indicate that human GST activity towards DCM is much
less than the estimate used here, then a reevaluation of the
assumptions would be necessary. More confidence in the
assumption that the MFO path does not contribute to
carcinogenicity and greater certainty in the values of the
appropriate metabolic parameters will be required before the
concomitant reduction in the risk estimates would be accepted as
appropriate.
Another question that arises, regardless of method, is upon
which organs are risk estimates to be based? The pharmacokinetic
approach gives information regarding specific organs. Site
concordance of tumor production between animals and humans is not
normally assumed in performing risk estimates. It is not clear
how to extrapolate for the entire human (all organs) when risks
have been calculated for specific organs by using pharmacokinetic
knowledge. One possible solution is to select the organ with the
highest risk number and apply this to the whole body. This could
result in an overestimation of the risk for many organs but would
ensure that no underestimation would occur due to a lack of
knowledge about an oversensitive or a highly metabolic tissue.
Alternately, if individual organ risks occur independently, they
76
-------
could be mathematically summed. Both Methods 1 and 2 could do
either of these. However, in the case of DCM, because of the
comparative insensitivity of Method 2 to the GST metabolic
parameter, even organs with several fold greater metabolic
activity than the lung would not be expected to have a risk far
different from that calculated here. The results of Method 1,
however, are more difficult to apply to other organs. A several
fold change in the GST level (as might be observed in other
organs) would result in a different value for the risk number.
Without knowing specific GST activities towards DCM in other
tissues, it is difficult to ascertain the impact of such an
uncertainty. Although there is no clear evidence of
carcinogenicity in organs other than the lung or liver, there
are some findings that raise concern about this issue. Benign
mammary gland adenomas and salivary gland tumors developed in
rats (NTP, 1985, 1986). The HRAC (1987, Chapter 6) discussed
pancreatic tumors in workers exposed to DCM. Although these
tumors may not be significant, some note should be taken.
It is clear that, once estimates or measurements of internal
dose at the sites of toxic action are obtained, many difficult
issues must be resolved as to how to use such data in the
extrapolation of risk from experimental animals to humans. The
problem is not confined to DCM, nor does it result from any
shortcoming in the information on the pharmacokinetics of this
compound. It is a general problem, reflecting the lack of
understanding of the pharmacodynamics of carcinogenesis.
77
-------
As discussed earlier, there are many difficulties in using
metabolic differences in species to modify a carcinogenic risk
assessment. Extrapolation between species involves many factors,
including metabolism and pharmacokinetics. The ability to
elucidate a species difference in one contributing component does
not necessarily indicate what, if any, adjustments should be made
to the overall extrapolation. It does not necessarily provide
more certainty than the empirical process currently used; in
fact, making the necessary assumption may introduce new
questions.
Method 1 advocates the adjustment of the applied-dose risk
extrapolation by the degree to which humans (at lower doses) and
the bioassay rodents metabolize different proportions of their
applied doses at the internal site of carcinogenic action. In
the present case, this method leads to a risk reduction of 8.8-
fold from the level estimated in EPA's previous applied-dose risk
assessment. In Method 1, the observed pharmacokinetic
differences between species are to be compared with those
expected to emerge as a result of differences in physical size
and the rates of physiological processes in rodents and humans.
Method 2, which leads to a risk reduction of 2.1-fold,
advocates the adjustment of the applied-dose risk extrapolation
only by the degree to which the proportion of the applied dose
that is metabolized differs from high human doses to low human
doses; any species difference in the proportion of the dose that
is metabolized is ignored as a basis for determining human
78
-------
carcinogenic potency. Instead, the interspecies component of
extrapolation is carried out as would be done if using applied
dose. The reasoning is that, in addition to the effect of
species differences in metabolism, there are expected (but
unknown) differences in the carcinogenic responsiveness of the
tissues to a given delivered dose. Pharmacokinetic data
illuminate only the metabolic differences. It may be, for
example, that greater sensitivity to carcinogens in humans
"compensates" for lower metabolic activation of the applied dose.
The justification for using the surface area correction on the
applied dose during the species-to-species extrapolation rests on
tradition. Empirical comparison of carcinogenic potencies in
humans (determined directly from epidemiologic data) with those
from experimental animals shows the surface area scaling
relationship to be a reasonable estimator for many compounds,
although other chemicals show potencies that differ from the
expectation based on this relationship by orders of magnitude.
Because of the problem of specifying interspecies
differences in tissue responsiveness to carcinogens, both Methods
1 and 2 can only give a relative adjustment to the applied-dose
calculation of human risk. That is, incorporation of
pharmacokinetic information can only raise or lower the "dose
delivery" component of interspecies extrapolation relative to its
appearance in the applied-dose procedure, while the component
representing "responsiveness" or pharmacodynamic differences
between experimental animals and humans remains problematic and
79
-------
continues to be based on assumptions retained from the former
applied-dose procedure. Methods 1 and 2 differ chiefly in the
way that assumptions from the applied-dose extrapolation
procedure are retained when data on the pharmacokinetic component
are available.
Method 1 is based on the conclusion that, given the most
reasonable scaling of key physiological variables across species,
delivered dose is expected, a priori, to be the same proportion
of applied dose in rodents and in humans. That is, differences
in body size and physiological rates between rodents and humans
do not, in themselves, lead to an expectation of differences in
the delivered doses of metabolically activated carcinogens. If
the proportion of a dose that is metabolized is in fact the same
across species, then the applied dose serves as a good surrogate
measure for the delivered or internal dose of a carcinogen, and
both dose measures will result in the same risk extrapolation.
Thus, the surface area correction, as traditionally used in the
applied-dose procedure at EPA, corresponds to an assumption about
(and correction for) interspecies sensitivity differences rather
than about metabolic differences. It is the factor by which
human risk is assumed to exceed mouse risk for a given dose
(applied or internal). This same assumption about relative
sensitivity is retained when a pharmacokinetic analysis of the
proportion of a dose that is metabolized replaces the prior
assumption of equality of dose delivery across species, implicit
in the use of applied dose in extrapolation.
80
-------
In the present case, for example, the surface area
correction between mouse and human doses is a factor of 12.7.
According to the model used by Andersen and Reitz, the species
difference in metabolism is such that humans (at low exposure
levels) metabolize about one-ninth as much of their applied dose
via the GST pathway as do mice at 2000 or 4000 ppm. (Most of
this difference is due to high- to low-dose differences that
result from the saturation of the competing MFO pathway at the
high bioassay exposures experienced by mice—the interspecies
difference at the same applied dose is quite small.) According
to Method 1, the lower metabolic activation of DCM in humans
implies that the carcinogenic potency difference between humans
and mice is only one-ninth as large as it was previously thought
to be, before the metabolism data were available. The
carcinogenic potency in humans (expressed in units of applied
dose) is only one-ninth as large as the value based on applied
dose.
Method 2, in contrast, suggests that no reasonable
assumption can be made about the effect of allometric scale on
metabolic differences among species. Under this view, any
magnitude of species difference in metabolism seems equally
probable a priori, and so there is no prior assumption against
which to compare empirical data on the actual difference.
Instead, it is presumed that, for a given dose level, the
combined effect of metabolic and sensitivity differences is given
by the surface area correction on applied dose. No explicit
81
-------
assumption about the species difference in sensitivity is made;
in fixing the magnitude of the combined effect, however, a value
of the sensitivity component is assumed implicitly. For example,
in the present case the pharmacokinetic model estimates that
humans at high doses metabolize 4.5-fold less of their delivered
dose in the lung and 1.5-fold less in the liver than do mice at
equally high doses. By assuming that the overall interspecies
factor is 12.7, Method 2 implicitly assumes that these metabolic
deficits are compensated for by greater human sensitivity of
57.2-fold in lung (1/4.5 x 57.2 - 12.7) and 19.1-fold in liver
(1/1.5 x 19.1 = 12.7). Low-dose human risks are adjusted by the
degree to which the proportion of the applied dose that is
metabolized via the GST pathway is different than at these high
human doses. That is, delivered dose is used only for the
extrapolation within species, where the question of interspecies
difference in sensitivity does not arise.
Thus, the crux of the difference between the two methods is
whether or not a reasonable prior assumption about the expected
species differences in metabolism can be made before
pharmacokinetic data are available. If a prior expectation can
be specified, when pharmacokinetic data become available, one may
replace that assumption with data (which may show the assumption
to have been inappropriate for that compound). The same
assumption about species differences in sensitivity is applied in
all cases. If, on the other hand, no prior expectation about
pharmacokinetic differences between species can be specified,
82
-------
there is no way to know whether the observed differences are
bigger or smaller than usual. The applied dose is therefore
used to extrapolate across species, and the sensitivity
assumption is adjusted to make its combined effect with the
observed metabolic differences come out to be equal to the
surface area correction, since it is assumed that the combined
effect scales in this way.
The choice between Method 1 and Method 2 has not been an
easy one, and has been made only after considerable debate and
discussion both within EPA and with representatives of other
federal regulatory agencies. The attributes of each method that
have been considered include their relative conservatism in the
face of uncertainty, their sensitivity to errors in the
underlying assumptions and estimates of the pharmacokinetic model
used by Andersen and Reitz, their correspondence to previous
practice, their ability to incorporate current understanding of
metabolism, however imperfect, into the risk extrapolation
process, and, of course, the plausibility of the assumptions upon
which they are founded.
EPA concludes that Method 1, which extrapolates risk across
species and from high to low doses based on the amount of
metabolism of DCM by the GST pathway, is the most advisable basis
for use of current pharmacokinetic information. 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 in the estimation of human
83
-------
risk. While acknowledging that many factors in addition to
pharmacokinetics influence •pecies difference* in carcinogenic
potency, EPA concludes that it is reasonable to modify risk
extrapolation from experimental animals to humans by the degree
to which the species manifest different degrees of metabolic
activation of their applied doses at the site of carcinogenic
action. The absolute levels of human risk that are estimated
remain uncertain, as always, owing to the lack of knowledge about
the contributions of the other, non-pharmacokinetic factors to
the relative carcinogenicity of DCM in rodents and humans. The
need to retain assumptions about the role of such factors should
not, in EPAfs opinion, dissuade us from examining the potential
contribution of such factors as can be experimentally examined.
The choice of Method 1, the choice of the GST pathway as the sole
route to carcinogenic activation, and the choice of the model
used by Andersen and Reitz as a means of its estimation have been
made because, in EPA's judgment, they represent the most likely
and plausible interpretation of the data at hand. Each choice is
made in the face of some uncertainty, and the interpretation of
the resulting estimate of the carcinogenic potency of DCM in
humans must be tempered with the knowledge that further data may
lead to other choices and different risk estimates becoming more
defensible.
The unit risk for continuous inhalation of 1 ug/m3 of DCM is
thus estimated as 4.7 x 10""7. For comparison, the applied-dose
extrapolation leads to a value of 4.1 x 10"6 (which is 8.8 times
84
-------
higher), and the use of the same metabolic data, but
extrapolating to human risk using Method 2, results in a value of
1.8 x 10""6 (which is 2.1-fold lower than the applied-dose method
and 3.8-fold higher than Method 1).
It should also be noted that both Methods 1 and 2 are
presented as modifications of the method of risk extrapolation to
humans commonly employed by the EPA and CPSC, that is, with the
surface area correction used as an interspecies correction
factor. If one instead uses the body weight basis for defining
equally risky applied doses in animals and humans (as is done by
the FDA), then the estimated human risk by all methods would be
12.7 times lower. That is, the applied dose procedure would lead
to a unit risk estimate of 3.2 x 10~7, while modifications of
this unit risk by accounting for metabolism would yield unit
risks of 3.7 x 10~8 for Method 1 and 1.4 x 10"*7 for Method 2.
These numbers are 12.7-fold, 8.8 x 12.7 » 111-fold, and 2.1 x
12.7 - 26.7-fold, respectively, below the published EPA unit risk
of 4.1 x 10~6 based on applied doses scaled by surface area.
Andersen et al. (1986, 1987) and Reitz et al. (1986) argue that,
because interspecies differences in metabolic and physiologic
parameters have been accounted for by the pharmacokinetic model,
there is no longer a need for any interspecies correlation
factor. Implicit in this view is the assumption that the
interspecies correction factor on applied dose is used solely to
account for species differences in metabolism, and that metabolic
differences completely account for differences in carcinogenic
85
-------
potency of a compound in animals and humans. The EPA takes the
view that this is not the case, since it ignores the contribution
of non-pharmacokinetic factors that influence a species*
responsiveness to a given internal dose.
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. However, the EPA's analyses of the model used by Andersen
and Reitz indicate that, 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 consequently have risks somewhat
underestimated.
Although EPA feels that it is warranted to use species-to-
species pharmacokinetic and metabolic information to adjust
estimates of human risk based on animal data, the absolute levels
of estimated human risk remain uncertain, owing to the unknown
86
-------
contribution of species differences in sensitivity to a given
internal dose of carcinogen. EPA recommends that intensive
efforts be made to develop information on the pharmacodynamics of
carcinogenesis that could be used in the risk assessment process
in the future. One approach, which may elucidate the magnitude
and variability of the pharmacodynamic factor for various species
comparisons, is to obtain pharmacokinetic information in both
animals and humans for known human carcinogens. This would allow
an implicit determination of the pharmacodynamics for humans
relative to various rodent species, since the contribution of
pharmacokinetics and the relative potencies of applied doses
could be estimated from available data.
87
-------
9. IMPACT OF CEFIC EXPERIMENTAL DATA ON RISK ESTIMATES
The following analysis provides an indication of the changes
in estimated risk implied by the results of CEFIC»s (1986e) in
vitro metabolism experiments. These studies found no detectable
activity of human liver cytosol toward DCM. The authors raised
the possibility that humans do not have the specific isozyme(s)
of GST that are active on DCM. Recently, however, scientists
from CEFIC and Dow Chemical Company have reported that they have
each independently detected low GST activity in human tissues
using a more sensitive assay. (This information was reported by
Green and Reitz in a May 1987 letter to EPA). No results have
been made available to the federal regulatory agencies at this
time, however. The following analysis is, therefore, based on
the limit of detection of the original CEFIC (1986e) work.
If the human GST isozyme(s) does act on DCM, it is no more
than one-sixtieth as active as the mouse GST system (expressed as
nmoles/min/mg). This assumption is based on the limit-of-
detection reported by CEFIC for their assay, i.e., the mouse
value of approximately 36 nmoles/min/mg divided by the limit-of-
detection, 0.6 nmoles/min/mg. Determining the impact of such a
level is extremely complex and requires a number of additional
assumptions in order to derive an idea of the in vivo effect on
risk of assuming that human GST is one-sixtieth as active as
mouse GST on DCM in vitro. By way of example, one set of
assumptions is adopted below to give an idea of the potential
88
-------
effect on risk. Alternative assumptions, which would give
different results, are still under consideration by the HRAC
(1987) .
The CEFIC data on in vitro GST metabolism by the human liver
cell fractions show no detectable activity toward DCM (four
livers were tested). On the basis of several assumptions, the
limit-of-detection of GST activity reported in the CEFIC study
can be used to set an upper limit on the parameter describing the
rate of human GST metabolism (kp) in the pharmacokinetic model
used by Andersen and Reitz. The CEFIC analysis treats the GST
pathway as saturable at high substrate concentrations (higher
than experienced in vivo), and hence represented by
Michaelis-Menton kinetics. In determining an upper limit on
human GST metabolism, the limit-of-detection CEFIC reported
(0.6nmol/min/mg protein) can be regarded as the maximum rate
attainable by the human GST pathway (Vmax). This rate is 60-fold
lower than the maximum in vitro rate CEFIC observed for mouse
liver (36 nmol/min/mg protein)
At lower doses (well below saturation) the rate becomes
directly proportional to the DCM concentration and thus directly
comparable to the rate in the model used by Andersen and Reitz.
The single rate constant used by Andersen et al. (1986, 1987) to
represent GST metabolism (kp) is, at low doses, essentially equal
to CEFICfs maximum rate (Vmax) divided by the Michaelis constant
(KM). This constant reflects the interaction of the GST enzyme
with its substrate. The human value of K^ cannot be estimated
89
-------
from the limit-of-detection, but if it is assumed that human KM
is equal to the CEFIC estimate of KM for mice, the GST pathway in
vitro rate constants (kF's) for mice and humans will be in the
same ratio as the Vmax»s; that is, the human value will be
60-fold lower than the mouse value.
Extrapolating from such in vitro estimates to the expected
rate of metabolism by a whole intact organ is also problematic.
If it is assumed that the mouse/human ratio of kF»s in vivo is"
the same as the ratio of the in vitro estimates developed above,
then the human whole-liver kF should be one-sixtieth that of the
mouse, or 4.30/60 - 0.072 hr'1. This value can then be compared
to the value of 0.53 hr"1 in the model used by Andersen and
Reitz. in other words, if the human GST metabolism constant in
the model used by Andersen and Reitz is replaced by one estimated
from the limit-of-detection of the CEFIC in vitro data, the rate
of human GST metabolism predicted by the modified model is
decreased by about sevenfold, assuming the Andersen and Reitz
value for mouse kp is correct.
To use this result in risk estimation, the assumption must
be made that GST metabolism in human lung is reduced by the same
amount as that for the liver. At present, no human in vitro data
on human lung are available to make this estimate directly.
Furthermore, it is only a rough approximation to say that a
sevenfold change in the model's parameter describing GST will
lead to risk estimates sevenfold lower. Revising the GST kinetic
parameter in the pharmacokinetic model used by Andersen and Reitz
90
-------
is a major change which should involve reconsideration of other
parameters, particularly those representing MFO metabolism.
Parameter adjustment reopens the issue of model validation. A
constructed model could give results indicative of something
quite different from a sevenfold reduction in risk.
In summary, an analysis of the implications of the results
of the CEFIC in vitro metabolism experiments, suggests that risks
from exposure to DCM could be some sevenfold lower than estimates
from the unmodified model used by Andersen and Reitz, based on a
rough estimation procedure. Green and Reitz reported, in a May
1987 letter to EPA, preliminary results of new studies that use
36C1-DCM; these results should greatly increase the level of
detection of GST activity. Furthermore, the CEFIC experiments
designed to provide new values for several model parameters,
including partition coefficients and kinetic parameters, may lead
to a modified pharmacokinetic model that is better validated and,
thus, could more confidently be used in estimating human risks.
91
-------
10. IMPACT OF EPIDEMIOLOGIC EVIDENCE ON HUMAN RISK ESTIMATION
Two historical cohort studies examine the mortality
experience of workers occupationally exposed to DCM; one
evaluates triacetate fiber extrusion workers (Ott et al., 1983)
and the other, Kodak film-casting workers (Friedlander et al.,
1978; Hearne and Friedlander, 1981). Neither study reports a
statistically significant increase in deaths from cancer among
workers exposed to DCM. The Kodak study, however, which is the
more powerful of the two studies, with far better exposure
information, was used for EPA's quantitative analysis. This
study has recently been updated by the addition of 262 men, an
increase of up to 8 years in the length of follow-up on the
original cohort members, and an investigation of the possible
contribution of potential confounding factors, such as smoking
(Hearne et al., 1986).
The updated Kodak study examines a 1964-1970 cohort of 1013
male film-casting employees evaluated through 1984. Time-
weighted exposures over the duration of the workday averaged 26
ppm, and ranged from 20 to 140 ppm, depending on job category.
Exposure duration averaged 22 years. Extensive monitoring and
detailed job history information permitted the calculation of
individual exposure histories. The mortality experience of these
workers was compared against two sets of standard rates: New
York State males and nonexposed male hourly employees at Kodak's
Rochester facilities. The second comparison is more informative,
92
-------
as it largely eliminates the "healthy worker effect."
The results show no elevation of total deaths from malignant
neoplasms (41 observed versus 52.7 expected), nor of respiratory
cancer deaths (14 versus 16.6), nor of liver cancer deaths (0
versus 0.5). Neither was there a trend in incidence of
respiratory cancer deaths with increasing exposure or with
increasing time-since-first-exposure. Hearne et al. (1986) state
that no other cause of death was statistically elevated, but
noted a nonsignificant excess of pancreatic cancer deaths (8
versus 3.1 expected) which they attribute to expected statistical
fluctuation when several end points are examined simultaneously.
Hearne et al. (1986) interpret their study as showing that
DCM is safe for humans at the occupational exposure levels
experienced in the study. They have calculated the number of
excess cancers that would have been expected in the cohort if the
EPA's published incremental cancer risk (U.S. EPA, 1985b) were
true, and claim greater than 99% power to detect such an
elevation if it were manifested as lung cancer deaths. Hearne et
al. concluded from their analysis that the federal regulatory
agencies have overestimated DCM cancer risks to humans.
Two reviews of the updated Kodak study, an EPA evaluation
(HRAC, 1987) and an independent analysis by Dr. G. Matanowski
(Batelle, 1986), found fault with some aspects of the comparison
of exposed workers to nonexposed workers. Both reviews also
criticize the study on the grounds that the Kodak cohort includes
a large number of men with extensive exposure before the
93
-------
enrollment date; these men thus represent survivors of previous
exposure, and the sample could be biased away from sensitive
individuals who may have left their film-casting jobs due to
illness.
Despite the criticisms of the Kodak study, EPA finds it to
be generally well conducted, but the study's apparent
disagreement with the carcinogenic effects of DCM as observed in
the NTP bioassay must be analyzed. A CPSC review (Conn and Rock,
1986) points out that the pancreatic tumors observed in the Kodak
study are significantly elevated if a 5% level of significance is
used rather the 1% level that Hearne et al. apply to
"non-hypothesized" causes of death, i.e., causes not indicated by
animal studies. EPA recognizes that the increase in pancreatic
tumors may merely reflect the fact that a few apparent increases,
even statistically significant increases, can be expected (even
when no excess exists), due to chance alone. The increase in
pancreatic tumors cannot be considered an unequivocal positive
response and should not be interpreted as evidence that DCM is a
human carcinogen. Nevertheless, an increase of this magnitude
raises some concern about possible human response in tissues
other than those found to respond in animals, a not uncommon
phenomenon.
Quantitative analyses have been conducted to determine
whether the results of the Kodak study, which show no statistical
increase in cancer deaths with the exception of a marginally
significant increase, at most, in deaths from pancreatic cancer,
94
-------
refute the magnitude of risk estimated from the animal data.
CPSC (Cohn and Rock, 1986) determined that CPSC's animal-based
incremental risk estimate for DCM (which is slightly lower than
the EPA applied dose estimate) predicts an excess of only 8.7
cancers in the Kodak cohort; such an excess of liver cancers
could be detected, but the statistical power to detect 8.7 excess
lung cancers is only 55%. CPSC notes that the observed excess of
pancreatic tumors in the Kodak study of almost 5 deaths is of the
same order as their prediction of 8.7 excess cancers, especially
since the cohort is rather young, with only 18% having died from
all causes, while the animal-based predictions are for lifetime
risks.
EPAfs analysis (HRAC, 1987) takes a somewhat different
approach: the Kodak study is used to directly generate a human
risk estimate, and a 95% upper bound to that estimate, which may
then be compared to the animal-based upper-bound incremental risk
estimates. This analysis uses age-specific data on cumulative
DCM exposure and on observed cancer risk. Two models are
applied. The first is an additive excess risk model, in which
one assumes that the excess cause-age-specific death rate due to
DCM exposure is increased in an additive way by an amount
proportional to the cumulative exposure up to that age. The
second is a multiplicative or relative risk model, in which the
cause-and-age-specific background rate at any given age is
increased by a multiplicative factor proportional to the
cumulative dose up to that age. An adjustment for a latent
95
-------
period of about 20 years is made by examining exposure and risk
only after age 45.
When applied to the data on pancreatic cancer deaths in the
Kodak study, the additive excess risk model procedure and the
relative risk model procedure yield a maximum likelihood estimate
of incremental risk from a lifetime exposure to 1 ug/m3 of
3.4 x 10~6 and 1.4 x 10""6, respectively. The 95% upper bounds on
the Kodak-based incremental risks are 7.1 x 10~6 and 2.8 x 10~6.
EPA applied the same analysis to respiratory cancers among
Kodak workers. Since fewer cancers were observed among exposed
workers than expected based on nonexposed workers, the best
estimate of incremental risk is zero. The 95% upper bound on
this incremental risk is 7.6 x 10"7 for a lifetime exposure to
1 ug/m3 according to the multiplicative model. (The additive
model could not be estimated for these data.)
The HRAC's (1987) calculation of an upper bound on DCM
potency based on the Kodak respiratory deaths is about fivefold
below the EPA applied dose extrapolation from mice; this result
is compatible with the Hearne et al. conclusion that their study
has over 99% statistical power to detect an increase in cancer in
the Kodak cohort of the magnitude predicted by EPA's (U.S. EPA,
1985a, b) unit risk. When the Kodak pancreatic cancer deaths are
used as the basis of human risk estimation, the maximum
likelihood estimate of risk (3.4 x 10"6) and the upper bound
estimate (7.1 x 10~6) are quite similar to the EPA applied dose
value (4.1 x 10"6). The Hearne et al. conclusion that the Kodak
96
-------
results refute EPA's original risk estimates depends on the
expectation of site concordance across species.
The issue of lack of compatibility of risks estimated from
the NTP bioassay data and those made directly from human data is
largely resolved when risks extrapolated from animal data are
estimated from internal dose. Animal-based internal dose risk
estimates, when extrapolated using Method 1, are 8.8-fold below
EPA's applied-dose risk estimates and lower than all of the
estimates made from the Kodak human data as well. The upper
bound on carcinogenic potency based on the Kodak respiratory
cancer deaths (7.6 x 10~7) is more than twice as high as the
lung-specific potency extrapolated by Method 1 from the internal
dose in the NTP mice (3.3 x 10~7). The maximum likelihood
estimates of potency based on the Kodak pancreatic cancer deaths
(3.4 x 10~6) is higher than the combined liver and lung risks
extrapolated by Method 1 from mice (4.7 x 10"7). (If the
pancreatic cancer deaths were to be taken as a clearly positive
response in the Kodak study, which they are not, this fact would
be grounds for using the human-based potency in preference to the
lower animal-based estimate).
The human risk estimate that emerges by extrapolating from
the NTP mice using Method 2 (1.8 x 10"6 per ug/m3) is only 2.1-
fold below the EPA applied-dose extrapolation. This animal-based
risk estimate is of the same order as the maximum likelihood
estimates based on the Kodak pancreatic cancer deaths (3.4 x 10"6
97
-------
and 1.4 x 10"6 for the additive and relative risk models,
respectively). The 95% upper confidence limit on risk based on
the Kodak respiratory cancer deaths (7.6 x 10~7) is still some
2.1-fold below the animal-based risk when extrapolating by
Method 2.
It can be argued that the Kodak study does not support the
use of either pancreatic tumors or respiratory tumors to quantify
risk, since these tumors were not clearly elevated to a
statistically significant level. Lack of statistical
significance does affect the qualitative weight of evidence
regarding DCM's human cancer potential; as noted previously, the
increase in pancreatic cancers cannot be interpreted as evidence
that DCM causes cancer in humans. When looking at whether or not
the Kodak study refutes the animal studies as a basis for human
risk estimation, however, EPA believes it to be appropriate to
compare quantitative estimates of cancer risks suggested either
by tumors corresponding to sites found in the animal study (lung
cancers) or by tumors with pronounced elevation.
Whether human risks are extrapolated from mice by Method 1
or by Method 2, the estimates are close to those derived directly
from the Kodak study. Thus, the Kodak study does not contradict
the conclusions that have been drawn from the animal studies.
Because of the lack of positive responses, the epidemiologic
studies do not add weight to the evidence of DCM's
carcinogenicity in humans. However, these studies do not
constitute evidence of its safety in humans either, nor do they
98
-------
indicate that the animal-based risk estimates must be too high.
99
-------
11. EPA'S CONCUJSIONS CONCERNING THE RISKS TO HUMANS FROM
EXPOSURE TO DICHLOROMETHANE
Data from the 1985 NTP inhalation cancer bioassay
demonstrate that DCM is oncogenic in two species of laboratory
animals, rats and mice. In rats, tumors were benign
fibroadenomas of the mammary gland. The literature (Russo et
al., 1982) suggests that adenomas of this type are histogenically
different from malignant adenocarcinomas and do not have a high
potential for progressing to malignant tumors. As the rat tumors
were not of a type with known malignant potential, the relevance
of these tumors to human health is unclear. The response in rats
cannot be entirely discounted but is considered to carry less
weight than the response in mice.
In mice, the response was unequivocally carcinogenic.
Administration of DCM via inhalation at doses of 2000 or 4000 ppm
caused a statistically significant increase in malignant tumors
in two organs, the liver and lung, in B6C3F1 mice. Although the
tumors were of a type that have occurred at a high and/or
variable background frequency in the strain of mouse tested,
there is no question of their statistical significance in the NTP
study. The tumor increases in mice were dramatic: at 4000 ppm
40 out of 50 male mice and 41 out of 48 female mice developed
lung tumors; 33 out of 49 male mice and 40 out of 48 female mice
developed liver tumors. Moreover, the tumors occurred in a
pattern which meets the criteria for determining whether commonly
100
-------
occurring tumors provide sufficient evidence of animal
carcinogenicity; an excess of malignant tumors was observed in
both sexes of mice, and the proportion of malignancies to benign
tumors, as well as the time to first appearance of tumor, were
dose related. The NTP data thus provide sufficient evidence of
animal carcinogenicity.
The EPA Guidelines for Carcinogen Risk Assessment (U.S. EPA,
1986) and the Office of Science and Technology Policy (OSTP,
1985), generally accepted by the scientific community, recognize
that chemical agents for which there are sufficient evidence of
carcinogenicity in animals should be regarded as probable
carcinogens for humans. The OSTP policy states that a "finding
of carcinogenicity in rodents is proof that the chemical is
carcinogenic in a mammalian species. Such a finding must be
taken as strong evidence that the chemical can be a carcinogen in
man, unless there is compelling. . .evidence to the contrary."
The varied response in chronic animal studies on DCM—a
clear carcinogenic response in mice exposed at high doses, a
negative response in hamsters, a primarily benign tumor response
in rats, and the lack of a statistically significant increase in
malignant tumors in mice exposed at low doses—are grounds for
investigating whether there are biological differences between
species that could lead to species (and/or dose related)
differences in risk.
EPA has been concerned with evaluating whether the new data
on DCM regarding species differences in metabolism and the
101
-------
carcinogenic mechanism of action lead to the conclusion that DCM
does not present a cancer risk to humans. EPA believes that to
make such a statement, the data would have to indicate with
reasonable certainty that:
1. the mechanism of carcinogenic action in mice is not
expected to occur in humans; or
2. the carcinogenic biochemical pathway is inactive in
humans; or
3. the epidemiologic data are sufficient to clearly
indicate that DCM does not cause cancer in humans.
After a critical analysis of the evidence, EPA has concluded
that the data fall short of meeting any of these criteria. The
carcinogenic mechanism of action of DCM remains unidentified.
There is little evidence supporting either an epigenetic or a
genotoxic mechanism. Results from a host of genotoxicity studies
are mixed. Unequivocal positive results were obtained chiefly in
prokaryotic systems; positive results were fewer as the
complexity of the test organism increased and test systems
progressed from in vitro to in vivo. The insensitivity of the
available mammalian in vivo studies and the lack of supporting
evidence for an alternative mechanism allow the suspicion,
however, that DCM may be a weak genotoxin in mammals.
Data on a possible cytotoxic mechanism are limited and
difficult to interpret. Exposure to DCM brought on transient
cytotoxic effects in specialized cells of the mouse lung, but the
significance of this response to the induction of carcinogenicity
102
-------
is unknown. Further, given that the human lung contains cells
biochemically similar to the affected cells in the mouse lung,
there is reason to believe that the cytotoxic response in mice
could be repeated in humans.
Evidence of an epigenetic mechanism in the liver is still
less certain. Data to support the hypothesis that the
carcinogenic response in the mouse liver results from increased
cell turnover are scant. An S-phase hepatocyte study registered
small, variable increases in the incidence of S-phase cells after
DCM exposure, but the authors of the study concluded that the
biological significance of these changes is unclear.
Current evidence is simply not sufficient either to identify
with reasonable certainty the mechanism of action of DCM or to
indicate that said mechanism would not be expected in humans.
Questions concerning the mechanism may be answered by additional
research. CEF1C has underway studies to better define the role
of the effects of DCM on the Clara cell in the development of
mouse tumors. The results of this research are expected in the
summer of 1987. NIEHS has planned investigations of the role of
cell replication and the pattern of oncogene activation in DCM
tumorigenesis. Preliminary results of NIEHS1 experiments may be
available in 1988.
It does seem likely from the available data that DCM
produces a carcinogenic response via GST-mediated metabolism.
The correlation between GST activity and the presence of tumors
is strong, and there is little evidence to implicate either
103
-------
parent DCM or the alternative MFC pathway. The level of GST
metabolism in the mouse lung and liver has been shown to be high,
but the degree to which humans metabolize DCM via the GST pathway
is uncertain, at present.
The pharmacokinetic model used by Andersen and Reitz
estimates human GST activity toward DCM to be not too far below
the level in mice. The Andersen and Reitz values may be in
conflict with CEFIC's estimates from in vitro measurements of GST
activity in human liver samples, however. The CEFIC data
indicate an absence of GST activity above the limit-of-detection
of the test system. At present, it seems reasonable to assume
that humans metabolize DCM via the GST pathway at a rate below
that of the mouse, possibly far below. In fact, recent communi-
cations received from CEFIC and the Dow Chemical Company reveal
that the preliminary results of in vitro studies using 36C1-DCM
show that mouse liver has relatively greater GST activity towards
DCM than does the liver of hamsters, rats, or humans. Rat and
human liver tissues were judged to be the least active.
Neither do the epidemiologic data rule out a risk to humans,
although the two studies of humans exposed to DCM in the
workplace show no statistically significant increase in deaths
from either liver or lung cancer. The study of Kodak
film-casting workers, a particularly well-documented analysis,
recorded no statistically significant excess of deaths from any
type of cancer with the possible exception of pancreatic cancer.
The Kodak cohort had an elevation in pancreatic cancer deaths
104
-------
which is marginally significant if a 5% level of significance is
used. The increase in pancreatic cancers is not considered an
unequivocal positive response, but gives some weight to the
possibility that a carcinogenic response in humans to DCM
exposure could occur at sites others than those found to respond
in animals. Further, a quantitative analysis of cancer risks
estimated directly from the human data indicates that the results
of the Kodak study do not refute the magnitude of risk estimated
from the animal data, when animal-based risks are estimated on
the basis of internal dose. The epidemiologic data are thus
judged to be insufficient to clearly indicate that DCM does not
cause cancer in humans.
EPA believes that the currently available data on mechanism
of action, carcinogenic metabolic pathway, and epidemiology do
not support a finding of zero cancer risk to humans from DCM
exposure. If the risk to humans is not zero then what is the
best estimate of risk?
Principle 26 of the OSTP (1985) states:
No single mathematical procedure is recognized as
the most appropriate for low-dose extrapolation in
carcinogenesis. -When relevant biological evidence on
mechanism of action exists (e.g., pharmacokinetics,
target organ dose), the models or procedures employed
should be consistent with the evidence.
The OSTP principle is echoed by the EPA Guidelines which
state that
When pharmacokinetic or metabolism data are available,
or when other substantial evidence on the mechanistic
aspects of the carcinogenesis process exists, a low-
dose extrapolation model other than the linearized
multistage procedure might be considered more appro-
105
-------
priate on biological grounds.
DCM has been shown to be metabolized in mice by two
pathways—one of which, the MFO pathway, is saturated at high
doses. Further, the second pathway, which is mediated by GST,
appears to be less active in humans than it is in mice. Andersen
et al. have developed a physiologically based pharmacokinetic
model which provides a framework for estimating the nonlinearity
inherent in metabolism by two pathways used to different extents
at high and low doses, and, possibly, for incorporating species
differences into the risk estimation procedure.
The pharmacokinetic model used by Andersen and Reitz is yet
to be fully validated, and some of its parameters, in particular
the partition coefficients and kinetic parameters, are considered
to be uncertain. Further, there is debate within the scientific
community over the best way to use the model's results in
developing estimates of risk. Nonetheless, the pharmacokinetic
model provides a means of taking into account metabolic data
which would otherwise be ignored in the applied-dose procedure.
Despite its uncertainties, the model allows the development of
preliminary estimates of risk based on metabolized dose.
When the pharmacokinetic model is used with the kinetic
parameters estimated by Andersen et al., it leads to a risk
estimate that is 8.8-fold lower than that calculated under the
applied-dose method, when extrapolation is done by a method
called Method 1 herein. This estimate reflects the assumptions
that the model can be used for both cross-species and cross-dose
106
-------
extrapolation of internal dose. A somewhat different
extrapolation method, referred to as Method 2, results in a
lowering of risk of 2.1-fold. This method assumes that
pharmacokinetics can at present be used only for high- to low-
dose extrapolation, while species-to-species extrapolation is
done on applied dose. Both methods assume that EPA breathing
rate factors should be employed in the model used by Andersen et
al. (1986, 1987), and that a "surface area correction" factor of
12.7 should be applied to account for certain expected diferences
in carcinogenic potency between species. Application of the
surface area correction factor is controversial, however. It can
be argued that the pharmacokinetic model leads to risk estimates
111-fold lower by Method 1 (8.8 x 12.7) or 27-fold lower by
Method 2 (2.1 x 12.7), than under the applied-dose method. The
fact that there is no clear basis for choosing to use surface
area correction or not (or for choosing some other method of
cross-species extrapolation of risks from internal dose) is a
weakness of the current state of the art of quantitative risk
assessment.
In vitro data supplied by CEFIC (1986e) suggest that the
kinetic parameter for the pathway in the model used by Andersen
and Reitz may be too high. A lower value would result in further
lowering the estimated human risk. EPA cannot place a great deal
of confidence in any quantitative use of these in vitro data in
risk estimation, however, because the method for estimating in
vivo metabolic kinetic parameters from in vitro data is not well
107
-------
defined, and such risk estimates do not, as yet, take into
account changes in the pharmacokinetic model required by the
adjustment of a particular kinetic parameter.
A comparison of risk estimates made directly from the human
data provided by the Kodak epidemiology study to risk estimates
derived from the results of the pharmacokinetic model used by
Andersen and Reitz does not show the animal-based risk estimates
to be overestimates, using upper-bound risk estimates from
respiratory cancer deaths or using either maximum likelihood
estimates or upper-bound estimates from the pancreatic cancer
deaths in the Kodak study.
In summary, EPA concludes that the animal evidence of
carcinogenicity conforms to the definition for "sufficient" in
the EPA Guidelines for Carcinogen Risk Assessment. The
epidemiology studies, while showing no evidence of either liver
or lung cancer attributable to DCM, are not sufficient to rule
out a risk to humans; the data on deaths from pancreatic cancer
give some weight to the possibility that DCM may cause cancer in
humans at sites other than those found in animal species.
Overall, the epidemiologic data conforms to the definition in the
Guidelines for "inadequate" insofar as the pancreatic cancer
deaths cannot be used to establish a connection between exposure
to DCM and human carcinogenicity, yet neither can the possibility
of a such a connection be entirely discounted. Thus, DCM meets
the Guidelines criteria for Group B2, probable human carcinogen.
The available body of evidence on the carcinogenic mechanism
108
-------
of action of DCM and on species differences in utilization of the
carcinogenic metabolic pathway are not sufficient to support an
estimate of zero cancer risk to humans. An evaluation of the
weight of evidence does lead to the conclusion, however, that
risks should be estimated on the basis of internal dose of the
GST metabolite(s). A comparison of the results of the available
studies indicates that the GST pathway is the most likely source
of the excess tumorigenesis observed in the NTP mouse bioassay.
Additional research on pharmacokinetic model parameters and
on the carcinogenic mechanism of action underway by CEFIC are
expected to lead to refinement of the risk estimates presented in
Chapter 8. It should be noted that data from the experiments may
lead to human risk estimates below those estimated from the
currently available data. To the degree that the estimates of
relative metabolism by the GST pathway change as a result of
these data, the extrapolated risks will change. (Method 2, which
does not use metabolism to extrapolate across species, would be
expected to give virtually the same risk estimates as before,
irrespective of interspecies differences in metabolism that may
be discovered.)
Using the pharmacokinetic model with its original kinetic
parameters to estimate the internal dose of the GST metabolite,
then following Method 1, and correcting internal dose for
interspecies differences in sensitivity by using the surface area
correction factor, leads to a unit risk estimate for continuous
inhalation exposure to 1 ug/m3 of 4.7 x 10~7.
109
-------
It would be unwise to read too much importance or
significance into changes in the unit risk of a few fold when
pharmacokinetic data are employed by either Method 1 or Method 2.
The previous chapters have outlined uncertainties in the
structure and parameter values of the model formulated by
Andersen et al. (1986, 1987). Although it is difficult to define
these uncertainties in quantitative terms (such as confidence
limits), it is clear that model projections of internal doses
could vary, perhaps by up to several fold, without contradicting
currently available model validation data. Moreover, there are
large uncertainties as to the biological effects of those
internal doses that overshadow any error in their estimation.
Species differences in responsiveness—and within-species
differences in susceptibility of various tissues—are unclear.
Perhaps the largest uncertainty lies in the question of the
relative carcinogenicity of high and low doses, owing to the lack
of knowledge about the mechanism of DCM's carcinogenic action.
It is somewhat ironic that the area of risk extrapolation that
has the least uncertainty as far as pharmacokinetics is
concerned—relative internal doses at high and low exposures—
also has the greatest uncertainty in terms of the degree of
carcinogenic response that those internal doses can be expected
to engender. (Such uncertainty continues to be accommodated by
the use of an upper-bound, linearized multistage model for low-
dose extrapolation, which recognizes that the true dose-response
curve may fall off more rapidly at low doses.)
110
-------
In view of the uncertainties involved, the changes in DCM's
carcinogenic potency that result from different uses of the
available pharmacokinetic information are not, in practical
terms, very distinct. Discussion of the issues has been
worthwhile because of their theoretical importance rather than
their practical significance in the present case. For other
compounds (or for DCM itself, upon the introduction of new data),
the distinction among extrapolation methods may have much greater
practical consequences.
Rather than focusing on exactly how much the risk
extrapolation has been changed by the use of pharmacokinetic
information, it is instructive to examine how little it has been
changed. Perhaps the most important result of the foregoing
analysis is that, in the case of DCM, pharmacokinetic
considerations have not revealed a great error inherent in using
applied dose as a surrogate for internal or delivered dose.
According to current understanding as expressed in the
pharmacokinetic model used by Andersen and Reitz, there is little
difference between mice and humans in the proportion of a given
applied dose that is metabolized. There are differences in this
proportion from high to low doses, but they are not especially
large. Having uncovered these pharmacokinetic factors, it is
well to incorporate our best understanding of them into the risk
extrapolation process, despite remaining questions as to their
exact magnitudes. The uncertainty in the resulting potency
estimates is reduced (compared to the extrapolation based on
111
-------
applied dose) because the potential for the influence of
pharmacokinetic factors has been markedly circumscribed.
112
-------
REFERENCES
Ahmed, A.E.; Anders, M.W. (1978) Metabolism of dihalomethanes
to formaldehyde and inorganic halide. II. Studies on the
mechanism of the reaction. Biochem. Pharmacol. 27:2021-
2025.
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.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.
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 B6C3F1 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.
113
-------
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.
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.
Boxenbaum, H. (1984) Interspecies pharmacokinetic scaling and
the evolutionary-comparative paradigm. Drug Metab.
Rev. 15(5 & 6):1071-1121.
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.
114
-------
CEFIC (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
metabolism in F344 rats and B6C3F1 mice. ICI Central
Toxicology Laboratory, Report no. TL/R/880. September 22,
1986.
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. ICI Central Toxicology Laboratory, Report No.
CTL/R/885. September 22, 1986.
CEFIC (European Council of Chemical Manufacturer's Federation).
(1987a) The assessment of carcinogenic hazard for humans
exposed to methylene chloride. Technical Report No. 26.
January 29, 1987.
CEFIC (European Council of Chemical Manufacturer's Federation).
(1987b) Methylene chloride (dichloroethane). Further
experimental data. Telex from M. Harris, ICI Americas to J.
Hopkins, Office of Toxic Substances, U.S. Environmental
Protection Agency. March 9, 1987.
CEFIC (European Council of Chemical Manufacturer's Federation).
(1987c) Telex from M. Harris, ICI Americas to J. Hopkins,
Office of Toxic Substances, U.S. Environmental Protection
Agency. April 16, 1987.
Cohn, M.S., and Rock A.R. (1986) Review of epidemiology study
on methylene chloride by Friedlander. U.S. Consumer Product
Safety Commission. April 25, 1986.
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.
Davidson, I.W.F.; Parker, J.C.; Beliles, R.P. (1986) Biological
basis for extrapolation across mammalian species.
Regul. Toxicol. Pharmacol. 6:211-232.
Dedrick, R.L. (1973) Animal scale-up. J. Pharmacokinet.
Biopharm. 1(5):435-461.
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.
115
-------
Dow Chemical Company. (1980) Methylene chloride: a two-year
inhalation toxicity and oncogenicity study in rats and
hamsters. Follow-up response A. FYI-OTS-0281-0097.
Washington D.C.: U.S. Environmental Protection Agency,
Office of Toxic Substances, Information Management
Division.
Dow Chemical Company. (1982) Methylene chloride: a two-year
inhalation and oncogenicity study in rats. Toxicology
Research Laboratory, Health and Environmental Sciences, Dow
Chemical Company, Midland, MI.
Food and Drug Administration (FDA). (1985) Cosmetics; proposed
ban on the use of methylene chloride as an ingredient of
aerosol cosmetic products. Federal Register 50(243):51551-
51559.
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.
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):277-288.
Hatch, G.G.; Mamay, P.O.; Ayer, M.L.; Casto, B.C.; Nesnow, S.
(1983) Chemical enhancement of viral transformation in
Syrian hamster embryo cells by gaseous and volatile
chlorinated methanes and ethanes. Cancer Res. 43:1945-1950.
Hazard/Risk Assessment Committee (HRAC). (1987) Technical
analysis of new methods and data regarding dichloromethane
hazard assessments. Prepared by an interagency workgroup of
the Integrated Chlorinated Solvents Project.
EPA/600/8-87/029.
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.
116
-------
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.
McConnell, F.E.; Solleveld, H.A.; Swenberg, J.A.; Boorman, G.A.
(1986) Guidelines for combining neoplasms for evaluation of
rodent carcinogenesis studies. J. Natl. Cancer Inst.
76(2):283-289.
Mordenti, J. (1986) Man versus beast: pharmacokinetic scaling
in mammals. J. Pharm. Sci. 75(11):1028-1040.
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.
Occupational Safety and Health Administration (OSHA). (1986)
Occupational safety of methylene chloride. Federal Register
51(226):42257-42266.
Office of Science and Technology Policy (OSTP). (1985, February)
Chemical carcinogens: a review of the science and its
associated principles. Federal Register 50:10372.
Ott, M.G.; Skory, L.K.; Holder, B.B.; Bronson, J.M.; Williams,
P.R. (1983) Health evaluation of employees occupationally
exposed to methylene chloride. Scand. J. Work. Environ.
Health 9(Suppl. 1):8-16.
117
-------
Ramsey, J.C.; Andersen, M.E. (1984) A physiologically based
description of styrene in rats and humans. Toxicol. Appl.
Pharmacol. 73:159-175.
Ramsey, J.C.; Gehring, P.J. (1980) Application of
pharmacokinetic principles in practice. Federation Proc.
39:60-65. * * * i
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.H. (1982) Differentiation of the
mammary gland and susceptibility to carcinogenesis. Breast
Cancer Res. Treat. 2(1):5-73.
Schmidt-Nielsen, K. (1984) Scaling: why is animal size so
important? New York, NY: Cambridge University Press.
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.
Thilagar, A.K.; Kumaroo, V. (1983) Induction of chromosome
damage by methylene chloride in CHO cells. Mutat. Res. 116:
361-367.
Thilagar, A.K.; Back, A.M.; Kirby, P.E.; Kumaroo, V.; Pant, K. J.;
Clarke, J.J.; Knight, R.; Haworth, S.R. (1984) Evaluation
of dichloromethane in short term in vitro genetic toxicity
assays. Environ. Mutagen. 6:418-419.
United Kingdom Health and Safety Executive (UKHSE). (1985)
Toxicity review 12: dichloromethane (methylene chloride).
Illing, H.P.A.; Shillaker, R.O. HMSO 1985.
U.S. Consumer Product Safety Commission (CPSC). (1986)
Household products containing methylene chloride; status as
hazardous substances. Federal Register 51(161):29778-
29809.
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.
118
-------
U.S. Environmental Protection Agency (EPA). (1985c) Analysis
of the applicability of TSCA section 4(f) to methylene
chloride. EPA Docket No. OPTS-62045. Federal Register
50:202126.
U.S. Environmental Protection Agency (EPA). (1985d) Initiation
of regulatory investigation for methylene chloride. Federal
Register 201:42037-42047.
U.S. Environmental Protection Agency (EPA). (1986) Guidelines
for carcinogen risk assessment. Federal Register 51(185):
33992-34003.
119
-------
-------
-------
3 r;
to o
^ c
75
n
C
.*
o.? 3
8 2 *"
q€ *
2 •< 01
* o a>
^
o>
«
O 2
I
If
> me
-»
0) 0»
a «
QJ tf>
T)
O
g|?,
I i 5
01 o
ro 3
o> 3
00 (D
(D
(A
(D
Q)
T
O
3-
------- |