United States
Environmental Protection
Agency
Office of Research
and Development
Washington, DC 20460
EPA/600/P-01/002A
August 2001
External Review Draft
wEPA
Trichloroethylene
Health Risk
Assessment:
Synthesis and
Characterization
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 merit and policy implications.
-------
DRAFT-DO NOT EPA/600/P-01/002A
CITE OR QUOTE August 2001
External Review Draft
Trichloroethylene Health Risk Assessment:
Synthesis and Characterization
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 merit and policy
implications.
National Center for Environmental Assessment-Washington Office
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
-------
DISCLAIMER
This document is a 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.
ABSTRACT
This assessment presents EPA's most current evaluation of the potential health risks from
exposure to TCE (trichloroethylene). TCE exposure is associated with several adverse health
effects, including neurotoxicity, immunotoxicity, developmental toxicity, liver toxicity, kidney
toxicity, endocrine effects, and several forms of cancer. Mechanistic research indicates that
TCE-induced carcinogenesis is complex, involving multiple carcinogenic metabolites acting
through multiple modes of action. Under EPA's proposed (1996) cancer guidelines, TCE can be
characterized as "highly likely to produce cancer in humans."
For effects other than cancer, an oral RfD of 3 x 1CT4 mg/kg-d was based on critical effects
in the liver, kidney, and developing fetus. An inhalation RfC of 4x 1CT2 mg/m3 was based on
critical effects in the central nervous system, liver, and endocrine system. Several cancer slope
factors were developed, with most between 2x 1CT2 and 4x 1CT1 per mg/kg-d. Several sources of
uncertainty have been identified and quantified.
The mechanistic information suggests some risk factors that may make some populations
more sensitive. There are suggestions that TCE could affect children and adults differently. In
addition, several chemicals have the potential to alter TCE's metabolism and clearance and
subsequent toxicity; conversely, TCE exposure can augment the toxicity of other chemicals.
Widespread environmental exposure to some of TCE's metabolites makes it important to
consider the cumulative effect of TCE along with other environmental contaminants.
8/1/01 ii DRAFT-DO NOT CITE OR QUOTE
-------
CONTENTS
LIST OF TABLES vi
LIST OF FIGURES vi
LIST OF ABBREVIATIONS vii
PREFACE ix
AUTHORS AND CONTRIBUTORS xi
1. SUMMARY AND CONCLUSIONS 1-1
1.1. OVERVIEW OF TCE HEALTH HAZARDS 1-1
1.2. MODE OF ACTION 1-2
1.3. WEIGHT OF EVIDENCE UNDER THE PROPOSED AND CURRENT
CANCER GUIDELINES 1-4
1.4. QUANTITATIVE HEALTH REFERENCE VALUES 1-5
1.5. UNCERTAINTY 1-8
1.5.1. Model Uncertainty 1-9
1.5.2. Parameter Uncertainty 1-10
1.5.3. Consideration of Multiple Areas of Uncertainty 1-10
1.6. HUMAN VARIATION AND SENSITIVE POPULATIONS 1-11
1.6.1. Metabolic Differences 1-11
1.6.2. Disease 1-12
1.6.3. Gender Differences 1-13
1.6.4. Some Implications 1-13
1.7. DIFFERENTIAL RISKS TO CHILDREN 1-14
1.7.1. Exposure Differences 1-14
1.7.2. Pharmacokinetic Differences 1-14
1.7.3. PharmacodynamicDifferences 1-15
1.7.4. Some Implications 1-15
1.8. CUMULATIVE RISKS INVOLVING TCE 1-15
2. EXPOSURE CHARACTERIZATION 2-1
2.1. ENVIRONMENTAL SOURCES 2-1
2.2. ENVIRONMENTAL FATE 2-1
2.3. EXPOSURE CONCENTRATIONS 2-2
8/1/01 iii DRAFT-DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
2.4. EXPOSURE PATHWAYS AND LEVELS 2-3
3. HAZARD ASSESSMENT AND CHARACTERIZATION 3-1
3.1. GENERAL APPROACH 3-1
3.2. METABOLISM 3-2
3.3. SUSCEPTIBILITY 3-4
3.3.1. Intrinsic Factors Affecting Susceptibility 3-5
3.3.2. Extrinsic Factors Affecting Susceptibility 3-8
3.4. TOXICITY 3-9
3.4.1. Neurotoxicity 3-9
3.4.2. Liver 3-10
3.4.3. Kidney 3-12
3.4.4. Immune and Lympho-Hematopoietic Systems 3-15
3.4.5. Developmental and Reproductive Toxicity, Endocrine Effects 3-17
3.4.6. Lung 3-21
3.4.7. Genetic Toxicity 3-22
3.5. MODES OF ACTION OF TCE TOXICITY 3-23
3.5.1. Liver Cancer 3-24
3.5.2. Kidney Cancer 3-36
3.5.3. Lung Cancer 3-40
3.5.4. Toxicity in Other Organs 3-41
3.6. HAZARD CHARACTERIZATION 3-44
3.6.1. Consideration of Causality in the Epidemiologic Evidence 3-44
3.6.2. Weight of Evidence under the Current and Proposed Cancer
Guidelines 3-47
4. DOSE-RESPONSE ASSESSMENT AND CHARACTERIZATION 4-1
4.1. GENERAL CONSIDERATIONS AND APPROACH 4-1
4.2. DOSIMETRY MODELING 4-2
4.2.1. Pharmacokinetic Modeling 4-3
4.2.2. Uncertainty Analyses 4-4
4.2.3. Route Extrapolation 4-5
4.3. ORAL REFERENCE DOSE FOR EFFECTS OTHER THAN CANCER 4-5
4.3.1. Critical Effects 4-6
8/1/01 iv DRAFT-DO NOT CITE OR QUOTE
-------
CONTENTS (continued)
4.3.2. Point of Departure 4-6
4.3.3. Uncertainty Factors 4-7
4.3.4. Calculation and Characterization of the Oral Reference Dose 4-9
4.4. INHALATION REFERENCE CONCENTRATION FOR EFFECTS
OTHER THAN CANCER 4-11
4.4.1. Critical Effects 4-11
4.4.2. Point of Departure 4-11
4.4.3. Uncertainty Factors 4-12
4.4.4. Calculation and Characterization of the Inhalation Reference
Concentration 4-13
4.4.5. Comparison of Reference Dose and Reference Concentration 4-13
4.5. SLOPE FACTOR AND INHALATION UNIT RISK FOR CANCER 4-15
4.5.1. Risk Estimates From Cancer Epidemiology Studies 4-15
4.5.2. Risk Estimates From Liver Tumors in Mice 4-18
4.5.3. Risk Estimates From Lung Tumors in Mice 4-23
4.5.4. Risk Estimates From Kidney Tumors in Rats 4-25
4.5.5. Risk Estimates from Testicular Tumors in Rats 4-26
4.5.6. Discussion of Alternative Cancer Risk Estimates 4-27
4.5.7. Sources of Uncertainty in the Dose-response Assessment 4-30
5. RESEARCHNEEDS 5-1
5.1. RESEARCHNEEDS FORHAZARD ASSESSMENT 5-1
5.2. RESEARCHNEEDS FOR DOSE-RESPONSE ASSESSMENT 5-2
5.3. RESEARCHNEEDS FOR EXPOSURE ASSESSMENT 5-3
REFERENCES R-l
8/1/01 v DRAFT-DO NOT CITE OR QUOTE
-------
LIST OF TABLES
Table 2-1. Preliminary dose estimates of TCE and TCE-related chemicals 2-5
Table 3-1. Liver tumors in mice exposed to TCE 3-50
Table 3-2. Kidney tumors in rats, adjusted for reduced survival 3-51
Table 3-3. Malignant lymphomas in mice exposed to TCE 3-52
Table 3-4. Testicular tumors in male rats exposed to TCE, adjusted for reduced survival 3-53
Table 3-5. Lung tumors in mice exposed to TCE 3-54
Table 4-1. Distributions of potential dose metrics from the uncertainty analysis of the
pharmacokinetic models 4-32
Table 4-2. Oral studies supporting development of an RfD 4-33
Table 4-3. Inhalation studies supporting development of an RfC 4-34
Table 4-4. Human-equivalent LED10s (mg/kg-d) derived from liver tumor incidences in four
mouse studies (NTP males or females, NCI males or females), using two dose
metrics (TCA area-under-the-curve, DCA area-under-the-curve) and four
pharmacokinetic models 4-35
Table 4-5. Extending the observed range with a study of less-than-lifetime dosing
of TCE 4-36
Table 4-6. Extending the observed range with studies of TCE's metabolites 4-37
Table 4-7. Human-equivalent LEC10s (|ig/m3) derived from lung tumor incidences in four
mouse studies, using RfC dosimetry 4-38
Table 4-8. Human-equivalent LED01s (mg/kg-d) derived from kidney tumor incidences
in rat studies, using 3/4-power scaling 4-39
Table 4-9. Compilation of cancer estimates 4-40
LIST OF FIGURES
Figure 3-1. Metabolism of TCE 3-55
Figure 4-1. Extending the observed range with less-than-lifetime dosing 4-41
Figure 4-2. Extending the observed range with studies of TCE's metabolites 4-42
Figure 4-3. TCE health risk values 4-43
8/1/01
VI
DRAFT-DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS
ADH alcohol dehydrogenase
AIRS Aerometric Information Retrieval System
ALDH aldehyde dehydrogenase
ATSDR Agency for Toxic Substances and Disease Registry
AUC area under the curve
Cal/EPA California Environmental Protection Agency
CEP Cumulative Exposure Project
CH chloral hydrate
CI confidence interval
CYP2E1 cytochrome P450 2E1
CYP450 cytochrome P450
DCA dichloroacetic acid, dichloroacetate
DCVC dichlorovinyl cysteine
DCVG dichlorovinyl glutathione
DEHP diethylhexyl phthalate
DHEA dehydroepiandrosterone
EPA United States Environmental Protection Agency
GST glutathione S-transferase
HPT hypothalo-pituitary-testis (axis)
LOAEL lowest-observed-adverse-effect level
mg/kg milligrams per kilogram body weight
mg/kg-d milligrams per kilogram body weight per day
mg/L milligrams per liter
mg/m3 milligrams per cubic meter
mM millimolar (concentration)
mmol millimoles
mRNA messenger RNA
N number (of cases)
NAT N-acetyl transferase
NCI National Cancer Institute
NHANES in Third National Health and Nutrition Examination Survey
NOAEL no-observed-adverse-effect level
NTP National Toxicology Program
8/1/01
vn
DRAFT-DO NOT CITE OR QUOTE
-------
LIST OF ABBREVIATIONS (continued)
OR odds ratio
PDH pyruvate dehydrogenase
PPAR peroxisome proliferator activated receptor
ppm parts per million
RfC reference concentration
RfD reference dose
RR relative risk
RXR retinoid-X receptor
SHBG sex hormone binding globulin
TCA trichloroacetic acid, trichloroacetate
TCE trichloroethylene
TCOH trichloroethanol
TEAM Total Exposure Assessment Methodology
TLV threshold limit value
TRI Toxics Release Inventory
ug/L micrograms per liter
ug/m3 micrograms per cubic meter
VHL von Hippel-Lindau tumor suppressor gene
8/1/01 viii DRAFT-DO NOT CITE OR QUOTE
-------
PREFACE
This assessment presents EPA's most current evaluation of the potential health risks from
exposure to trichloroethylene (TCE). TCE is a chemical solvent used for vapor degreasing of
metals and as an intermediate in the manufacture of other chemicals. One of the chemicals most
released into the environment, TCE is highly volatile, with most TCE released into the
environment finding its way into the air. An important exception occurs when TCE is released
into groundwater, where TCE can persist for years because of the limited contact between
groundwater and air. When people are exposed to TCE, it is readily absorbed by all exposure
routes and is widely distributed throughout the body.
The potential health risks of TCE in the environment have caused it to be listed as a
chemical of concern across several environmental programs. TCE is listed as a hazardous air
pollutant under the Clean Air Act, a toxic pollutant under the Clean Water Act, a contaminant
under the Safe Drinking Water Act, a hazardous waste under the Resource Conservation and
Recovery Act, and a hazardous substance under the Comprehensive Environmental Response,
Compensation, and Liability Act (Superfund). It is a toxic chemical with reporting requirements
under the Emergency Planning and Community Right-to-Know Act, and under the Toxic
Substances Control Act certain releases must be reported to the Toxics Release Inventory.
This assessment draws on 16 state-of-the-science papers published as a supplemental
issue of Environmental Health Perspectives (Wartenberg et al., 2000; Lash et al., 2000a; Pastino
et al., 2000; Moore and Harrington-Brock, 2000; Bull, 2000; Green, 2000; Fisher, 2000; Bois,
2000a,b; Clewell et al., 2000; Boyes et al., 2000; Barton and Clewell, 2000; Chen, 2000;
Rhomberg, 2000; Wu and Schaum, 2000), plus some other key references. Accordingly, this
assessment focuses on analysis and interpretation rather than a compilation of study results.
More detailed information on the epidemiologic and experimental studies of TCE can be found
in the state-of-the-science papers and in comprehensive reviews compiled by ATSDR (1997) and
EPA (1985, 1987).
As this assessment is written, it is being shaped by several new developments in risk
assessment. The practice of risk assessment is evolving from a focus on one toxic effect of one
pollutant in one environmental medium toward integrated assessments covering multiple effects
and multiple media that incorporate information about mode of action, uncertainty, human
variation, and cumulative effects of multiple pollutants in different media. This evolution
responds to recommendations of the National Research Council (NRC, 1994), which have been
embraced in EPA's proposed cancer guidelines (U.S. EPA, 1996).
Mode-of-action information is used throughout this assessment, both to understand the
cancer hazard and to suggest common mechanisms with some of the effects other than cancer.
8/1/01 ix DRAFT-DO NOT CITE OR QUOTE
-------
Mode-of-action information is also used to assess the relevance of the animal results and to
identify risk factors for sensitive human populations.
Uncertainty analyses are used to quantitatively characterize the range of plausible risk
estimates and to guide the choices made during the assessment. Uncertainty analyses can be
particularly useful for explicitly characterizing the ability of the data and the models to provide
stable, credible risk estimates.
Humans vary in their response to environmental contaminants, and risk assessors have
been charged with using mode-of-action information to identify susceptible populations. Of
special concern has been the potential for disproportionate risks to children. Executive Order
13045 (1997) requires "each Federal Agency shall make it a high priority to identify and assess
environmental health and safely risks that may disproportionately affect children, and shall
ensure that their policies, programs, and standards address disproportionate risks that result from
environmental health risks or safety risks."
Human response to an environmental contaminant can be affected by other exposures and
conditions that occur in a population. EPA (1997) guidance directs "each office to take into
account cumulative risk issues in scoping and planning major risk assessments and to consider a
broader scope that integrates multiple sources, effects, pathways, stressors, and populations for
cumulative risk analyses in all cases for which relevant data are available."
This assessment evaluates the potential health risks of TCE in light of the available
information and analyses pertinent to mode of action, uncertainty, human variation, and
cumulative risks.
8/1/01 x DRAFT-DO NOT CITE OR QUOTE
-------
AUTHORS AND CONTRIBUTORS
Authors
The final draft of this assessment was developed and written by V. James Cogliano,
Cheryl Siegel Scott, and Jane C. Caldwell of EPA's National Center for Environmental
Assessment (NCEA), a part of the Office of Research and Development. This work, however,
would not have been possible without the significant contributions of many others, for which the
authors are most appreciative.
Contributors
Quantitative analyses were contributed by Chao W. Chen (NCEA), Jerry Blancato
(EPA's National Exposure Research Laboratory), and Gina M. Pastino (formerly with NCEA as
a fellow of the American Association for the Advancement of Science). Earlier drafts of this
assessment included contributions by Jean C. Parker (NCEA), Robert P. Bellies (NCEA), and
Kim-Chi Hoang (formerly with NCEA, now with EPA Region 9 in San Francisco).
The authors would like to thank William Farland, Vanessa Vu, and Michel Stevens of
NCEA for management support and scientific comments through all phases of this assessment.
Special thanks are also due to John C. Lipscomb (NCEA), Robert E. McGaughy (NCEA),
Gregory M. Blumenthal (NCEA), and Hugh A. Barton (EPA's National Health and Ecological
Effects Research Laboratory) for constructive comments and suggestions on earlier drafts.
Antoinette Johnson (NCEA) deserves appreciation for her work in compiling and formatting
some of the detailed tables. In addition, the authors would like to thank David Kelly (NCEA)
and Melissa Revely-Brown (NCEA) for effectively managing contracts to facilitate workshops
for peer involvement and peer review.
This assessment has benefited from extensive peer involvement. Its conclusions draw
from 16 state-of-the-science papers written by experts who have conducted research related to the
assessment of TCE's health risks. These papers have been published as a supplemental issue of
Environmental Health Perspectives (volume 108, supplement 2, May 2000). Cheryl Siegel
Scott served as editor of this special issue and managed the overall development of the state-of-
the-science papers. Funding was provided by EPA, the U.S. Air Force, the Occupational Safety
and Health Administration, and private industry. The lead authors, affiliations listed in the state-
of-the-science papers, and topics of the state-of-the-science papers include:
Daniel Wartenberg, Robert Wood Johnson Medical School, Cancer epidemiology.
Lawrence H. Lash, Wayne State University School of Medicine, Metabolism.
8/1/01 xi DRAFT-DO NOT CITE OR QUOTE
-------
AUTHORS AND CONTRIBUTORS (continued)
Gina M. Pastino, U.S. Environmental Protection Agency (Fellow of the American
Association for the Advancement of Science), Human variation and susceptibility.
Martha M. Moore, U.S. Environmental Protection Agency, Mutagenicity.
Lawrence H. Lash, Wayne State University School of Medicine, Modes of action for
kidney tumorigenesis.
Richard J. Bull, Pacific Northwest National Laboratory, Modes of action for liver
tumorigenesis.
Trevor Green, Zeneca Central Toxicology Laboratory (United Kingdom), Modes of
action for lung tumorigenesis.
Jeffrey W. Fisher, Air Force Research Laboratory, Pharmacokinetic modeling.
Frederic Yves Bois, Institut National de I'Environnement Industriel et des Risques
(France), Statistical analysis of Fisher model.
Harvey J. Clewell III, ICF Consulting, Pharmacokinetic modeling.
Frederic Yves Bois, Institut National de I'Environnement Industriel et des Risques
(France), Statistical analysis of Clewell model.
William K. Boyes, U.S. Environmental Protection Agency, Neurotoxicity and
pharmacokinetics.
Hugh A. Barton, ICF Consulting, Dosimetry and mode of action for effects other than
cancer.
Chao W. Chen, U.S. Environmental Protection Agency, Mechanism-based modeling for
liver tumorigenesis.
Lorenz R. Rhomberg, Harvard School of Public Health, Dose-response analyses for
liver, lung, and kidney tumorigenesis.
Chieh Wu, U.S. Environmental Protection Agency, Exposure assessment.
These authors have individually provided many constructive comments and suggestions for this
synthesis and characterization, but were not asked for consensus on its overall conclusions.
Peer involvement was also facilitated by an External Involvement Group, composed of
nine representatives from private industrial organizations, public interest groups, academic
research institutions, and State and Federal agencies. These representatives were asked to (1)
propose additional topics for state-of-the-science papers and secure expert scientists as authors,
(2) review these additional papers for balance and completeness, (3) propose topics for synthesis
and characterization, (4) suggest peer reviewers, and (5) keep scientists from their sector
informed. The External Involvement Group includes:
8/1/01 xii DRAFT-DO NOT CITE OR QUOTE
-------
AUTHORS AND CONTRIBUTORS (continued)
Melvin Andersen, Colorado State University.
Paul Cammer, Trichloroethylene Issues Group.
Christopher DeRosa, Agency for Toxic Substances and Disease Registry.
Paul Dugard, Halogenated Solvents Industry Alliance.
Jodi Flaws, Environmental Defense.
Philip Landrigan, Mt. Sinai School of Medicine.
Elizabeth Maull, United States Air Force.
Margaret Round, Melinda Treadwell, Northeast States for Coordinated Air Use
Management.
Lauren Zeise, California Environmental Protection Agency.
These representatives have also provided many constructive comments and suggestions for this
synthesis and characterization, but were not asked for consensus on its overall conclusions.
Elizabeth Maull, Paul Cammer, and Paul Dugard were instrumental in identifying authors and
providing funding for some state-of-the-science papers.
8/1/01 xiii DRAFT-DO NOT CITE OR QUOTE
-------
1. SUMMARY AND CONCLUSIONS
1 1.1. OVERVIEW OF TCE HEALTH HAZARDS
2 An extensive database—comprising epidemiologic studies, animal bioassays, metabolism
3 studies, and mechanistic studies—shows that trichloroethylene (TCE) exposure is associated
4 with a wide array of adverse health effects (see Section 3.4). TCE has the potential to induce
5 neurotoxicity, immunotoxicity, developmental toxicity, liver toxicity, kidney toxicity, endocrine
6 effects, and several forms of cancer. Overall, similar effects are seen by oral and inhalation
7 exposures, suggesting that TCE or its active metabolites reach their target sites after absorption
8 from either exposure route. Special attention has been given to investigating TCE's potential to
9 cause cancer. Epidemiologic studies, considered as a whole, have associated TCE exposure with
10 excess risks of kidney cancer, liver cancer, lympho-hematopoietic cancer, cervical cancer, and
11 prostate cancer. TCE has been extensively tested in animals, with mice developing liver tumors,
12 lung tumors, and lymphomas, and rats developing kidney tumors and testicular tumors. The
13 epidemiologic evidence is strongest at sites where the animals develop cancer, with site
14 concordance for kidney cancer (in rats and humans), liver cancer (in mice and humans), and
15 lympho-hematopoietic cancer (in mice and humans). TCE is also associated with cervical cancer
16 and prostate cancer in humans, sites for which there are no corresponding animal models.
1 7 Extensive metabolic studies show that exposure to TCE results in internal exposure to a
18 complex mixture of TCE's metabolites (see Section 3.2). Much of TCE-induced toxicity may be
19 attributable to its metabolites, as toxicity tests show that some metabolites cause effects similar
20 to those of TCE and, in some assays, TCE is inactive in the absence of its metabolites.
21 Depending on the effect, the pattern of TCE-induced toxicity does not match any single
22 metabolite, but reflects a mixture. The metabolic studies also show that the pattern of metabolite
23 formation is different at high and low doses of TCE, and that metabolism can be altered by levels
24 of metabolites already in the body.
25 This last point takes on added significance in view of widespread environmental exposure
26 to some of TCE's metabolites (see Section 2). These exposures can be either direct (for example,
27 TCE's metabolites trichloroacetic acid [TCA] and dichloroacetic acid [DCA] are also major
28 byproducts of drinking water disinfection) or indirect (for example, the dry cleaning solvent
29 perchloroethylene can be metabolized to TCA). These common exposures give rise to a
30 background level of some metabolites through which TCE may act. Accordingly, these other
31 exposures should not be regarded as confounding exposures in the traditional sense. Rather, they
32 may help explain why TCE is associated with toxic effects in one study but not in another.
33 Determining background exposure to TCE and its metabolites is important to understanding
34 TCE-induced toxicity (see Section 1.8).
8/1/01 1-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 Many of TCE' s toxic metabolites are formed through the enzyme system that also
2 metabolizes ethanol, acetaminophen, and many other drugs and environmental pollutants (see
3 Section 3.2). Exposure to these chemicals (for example, recent alcohol consumption) can affect
4 enzyme levels, temporarily altering or enhancing TCE's metabolism and toxicity in a manner
5 analogous to drug interactions known for many pharmaceutical s (see Sections 1.6 and 3.3). In
6 addition, enzyme activity can vary significantly between individuals, through both intrinsic
7 factors (for example, genetics) and acquired factors (for example, disease) (see Sections 1.6 and
8 3.3). Thus, the effects of TCE can vary, depending on exposures and conditions specific to each
9 person.
10
11 1.2. MODE OF ACTION
12 Mechanistic research on a few TCE metabolites has yielded insights into TCE's possible
13 modes of action.1 The research to date indicates that TCE-induced carcinogenesis is complex,
14 involving multiple carcinogenic metabolites acting through multiple modes of action. Past
15 explanations, such as the hypothesis linking mouse liver tumors to peroxisome proliferation, are
16 not consistent with the whole of the data, and more complex hypotheses have been formulated
17 (see Section 3.5).
18 TCE-induced mouse liver tumors may arise through multiple metabolites and multiple
19 modes of action (see Section 3.5.1). The preponderance of evidence suggests that formation of
20 TCE's CYP450 metabolites2 is sufficient to explain the liver tumors caused by TCE (Bull, 2000;
21 Chen, 2000). Exposure to the CYP450 metabolites chloral hydrate (CH), TCA, or DCA can
22 cause liver tumors in mice, and DCA can also cause liver tumors in rats. A plausible mode of
23 action is that TCE induces liver tumors through TCA and DCA modifying cell signaling systems
'EPA's proposed cancer guidelines define mode of action as "a series of key events and processes starting with
interaction of an agent with a cell, and proceeding through operational and anatomical changes resulting in cancer
formation." A key event is "an empirically observed precursor step consistent with a mode of action." "Mode of
action" is contrasted with "mechanism of action," which implies a more detailed understanding and description of
events, often at the molecular level, than is meant by mode of action. An agent may work by more than one mode of
action, both at different sites and at the same tumor site (U.S. EPA, 1996, 1999). Mode-of-action data can be used to
(1) assess the relevance of laboratory animal results to human environmental exposures, (2) identify risk factors and
sensitive populations, (3) provide insight into the shape of the dose-response curve, and (4) quantify the relative
sensitivity of laboratory animals and human populations.
2Cytochmme P450 (CYP450) is a major system of enzymes that metabolize chemicals through oxidation. TCE
metabolism is most strongly dependent on activity of one particular enzyme, CYP2E1. CYP2E1 activity has been
found to vary among humans through both genetic and acquired factors, leading to differences in metabolism and
sensitivity to TCE-induced toxicily. Involvement of CYP2E1 in TCE metabolism also raises the possibility that
TCE metabolism and toxicity can be affected by exposure to other chemicals, as CYP2E1 activity can be induced by
its role in metabolizing many drugs and environmental pollutants, including ethanol, acetaminophen, benzene,
1,3-butadiene, styrene, methylene chloride, chloroform, carbon tetrachloride, vinyl chloride, vinylidene chloride, and
perchloroethylene (Pastino et al., 2000).
8/1/01 1-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 that control rates of cell division and cell death (Bull, 2000). Tumors from the TCA and DCA
2 bioassays apparently arise from different mechanisms that can be distinguished from each other
3 (see Section 3.5.1). Characterizing the tumors from the TCE bioassays as either TCA-mediated
4 or DCA-mediated is a critical research need that limits understanding of the respective roles and
5 interaction of TCA and DCA in the TCE-induced tumors (see Section 5).
6 There has been much less investigation of the mode of action of TCE-induced mouse
7 lung tumors. It has been suggested that the lung tumors arise from rapid CYP2E1 metabolism of
8 TCE in the Clara cells of the mouse lung that leads to an accumulation of CH and causes cell
9 damage and compensatory cell replication that in turn leads to tumor formation (Green, 2000).
10 Although this hypothesis is consistent with the higher level of CYP2E1 in mouse lung, it has not
11 yet been determined whether accumulation of CH is a key event causing tumors or a coincidental
12 event unrelated to the tumors (see Section 3.5.3). Moreover, several questions about this
13 hypothesis remain unresolved. In addition, CH is clearly clastogenic and mutagenic at high
14 doses (Moore and Harrington-Brock, 2000), raising the possibility of multiple modes of action.
15 TCE-induced rat kidney tumors may arise through multiple modes of action (see Section
16 3.5.2). The preponderance of evidence suggests that TCE's glutathione-S-transferase (GST)
1 7 metabolites3 are responsible for kidney toxicity and kidney tumors (Lash et al., 2000b).
18 Although these metabolites have not been tested in cancer bioassays, the GST metabolite
19 dichlorovinyl cysteine (DCVC) is mutagenic (Moore and Harrington-Brock, 2000). Cytotoxicity
20 may also be involved, as cytotoxicity is observed in both rats and humans exposed to high levels
21 of TCE (Lash et al., 2000b). Information is lacking on whether the CYP450 metabolites also
22 have a role in kidney tumor development.
23 Although these possible modes of action are consistent with much of the mechanistic
24 research that has been conducted, the explanations so far fall short of identifying the active
25 metabolites and sequence of key events involved at each cancer site (see Section 3.5). Ambiguity
26 about active metabolites and key events leads to uncertainty in extrapolating the proposed modes
27 of action across species. For some suggested modes of action, it is possible that most humans
28 may be less sensitive than rats or mice, and this differential sensitivity has been addressed in the
29 dose-response assessment (see Sections 4.5.2, 4.5.3). Species-dependent differences in the
30 kinetics of TCE's metabolism can bring about a different proportion of metabolites that depends
31 on dose. This, in turn, can lead to different apparent modes of action across species and across
32 dose ranges. For example, rat kidney tumors at high experimental doses may involve high-dose
33 cytotoxicity, while rat and human kidney tumors at lower doses may arise through a mutagenic,
^Glutathione-S-transferase (GST) is an enzyme system that produces a minor quantity of TCE's metabolites. The
potentially higher toxicity of these metabolites, however, may make them important agents of TCE's toxicity.
8/1/01 1-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 not cytotoxic, mode of action. Accordingly, it may not be appropriate to assume concordance of
2 tumor site or mode of action across species or across dose ranges.
3 Recent mechanistic research has begun to investigate a more fundamental level of cellular
4 activity (see Section 3.5), for example, TCE's effects on cell signaling and carbohydrate
5 metabolism (Bull, 2000). Effects at this level could indirectly accelerate ongoing tumor
6 development specific to each species. In this way, TCE could be an "opportunistic carcinogen"
7 that can contribute to ongoing processes or conditions that induce different forms of cancer in
8 different species or populations. This would imply that TCE poses a cancer risk to humans,
9 although mechanism-specific animal models might not predict human cancer risks with much
10 confidence.
11
12 1.3. WEIGHT OF EVIDENCE UNDER THE PROPOSED AND CURRENT CANCER
13 GUIDELINES
14 Under EPA's proposed (1996, 1999) cancer guidelines, TCE can be characterized as
15 "highly likely to produce cancer in humans" (see Section 3.6.2). Support for this characterization
16 is strong, including (1) association of TCE exposure with increased risks of human kidney
1 7 cancer, liver cancer, lympho-hematopoietic cancer, cervical cancer, and prostate cancer; (2)
18 induction of multiple von Hippel-Lindau (VHL) gene mutations in renal cell carcinoma patients
19 who had been exposed to TCE; (3) induction of kidney tumors and testicular tumors in rats and
20 liver tumors, lung tumors, and lymphomas in mice; and (4) mechanistic information suggesting
21 that the animal tumors arise through processes that may be relevant to humans. Under EPA's
22 current (1986) cancer guidelines, TCE would be classified as a "probable human carcinogen"
23 (group Bl), with "limited" human evidence and "sufficient" animal evidence of carcinogenicity.
24 Two areas of much past discussion are the weight of the human evidence and the
25 relevance of the animal tumors. Compared with previous TCE assessments, the weight of the
26 epidemiologic evidence has become stronger with the state-of-the-science analysis by
27 Wartenberg et al. (2000) (see Sections 3.4, 3.6.1). The epidemiologic studies have recently been
28 augmented by molecular information, in which multiple mutations of the VHL tumor suppressor
29 gene,4 primarily C-to-T changes including nucleotide 454, were found in renal cell carcinoma
30 patients with high, prolonged TCE exposure (Briining et al., 1997b) (see Section 3.5.2). The
4The von Hippel-Lindau gene normally suppresses renal cell carcinomas. Mutations to this gene have been noted in
kidney cancers and may be an important risk factor and mode of action for chemically induced renal cell cancer
(Lash et al., 2000b). To illustrate the difference between "mode" of action and "mechanism" of action, knowledge
that VHL gene mutations are involved may be enough to identify a plausible "mode" of action, whereas knowing
how such modifications induce subsequent events leading to kidney cancer would be needed to identify the
"mechanism" of action.
8/1/01 1-4 DRAFT-DO NOT CITE OR QUOTE
-------
1 mechanistic research into the mode of action for each animal tumor site has begun to link TCE
2 with disturbances in cell signaling and carbohydrate metabolism, which can lead to human cancer
3 and other diseases (see Section 3.5). Subject to dose-response adjustments for relative human-to-
4 animal sensitivity (see Sections 4.5.2, 4.5.3), this research makes it plausible that TCE acts
5 through mechanisms that can cause cancer in humans.
6 These characterizations are consistent with recent assessments by other authoritative
7 health agencies, including IARC, which classifies TCE as a "probable human carcinogen" (IARC
8 1995) and the National Toxicology Program (NTP), which classifies TCE as "reasonably
9 anticipated to be a human carcinogen" (NTP, 2000a).5 The NTP recently considered upgrading
10 TCE to "known to be a human carcinogen," making a case for a stronger classification in the
11 future (NTP, 2000b). Key research to support this classification includes further investigation of
12 VHL gene mutations in other cohorts exposed to TCE (see Section 5). On the other hand, a
13 sizable faction of scientists is not convinced of the relevance of the animal tumors or the strength
14 of the epidemiologic studies, which only recently have accumulated enough power to detect
15 associations between moderate occupational or residential TCE exposures and some relatively
16 common cancers. After considering this alternative view, these health agencies have now
1 7 converged in their assessment that TCE is "highly likely," "probable," or "reasonably anticipated
18 to be" a human carcinogen.
19
20 1.4. QUANTITATIVE HEALTH REFERENCE VALUES
21 For effects other than cancer, an oral reference dose (RfD)6 of 3 x 1CT4 mg/kg-d was based
22 on critical effects in the liver, kidney, and developing fetus (see Section 4.3.1). These effects
23 were observed in subchronic studies in mice and rats at doses as low as 1 mg/kg-d (see Section
24 4.3.2). Because laboratory conditions do not necessarily represent lifetime human environmental
25 exposure, including sensitive populations, this dose was adjusted from subchronic to lifetime
26 exposure, from animals to humans, and from a general to a sensitive population (see Section
27 4.3.3). In addition, humans start higher on the dose-response curve than do animals, as humans
28 are exposed to TCE's active metabolites from sources other than TCE, and this cumulative
5IARC's classification of TCE as a "probable human carcinogen" (their group 2 A) is based on "limited evidence" of
carcinogenicity in human studies and "sufficient evidence" in animal studies (IARC, 1995). The NTP's listing of
TCE as "reasonably anticipated to be a human carcinogen" is based on "limited evidence of carcinogenicity from
studies in humans, sufficient evidence of malignant tumor formation in experimental animals, and convincing
relevant information that trichloroethylene acts through mechanisms indicating it would likely cause cancer in
humans" (NTP, 2000a).
6A reference dose (RfD) is an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure
to the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious
noncancer effects during a lifetime (Barnes and Dourson, 1988).
8/1/01 1-5 DRAFT-DO NOT CITE OR QUOTE
-------
1 exposure can alter TCE's metabolism and toxicity (see Sections 1.8, 3.3, 4.3.3). Differences
2 between average and sensitive humans were estimated in an uncertainty analysis to be about 50-
3 fold, and a further 100-fold overall adjustment was considered appropriate for the other
4 differences (see Section 4.3.3). This latter 100-fold uncertainty indicates the potential for future
5 research to reduce uncertainty and improve this assessment's accuracy (see Section 5). In the
6 end, however, the RfD was calculated by reducing the 1 mg/kg-d dose by a factor of 3,000,
7 representing the largest divisor used by EPA in the presence of substantial uncertainty (see
8 Section 4.3.4).
9 An inhalation reference concentration (RfC)7 of 4x 10~2 mg/m3 was based on critical
10 effects in the central nervous system, liver, and endocrine system (see Section 4.4.1).
11 Occupational studies showed adverse central nervous system effects in workers at a subchronic
12 exposure to 7 ppm (38 mg/m3) (see Section 4.4.2). To convert this to an RfC for lifetime
13 exposure, this concentration was divided by a composite uncertainty factor of 1,000, representing
14 a default of 10 for human variation and 100 for uncertainty in extrapolating from subchronic to
15 lifetime exposure and effect levels to NOAELs (see Section 4.4.3).
16 The 3,000-fold and 1,000-fold uncertainty factors used in deriving the RfD and RfC,
17 respectively, are not large, considering that the Agency for Toxic Substances and Disease
18 Registry (ATSDR) (1997) has characterized the chronic studies as inadequate for supporting
19 chronic health reference values.8 The key uncertainty about the RfD and RfC is the relative lack
20 of chronic studies, necessitating reliance on subchronic studies and the numerous attendant
21 uncertainties in using subchronic studies to estimate health reference values for lifetime
22 exposure. This assessment's use of pharmacokinetic modeling and route extrapolation allows
23 comparison of a wider range of oral and inhalation studies, revealing that critical effects for both
7 A reference concentration (RfC) is an estimate (with uncertainty spanning perhaps an order of magnitude) of a
continuous inhalation exposure to the human population (including sensitive subgroups) that is likely to be without
an appreciable risk of deleterious noncancer effects during a lifetime (U.S. EPA, 1994).
8 ATSDR (1997) concluded that the available chronic-duration studies had inadequate characterization of exposure,
inadequate quantification of results, and lack of endpoints suitable for deriving chronic levels. ATSDR did calculate
an intermediate-duration inhalation level of 0.1 ppm, based on neurologic effects (decreased wakefulness and
decreased slow-wave sleep) in rats observed at a LOAEL of 50 ppm for 6 weeks (Arito, 1994). The corresponding
human-equivalent concentration (44 ppm) was divided by a composite uncertainty factor of 300 (10 for human
variability, 3 for extrapolation from animals to humans, and 10 for use of a LOAEL). ATSDR also calculated an
acute-duration inhalation level of 2 ppm, based on neurologic effects (fatigue and drowsiness) in six human
volunteers observed at a LOAEL of 200 ppm for 5 days (Stewart, 1970). The corresponding daylong concentration
(58 ppm) was divided by a composite uncertainty factor of 30 (10 for human variability and 3 for use of a minimal
LOAEL). ATSDR calculated an acute-duration oral level of 0.2 mg/kg-d, based on developmental effects
(decreased rearing) in mice observed at a LOAEL of 50 mg/kg-d (Fredriksson, 1993). This was divided by a
composite uncertainty factor of 300 (3 for human variability in TCE metabolism, 10 for extrapolation from animals
to humans, and 10 for use of a LOAEL).
8/1/01 1-6 DRAFT-DO NOT CITE OR QUOTE
-------
1 exposure routes fall within a narrow dose range and providing strength to both the RfD and RfC
2 determinations.
3 Several cancer slope factors9 were developed, with most slope factors falling between
4 2x 1CT2 and 4x 1CT1 per mg/kg-d. This range is supported by estimates derived for kidney cancer
5 and liver cancer from occupational exposure to TCE, non-Hodgkin's lymphoma from exposure
6 to drinking water contaminated with an average of 23 |ig/L TCE, and liver cancer in laboratory
7 mice exposed to TCE (see Sections 4.5.1, 4.5.2). Because this range is supported by diverse
8 studies and does not include the highest or lowest estimates that were calculated, it can be
9 considered rather robust and independent of the results or methods used to analyze any single
10 study (see Section 4.5.6).
11 The range of cancer slope factors has not been reduced to a single number. A range is
12 reasonable in view of the risk factors that can modify the effects of TCE in different populations
13 (see Sections 1.6, 3.3). (Ranges were not developed for the RfD and RfC, because they include
14 sensitive individuals by applying a factor to cover human variation.) For most cancer risk
15 factors, however, data that would allow differential risks to be quantified are lacking. Only for
16 the GST polymorphism that modifies the kidney cancer risk are data available, indicating a
1 7 fourfold increased risk for the more sensitive population. Because the modifying effect of most
18 risk factors cannot be quantified at this time, this assessment proposes instead that risk assessors
19 use the upper end of the slope factor range for susceptible populations having risk factors for
20 TCE-induced cancer. Although the extremes of the slope factor range are not based on data from
21 more- or less-susceptible populations, this approach emphasizes the possibility of different risks
22 in different circumstances, identifies risk factors that may increase susceptibility to TCE's
23 effects, and provides a practical way to adjust risk estimates to reflect differential susceptibility.
24 The purpose of a risk assessment can also affect the choice of a value from the slope
25 factor range. An assessment of maximum individual risk would use the upper end of the slope
26 factor range, while an assessment of the number of cancer cases in a general population could use
27 the midpoint of the range. This implies that the high end of the slope factor range is appropriate
9 A slope factor is an estimate of a carcinogen's potency, characterized as a plausible upper bound on the increased
human cancer risk from lifetime exposure to an average dose of 1 mg/kg-d. That is, the slope factor estimates a
bound on the risk per mg/kg-d, accordingly, the slope factor is expressed in units of inverse lifetime-average dose, or
(mg/kg-d)"1. Multiplying a slope factor by a lifetime-average dose (in mg/kg-d) yields a plausible upper bound on
the increased probability of developing cancer from exposure to the carcinogen. A unit risk is analogous to a slope
factor, but expressed in units of inverse lifetime-average ambient air concentration (ug/m3)"1 or inverse lifetime-
average drinking water concentration (ug/L)"1 instead of inverse lifetime-average dose (mg/kg-d)"1. Unit risks are
convenient when exposure is expressed in terms of environmental concentrations (• g/m3 or • g/L). Unit risk
estimates are based on particular exposure assumptions, specifically, a 70-kg adult drinking 2 L/d and breathing 20
m3/d. When applied to other populations with different exposure factors—for example, children—unit risk estimates
should be adjusted accordingly.
8/1/01 1-7 DRAFT-DO NOT CITE OR QUOTE
-------
1 for susceptible individuals but is less descriptive of the rest of the population. The 20-fold span
2 of the slope factor range is smaller than the uncertainty in dose estimates from the
3 pharmacokinetic models (see Table 4-1), suggesting that the range may, in some cases,
4 underestimate high-end risk.
5 For less-than-lifetime exposure, current risk assessment practice typically assumes that
6 cancer risk is proportional to exposure duration, for example, exposure for 7 years of a 70-year
7 human lifespan would carry one-tenth the risk of lifetime exposure.10 There is some evidence for
8 TCE to suggest that this practice may underestimate the risk from less-than-lifetime exposures
9 (see Section 4.5.2). Accordingly, tempering the assumption of duration-proportionality may be
10 warranted in applications involving less-than-lifetime exposure.
11 The continuing development of revised cancer guidelines (U.S. EPA, 1996, 1999) has
12 raised expectations that nonlinear approaches11 may replace the linear extrapolation used to
13 estimate cancer slope factors for TCE. This assessment has pursued nonlinear approaches for
14 several tumor sites. The key limitation to the nonlinear analyses is the uncertain identity of the
15 active metabolites and key events involved in TCE-induced cancer. This assessment's
16 consideration of potential nonlinear approaches indicates that concern for cancer would extend
1 7 down to doses of 3 x 10~4 and 8* 10~4 mg/kg-d for liver tumors and testicular tumors, respectively
18 (see Sections 4.5.2, 4.5.5). These doses are comparable to the RfD for effects other than cancer,
19 indicating the importance of considering all health effects in any risk assessment of TCE.
20
21 1.5. UNCERTAINTY
22 Several sources of uncertainty have been identified and quantified for TCE.12
23 Consideration of these uncertainties has shaped the choices made in this assessment and
'"Cancer risk is calculated by multiplying the slope factor by the lifetime average daily dose (LADD), defined as
LADD = (C x IR x ED) I (BW x LT), where C is the concentration, IR the intake rate, ED the exposure duration, BW
the body weight, and LTtiae lifetime (U.S. EPA, 1992a).
"Because nonlinear dose-response curves can have substantial model uncertainty below the experimental doses,
these curves are not extrapolated below the observed data. Instead, the risk assessment discusses current
understanding of the phenomena that may be occurring at lower doses. Factors to be considered include (1) nature
of the observed response, (2) shape of the observed dose-response curve, (3) human sensitivity compared with
experimental animals, (4) human variation in sensitivity, and (5) human exposure (U.S. EPA, 1996, 1999).
12Uncertainty arises because study conditions differ from conditions of human environmental exposure.
Consequently, risk assessments typically involve several extrapolations: from laboratory animals to humans, from
experimental doses to environmental levels, from one exposure route to another, from one exposure pattern to
another, and from small samples to large, more heterogeneous populations.
One type of uncertainty stems from different interpretations of the available data (known as model
uncertainty), another from estimation errors due to incomplete or imprecise data (parameter uncertainty).
An understanding of uncertainty can be used (1) to make more informed choices during the conduct of a
risk assessment, and (2) to characterize the range of plausible risk estimates for different populations.
8/1/01 1-8 DRAFT-DO NOT CITE OR QUOTE
-------
1 described the range of estimates that are consistent with existing data and current scientific
2 understanding.
3
4 1.5.1. Model Uncertainty13
5 In this assessment, the two major areas of model uncertainty are pharmacokinetics and
6 low-dose extrapolation.
7 The two pharmacokinetic models (Fisher, 2000; Clewell et al., 2000) reveal the existence
8 of substantial model uncertainty, with risk estimates that differed by 15-fold (see Table 4-4).
9 Fitting these models to additional data sets (Bois, 2000a, 2000b) produced calibrated models that
10 yielded more compatible results and reduced this source of model uncertainty (see Section 4.2.1).
11 The choice between linear and nonlinear extrapolation14 to low doses creates two distinct
12 classes of estimates (see Sections 4.5.2, 4.5.3). This dichotomous uncertainty stems from
13 different strongly held and widely held interpretations of the information on TCE's mode of
14 action (see Section 3.5). Accordingly, this assessment used both linear and nonlinear
15 approaches, the former to bound the risks and the latter to demonstrate the extent of uncertainty
16 and value of further research into mode of action and human variation.15 In addition, this
17 assessment made efforts to develop risk estimates from human studies as an alternative to
18 extrapolating from animal models. Average exposure in one human study was close to the
19 occupational standard, reducing the uncertainty of high- to low-dose extrapolation.
20
l?>Model uncertainty refers to a lack of knowledge needed to determine which scientific theory on which a model is
based is correct. In risk assessment, model uncertainty is reflected in alternative choices for model structure, dose
metrics, and extrapolation approaches. Other sources of model uncertainty concern whether surrogate data are
appropriate, e.g., using data on adults to make inferences about children. The full extent of model uncertainty cannot
be quantified, only models that have been analyzed. Model uncertainty is expressed through separate analyses
yielding alternative estimates, not as a single estimate or distribution made by combining estimates from
incompatible models. Combining results of incompatible models to create hybrid risk estimates can undermine the
value of having preferred models (NRC, 1994, p. 174).
^Linear extrapolation implies that risk decreases proportionally with dose below the experimental data. Sublinear
extrapolation (often called by the less specific term nonlinear extrapolation) implies that risk decreases more than
proportionally.
Toxicologists often speak of a threshold dose below which an individual does not respond. Risk
assessments generally do not use this term, as experimental data are not able to estimate thresholds with much
confidence. For example, a response that is not statistically significant does not indicate a threshold, rather it can be
consistent with a small risk that falls below the experiment's power of detection.
15Mode of action is not the only determinant of the shape of a dose-response curve. Human variation also affects the
shape by spreading the responses in a heterogeneous population over a wider dose range (see footnotes 17, 21).
Another factor is background exposure, which determines where to place an incremental environmental exposure on
the overall dose-response curve (see footnote 26).
8/1/01 1-9 DRAFT-DO NOT CITE OR QUOTE
-------
1 1.5.2. Parameter Uncertainty16
2 Uncertainty analyses and confidence intervals were developed for some of this
3 assessment's key pharmacokinetic and dose-response parameters. Each description of parameter
4 uncertainty assumes that the underlying model is valid.
5 Uncertainty in the pharmacokinetic data was analyzed and used to gauge which
6 applications could be supported by stable models. Confidence intervals spanned approximately
7 100-, 5,000-, and 14,000-fold for doses in the liver, kidney, and lung, respectively (see Table 4-1,
8 Section 4.2.2). The unstable model estimates for kidney and lung resulted in use of default
9 scaling methods for these sites. For all sites, median dose estimates were used in subsequent
10 calculations.
11 Uncertainty in the animal dose-response data is reflected by the ratio of ED10s to LED10s.
12 These generally do not exceed a factor of 2 (Rhomberg, 2000).
13
14 1.5.3. Consideration of Multiple Areas of Uncertainty
15 How to combine these individual aspects of uncertainly is an unresolved question in risk
16 assessment. Multiplying the sizes of the individual confidence intervals is not recommended, as
17 this would yield an overall confidence interval whose extremes are too improbable and whose
18 range is too wide to be of practical use. There are more defensible statistical approaches for
19 combining multiple uncertain quantities, but these depend on understanding the interrelationships
20 between the uncertain quantities and on addressing model uncertainty by assigning numerical
21 weights to the different models. Because this question is unresolved, this assessment has adopted
22 EPA's 3,000-fold limit on uncertainty factors for RfDs and RfCs and has developed a cancer
23 slope factor range that reflects a middle range of estimates that are supported by several sources
24 of data and several lines of reasoning. This implies, however, that the range may, in some cases,
25 underestimate high-end risk.
26 It has been suggested that to facilitate probabilistic risk assessment, slope factors be
27 replaced by slope factor distributions. The quantifiable uncertainties (for example, the twofold
28 LED10-to-ED10 range) are less important than others that have not been quantified (for example,
29 pharmacokinetic model uncertainty and parameter uncertainties). To avoid misleadingly narrow
30 uncertainty distributions that ignore the most important sources of uncertainty, this assessment
31 does not recommend distributions based on uncertainties of secondary importance.
^Parameter uncertainty refers to lack of knowledge about the values of a model's parameters. This leads to a
distribution of values for each parameter. Common sources of parameter uncertainty include random measurement
errors, systematic measurement errors, use of surrogate data instead of direct measurements, misclassification of
exposure status, random sampling errors, and use of an unrepresentative sample. Most types of parameter
uncertainty can be quantified by statistical analysis.
8/1/01 1-10 DRAFT-DO NOT CITE OR QUOTE
-------
1 Consideration of nonlinear dose-response models draws attention to the uncertainty in
2 estimating the point where the dose-response curve begins to turn sharply upward.
3 Overestimating the dose where the curve turns sharply upward could lead to the error of
4 declaring a dose "safe" when the risk actually could be quite high. Another consequence of using
5 nonlinear dose-response curves is the need to determine background exposure. In the case of
6 TCE, there is substantial background exposure to some of its toxic metabolites. If their dose-
7 response curves are steep (as for DCA-induced liver tumors in mice), the incremental risk can be
8 greater in a population with high background exposure (see Section 1.8).
9 The uncertainties quantified for TCE may be similar to what would be found for other
10 chemicals if pertinent analyses were performed. Thus, the practical importance of describing
11 TCE uncertainty is not to decrease confidence in the TCE estimates, but to understand that all
12 risk estimates are uncertain, and that the true risks can be either higher or lower. This assessment
13 uses central estimates, not extreme values, for several major sources of uncertainty. The
14 resulting risk estimates are supported by several sources of data and several lines of reasoning,
15 providing some measure of confidence that the risk estimates are robust and not likely to be
16 substantially changed by a single new study or analysis.
17
18 1.6. HUMAN VARIATION AND SENSITIVE POPULATIONS
19 The mechanistic information on TCE indicates a potential for considerable human
20 variation17 and suggests some risk factors18 that would make a population more sensitive. Further
21 research is needed to estimate the magnitude of the increased risk (see Section 5).
22
23 1.6.1. Metabolic Differences
24 Because TCE's metabolites contribute to its toxicity, differences in metabolism can lead
25 to differences in response (see Sections 3.2, 3.3). TCE's metabolism is extremely complex, and
26 variation in one metabolic reaction can shift activity from one metabolic pathway to another,
11 Human variation refers to person-to-person differences in susceptibility due to differences in exposure or
differences in biological sensitivity. Differences in biological sensitivity can be linked to inherited factors (e.g.,
genetics) or acquired factors (e.g., disease). Although human variation and uncertainty both can be characterized as
ranges or distributions, they are fundamentally different concepts. Uncertainty can be reduced by further research
that supports a model or improves a parameter estimate, but human variation is a fact that can be better
characterized, but not reduced, by further research. Fields other than risk assessment use "variation" or "variability"
to mean dispersion about a central value, including measurement errors and other random errors that risk assessors
address as uncertainty.
18A risk factor is a condition (e.g., genetics or disease) or exposure (e.g., alcohol or chlorinated solvents) associated
with a particular risk. For example, in the context of traffic accidents, known risk factors include alcohol
consumption, excessive speed, and bad weather. In a particular case, causation is often attributed to a combination
of risk factors.
8/1/01 1-11 DRAFT-DO NOT CITE OR QUOTE
-------
1 yielding different proportions of metabolites that, in turn, yield different toxicity. Because of this
2 complexity, metabolic differences may not lead to proportionate differences in risk, but do
3 indicate a potential for differential risks.
4 For GST metabolism, genetic polymorphisms19 have been identified and linked to an
5 increased risk of TCE-induced kidney cancer (see Sections 3.3, 3.4.3). The California
6 Environmental Protection Agency (Cal/EPA) (1999) has estimated that these may lead to
7 fourfold differences in kidney cancer risk (see Section 4.5.1). For CYP2E1 metabolism, several
8 investigators have found up to 50-fold variations in small groups of humans (see Section 3.3).
9 This suggests a potential for considerable differences in response.
10 It has been asserted that increases in metabolic rate due to enzyme induction have little
11 effect on the amount of metabolite formed in the liver (Kedderis, 1997).20 This assertion was
12 derived from physiologically based pharmacokinetic modeling exercises reported in a poster
13 abstract (Kedderis, 1996). A consequence of this assertion is that first-pass metabolism from oral
14 dosing would be virtually 100%, leaving no TCE to circulate throughout the body. More
15 complex multiple-pathway pharmacokinetic models (Fisher, 2000; Clewell et al., 2000) show
16 that this is not the case with concentrations of TCE circulating throughout the body being stored
17 as fatty tissue. Moreover, even if the total amount of metabolites formed is unchanged, the
18 balance between the rates of formation of GST metabolites and CYP450 metabolites would
19 clearly be altered by a relative induction of one enzyme as opposed to another. These
20 considerations suggest that enzyme induction and inhibition may be involved in susceptibility to
21 the effects of TCE.
22
23 1.6.2. Disease
24 Some of TCE's effects may result from its ability to disturb carbohydrate metabolism and
25 cell signaling (Bull, 2000). Diabetes is a disease caused by such disturbances, and diabetics are
26 at greater risk for liver cancer, kidney cancer, and many other effects. Further, uncontrolled
27 diabetes causes induction of CYP2E1, favoring increased formation of toxic metabolites. This
28 suggests that diabetics may be sensitive to effects from TCE (see Section 3.3).
1E'A polymorphism is the presence of two or more distinct types due to genetic variation in a population. A familiar
example is the polymorphism of blood types O, A, B, and AB. A polymorphism of metabolic enzymes indicates
variation in metabolic activity across groups of people. The genetic basis of polymorphisms implies that they can be
differently distributed across racial or ethnic groups.
20Under this hypothesis, known as blood flow limitation, the metabolic rate without enzyme induction is so rapid that
all TCE would be metabolized, hence metabolite production would be limited only by the amount of TCE present in
the blood flowing into the liver. Increasing the metabolic rate would have no effect, as there would be no TCE left to
metabolize.
8/1/01 1-12 DRAFT-DO NOT CITE OR QUOTE
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Alcohol, like TCE, is metabolized by CYP2E1, alcohol dehydrogenase (ADH), and
aldehyde dehydrogenase (ALDH). Concurrent exposure to TCE and alcohol can exacerbate the
metabolism and acute effects of both in a manner similar to acute drug-alcohol interactions
known for many pharmaceuticals. More importantly, chronic alcohol consumption can induce
long-term 20-fold increases in CYP2E1 levels, favoring increased formation of toxic metabolites
(see Section 3.3).
1.6.3. Gender Differences
TCE exposure has been linked to autoimmune diseases that occur mostly in women. In
addition, recent epidemiologic studies have linked TCE exposure to cervical cancer. These
differences in risk may be related to either pharmacokinetic or pharmacodynamic differences (see
Section 3.3).
1.6.4. Some Implications
Identifying risk factors that vary across the human population allows us to begin
identifying who is likely to be more sensitive. Human variation is a determinant of the shape of a
population dose-response curve, which can be different from the dose-response curve for an
individual.21 Understanding the functional relationship between a risk factor and the associated
disease is a critical research need that presently prevents estimating the differential risk faced by
sensitive populations (see Section 5).
21To illustrate, consider two sensitivity distributions, centered
at the same dose, for a homogeneous population (see thin line
in upper figure) and a heterogeneous population (see thick
line in upper figure). The corresponding population dose-
response curves have different shapes, with the curve for the
heterogeneous population spread over a wider range (see
lower figure). This occurs even when the mode of action
suggests a threshold, because the threshold would vary across
a heterogeneous population and the population dose-response
curve would be indistinguishable from those associated with
nonthreshold models. In the human population, genetic and
lifestyle factors contribute to variation in sensitivity that
spreads the dose-response curve over a wider range (Lutz,
1990).
8/1/01
1-13
DRAFT-DO NOT CITE OR QUOTE
-------
1 1.7. DIFFERENTIAL RISKS TO CHILDREN
2 There are several reasons to suspect that TCE could affect children and adults
3 differently.22
4
5 1.7.1. Exposure Differences
6 Nursing is an exposure pathway unique to children. From its lipophilic nature, TCE
7 would be expected to be present in milk. Limited data indicate that TCE has been detected in
8 each of eight samples of human milk from four U.S. urban areas, but levels were not reported
9 (ATSDR, 1997). Quantification of levels of TCE and its metabolites in milk, in both highly
10 exposed and relatively unexposed populations, is a critical research need that presently prevents
11 comparing levels in milk with levels allowed in drinking water (see Section 5). The widely
12 accepted benefits of nursing highlight the value of minimizing the exposure of nursing mothers
13 to TCE.
14 The cancer unit risk estimates (see Section 4.5) assume that mg/kg-d is an appropriate
15 measure of dose and incorporate exposure factors that are based on adults (specifically, 2 L/d
16 drinking water intake, 20 m3/d air intake, and 70 kg body weight). Relative to body weight,
1 7 however, children drink more water, eat more food, and breathe more air than do adults (U.S.
18 EPA, 1999). When assessing risks from less-than-lifetime exposure that occurs during
19 childhood, good exposure assessment practice would replace these adult exposure factors with
20 childhood exposure factors (U.S. EPA, 1999).23
21
22 1.7.2. Pharmacokinetic Differences
23 Children—especially infants—and adults can have different levels of metabolizing
24 enzymes (see Sections 3.2, 3.3). CYP2E1 is present in children, although levels are lower than in
25 adults. The same is true for some enzymes that clear TCE and its metabolites from the body.
26 Such differences may be responsible for early-life persistence in the body of trichloroethanol
27 (TCOH), the only TCE metabolite studied in infants, children, and adults. Half-life is 8 days in
28 adults, 10 days in children, 28 days in full-term infants, and 40 days in preterm infants (Renwick,
29 1998). This trend runs counter to the general expectation that smaller organisms metabolize and
22Children are not only a subpopulation, rather childhood is a sequence of life stages that affect the whole population.
EPA has long been concerned about environmental risks to children and has identified areas where risks can be
different in children and adults (Farland, 1992: (1) different exposures, (2) different remaining life expectancies
(early exposures leave more time for latent effects to develop), (3) different internal doses from the same external
dose (pharmacokinetic differences), and (4) different responses to the same internal dose (pharmacodynamic
differences).
23For example, to adjust the drinking water unit risk for a 9-kg infant who drinks 1 L/d (instead of a 70-kg adult who
drinks 2 L/d), multiply the unit risk by (1 L/d / 2 L/d) / (9 kg / 70 kg) = 3.9.
8/1/01 1-14 DRAFT-DO NOT CITE OR QUOTE
-------
1 clear chemicals more rapidly; instead, TCOH persists longer in children than in adults, longer in
2 infants than in children, and longer in preterm than in full-term infants. These differences do not
3 necessarily imply greater sensitivity, as we do not know which metabolites are responsible for
4 each adverse effect, but it does indicate a need for caution, as limited evidence suggests that TCE
5 metabolism and clearance are different in infants and children compared to adults.
6
7 1.7.3. Pharmacodynamic Differences
8 TCE's neurotoxic potential suggests an increased risk during childhood, a period of rapid
9 brain development (see Sections 3.4.1, 3.3). TCE's potential for developmental and endocrine
10 effects also supports a concern for childhood exposure (see Section 3.4.5). Cancer, too, is an
11 effect where children may be at higher risk. An accepted concept in carcinogenesis is that young
12 animals are usually more susceptible to the carcinogenic activity of a chemical than are adults
13 (McConnell et al., 1992).24 Because TCE's metabolites can cross the placenta, the period of
14 concern would extend to prenatal exposure. This concern is supported by a larger risk of
15 childhood leukemia observed after maternal exposure to TCE-contaminated drinking water
16 during the prenatal period (MA-DOH, 1997).
17
18 1.7.4. Some Implications
19 The health reference values developed in this assessment (see Sections 4.3-4.5) do not
20 address the exposure, pharmacokinetic, and pharmacodynamic considerations discussed here.
21 For example, the uncertainty factor for human variation is based on data collected in healthy
22 adults. Some assurance that this assessment does not understate risks to children comes from
23 consideration of sensitive test systems such as mouse liver. This sensitivity, though, is not
24 related to the reasons that children and adults may differ. Consequently, it may be appropriate to
25 consider a further factor to address the potential for differential risks to children (see Section 5).
26
27 1.8. CUMULATIVE RISKS INVOLVING TCE
28 Several chemicals have the potential to alter TCE's metabolism and clearance and
29 subsequent toxicity. Much TCE-induced toxicity is attributed to its metabolites, and risks from
30 TCE exposure can depend on background exposure to its metabolites. Some metabolites are
31 highly present in the environment (see Section 2), from both direct sources (for example,
24Although much has been made of the qualitative finding that perinatal exposure rarely induces tumors not found
with adult exposures, quantitatively, perinatal exposure in conjunction with adult exposure usually increases the
incidence and reduces the latent period of a given tumor. For example, in a survey of 22 chemicals tested for
perinatal exposure (TCE and its metabolites have not been so tested), a higher incidence of neoplasms was found
when animals were exposed during the last third of pregnancy (McConnell et al., 1992).
8/1/01 1-15 DRAFT-DO NOT CITE OR QUOTE
-------
1 ingestion of TCA and DCA as byproducts of drinking water disinfection) and indirect sources
2 (for example, metabolism of other chlorinated solvents to TCA). Hence, these exposures and
3 conditions can modify the effects of TCE on a population.25
4 At the same time, TCE exposure can augment toxicity induced by other chemicals.
5 Among these, support is strongest for TCE to exacerbate the effects of concurrent exposure to
6 alcohol or chlorinated solvents by inducing shared metabolic enzymes (see Section 3.3).
7
25Known as cumulative risk assessment, this area of inquiry focuses on a population and considers all chemicals and
stressors that can affect the population. A cumulative risk assessment goes beyond the effects of each chemical
individually to consider (1) how exposure to other chemicals and stressors can alter the toxicity of a chemical, (2)
how exposure to a chemical can alter the toxicity of other chemicals and stressors, and (3) how genetics, life stage,
lifestyle, diseases, and other conditions in an individual—factors associated with sensitivity—can affect expression
of a chemical's toxicity. In this way, cumulative risk assessments consider human-chemical interactions as well as
chemical-chemical interactions.
8/1/01 1-16 DRAFT-DO NOT CITE OR QUOTE
-------
1 When a nonlinear dose-response curve is assumed, high background exposure can imply
2 greater susceptibility to incremental exposures.26 Consequently, an estimate of background
3 exposure is needed to determine where to place an incremental exposure to TCE on the dose-
4 response curve, as the size of the increased risk is different in different regions of the dose-
5 response curve. Finding a safe dose in an otherwise unexposed population does not mean that
6 that dose is safe when background exposures are considered. This aspect of susceptibility is a
7 new feature that has not been addressed in past risk assessments, but will become important in
8 future assessments that consider nonlinear dose-response curves (see Section 5).
9 This assessment has begun to address cumulative risk in several ways:
10 *- Developing a pharmacokinetic model (at EPA's National Exposure Research Laboratory)
11 for concurrent exposure to TCE, TCA, and DCA. This will enable risk assessors at
12 different geographic sites to tailor the model to the cumulative exposures specific to their
13 sites (see Section 4.2.1).
14 > Identifying diseases and other risk factors that may make some individuals more sensitive
15 to TCE's effects (see Sections 1.6, 3.3).
16 + Including a modifying factor in the RfD to reflect that background exposures to TCE and
1 7 its metabolites can be higher in humans than in laboratory animals (see Sections 1.4,
18 4.3.3). An analogous modifying factor was not used in the RfC because the exposed
19 workers on which the RfC was based, unlike the laboratory animals on which the RfD
20 was based, are likely to have had background exposures comparable to those of the
21 general human population.
22 > Developing a range of cancer slope factors to permit risk assessors to choose a slope
23 factor based on a population's risk factors and background exposures (see Sections 1.4,
24 4.5.6).
26The correlation between background exposure and
susceptibility is a property of sublinear dose-response curves
in general. A familiar analogy is how federal income tax
rates increase with income. For example, the increased tax
on $10,000 additional income can be (see figure at right):
$0 if there is no other income.
$1500 if added to $20,000 income.
$2800 if added to $40,000 income.
In this analogy, the same incremental income ("dose") results
in a tax ("effect") that depends on background income.
>$2SOO
8/1/01
1-17
DRAFT-DO NOT CITE OR QUOTE
-------
2. EXPOSURE CHARACTERIZATION
1 This exposure characterization covers TCE, several metabolites of TCE, and other parent
2 compounds that produce a similar profile of metabolites (see Figure 2-1). Direct exposure to
3 TCE's metabolites or to parent compounds that produce these metabolites can alter or enhance
4 TCE's metabolism and toxicity by generating higher internal metabolite concentrations than
5 would result from TCE exposure by itself.
6 This characterization draws from the state-of-the-science paper on exposure (Wu and
7 Schaum, 2000). More detailed exposure information can also be found in a supplemental report
8 (Wu and Schaum, 2001).
9
10 2.1. ENVIRONMENTAL SOURCES
11 The major use of TCE is as a degreaser for metal cleaning operations. It is also used as a
12 paint stripper, adhesive solvent, ingredient in paints and varnishes, and in the manufacture of
13 organic chemicals. Releases from nonanthropogenic activities are negligible. TCE is on the list
14 for reporting to the Toxics Release Inventory (TRI). Reported releases into air predominate over
15 other types. Reported releases decreased by a third between 1987 and 1990 and have been stable
16 since then.
17 TCE's metabolites and their other parent compounds are also generally used in a wide
18 variety of manufacturing industries. Exceptions to this pattern are TCA and DC A, which are
19 formed as byproducts of drinking water chlorination; 1,1-dichloroethane, which can be formed
20 by biodegradation of 1,1,1-trichloroethane in landfills and other anaerobic environments; and
21 perchloroethylene, whose releases have come primarily from the large number of local dry
22 cleaning operations.
23 The TRI is the primary source of this information and can be considered recent and well
24 documented. The major uncertainty is the number of minor releases that occur, which are
25 exempt from reporting but, collectively, may be important.
26
27 2.2. ENVIRONMENTAL FATE
28 The chemical properties of TCE are well documented and widely accepted. Because of
29 its high vapor pressure, TCE in the atmosphere is expected to be present primarily in the vapor
30 phase rather than adsorbed to particles. Some removal from the atmosphere during wet
31 precipitation is expected because of the moderate solubility of TCE in water (1.1 g/L). The
32 major degradation process affecting vapor-phase TCE is photooxidation by hydroxyl radicals
33 (half-life of 1-11 days).
8/1/01 2-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 The dominant fate of TCE released to surface soils or surface waters is volatilization to
2 the air. Because of its moderate water solubility, TCE in soil (for example, landfills) has the
3 potential to migrate into groundwater. The relatively frequent detection of TCE in groundwater
4 confirms its mobility in soils. Biodegradation in soil and groundwater is thought to be slow
5 (half-life on the order of months to years). For releases to surface waters, bioconcentration,
6 biodegradation, and adsorption to sediments and suspended solids are thought to be insignificant.
7
8 2.3. EXPOSURE CONCENTRATIONS
9 No statistically based national sampling programs have been conducted that would allow
10 estimates of true national means for any environmental medium. A substantial amount of air
11 data, however, has been collected and summarized. TCE has been detected in the air throughout
12 the United States. Ambient air measurement data can be obtained from the Aerometric
13 Information Retrieval System (AIRS)27, which contains about 1,200 measurements from 25
14 States during 1985-1998. Air levels in 1998 can be summarized as follows: range=0.01-3.9
15 Hg/m3, mean=0.88, median=0.32, 90th percentile=l.76.
16 TCE ambient air concentrations in 1990 were modeled for all census tracts of the
1 7 continental United States as part of EPA's Cumulative Exposure Project (CEP)28. The modeling
18 suggests that 97% of the census tracts have TCE concentrations ranging from 0 to 1.5 |ig/m3.
19 The average level was estimated as 0.37 |ig/m3 and the maximum as 32 |ig/m3. The averages and
20 percentiles can be interpreted as population-weighted values because all census tracts have
21 roughly equal populations (but vary in geographic size). The modeling uses population
22 weighting for its estimates of concentration within each census tract. CEP data suggest a pattern
23 of generally low TCE levels in rural areas, higher levels in urban areas, and highest levels in
24 small commercial or industrial sectors across most States. The pattern is consistent with the
25 monitoring data.
26 These modeled values should be interpreted with caution. They are not as reliable as
27 measured values for specific locations. AIRS data show an average for 1990 across 59
28 monitoring stations of 1.84 |ig/m3. This is much higher than the national average from CEP of
29 0.37 |ig/m3. An important difference, though, is that the CEP estimate represents all areas of the
30 continental United States, whereas the 1990 AIRS data for TCE represent only 59 monitors
31 located in eight States.
32 TCE has been reported in rainwater, surface water, groundwater, drinking water, and
33 seawater. ATSDR reports that TCE is the most frequently reported organic contaminant in
27www.epa.gov/airsdata
28www.epa.gov/CumulativeExposure/air/air.htm
8/1/01 2-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 groundwater and has estimated that between 9% and 34% of drinking water supply sources have
2 some TCE contamination. Most municipal water supplies that do not use contaminated
3 groundwater, however, are in compliance with the maximum contaminant level of 5 |ig/L.
4 TCE has been reported in marine sediments, marine invertebrates, marine mammals,
5 foods, human milk, and human urine and blood. The food with the highest reported mean
6 concentration is butter and margarine (73.6 ppb).
7 Biological monitoring is an important supplement to environmental monitoring for
8 characterizing human exposure, as environmental monitoring data (especially for air) may not be
9 representative of actual exposures. TCE has been most frequently detected in persons exposed
10 through occupational degreasing operations. In 1982, TCE was detected in each of eight human
11 milk samples from four U.S. urban areas. NHANES IE reported that, from 1988 to 1994, TCE
12 levels in whole blood were below the detection limit of 0.01 mg/L for about 90% of the people
13 sampled.
14
15 2.4. EXPOSURE PATHWAYS AND LEVELS
16 Most people are exposed to TCE through drinking water, air, or food. TCE has been
1 7 detected in human milk samples from the general population; consequently, nursing infants may
18 be exposed. Also, because TCE can be present in soil, children may be exposed through
19 activities such as playing in or ingesting soil.
20 AIRS monitoring data for 1998 indicate a mean outdoor air level of about 0.8 ng/m3.
21 Based on an inhalation rate of 20 m3/d, intake by this pathway is about 18 |ig/d. This is
22 consistent with ATSDR, which reported an average daily air intake for the general population of
23 11-33 (ig/d. Total inhalation exposures are likely to be higher, because limited studies suggest
24 that indoor air may contribute more to overall exposure than outdoor air.
25 A survey of large water utilities in California for 1984 found a median concentration of 3
26 Hg/L- Based on a water consumption rate of 2 L/d, intake by this pathway is about 6 |ig/d. This
27 is consistent with ATSDR, which reported an average daily water intake for the general
28 population of 2-20 • g/d.
29 TCE can be found in groundwater. It is not known how often TCE reaches levels of
30 concern, but the highest potential for this to happen would be at wells located near TCE disposal
31 or contamination sites where leaching occurs. TCE is a common contaminant at Superfund sites.
32 It has been identified in at least 852 of the 1,416 hazardous waste sites proposed for inclusion on
33 the EPA National Priorities List. About 41 million people live less than 4 miles from a
34 Superfund site.
35 TCE exposures may be elevated for people living near waste sites, residents of some
36 urban or industrialized areas, people exposed at work, and people using products containing TCE
8/1/01 2-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 with poor ventilation. In addition, TCE is transferred from shower water to air (mean efficiency,
2 61%); ATSDR reported that a 10-minute shower in TCE-contaminated water could result in
3 inhalation exposure comparable to that from drinking TCE-contaminated tap water.
4 Table 2-1 presents adult exposure estimates for TCE and several of its metabolites and
5 other parent compounds that also produce those metabolites. Of special note is that general-
6 population exposures to TCA and DC A are each approximately 10-fold higher than exposure to
7 TCE. This indicates that there is a high background level of these metabolites that not only can
8 alter and enhance the metabolism of TCE, but also have toxic effects themselves. This topic is
9 further discussed in Section 3.
10 The general population exposure estimates are derived directly from environmental media
11 data and, therefore, have the same uncertainties as described above. In addition, inhalation
12 exposures are probably low, because they are based on ambient air measurements, which do not
13 reflect indoor levels. It is likely that indoor levels are generally higher and contribute more to
14 inhalation exposure than does outdoor air. For example, the 1987 EPA TEAM (Total Exposure
15 Assessment Methodology) study shows that the ratio of indoor to outdoor TCE concentrations
16 for Greensboro, NC, was about 5:1. The prevalence of TCE exposure on the basis of blood
1 7 levels is based on NHANES HI, which was recent (1988 to 1994) and fairly extensive (644
18 individuals).
8/1/01 2-4 DRAFT-DO NOT CITE OR QUOTE
-------
Table 2-1. Preliminary dose estimates of TCE and TCE-related chemicals
Chemical Population
Trichloroethylene
Tetrachl oroethyl ene
1,1,1 -Trichloroethane
1,2-Dichloroethylene
cis- 1 ,2-Dichloroethylene
1,1,1 ,2-Tetrachloroethane
1, 1-Dichloroethane
Chloral
Monochloroacetic acid
Dichloroacetic acid
Trichloroacetic acid
General
General
Worker
General
General
Worker
General
General
General
General
General
General
General
General
General
General
General
General
General
Medium
Air
Water
Air
Air
Water
Air
Air
Water
Air
Water
Air
Water
Air
Air
Water
Water
Water
Water
Water
Range of
estimated adult
exposures
C §/day)
11-33
2-20
2,232 - 9,489
80 - 200
0.1-0.2
5,897-219,685
10.8-108
0.38-4.2
1-6
2.2
5.4
0.5-5.4
142
4
2.47-469.38
0.02-36.4
2-2.4
10-266
8.56-322
Range of estimated
adult doses
(mg/kg/day) Data sources
1.57E-04-4.71E-04
2.86E-05-2.86E-04
3.19E-02-1.36E-01
1.14E-03-2.86E-03
1.43E-06-2.86E-06
8.43E-02-3.14
1.54E-04-1.54E-03
5.5E-06-6.00E-05
1.43E-05-8.57E-05
3.14E-05
7.71E-05
7.14E-06-7.71E-05
2.03E-03
5.71E-05
3.53E-05-6.71E-03
2.86E-07-5.20E-04
2.86E-05-3.43E-05
1.43E-04-3.80E-03
1.22E-03-4.60E-03
(3)
(3)
(3)
(13}
(13}
(13}
(16}
(16)
(18}
(18}
(6}
(6}
(6}
(7}
(7}
(6)
(19}
(19}
(19}
8/1/01
2-5
DRAFT-DO NOT CITE OR QUOTE
-------
3. HAZARD ASSESSMENT AND CHARACTERIZATION
1 3.1. GENERAL APPROACH
2 The database for TCE is relatively rich in toxicity information that includes
3 epidemiologic studies, animal bioassays, metabolism studies, genetic toxicity studies, and
4 mechanistic studies. Interpretation of this information, especially the cancer studies, has sparked
5 controversies concerning the strength of the human evidence and the human relevance of the
6 animal responses. These controversies have prompted some new epidemiologic studies and
7 reviews, plus a wealth of mechanistic research committed to developing and supporting
8 hypotheses about TCE's potential modes of action in animals.
9 To help provide a sound scientific basis for addressing these and other issues, this
10 assessment commissioned a set of state-of-the-science papers that appears in a May 2000
11 supplement of Environmental Health Perspectives. One paper addresses the controversy over the
12 weight of the human evidence. Wartenberg et al. (2000) conducted a joint analysis of the cancer
13 epidemiology studies, using a statistically based weight-of-evidence approach that stratified the
14 available studies into tiers according to how well each study's results can be associated with TCE
15 exposure specifically. Such an analysis affords an opportunity to put differing results in
16 perspective by evaluating whether the positive or the nonpositive results overall have greater
1 7 weight, and whether the overall result is statistically significant.
18 Other state-of-the-science papers focus on describing the effects of TCE and its
19 metabolites as well as the human relevance of observed animal responses. Lash et al. (2000a)
20 describe the metabolism of TCE, providing a foundation for the consideration of various
21 metabolites and target sites throughout the body. Pastino et al. (2000) discuss factors that can
22 cause humans to vary in their susceptibility to adverse health effects from TCE exposure and
23 identify populations that are likely to be more susceptible. Moore and Harrington-Brock (2000)
24 discuss the genetic toxicity of TCE and its metabolites, a critical step in understanding TCE's
25 mode of action at different sites. Bull (2000), Green (2000), and Lash et al. (2000b) discuss
26 potential modes of action for tumorigenesis in the liver, lung, and kidney, respectively. These
27 latter papers reflect the considerable research effort that has gone into developing the modes of
28 action for cancer for their respective tumor sites, described as targets in past TCE assessments
29 (U.S. EPA, 1985, 1987). The paper of Barton and Clewell (2000) reviews the evidence in the
30 rodent for effects other than cancer (discussed in this section), and also provides support for the
31 development of an inhalation reference concentration and oral reference dose (see Section 4).
32 Bull's state-of-the-science paper (Bull, 2000) has turned attention toward cell signaling to
33 give insight into possible modes of action for TCE. This in turn, has helped provide focus for
34 common modes and targets for cancer and other forms of toxicity. Many potential targets of
8/1/01 3-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 TCE toxicity (e.g., heart, brain, testes, and the developing fetus) may be affected by disturbances
2 in carbohydrate metabolism and alterations in cell signaling, thus, common themes may
3 ultimately emerge for different targets and effects. Particularly relevant is the work of Bannasch
4 (Bannasch et al., 1984, 1986, 1997; Bannasch, 1986, 1996) demonstrating that disturbance of
5 carbohydrate metabolism and cell signaling may be an early marker and common feature of
6 carcinogenesis. Discussions with the state-of-the-science authors have also directed attention to
7 the potential role in the toxicity of TCE of activation of the peroxisomal proliferator-activated
8 receptor by TCE or its metabolites. They also suggested a focus on the rationale supporting
9 linear or nonlinear carcinogenic responses in the liver. Accordingly, an expanded mode-of-
10 action section includes a detailed discussion of potentially relevant epigenetic mechanisms of
11 carcinogenesis.
12 This hazard assessment focuses on analysis and interpretation rather than a compilation of
13 study results. More detailed information on the epidemiologic and experimental studies on TCE
14 can be found in the state-of-the-science papers and in comprehensive reviews compiled by
15 ATSDR (1997), IARC (1995), and EPA (1985, 1987). Considerable new literature has become
16 available on TCE's potential toxicity and modes of action, including several effects other than
17 cancer. Some of the new information postdates the publication of the state-of-the-science papers
18 and is integrated into this hazard assessment. Thus, the discussion of toxicity and mode of action
19 is considered more current and the recent references expand the scope beyond cancer. Section
20 3.2 examines TCE's metabolism and clearance, because TCE-induced toxicity is often attributed
21 to TCE's metabolites. This leads to a discussion in Section 3.3 of susceptibility, because several
22 features of TCE metabolism and clearance have been found to be differentially distributed across
23 the general population, suggesting some potentially susceptible populations. Section 3.4
24 summarizes the toxic effects reported from TCE exposure. These effects are grouped by organ
25 system, and for each organ system, cancer and other effects are discussed together. Each
26 summary of TCE-induced effects is followed by a summary of available information about
27 similar effects reported in studies of TCE's metabolites. Section 3.5 discusses some potential
28 modes of action through which TCE may induce these effects. A mode of action can involve
29 several systems, and conversely, effects in one system may arise from several distinct modes of
30 action. Much of the research supporting the hypothesized modes of action has concentrated on a
31 few TCE metabolites rather than on TCE itself.
32
33 3.2. METABOLISM
34 The state-of-the-science paper by Lash et al. (2000a) discusses the absorption,
35 distribution, metabolism, and excretion of TCE and its metabolites. TCE is rapidly and
36 extensively absorbed by all routes of environmental exposure—ingestion, inhalation, and skin
8/1/01 3-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 contact. Once absorbed, TCE is distributed via the circulatory system throughout the body,
2 where it can accumulate in fat and other tissues. Storage of TCE in fat represents an internal
3 source of exposure that can later release TCE again into the circulation. The pathways of
4 biotransformation observed in humans and animals are thought to be similar qualitatively
5 (Pastino et al., 2000).
6 TCE is metabolized primarily in the liver, but metabolism can also occur in other tissues.
7 Metabolism in these other tissues may be important to TCE-induced toxicity. The kidney is the
8 only other tissue that has been extensively studied to any extent. Metabolism occurs through two
9 main pathways (oxidation or conjugation with glutathione) (see Figure 3-1). Although several
10 metabolic products derived from these pathways have been identified for TCE, only a few have
11 been characterized pharmacokinetically (Lash et al., 2000a). The first major pathway begins with
12 oxidation of TCE by at least four isozymes of the cytochrome P450 (CYP450) system to CH,
13 which is then acted upon by ADH and ALDH to form trichloroethanol (TCOH) and TCA,
14 respectively. CYP450 is also most likely involved in the reduction of CH to TCOH with TCA as
15 a product. TCOH undergoes glucuronidation and is either excreted in the urine or persists in the
16 body through enterohepatic recirculation.29 TCOH can then be excreted from the liver to the
1 7 small intestines, where it is reabsorbed into the circulation. In general, enterohepatic
18 recirculation is more extensive in humans than rodents (Barton and Clewell, 2000), with the
19 result of a larger area-under-the-curve (AUC) in humans than rodents for an equivalent oral dose.
20 Overall, the rate-limiting step for this pathway is oxidation of TCE to CH (Lash et al., 2000a).
21 DC A is another metabolite detected in mice and humans, but the source of DCA
22 formation is not clear. DCA may result from the oxidative pathway by further oxidation of both
23 TCA and TCOH or by rearrangement directly from dichloroacetyl chloride, from glutathione
24 conjugation, or, likely, from both pathways (Lash et al., 2000a; Cai and Guengerich, 2000;
25 Volkel et al., 1998). The estimation of the amount of DCA formed in humans contains
26 uncertainty because there are problems associated with the analytical methodology, particularly
27 in the presence of large amounts of TCA, and because of DCA's relatively rapid metabolism to
28 oxalic acid, monochloroacetic acid, glycolic acid, and glyoxylic acid. Conversion to glyoxylic
29 acid is thought to be by a newly described isoform of glutathione S-transferase (GST), identified
30 as GST zeta (GSTZ) (Lash et al., 2000a). GSTZ enzyme activity in vitro differs between rats and
31 humans and is relatively more sensitive to inhibition by DCA in rats (Cornett et al., 1999). In
32 general, GSTs are present in multiple forms in mammals and have a large degree of overlapping
29Enterohepatic recirculation is the process by which metabolites formed from the biotransformation of the parent
compound in the liver are excreted directly into the bile without first entering systemic circulation. From the bile,
the metabolites pass into the small intestine and are available for reabsorption (Klaussen, 1980).
8/1/01 3-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 substrate specificity among the differing forms. They also exhibit polymorphisms and species
2 differences in expression and catalytic activities. Hence, the further characterization of this
3 pathway as it relates to TCE is important.
4 The second major pathway of TCE metabolism also involves GST and begins with
5 glutathione conjugation, primarily in the liver, to S-(l,2-dichlorovinyl)glutathione (DCVG).
6 DCVG formed in the liver is excreted in the bile and converted in the biliary tract and intestines
7 to S-(l,2-dichlorovinyl)-L-cysteine (DCVC). DCVC is translocated to the kidneys, where its
8 concentration modulates the amount of toxic metabolites generated in the kidney. DCVC can be
9 detoxified by N-acetyltransferase (NAT) in the kidney and excreted in the urine, or it can be
10 metabolically activated to a thioacetylating agent by • *lyase. Although the capacity of this
11 pathway is small compared with the CYP450 pathway, the chemically unstable and highly
12 reactive nature of the metabolites produced by • *lyase metabolism make it likely that relatively
13 small amounts can induce significant responses (Lash et al., 2000a).
14 Quantitatively, the comparative rate of TCE metabolic reactions between species or
15 between potential target organs is a complex question. The interspecies relationship of TCE
16 metabolic pathways has been investigated using pharmacokinetic models (Clewell et al., 2000;
1 7 Fisher, 2000) and analyses that look at uncertainty in those models (Bois, 2000a,b). Peak blood
18 levels of TCE's oxidative metabolites (e.g., TCA) are higher in mice and rats than in humans
19 administered equivalent doses, whereas blood levels of TCE in humans generally remain for a
20 longer time period, i.e., longer half-life. Differences in enterohepatic recirculation and
21 metabolism could be factors in interspecies differences. These factors result in greater
22 persistence in the body and a larger AUC for some TCE metabolites in humans compared to
23 rodents. For the GST conjugation pathways in the kidney, in vitro work suggests that DCVG
24 formation in human liver and kidney was only slightly lower than in the rat, with DCVG
25 formation in mice shown to be markedly faster than in either humans or rats (Lash et al., 2000a).
26 However, DCVG formation is only the initial step in the generation of nephrotoxic species, with
27 additional studies necessary to characterize this metabolic pathway (Lash et al., 2000a). A
28 gender difference also is apparent, with an overall higher GSH conjugation in male rats than in
29 female rats (Lash et al., 2000a).
30
31 3.3. SUSCEPTIBILITY
32 The state-of-the-science paper by Pastino et al. (2000) discusses physiologic factors (e.g.,
33 genetics, gender, and age) and acquired factors (e.g., disease, alcohol consumption, and exposure
34 to other solvents) that may cause humans to vary in their susceptibility to adverse health effects
35 from TCE exposure. Another way to view the factors that can influence susceptibility to
36 potential TCE-related toxicity is to categorize them as either intrinsic or extrinsic. Inherent
8/1/01 3-4 DRAFT-DO NOT CITE OR QUOTE
-------
1 differences in an individual's genetic makeup, gender, age, or disease state may affect
2 susceptibility intrinsically, whereas coexposure to a pollutant that alters their metabolism or
3 clearance, or that adds on to background levels of metabolites, may affect susceptibility
4 extrinsically. Through either mechanism, increased susceptibility to potential TCE-related
5 toxicity may occur by altering the body burden of TCE or its metabolites, sensitizing the
6 individual to other agents that act in common metabolic pathways, or augmenting underlying
7 conditions that share common effects with TCE (e.g., effects on cell signaling).
8 The variation observed for specific metabolic reactions or responses within individual
9 populations suggests potential for considerable overall differences in response between
10 individuals in the whole population. It is, however, difficult to describe quantitatively the overall
11 extent of human variability, as the range of variation in a single process or metabolic reaction, as
12 well as variation within a specific ethnic group, may not encompass the span of variation across
13 the entire population or take into account the multiplicity of factors that can determine the
14 toxicity of TCE. Nonetheless, the variation observed for specific metabolic reactions or groups
15 not only suggests a potential for considerable overall differences in response between
16 individuals, but can be used to identify populations and conditions that may be associated with a
1 7 higher risk from TCE exposure.
18 Human variation is also a determinant of the shape of a population dose-response curve.
19 The dose-response curve for a susceptible population can be different and changed when
20 compared to the dose-response curve for the heterogeneous population. Additionally, direct or
21 background exposure to some of TCE's metabolites can be considerable and, for individuals with
22 these exposures, can aggravate the effect of small incremental exposures of TCE. That is,
23 background and cumulative exposures place the susceptible population on a different part of the
24 dose-response curve. For example, given a metabolite with an "S"-shaped dose-response curve,
25 a small dose would induce a minimal response in the absence of background exposure, yet the
26 same dose could induce a greater response in the presence of background exposure that places an
27 individual higher on the "S"-shaped curve, in the range where the risk increases rapidly with
28 dose.
29
30 3.3.1. Intrinsic Factors Affecting Susceptibility
31 3.3.1.1. Differences in Metabolism and Clearance
32 Adverse effects resulting from TCE exposure are believed to be attributed mainly to its
33 metabolites. Hence, individual differences in metabolism can lead to individual differences in
34 response. The influence of physiologic and acquired factors on metabolism, particularly on
8/1/01 3-5 DRAFT-DO NOT CITE OR QUOTE
-------
1 CYP2E1,30 which may be responsible for more than 60% of TCE metabolism, is important
2 (Lipscomb et al., 1997). Rates for TCE metabolism by the oxidative pathway are believed to
3 vary across the population. In vitro, an eightfold variation in metabolic activity of TCE via
4 CYP2E1 was seen in 23 human liver samples (Lipscomb et al., 1997); several investigators have
5 found activity of CYP2E1, in general, to vary up to 50-fold in humans (Stevens et al., 1994;
6 Peters et al., 1990; Yoo et al., 1998; Raucy, 1995; Lieber 1997; Pastino et al., 2000).
7 For the GST pathway,31 genetic polymorphisms may not only influence metabolism of
8 TCE and its target of toxicity. One epidemiologic study has linked polymorphisms of the GST
9 pathway (GSTT and GSTM) to increased risk of kidney cancer from TCE exposure (Briining et
10 al., 1997a). Additionally, one isozyme, GSTZ, is important to the further metabolism of DCA
11 (Tong et al., 1998), and in vitro has been shown to be irreversibly inhibited in mice, rats, and
12 humans with chronic DCA exposure (Tzeng et al., 2000). In vitro studies of cytosolic GST
13 conjugation of TCE in humans have revealed considerable (3-8 fold) variation in relatively small
14 samples (Lash et al., 2000a). In general, regulation of GST expression differs among tissues,
15 such that not all GST isoforms are expressed in every tissue. Partial deletions are also found to
16 varying degrees in specific ethnic populations (15%-20% of Caucasians and 60% of Asians
17 having partial deletions). Therefore, it is difficult to accurately predict expression of GST genes
18 in a given tissue (Eaton and Bammler, 1999).
19
20 3.3.1.2. Disease States
21 Certain diseases may increase susceptibility to TCE's adverse effects through induction
22 of metabolic enzymes, favoring increased formation of toxic metabolites of TCE, or by adding to
23 ongoing physiological processes. For example, uncontrolled diabetes causes induction of
24 CYP2E1 (Wang, 2000; Nakajima, 2000). Furthermore, diabetics may be a population
25 susceptible to TCE's adverse effects on the basis of recent mechanistic understanding about the
30CYP2E1 is one of the CYP450 enzymes and is the primary catalyst of hydrocarbon bioactivation in animals and,
most likely, in humans as well (Raucy et al., 1993). CYP2E1 also metabolizes many drugs and environmental
pollutants, including ethanol, acetaminophen, benzene, 1,3-butadiene, styrene, methylene chloride, chloroform,
carbon tetrachloride, vinyl chloride, vinylidene chloride, and perchloroethylene. Exposure to these chemicals can
affect enzyme levels or activity and, in so doing, alter or enhance TCE's toxicity. CYP2E1 is found in greatest
concentration in liver, but is also expressed in parenchymal and non-parenchymal cells of the liver (e.g., Kupffer
cells) and in many extrahepatic tissues (e.g., kidney, brain, testes, lung, and bone marrow).
3'GST enzymes are involved in metabolizing many drugs and environmental pollutants, including poly cyclic
aromatic hydrocarbons, phenolic antioxidants, reactive oxygen species, isothiocyanate, trivalent arsenicals,
barbiturates, and synthetic glucocorticoids (Hayes and Pulford, 1995). GST metabolism aids in the clearance of
xenobiotics and is generally considered a detoxification process. This is not the case for trichloroethylene, where
highly reactive metabolites can be produced.
8/1/01 3-6 DRAFT-DO NOT CITE OR QUOTE
-------
1 relationship between TCE and its metabolites and carbohydrate handling and cell signaling (Bull,
2 2000) (see Section 3.5.1).
3
4 3.3.1.3. Gender
5 Gender-related differences in metabolism may influence susceptibility by shifting
6 production of metabolites from one toxic pathway to another, or by providing differing targets for
7 TCE's toxicity. Males have been reported to have higher rates of DCVG formation from acute
8 exposure (Lash et al., 2000a). There may also be gender-related differences in target organs.
9 One recent case-control study (Dosemeci et al., 1999) observed an elevated risk among females,
10 but not males, between kidney cancer and occupational TCE exposure. This study was not able
11 to address potential gender differences in level of exposure, however. Females have also shown
12 a greater susceptibility than males to some of TCE's immune system effects (Sanders et al.,
13 1992; Barton and Clewell, 2000). Moreover, gender-specific organs may be targets of TCE's
14 effects. TCE exposure has been associated with excess risks of cervical cancer in occupationally
15 or environmentally exposed women (Blair et al., 1998; Anttila et al., 1995; Burg, 1997), and in
16 male rats, has been shown to concentrate in the testes (Zenick et al., 1984), decrease serum
17 testosterone, and alter testicular steroid precursors (Kumar et al., 2000).
18
19 3.3.1.4. Age
20 Several inherent aspects of uptake, distribution, and clearance of TCE in children can
21 predispose children to its toxicity. In general, children are believed to metabolize and clear
22 xenobiotics faster than adults, although rate differences vary across chemicals and the patterns of
23 metabolism may be different from those of adults. In contrast to this general expectation, the
24 half-life of TCOH in the body has been reported to be 20% higher in children, 3.5-fold higher in
25 full-term infants, and fivefold higher in preterm infants compared with healthy adults (Renwick,
26 1998). This finding, which could extend to other TCE metabolites, may be due to differing
27 enzyme activities in children. For example, glucuronidation does not reach adult levels until 3 to
28 6 months of age. Individual children may also have low levels of a particular isoform, as shown
29 by their inability to efficiently carry out glucuronidation and their increased risk for developing
30 hyperbilirubenemia. Such children may also be less able to glucuronidate TCOH. Furthermore,
31 ADH activity has been shown to be decreased in the neonate (Pastino et al., 2000). Other life
32 stage differences, such as the amount of body fat in an organ, can also lead to differences in TCE
33 distribution and accumulation. The brain of the newborn contains more lipid than that of adults,
34 thus affecting the distribution of TCE.
35 The prenatal and postnatal periods of development may be critical for TCE's potential
36 carcinogenic and noncarcinogenic toxicity. Although the modes of action of the carcinogenicity
8/1/01 3-7 DRAFT-DO NOT CITE OR QUOTE
-------
1 of TCE are unknown, if TCE behaves similarly to the limited number of chemicals tested during
2 perinatal exposure as well as adult exposure, the carcinogenic potential of TCE may also be
3 greater in the young (McConnell, 1992). Moreover, TCE exposure to mice during the perinatal
4 period has been reported to be associated with long-lasting alterations in spontaneous motor
5 activity (Fredriksson et al., 1993), although this report is only suggestive because of its lack of
6 statistical analysis on the litter. Exposure during the prenatal period has been shown to induce
7 toxicity. For example, animal studies report associations between maternal exposure to TCE in
8 drinking water and cardiac malformations in offspring (Johnson et al., 1998a,b; Dawson et al.,
9 1993; Epstein et al., 1993). Finally, the longer life expectancy of children compared with adults
10 may lead to a longer period for expression of toxicity.
11
12 3.3.2. Extrinsic Factors Affecting Susceptibility
13 3.3.2.1. Alteration of Metabolism and Clearance Through Coexposure to Other Chemicals
14 Exogenous exposures that induce CYP450 enzymes can enhance TCE metabolism and
15 toxicity. Cytochrome P450, particularly CYP2E1, is important to the metabolism of many
16 exogenous chemicals, including solvents and ethanol; CYP2E1 is also readily induced by a large
17 number of these substrates. For example, ethanol is metabolized by CYP2E1, alcohol
18 dehydrogenase, and aldehyde dehydrogenase. Although interactions have been demonstrated in
19 the toxicity of both ethanol and TCE given simultaneously at relatively high levels for short
20 periods, more profound effects on TCE toxicity are liable to occur from chronic ethanol
21 exposure. Chronic ethanol consumption can result in 20-fold induction of CYP2E1 (Pastino et
22 al., 2000; Fisher, 2000), significant because large numbers of people consume ethanol. In
23 addition, ethanol is a weak inducer of peroxisomes, with ethanol consumption linked with
24 increased risk of liver cancer (Falk, 1982; NIH, 1985). Thus, exposure to both ethanol and TCE
25 may create conditions that increase risk through shared pathways of metabolism, induction of
26 metabolic enzymes, and potential targets of toxicity. Methanol is often a concurrent exposure for
27 TCE and shows affinity for many of the same enzymes. Furthermore, TCE exposure can
28 increase the toxicity associated with methanol and ethanol exposure by altering not only
29 metabolism to aldehydes but also their detoxification. For example, TCE exposure has been
30 reported to induce an "alcohol-flush," similar to those who possess an inactive ALDH allele (as
31 seen in half of individuals of Asian heritage), which may be due to TCE's ability to decrease
32 ALDH activities at relatively low concentrations (Wang et al., 1999). In addition, the induction
33 of P450 metabolism or aldehyde clearance by TCE may increase aldehyde production as well
34 (Wang et al., 1999).
35
36 3.3.2.2. Additive Effects from Coexposure to Other Chemicals
8/1/01 3-8 DRAFT-DO NOT CITE OR QUOTE
-------
1 The addition of TCE exposure to other background exposure has the potential to increase
2 susceptibility to TCE's adverse effects. Exposure to perchloroethylene, the dry cleaning solvent,
3 can be considerable in some populations, and TCA is a primary metabolite of perchloroethylene.
4 Moreover, TCA and DCA may be found in drinking water, with background exposures to TCA
5 and DCA each about 10-fold higher than background exposure to TCE (Wu and Schaum, 2000).
6
7 3.3.2.3. Age
8 Children and infants may be more susceptible to TCE's toxicity because they may receive
9 higher exposures via the mother either through systemic circulation during pregnancy or
10 mother's milk, as well as from pharmacokinetic differences between a developing child and
11 adult. Limited data collected from residents in U.S. urban areas have indicated that TCE is
12 present in human milk (ATSDR, 1997); nursing infants may receive a larger exposure to TCE
13 owing to its presence in mother's milk (Fisher et al., 1989). The fetus may also serve as a
14 reservoir for maternally derived metabolites; CYP2E1 has been reported to be expressed in full-
15 term placentas (Vieira et al., 1998). TCA has been reported to accumulate in mouse fetuses and
16 amniotic fluid (Danielsson, 1990; Ghantous et al., 1986). DCA also readily crosses the placenta
1 7 to the fetus (Smith et al., 1992).
18
19 3.4. TOXICITY
20 TCE affects many organs and systems of the body, consistent with its lipophilic nature
21 and ability to distribute widely throughout the body. This section summarizes effects reported
22 for exposure to TCE and its metabolites, presented by organ system. A discussion of
23 mechanisms and modes of action are discussed in Section 3.5 because they are, in several cases,
24 applicable to more than one system.
25
26 3.4.1. Neurotoxicity
27 The ability of TCE to cause neurotoxic effects is well established. In general, TCE
28 produces a "solvent narcosis" that may be related to effects on membrane fluidity and may
29 include anesthetic effects. TCE was formerly used as a general anesthetic and induces this effect
30 at about 2,000 ppm (ASTDR, 1997). The array of symptoms observed with TCE exposure is
31 similar to other solvents, such as alcohol, ethers, petroleum distillates, and other halogenated
32 solvents.
33 Acute or short-term inhalation of TCE has been associated with dizziness, headache,
34 sleepiness, nausea, confusion, blurred vision, and weakness in several human studies cited by
35 Barton and Das (1996). TCE is distributed to the brain, and brain concentrations of TCE in the
36 rat tend to mimic blood concentrations (Boyes et al., 2000). Acute neurologic effects of TCE in
8/1/01 3-9 DRAFT-DO NOT CITE OR QUOTE
-------
1 both humans and rats are more closely associated with peak concentrations rather than AUC
2 (Boyes et al., 2000), but rats require a higher peak concentration to elicit these effects (Bushnell,
3 1997). Hence, humans appear to be a sensitive species for the neurotoxic effects of TCE.
4 Longer term or chronic inhalation exposure is associated in a number of human studies
5 with a similar array of neurological symptoms as those observed with acute exposure (ATSDR,
6 1997). One study (Vandervort et al., 1973), used by Cal/EPA (1999) to derive a reference
7 exposure level (similar to an RfC), reported eye irritation, drowsiness, heart palpitations, cough,
8 weakness, and dizziness among workers exposed to a wide range of TCE concentrations. Other
9 studies of occupational exposure to TCE have reported similar LOAELs (see Table 4-3).
10 Tremors have been reported in several inhalation studies (Grandjean et al., 1955; Liu et al., 1988;
11 Bardodej and Vyskocil, 1956) and in one study of a population exposed to a mixture of TCE and
12 perchloroethylene in drinking water (White et al., 1997). Long-term inhalation exposure at an
13 average concentration of 44 ppm TCE has also been associated with early nerve function
14 impairment (Ruitj en et al., 1991).
15 It is not clear whether some of TCE's neurotoxic effects may be attributed to TCE as the
16 parent or to one of its metabolites such as DCA, which has shown longer term effects on the
1 7 central nervous system. A recent study (Moser et al., 1999) in two strains of rats exposed to
18 DCA for up to 2 years showed neuromuscular toxicity and mild tremors with neurotoxicity being
19 route, duration, and strain dependent. Drinking water exposure was shown to be more toxic than
20 gavage, with effects seen at the lowest drinking water dose (16 mg/kg-d). In addition,
21 neurotoxicity from DCA was described as more pronounced and persistent, and occurring at
22 lower doses than previously reported (Moser et al., 1999). Subchronic exposures to DCA in
23 drinking water are associated with brain lesions in dogs at doses of 12.5 mg/kg-d or higher
24 (Cicmanec et al., 1991) and rats at doses of 125 mg/kg-d or higher (Bhat et al., 1991; Katz et al.,
25 1981). At much higher doses, focal reactive gliosis—proliferation of reactive astrocytes after
26 dead neurons have been cleared away—and vacuolization were observed in the brains of DCA-
27 treated rats (Bhat et al., 1991). DCA given orally is also reported to decrease nerve conduction
28 velocity in tibial nerves in terms of areas, perimeters, and diameter in rats (Yount et al., 1982)
29 and, in a human case report, to induce a peripheral neuropathy (Moore et al., 1979).
30
31 3.4.2. Liver
32 3.4.2.1. GeneralHepatotoxicity
33 TCE can induce hepatotoxicity in humans, as reported in several studies cited by ATSDR
34 (1997). This is not unexpected, as most metabolism of TCE occurs in the liver. TCE can affect
35 liver functions not reflected in a standard battery of liver function tests. Workplace exposure to
36 TCE has been reported to change serum cholesterol and increase bile acids (Nagaya et al., 1993;
8/1/01 3-10 DRAFT-DO NOT CITE OR QUOTE
-------
1 Driscoll et al., 1992). Alterations of plasma bile acids have been noted in workers at inhalation
2 exposures of less than 5 ppm (Driscoll et al., 1992; Neghab et al., 1997). Elevations of serum
3 bile acids, possibly an early sign of liver dysfunction, have also been seen in exposed rats (Wang
4 and Stacey, 1990; Hamdan and Stacey, 1993). Alterations of cholesterol metabolism have been
5 observed in workers with low-level chronic TCE exposure, effects that persisted after 2 years
6 (Nagaya et al., 1993). TCE exposure also can alter insulin and endocrine profiles in
7 occupationally exposed individuals (mean, 30 ppm) (Chia et al., 1997; Goh et al., 1998), effects
8 that may be important to responses of the liver.
9 Increased liver weight, primarily resulting from cytomegaly, is one of the most frequently
10 reported effects in animals dosed chronically with TCE by either the inhalation or oral route.
11 Barton and Clewell (2000) summarize several studies in which orally administered TCE changed
12 the liver-weight-to-body-weight ratio in rats and mice. These effects have been observed at
13 gavage doses as low as 50 mg/kg-d for 14 days (Berman et al., 1995). TCE was reported to
14 increase cell replication but not cytotoxicity or reparative hyperplasia. TCE metabolites can also
15 stimulate cell replication; TCA, DCA, and CH have all been associated with increased liver size
16 (Bull, 2000).
17 Although the state-of-the-science papers have focused on the effects of TCE on
18 parenchymal cells of the liver, nonparenchymal cells may also have a role in TCE hepatotoxicity.
19 TCE and DCA have been reported to alter Kupffer cell morphology, indicating activation, in
20 mice and dogs, respectively (Kjellstrand et al., 1983; Katz et al., 1981). Kupffer cells32 contain
21 CYP2E1, which is highly inducible, and have been shown to be critical in the toxicity of ethanol,
22 acetaminophen, phenobarbital, galactosamine, and carbon tetrachloride (Laskin et al., 1988,
23 1986, 1995; Knecht et al., 1995; Laskin, 1996; Lieber, 1997).
24
25 3.4.2.2. Liver Cancer
26 The review of the TCE cancer epidemiology finds an overall excess incidence of liver
27 cancer in the tier-I studies (RR=1.9, 95% CI=1.0-3.4, N=12) and the tier-II studies (RR=2.0,
28 95% CI=1.3-3.3, N=15), with null results in the tier-IE studies (Wartenberg et al., 2000).
29 Although none of the tier-I studies achieves statistical significance individually, the relative risks
30 exceed 1.0 in all three incidence studies, and in the only study that reported mortality for this
32Kupffer cells, which remove paniculate material from the blood and drain the digestive system before it enters
systemic circulation, can release potent mediators such as eicosinoids, toxic cytokines, proteolytic enzymes,
superoxide anion, and tumor necrosis factor and be involved in hepatotoxicity (Caldwell-Kenkel et al., 1991).
Biologically active mediators released by activated macrophages have been implicated in tumor promotion, as well
as regulation of CYP450-mediated drug biotransformation (Laskin et al., 1988). Kupffer cells may also have
implications for increased toxicity or increased likelihood of tumor progression if their role in immune and endocrine
function is altered.
8/1/01 3-11 DRAFT-DO NOT CITE OR QUOTE
-------
1 specific site. There is a biological gradient seen in the Anttila study (Anttila et al., 1995), as liver
2 cancer incidence increased with increasing exposure and with increasing time since first
3 exposure. Among those with the longest time since first exposure, relative risks are large
4 (RR=6.1, 95% CI=2.8-17.7). A more recent study (Ritz, 1999) supports the observations in the
5 Anttila study. Although the study was based on few deaths, risks increased with increasing
6 potential TCE exposure and with increasing time since first exposure. Case-control studies are
7 supportive, generally, of statistically significant associations with organic solvents as a class.
8 These findings are supported by the well-established and widely acknowledged result that
9 TCE causes liver tumors in male and female mice (see Table 3-1). The NTP conducted two
10 gavage studies in B6C3F1 mice (NCI, 1976; NTP, 1990), and Maltoni conducted inhalation
11 studies in Swiss mice and B6C3F1 mice (Maltoni et al., 1986, 1988). In each study, TCE
12 exposure caused a statistically significant increased incidence of liver tumors in male and female
13 mice. Neither series of studies, however, observed liver tumors in rats.
14 Previously, it had been suggested that the liver tumors might be caused by
15 epichlorohydrin, 1,2-epoxybutane, or other stabilizers present with the TCE (Henschler et al.,
16 1980). Subsequent testing, however, has ruled this out (Henschler et al., 1984). The state-of-the-
1 7 science papers (Lash et al., 2000a; Bull, 2000) suggest that TCE induces liver tumors through its
18 metabolites. TCA and DCA, as well as CH, have all been shown to cause liver tumors in mice
19 (Bull, 2000), with DCA also causing liver tumors in rats (DeAngelo et al., 1996).
20
21 3.4.3. Kidney
22 3.4.3.1. GeneralNephrotoxicity
23 A range of kidney toxicity following human exposure to TCE has been reported in several
24 studies cited by ATSDR (1997) and Lash et al. (2000b). Tubular damage, as assessed by altered
25 excretion of urinary proteins, has been noted in occupationally exposed individuals (ATSDR,
26 1997), as well as in renal cell carcinoma patients with high-level occupational exposure to
27 trichloroethylene (Briining et al., 1999a,b). TCE causes dose-related nephrotoxicity in male and
28 female rats and mice (NTP, 1988, 1990; Maltoni et al., 1986, 1988). The lesions are not the
29 chronic interstitial nephrosis usually observed in aged rats, but consist of cytomegaly,
30 kayomegaly, and toxic nephrosis of tubular epithelial cells in the inner renal cortex (Lash et al.,
31 2000b). As in the liver, TCE can change the kidney-weight-to-body-weight ratio in rats, mice,
32 and gerbils (Barton and Clewell, 2000; Maltoni et al., 1986, 1988; Kjellstrand et al., 1981, 1983).
33
34 3.4.3.2. Kidney Cancer
35 Consistency across epidemiological studies is strongest for an association between TCE
36 exposure and kidney cancer (Wartenberg et al., 2000). There is an overall excess of kidney
8/1/01 3-12 DRAFT-DO NOT CITE OR QUOTE
-------
1 cancer incidence in the tier-I studies (RR=1.7, 95% CI=1.1-2.7, N=21) and tier-H studies
2 (RR=3.7, 95% CI=1.7-8.1, N=6). Mortality was also elevated in the tier-IE studies (RR=2.3,
3 95% CI=1.5-3.50, N=20), but not incidence. The relative risk exceeds 1.0 in three of four
4 studies in tier 1 that reported incidence, one of which was a follow-up cluster report and achieved
5 statistical significance individually. It is not obvious that index cases were excluded from these
6 studies. Supporting the tier-I findings are the tier-II studies and case-control studies of kidney
7 cancer and TCE (Vamvakas et al., 1998; Dosemeci et al., 1999), degreasers (Schlehofer et al.,
8 1995), and solvents (Mellemgaard et al., 1994). Dosemeci (Dosemeci et al., 1999) and Blair
9 (Blair et al., 1998) observed differences in risk between men and women, with higher risk in
10 women. Dosemeci (Dosemeci et al., 1999) puts forth several hypotheses, including a possible
11 gender-related difference in the distribution of polymorphisms of the GST metabolic pathway or
12 differences in level of exposure. The report of Briining (Briining et al., 1997a) showed an
13 increased risk for kidney cancer among individuals with GST isozymes Ml and Tl, suggesting
14 the GST pathway and polymorphisms of this pathway as important contributors to kidney cancer
15 risk (Lash et al., 2000b).
16 Recently, occupational exposure to TCE has been linked to somatic mutations of the
1 7 VHL gene33 (Briining, 1997b; Brauch et al., 1999). Thirty-three of the 44 renal cell carcinomas
18 had one or more mutations in the VHL gene, predominately characterized as changes in cytosine
19 to thymidine (Brauch et al., 1999). An interesting observation was a mutation in cytosine at a
20 specific point in the VHL gene, at nucleotide (nt) 454, in 13 of these tumors; none of the 107
21 renal cell carcinomas from individuals not exposed to TCE displayed this mutation. High levels
22 of exposure were associated with a higher frequency of tumors with multiple mutations and with
23 the mutation at nucleotide 454 than lower exposure. This mutation was, moreover, found in
24 normal kidney tumor tissue that was adjacent to neoplastic tissue in four patients. This
25 observation suggests that a mutation at nucleotide 454 may precede tumor development. Both
26 the studies of Briining (1997b) and Brauch et al. (1999) point to an association of increased
27 occurrence of mutated sequence in the entire gene for TCE-exposed renal cancer patients, with
28 the work of Brauch indicating multiple mutations and a specificity in where and how mutations
29 occurred within the nucleotide sequence of this gene.
30 Another study (Schraml, 1999) assessed the entire genome of 12 renal cell tumors from
31 individuals with suspected exposure to TCE and other organic solvents (exposure was not
32 described in this report, nor was an exposure score assigned to individual cases) and observed
33The VHL gene, consisting of hundreds of nucleotides, normally suppresses renal cell carcinoma. Somatic
alterations to the gene are noted in sporadic kidney cancers, and mutation of the VHL gene is considered a risk factor
for kidney cancer (Lash et al., 2000b).
8/1/01 3-13 DRAFT-DO NOT CITE OR QUOTE
-------
1 mutations to genes on several chromosomes besides chromosome 3, the location of the VHL
2 gene. Mutations found on these other chromosomes included both DNA losses and DNA gains.
3 There were no differences between TCE and non-TCE exposed renal cell tumors regarding
4 which chromosomes were affected. Schraml (1999), however, observed a much lower frequency
5 of mutation of the VHL gene (25%, 4 tumors) among the 12 tumors from TCE-exposed
6 individuals compared to either the report of Brauch et al. (1999) or to background rates (around
7 60% of all renal cell carcinomas) (Brauch et al., 1999). This study provides only limited
8 information because of its absence of complete information on exposure and on tumor stage or
9 grade, its smaller sample size, and its lower than expected number of tumors with VHL
10 mutations.
11 The human results are supported by the animal studies, where TCE causes kidney cancer
12 in rats by both ingestion and inhalation exposure (see Table 3-2). Maltoni observed a
13 significantly increased incidence of kidney adenocarcinomas in male Sprague-Dawley rats
14 inhaling TCE (Maltoni et al., 1986, 1988). The NTP conducted parallel corn-oil gavage studies
15 in ACT, August, Marshall, and Osborne-Mendel rats (NTP, 1988), and a similar study in F344/N
16 rats (NTP, 1990). The NTP considered the former to be inadequate studies of carcinogenic
17 activity because of chemically induced toxicity, reduced survival, and deficiencies in the conduct
18 of the studies. Despite these limitations, however, the NTP concluded that there were tubular
19 cell neoplasms of the kidney in rats exposed to TCE. Consistent with these results, an earlier
20 NCI study also reported a tubular adenocarcinoma in one male Osborne-Mendel rat exposed to
21 TCE for 78 weeks (NCI, 1976).
22 As noted by the NTP (1988), the survival of exposed rats was substantially reduced
23 compared with controls. Consequently, the tumor incidences tabulated in the NTP report may be
24 misleading, because animals dying early may not be at full risk of developing tumors. To
25 ascertain the effect of this reduced survival on the tumor incidences, the individual animal tumor
26 pathology in the NTP report was used to adjust the tumor incidences to reflect the number of
27 animals at risk. Animals were discounted if they died before the first kidney tumor was
28 observed, at 57 weeks. Counting only animals alive at 57 weeks leads to the corrected incidences
29 reported in Table 3-2.
30 Rhomberg (2000) analyzed the findings of TCE bioassays in rats and suggests that
31 females are as sensitive as males to TCE-induced renal carcinogenicity. Previously, the kidney
32 tumors in male rats had received more attention than those in females (ATSDR, 1997).
33 Consideration of all strains together in Table 3-2, however, reveals that there is virtually no
34 difference between males and females in the overall incidence of these tumors.
35
8/1/01 3-14 DRAFT-DO NOT CITE OR QUOTE
-------
1 3.4.4. Immune and Lympho-Hematopoietic Systems
2 3.4.4.1. General Effects on the Immune and Lympho-Hematopoietic System
3 Immune system34 changes have been noted in individuals presumed to have been exposed
4 to TCE and other chlorinated solvents in drinking water (ATSDR, 1997). The ability of TCE to
5 alter immune responses has been observed in mice exposed to TCE via drinking water or
6 inhalation (Khan et al., 1995; Aranyi et al., 1986; Hobara et al., 1984; Parks et al., 1993; Sanders
7 et al., 1982; Tucker et al., 1982). Mice exposed to TCE in drinking water at 0.1 mg/mL and
8 higher showed inhibition of humoral and cell-mediated immunity, as well as effects on
9 macrophage function and monocyte-granulocyte progenitor cells (Sanders et al., 1982). Females
10 showed a greater susceptibility than males to these immunotoxic effects of TCE (Sanders et al.,
11 1982; Barton and Clewell, 2000).
12 TCE and one of its metabolites have also been shown to affect bone marrow function or
13 components. Sanders et al. (1982) observed an inhibition of stem cell colonization in both male
14 and female mice. Lock et al. (1996) reported that DCVC causes renal toxicity and fatal aplastic
15 anemia in calves at a single dose of 4 mg/kg, a dose much larger than that expected from TCE
16 metabolism, and that toxicity may be related to the • *lyase metabolic pathway (Anderson and
1 7 Schultze, 1965; Bhattacharya and Schultze, 1967). However, aplastic anemia has not been
18 reported from similar exposures in rat, guinea pig, dog, or cat (Lock et al., 1996).
19 TCE exposure may also accelerate an autoimmune response. Several studies of
20 occupational exposure have associated TCE and organic solvent exposure with several
21 autoimmune diseases:35 systemic sclerosis (scleroderma), fascitis, and systemic lupus
22 erythematosus (SLE) (Lockey et al., 1987; Saihian et al., 1978; Niertert et al., 1998; Bovenzi et
23 al., 1995; Goldman, 1996; Schaeverbeke et al., 1995; Waller et al., 1994). Furthermore, two
24 studies of populations exposed to TCE and other solvents in drinking water report changes in T-
25 lymphocytes and increased autoantibodies (Byers et al., 1988; Kilburn and Warshaw, 1992).
26 Autoimmune-disease-prone mice exposed to 2.5 or 5.0 mg/mL TCE in drinking water show an
27 increase in autoimmune antinuclear antibodies, elevated serum levels of several
28 immunoglobulins, activation of CD+4, and a cytokine profile consistent with a TH1 type
34Immune cells are derived from stem cells in the bone marrow or fetal liver and are widely distributed throughout
the body (Selgrade, 1995). They are also controlled by cytokines. The immune system is highly conserved across
species such that organs and cells of the immune system in humans, mice, and rats are similar. In instances where
controlled human studies have been possible, results of immune function studies in mice have been accurate
predictors of effects in humans.
35Autoimmunity can be defined as a loss of self-tolerance that results in an immune reaction against oneself or self-
antigens. Autoimmunity appears to have a multifactorial etiology that includes genetic and environmental factors
(chemicals and microbes) (D'Cruz, 2000).
8/1/01 3-15 DRAFT-DO NOT CITE OR QUOTE
-------
1 response36 (Gilbert, 1999; Griffin et al., 2000a,b). In another recent drinking water study (Griffin
2 et al., 2000c), these investigators noted many of these same observations, but these effects
3 occurred at much lower exposures to TCE than previously reported; doses in this latter study
4 were 0.1 mg/mL, 0.5 mg/mL, and 2.5 mg/mL. An early response to TCE exposure was a dose-
5 related and significant increase in serum antinuclear antibodies at the lowest exposure (0.1
6 mg/mL, or 4 mg/kg/d). Furthermore, histopathologic changes that included portal infiltration by
7 mononuclear cells were seen at the termination of the study, 32 weeks, and were consistent with
8 the induction of autoimmune disease in the liver (Griffin et al., 2000c). Additionally, hepatocyte
9 reactive changes were observed at all exposures including the lowest tested dose (0.1 mg/mL, or
10 4 mg/kg/d).
11
12 3.4.4.2. Lymphoid Cancer
13 Overall, an increased incidence of lympho-hematopoietic cancer (RR=1.4, 95%
14 CI=1.0-2.0, N=40) was noted in tier-I by Wartenberg et al. (2000), primarily a reflection of the
15 elevated, but not statistically significant, incidence of non-Hodgkin's lymphoma (RR=1.5, 95%
16 CI=0.9-2.3, N=22). Null results were seen in the tier-II and tier-Hi studies (Wartenberg et al.,
1 7 2000). Two drinking water studies of ecologic design provide some evidence of an association
18 between non-Hodgkin's lymphoma and drinking water that includes TCE and other chemically
19 similar solvents (Fagliano et al., 1990; Vartianen et al., 1993). Both studies were of ecological
20 design without individual exposure or residential histories. The case-control studies that
21 examined occupational exposures were not informative, primarily because of few cases and,
22 consequently, lower statistical power. However, case-control studies that evaluated lymphoid
23 cancer (leukemia and non-Hodgkin's lymphoma) and exposure to drinking water contaminated
24 with TCE report elevated risks, although not all risks were statistically significant (Cohn et al.,
25 1994; MA-DOH, 1997). One study of childhood leukemia (MA-DOH, 1997) observed a very
26 strong association with exposure during pregnancy (OR=13.2, 95% CI=0.9-205.2). For both
27 leukemia and non-Hodgkin's lymphoma, these studies report exposure-response gradients
28 supporting drinking water contamination as the etiologic agent (Cohn et al., 1994; MA-DOH,
29 1997). Exposures are not sufficiently specific to a single chemical, although TCE is often the
30 chemical found in highest concentration and the others have similar metabolism and targets.
31 TCE induces malignant lymphomas in mice in a number of inhalation or gavage
32 bioassays. Mice have been shown to be an excellent immunopathologic model for human
36The profile of cytokines secreted by activated CD4+ T cells is often measured as an indicator of the type of
immune response induced by a toxicant. A T-helper type 1 (TH1) type response reflects macrophage activation and
inflammation and is marked by an increased secretion of IFN-» "with a simultaneous decrease in the secretion of the
cytokine IL-4 (Griffin et al., 2000a).
8/1/01 3-16 DRAFT-DO NOT CITE OR QUOTE
-------
1 lymphoid cancers (Pattengale, 1986, 1994). Henschler found a statistically significant increased
2 incidence of malignant lymphomas in female Han:NMRI mice exposed by inhalation (Henschler
3 et al., 1980). The NTP found a significant increase in malignant lymphomas in female B6C3F1
4 mice exposed by gavage (/X0.05 by life-table test), although this result was not mentioned in the
5 abstract or summary of peer review comments (NTP, 1990). Consistent results were found in the
6 NCI study (NCI, 1976). Lymphosarcomas and malignant lymphomas were just beginning to
7 appear in exposed mice when the NCI study was terminated at 90 weeks, although the numbers
8 were too few for statistical significance. For comparison, in the 104-week NTP study only 1 of
9 the 13 malignant lymphomas in exposed female mice was observed before week 90, at week 85.
10
11 3.4.5. Developmental and Reproductive Toxicity, Endocrine Effects
12 3.4.5.1. Developmental Effects
13 Epidemiologic studies of women exposed to degreasing solvents, including TCE, have
14 reported elevated risks for cardiac anomalies in their offspring (Wilson et al., 1998; Goldberg et
15 al., 1990; Ferencz et al., 1997). Large, statistically significant excesses were observed for
16 specific cardiac defects: left-sided obstructive defects (OR=6.0, 95% CI=1.7-21.3) and
1 7 hypoplastic left heart (OR=3.4, 95% CI= 1.6-6.9) with an attributable risk37 of 4.6% (Wilson et
18 al., 1998; Ferencz et al., 1997). Neural tube defects have also been noted with either
19 occupational or drinking water exposure to solvents including TCE (Bove et al., 1992;
20 Holmberg, 1980a,b, 1982).
21 TCE or its metabolites, TCA and DC A, cause cardiac malformations and eye
22 malformations (anopthalmia, micropthalmia) in rat pups38 (Dawson et al., 1993; Smith et al.,
23 1989, 1992; Johnson et al., 1998a,b; Epstein et al., 1993) given drinking water exposure;
24 increases for both malformations were dose related (Dawson et al., 1993; Smith et al., 1989,
25 1992). These malformations have not been reported in inhalation studies, though there is a
26 question as to whether the ability of the methods or the number of pups in the studies were
27 adequate to detect these malformations (Barton and Clewell, 2000). The study of Dawson et al.
28 (1993) raises issues for quantitative analysis because of the wide spacing between low and high
29 dose (exposure to TCE was at 1.5 ppm and 1,100 ppm) and whether TCE or one of its
30 metabolites was responsible for the effect reported (Barton and Clewell, 2000).
^Attributable risk is the risk or rate difference that may be attributable to the exposure (Rothman, 1986).
38Although no battery of animal tests can provide complete assurance in predicting human developmental effects, the
similarity in timing and sequencing of events during embryogenesis and organogenesis, particularly cardiogenesis,
between humans and rats suggests that the rat is a suitable nonprimate choice for studying developmental effects.
8/1/01 3-17 DRAFT-DO NOT CITE OR QUOTE
-------
1 Graeter et al. (2000) reported that gavage exposure to TCE, DC A, or TCA did not
2 produce a statistically significant difference in cardiac malformations in an abstract lacking
3 information on number of animals examined and exposure levels. DCA exposure, however,
4 produced a statistically significantly increased number of external malformations, and both DCA
5 and TCA individually produced lower fetal body weights in Sprague-Dawley rats. TCA also
6 produced a statistically significant lower mean uterine weight in dams compared with vehicle
7 controls.
8 Recently Boyer (2000) reported TCE exposure in vitro produces a dose-dependent
9 inhibition of mesenchymal cell transformation (a critical event in development of the heart) in
10 progenitors of the valves and septa of the heart, and briefly mentions the findings of a concurrent
11 study that showed altered gene expression in embryo hearts with maternal exposure to drinking
12 water with 110 ppm TCE. The concentration in the in vitro study was in the millimolar (mM)
13 range, comparable to doses of TCE studied in vivo and, because of factors inherent with using an
14 in vitro paradigm, likely to be much less than those administered to the culture medium (Runyan,
15 2000). However, although direct quantitative extrapolation of in vitro doses to those studied in
16 vivo is problematic, these observations support a TCE effect on cardiac development. Moreover,
17 the finding of cardiac anomalies in rat pups exposed to two TCE metabolites, TCA and DCA,
18 supports a hazard for this effect with TCE (Smith et al., 1989, 1992; Johnson et al., 1998a,b;
19 Epstein et al., 1993).
20 Both TCA and DCA induce other effects besides cardiac and eye anomalies in offspring.
21 Both TCA and DCA exposure were associated with a substantial increase in implantation and
22 resorption sites39 and decreased fetal weight (Smith et al., 1989, 1992). These effects were
23 observed in the absence of maternal toxicity.
24 Exposures to TCE during the prenatal period have been reported to induce
25 neurobehavioral alterations in rat pups, including changes in long-term exploratory and
26 locomotor behavior, altered glucose uptake and metabolism in the neonatal brain, and decreases
27 in myelin (Taylor et al., 1985; Isaacson et al., 1990). In addition, changes in adult behavior are
28 suggested with perinatal exposure to 50 mg/kg TCE (Fredricksson et al., 1993). Statistical
29 analysis of the data used individual pups, possibly resulting in an overestimate of the statistical
30 significance of observed results.
31 Exposure to TCE can potentially modify responses to other chemical exposures, which is
32 an important consideration because TCE environmental exposure is usually not isolated but also
33 concurrent with other chemicals. TCE and DEHP (a peroxisome proliferator) together were
39Implantation sites are sites of initial attachment of a fertilized ovum, but with no further growth. Resorption sites
are sites of implantation with embryonic growth, but arrest and necrosis during development.
8/1/01 3-18 DRAFT-DO NOT CITE OR QUOTE
-------
1 reported to have synergistic effects with regard to prenatal loss, pup weight changes, and
2 anopthalmia (Narotsky et al., 1995a,b).
3
4 3.4.5.2. Endocrine System Effects and Reproductive Toxicity
5 Reports of TCE-induced effects on human sex hormones and their binding proteins are
6 just beginning to emerge in the published literature. For the most part, workplace exposure to
7 TCE has not been reported to change hormones, spermatogenesis, or semen quality in men (Goh
8 et al., 1998). Recently, however, low-level subchronic TCE exposure (mean 30 ppm) was
9 associated with changes in insulin and steroid level, decreased serum testosterone, and decreased
10 sex hormone binding globulin (SHBG40) in male workers (Goh et al., 1998). Exposure gradients
11 were observed for several parameters. In analyses adjusted for age, serum testosterone and
12 SHBG were observed to decrease whereas free androgen levels increased with increasing
13 duration of TCE exposure.41 Moreover, dehydroepiandrosterone (DHEA)42 increased with years
14 of TCE exposure (Chia et al., 1997), contradicting expectations as DHEA levels normally
15 decrease with age (Metzger et al., 1997; Chiu et al., 1999). A more complicated picture was
16 observed for insulin. Insulin levels were elevated for exposures of less than 2 years, but
17 decreased to below-normal levels for increasing durations of exposure. The normal
18 SHBG/insulin relationship appeared to be disturbed with longer durations of exposure (Goh et
19 al., 1998).
20 TCE exposure has been shown to affect the reproductive system of males; studies in
21 females are few. In another report of the same population of workers studied above, increase in
22 urinary levels of TCA was correlated with hyperzoospermia, which has been implicated in
23 infertility (Chia et al., 1996). A lower percentage of normal sperm, compared to referent values
24 identified by the World Health Organization, was also observed in this population (Chia et al.,
25 1996). TCE has been shown to accumulate in the male reproductive tract of rats and to decrease
26 testosterone levels (Zenick et al., 1984; Kumar et al., 2000). Kumar (2000) recently reported
27 significant decreases in total epididymal sperm count and motility, with the magnitude of loss
28 correlated with increased duration of exposure. A concurrent increase in total testicular
29 cholesterol and decreases in enzyme activities, which are rate limiting in the biosynthesis of
30 testosterone, was also reported. Of note is the use of young rats (starting weights of 50 g) in this
40SHBG has high affinity but low capacity for testosterone and estradiol 17p, and its level may be related to liver
toxicity, as it is produced in the liver.
41Free androgen is calculated by the division of testosterone levels by SHBG and is that portion free to be
biologically active. A decreased index was achieved by differential rates of reduction of SHBG and testosterone.
42DHEA is an adrenal steroid associated with insulin resistance, androgenic effects, peroxisome proliferation, and
induction of CYP4A (Mayer, 1998, 1999).
8/1/01 3-19 DRAFT-DO NOT CITE OR QUOTE
-------
1 subchronic (12 and 24 weeks) study. This report describes results from a preliminary study that
2 indicated TCE-induced decreases in fertility as well. DCA, a metabolite of TCE, is also a
3 testicular toxin to rats and to dogs at exposures as low as 12.5 mg/kg-d (Katz et al., 1981,
4 Cicmanec et al., 1991).
5 One report exists of increased menstrual disorders in female workers exposed to TCE
6 (Zielinski, 1973).
7
8 3.4.5.3. Cervical, Prostate, and Testicular Cancer
9 The human studies provide some limited evidence for increased cancer risk for certain
10 reproductive organs. Cervical cancer, though sparsely reported, is elevated in tier-I incidence
11 studies (RR=2.4, 95% CI=1.2-4.8, N=8) and in tier-HI (dry cleaner) mortality studies (RR=1.7,
12 95% CI=1.5-2.0, N=43) (Wartenberg et al., 2000). The incidence study by Anttila et al. (1995)
13 provided the greatest weight for an association and reported a statistically significant elevation of
14 cervical cancer among those with greatest urinary TCA levels. The comparison with internal
15 controls in Blair et al. (1998) also is suggestive that TCE may be the etiologic agent, and not
16 some other confounder such as socioeconomic class. Prostate cancer incidence and mortality are
1 7 also elevated in tier-I studies (RR=1.3, 95% CI=1.0-1.6, N=95; RR=1.2, 95% CI=1.0-1.4,
18 N=131).
19 In an NTP study (NTP, 1988), interstitial cell neoplasms of the testis were reported in
20 Marshall rats exposed to TCE by gavage. Increased incidences are apparent in the other rat
21 strains when counting the animals alive when the first testicular tumor was observed. In another
22 study, (Maltoni et al., 1986, 1988) reported significantly increased incidences of Leydig cell
23 tumors in male rats exposed to TCE by inhalation. Although testicular cancer is often thought of
24 as occurring spontaneously in some strains of rats late in life, the testicular cancers reported for
25 TCE were shown in multiple strains with lower background rates and at earlier times of
26 appearance (see Table 3-4).
27 DCA at levels of 50 mg/kg-d and higher for 13 weeks produced prostate gland atrophy
28 and testicular changes that were considered dose dependent (Katz et al., 1981).
29
30 3.4.5.4. Pancreatic Cancer
31 Pancreatic cancer incidence mortality is statistically significantly elevated in the dry
32 cleaner (tier-Ill) studies (RR=1.7, 95% CI=1.2-2.6, N=22; RR=1.3, 95% CI=1.0-1.7, N=42),
33 with mixed results in tiers I and n (Wartenberg et al., 2000). Disturbances in insulin levels as
34 noted by Goh et al. (1998) and Kato-Weinstein et al. (1998) and effects on carbohydrate handling
35 may have a role in the induction of effects on the pancreas by TCE in humans. For example,
36 diabetics have an increased risk of pancreatic cancer (Calle, 1998). In experimental animals,
8/1/01 3-20 DRAFT-DO NOT CITE OR QUOTE
-------
1 cancer in this organ has not been well studied in the carcinogenic assays carried out to date. Katz
2 et al. (1981) reported that DCA causes pancreatitis in dogs.
3
4 3.4.6. Lung
5 3.4.6.1. Pulmonary Effects
6 There are few human data reporting pulmonary toxicity for TCE (ATSDR, 1997; Barton
7 and Clewell, 2000). Mice appear to be more sensitive than rats to histopathological changes in
8 the lung via inhalation; pulmonary effects are also seen in rats with gavage exposure. Lungs of
9 mice exposed via inhalation to 500 ppm TCE for 30 minutes displayed vacuole formation and
10 endoplasmic reticulum dilation, specifically in Clara cells of the bronchial tree (Villaschi, 1991).
11 Additionally, a reduction in pulmonary CYP450 enzyme activity was observed after a 6-hour
12 exposure to 100 ppm TCE (Odum et al., 1992). TCE administered to rats via gavage at doses
13 1,000 mg/kg-d and higher caused rales and dyspnea (Narotsky et al., 1995) and pulmonary
14 vasculitis (NTP, 1990).
15 At relatively high doses (drinking water exposures of 80.5 mM DCA or 45.8 mM TCA),
16 lung inflammation (perivascular) is seen in rats exposed to both these TCE metabolites, but not
17 in controls (Bhat et al., 1991). At lower doses, a secondary effect of DCA exposure in dogs has
18 been reported to be pulmonary inflammation (Katz et al., 1981).
19
20 3.4.6.2. Lung Cancer
21 The epidemiologic studies are not generally supportive of lung cancer; lung cancer
22 mortality was statistically significantly elevated only in dry cleaners (tier-in) (RR=1.3, 95% CI:
23 1.1-1.5, N=137) (Wartenberg et al., 2000). TCE has been reported to significantly increase the
24 incidence of lung tumors in male Swiss mice and female B6C3F1 mice by inhalation (see Table
25 3-5) (Maltoni et al., 1986, 1988). Another study extended this result to female ICR mice (males
26 were not tested) (Fukuda et al., 1983). These studies did not observe lung tumors in rats.
27 Ingested TCE has not been observed to cause lung tumors in either rats or mice. This is the only
28 tumor site reported to have a positive carcinogenic response by one route but not another.
29
8/1/01 3-21 DRAFT-DO NOT CITE OR QUOTE
-------
1 3.4.7. Genetic Toxicity
2 TCE and a number of its metabolites (CH, DCA, and TCA) have been evaluated in a
3 large number of genotoxic assays.43 On the basis of the available data, definitive conclusions
4 cannot be drawn as to whether TCE is or is not a genotoxic carcinogen. Taken together, TCE
5 and its metabolites CH, DCA, and TCA are, at most, very poor genotoxicants. There is evidence
6 that the metabolite DCVC is more potent as a potential genotoxic agent. Evidence from a
7 number of different analyses and a number of different laboratories using a fairly complete array
8 of genetic endpoints indicates that TCE may have some potential to be genotoxic. However, the
9 evidence is not convincing that it is genotoxic, and is confounded by the fact that TCE is often
10 stabilized with a very low concentration of epichlorohydrin or 1,2-epoxybutane, both known to
11 be potent mutagens. A series of carefully controlled studies evaluating TCE (without mutagenic
12 stabilizers) found TCE incapable of inducing point mutations in standard Salmonella assays.
13 Furthermore, in vitro gross chromosome aberration assays were negative. In vivo, there is some
14 evidence that TCE or its metabolites bind to DNA and can induce single-strand DNA breaks in
15 both hepatic and kidney cells. However, the dose required to cause these DNA breaks was very
16 high (4-10 mM of TCE/kg body weight) and the response was very low. More recent data show
1 7 TCE to affect DNA methylation in whole-liver preparations at lower exposures with increased
18 expression of the two proto-oncogenes c-jun and c-myc (Tao et al., 1999). Moore and
19 Harrington-Brock's review of the potential genotoxi city of TCE (Moore and Harrington-Brock,
20 2000) notes that TCE is negative in a number of standard tests, and furthermore, in those tests
21 where positive responses were observed, the dose was very high and response low. Accordingly,
22 although it is not possible to totally eliminate the possibility that TCE would be a genotoxic
23 carcinogen, it is clear from these evaluations that TCE is not a potent genotoxicant (Moore and
24 Harrington-Brock, 2000).
25 CH has been extensively studied as a potentially genotoxic agent. It has been evaluated
26 in the recommended battery44 and several other assays, including genetic alterations in rodent
27 germ cells. CH is weakly positive at very high concentrations in in vitro genotoxicity assays that
28 detect point mutations, chromosomal mutations, and/or aneuploidy. There is a mixture of
29 positive and negative in vivo data, and because there is no reason to weigh any of the studies
43Genetic toxicity can be defined using mutational endpoints, cytogenetic analysis, and evaluation of primary DNA
damage with mutagenicity described as the ability to induce heritable mutations (Moore and Harrington-Brock,
2000). Primary DNA damage may occur via adduct formation, strand or chromosomal breakage, and unscheduled
DNA synthesis. Mutational damage consists of point mutations, chromosomal rearrangement, deletions, loss or gain
of chromosomes (aneuploidy), or changes in whole chromosome complements (polyploidy) (Moore and
Harrington-Brock, 2000).
44The core battery of recommended tests for genetic toxicity include the Salmonella assay, in vitro mouse lymphoma
mutation assay, and in vivo cytogenetic assay (Moore and Harrington-Brock, 2000).
8/1/01 3-22 DRAFT-DO NOT CITE OR QUOTE
-------
1 more than the others, it is not clear whether CH has the potential to be genotoxic in vivo (Moore
2 and Harrington-Brock, 2000).
3 DCA and TCA have also been extensively studied using a variety of genotoxicity assays,
4 including the recommended battery. DCA is genotoxic at very high concentrations in the
5 Salmonella assay, the in vitro mouse lymphoma assay, and in vivo cytogenetic and gene mutation
6 assays. DCA can cause DNA strand breaks in mouse and rat liver cells following in vivo
7 administration by gavage. As with TCE and CH, the concentration of DCA required to induce
8 damage is very high and the level of response is generally very low. Therefore, although one
9 cannot eliminate the possibility that DCA might induce tumors via a mutagenic mode of action,
10 the weight of the evidence argues that TCE-induced tumors would not be mediated by DC A-
11 induced mutation. TCA is the least mutagenic of the TCE metabolites discussed so far. It is
12 negative in the Salmonella assay, and although positive in the mouse lymphoma assay, it is
13 substantially less potent than DCA or CH. It is unclear whether TCA can induce chromosomal
14 damage in vivo; some assays are positive and some negative. TCOH is negative in the
15 Salmonella assay but has not been evaluated in the other recommended screening assays.
16 Therefore, it is unclear whether TCOH is genotoxic (Moore and Harrington-Brock, 2000).
17 Cysteine intermediates (DCVC and DCVG) formed by the GST pathway are capable of
18 inducing point mutations, as evidenced by the fact that they are positive in the Salmonella assay.
19 DCVC is the most potent of the TCE metabolites as a Salmonella mutagen; DCVG appears to be
20 similar in potency to DCA. Although DCVC and DCVG have not been evaluated in the other
21 EPA-recommended screening assays, there is some indication that they can induce primary DNA
22 damage in mammalian cells in vitro and in vivo. Long-term DCVC exposure can induce de-
23 differentiation of LLC-PK1 cell clones, supporting evidence of genetic toxicity (Vamavakas et
24 al., 1996). DCVC has also been shown to induce the expression to two proto-oncogenes, c-fos
25 and c-myc (Lash et al., 2000b). Moreover, chlorothioketenes, which are the intermediates formed
26 from the further metabolism of DCVC by • *lyase, have been shown in vitro to react with DNA
27 (Miiller et al., 1998a,b) and, in an aqueous solution, to bind with cytosine (Volkel et al., 1998).
28
29 3.5. MODES OF ACTION OF TCE TOXICITY
30 The previous section identifies TCE and its metabolites as eliciting toxicity in a number
31 of organs in both humans and laboratory animals. Possible modes of action for these observed
32 effects are discussed in this section, which focuses primarily on events occurring in the rodent so
33 as to put into perspective observations in chronic toxicity assays. A discussion of the human
34 relevance of these possible modes of action is also presented. Much of the discussion focuses on
35 events that may give rise to rodent liver, kidney, and lung tumors. The epidemiologic analysis,
36 however, suggests other organ systems may also be targets. Moreover, noncancer toxicity is
8/1/01 3-23 DRAFT-DO NOT CITE OR QUOTE
-------
1 observed and may arise through a common mode of action with the carcinogenic effects. Mode-
2 of-action hypotheses for these observations are presented last.
3 An understanding of how TCE may elicit toxicity in a number of organ systems can arise
4 through the identification of very specific events such as changes in gene regulation and cell
5 signaling, or through a discussion of more generalized events such as common changes to
6 physiological processes. Data supporting mode-of-action hypotheses are not uniform across
7 potential targets of toxicity. Identifying mode(s) of action for TCE toxicity entails obtaining
8 answers to several key questions concerning the relative amount of metabolites formed between
9 species, the ability to measure active metabolites in biological samples (e.g., DCA), the
10 determination of which metabolites may impart adverse effects, the sequence of biological
11 changes that can lead to tumor development, and the extrapolation of these events from rodents
12 to humans and from higher level exposure in the bioassay to lower environmental exposure.
13 The carcinogenicity data on TCE are complex in that a consistent site-specific response is
14 not seen across species. Rather, carcinogenicity is demonstrated in the mouse and the rat, but in
15 different organs. At the most basic level, the observation of rodent tumors indicates a potential
16 human cancer hazard, and the epidemiologic data are further suggestive of a cancer hazard.
17 Mode-of-action data have been generated to explain liver tumor development in the mouse and
18 are rich compared with observed effects in other organ systems; however, much uncertainty still
19 exists. In all cases, key pieces of information are still missing. The similarity between
20 carcinogenic effects induced by parent compound and metabolites supports the conclusion that
21 TCE metabolites are mostly responsible for the liver and kidney tumors observed in TCE
22 bioassays. Accordingly, much of the mechanistic data on the liver are derived from experiments
23 on the oxidative metabolites, TCA and DCA; parallel studies in the parent compound are few,
24 making it difficult to confidently assign the mode of action to a particular metabolite. The
25 complexity of TCE metabolism and clearance also complicate the identification of a
26 metabolite(s) that can be identified as responsible for TCE-induced effects. Not surprisingly,
27 more than one mode of action may explain TCE-induced carcinogenicity and several hypotheses
28 are discussed. In all likelihood, a number of events will be significant to tumor development in
29 the rodent under bioassay conditions. Uncertainty exists, however, as to which events may be
30 more relevant to human exposure to TCE at environmental levels.
31
32 3.5.1. Liver Cancer
33 Much attention has focused on understanding the development of liver tumors in the TCE
34 bioassay and the role of the oxidative metabolites, CH, DCA, and TCA. Each of these
35 metabolites has induced liver tumors in mice under bioassay conditions; DCA also induces liver
36 tumors in the rat. Mechanistic studies have concentrated on TCA and DCA. One of the largest
8/1/01 3-24 DRAFT-DO NOT CITE OR QUOTE
-------
1 uncertainties associated with mode-of-action hypotheses based on these metabolites is the
2 shortage of parallel evidence with TCE itself. Little experimental evidence is available on CH,
3 and its rapid metabolism to the haloacetic acids is believed to minimize a direct role of CH in
4 TCE-induced liver cancer (Bull, 2000). Several hypotheses have been put forth regarding the
5 development of liver tumors in the mouse. These include the role of the peroxisome proliferator-
6 activated receptor (PPAR) and peroxisome proliferation, disturbances in cell signaling, and
7 effects on DNA (Bull, 2000). A short discussion of how these events are consistent with other
8 physiologic conditions associated with hepatocarcinogenicity is presented at the end of the
9 section.
10
11 3.5.1.1. Peroxisome Proliferation
12 The prevailing view of TCE-induced mouse liver carcinogenesis has been that these
13 tumors arise in parallel with peroxisome proliferation in the liver by TCE metabolites (Elcombe,
14 1985; Elcombe et al., 1985; Goldsworthy and Popp, 1987; Melnick et al., 1987; DeAngelo et al.,
15 1989; Cattley et al., 1998). Specifically, TCA is most closely associated with this hypothesized
16 mode of action because a sustained response with chronic exposure has been noted, in contrast to
1 7 the brief response reported for DCA (Bull, 2000). How tumors may develop from peroxisome
18 proliferation remains unclear (Bull, 2000; Zhou and Waxman, 1998). It has been proposed that
19 oxidative damage caused by marked increases in free-radical generating enzymes and
20 peroxisomal • 'oxidation might initiate carcinogenesis. It was generally believed that as
21 peroxisome proliferation has not been observed in humans, agents that produced this result in the
22 rodent would not present a carcinogenic hazard to humans.
23 This view had merit in the past; however, emerging data on the peroxisome proliferator-
24 activated receptor (PPAR) implicate nonperoxisomal events such as PPAR-triggered gene
25 transcription as having important bearing on liver toxicity and subsequent tumor development.
26 This position is consistent with the view taken by an expert panel convened by ILSI, which
27 concluded that the ability to act as a peroxisome proliferator does not exclude the possibility that
28 other properties independently contribute to the development of cancer (Cattley et al., 1998; Bull,
29 2000). Moreover, this expert panel determined that there was insufficient basis for concluding
30 that these compounds did not represent a carcinogenic hazard to humans. More recent
31 experimental evidence suggests that chemicals that induce peroxisome proliferation also produce
32 a carcinogenic response in rodents through the activation of the PPAR by the chemical or its
33 metabolite(s). The activated PPAR will interact with the retinoid-X receptor (RXR) and other
34 specific response elements, resulting in increased synthesis of peroxisomal enzymes and
35 promotion of gene transcription including transcription of enzymes important in lipid
36 metabolism.
8/1/01 3-25 DRAFT-DO NOT CITE OR QUOTE
-------
1 The PPAR belongs to the steroid hormone receptor superfamily and includes several
2 distinct types: PPAR* ,*PPAR* ,*and PPAR* • PPAR* *is found in many species including mice,
3 rats, and humans. It has been mostly identified in the liver of rodents and more recently in the
4 human liver. PPAR* *is also found in other human tissues including muscle, liver, kidney, heart,
5 and testis (Youseff and Badr, 1999; Schultz et al., 1999). Human liver PPAR* *mRNA has been
6 reported to be 10-fold lower than that of mouse liver, whereas that of skeletal muscle is higher in
7 humans than in rats (Youseff and Badr, 1999). Interindividual variations are also evident
8 (Tugwood et al., 1998).
9 TCE and its metabolites have been shown to induce peroxisome proliferation in rodents
10 and its metabolites (TCA and DC A) to activate PPAR* "(Goldsworthy and Popp, 1987; Maloney
11 and Waxman, 1999; Zhou and Waxman, 1998). However, TCE, TCA, and DC A are considered
12 as weak peroxisome proliferators compared to the model drug, WY-14643. Recent in vitro
13 experimental evidence at high concentrations of both TCA and DC A shows activation of human
14 PPAR* *(Maloney and Waxman, 1999). No species differences were noted in the magnitude of
15 activation, which, for both human and mouse PPAR* *activation, appeared to increase with
16 increasing concentration of either TCA or DC A. High concentrations of TCA, but not DC A,
1 7 were further shown to activate PPAR* •from mice; no activity was demonstrated for human
18 PPAR* "(Maloney and Waxman, 1999).
19 PPAR activation is likely more important as a possible mode of action for TCA than for
20 DCA, even though both DCA and TCA are shown to activate mice and human PPAR* • As
21 discussed more thoroughly in the next section, TCA- and DCA-induced tumors are
22 characteristically very different. These observations support an inference of differing modes of
23 action for tumor induction by each of these metabolites. PPAR* *activation is seen in vitro with
24 TCA concentrations in mM range (Issemann and Green, 1990; Maloney and Waxman, 1999),
25 approximately the same range that would be reached in blood of mice by treatment with TCE at
26 the doses employed in the in vivo cancer bioassays (Bull, 2000). Like TCA, mM concentrations
27 of DCA are also shown to activate PPAR* • However, in contrast to TCA, systemic
28 concentrations of DCA associated with experimental doses that induce liver tumors in mice are
29 expected to be very small (Barton et al., 1999). Moreover, apoptosis and other effects seen with
30 DCA exposure (discussed in the following section) are expected to occur at DCA concentrations
31 much lower than the in vitro concentrations producing PPAR activation. Thus, PPAR activation
32 and the resulting pleotrophic responses associated with receptor activation are not expected to
33 contribute greatly to DCA's carcinogenic mode of action.
34 TCA's ability to activate PPAR* *has cross-species relevance. The observations of
35 Maloney and Waxman (1999) support a qualitative similarity between mice and humans; thus, a
36 PPAR hypothesized mode of action is relevant to both mice and humans. As identified above,
8/1/01 3-26 DRAFT-DO NOT CITE OR QUOTE
-------
1 quantitative differences in the expression of PPAR* "exist between mice and humans, with
2 humans in general having a lower expression of PPAR* •compared to mice. These quantitative
3 differences have import to the dose-response analysis of the mouse liver tumors and will be more
4 rigorously treated in Section 4, Dose-Response Assessment and Characterization.
5
6 3.5.1.2. Cell Signaling and Effect of Negative Selection andApoptosis
7 Bull (2000) suggests that modification of cell signal pathways by TCA and DCA
8 resulting in alterations in cell replication, selection, and apoptosis (programmed cell death) are
9 likely important contributions to the hepatocarcinogenicity of TCE and its metabolites. TCA's
10 ability to activate PPAR and the subsequent cascade of pleotrophic responses, including effects
11 on gene transcription, is a classic example of cell signaling. DCA exposure has additionally been
12 shown to influence other cell signaling pathways, and observed perturbations provide insight on
13 mode-of-action hypotheses regarding induction of DCA tumors.
14 Hepatomegaly, or increase in liver size, is a common feature of TCE, DCA, TCA, and
15 CH exposure to rodents. Hepatomegaly associated with TCE is observed at experimental
16 exposures well below those shown to induce liver tumors in mice (Bull, 2000). Hepatomegaly is
17 influenced by increases in either cell number or cell size. Cell number is affected by cell
18 replication rates and by rates of apoptosis. There is also some evidence that hepatomegaly
19 associated with PPAR* •activation results from suppressed apoptosis (Roberts et al., 1995);
20 Kupffer cells have been shown to contribute to this process (Chevalier and Roberts, 1998; Rose
21 1997). One of the roles of apoptosis appears to be the prevention of damaged cells from dividing
22 and progressing to tumors.
23 Hepatomegaly seen with TCA and DCA exposure is engendered by different processes.
24 Hepatomegaly associated with DCA is attributable to cytomegaly, increases in cell size, rather
25 than increases in cell numbers (Bull, 2000). Hepatocytes stain heavily for glycogen, which
26 accumulates in cells within 1 to 2 weeks of treatment and becomes more severe over time (Bull,
27 2000). These observations are quite different from the hepatomegaly induced by TCA. For
28 TCA, increases in liver weight are observed soon after exposure and are linear with experimental
29 dose (Bull, 2000); however, this is primarily a reflection of cell number and not cell size.
30 Experimental observations support the inference that DCA and TCA induce tumors that
31 vary in growth rate and develop from different clones of cells in mouse liver (Bull, 2000).
32 Tumors produced from DCA and TCA exposure display different characteristics. DCA produces
33 more benign foci that lead to small adenomas, although larger lesions include carcinomas. TCA
34 elicits fewer altered foci but more carcinomas. These two metabolites, moreover, show divergent
35 responses once exposure is terminated (Bull, 2000). Termination of DCA treatment halts
36 progression of the foci and adenomas to carcinomas. In contrast, carcinomas are still observed
8/1/01 3-27 DRAFT-DO NOT CITE OR QUOTE
-------
1 after termination of TCA exposure, although fewer in number (Pereira and Phelps, 1996).
2 Differences are also reported in the immunoreactivity of tumors induced by TCA and DC A in
3 regard to the proto-oncogenes c-jun and c-fos (Bull, 2000). DCA- and TCA-induced tumors also
4 appear to have differences as to the shape of their respective dose-response curves, with the
5 number of DCA-induced tumors appearing to increase disproportionately with dose while those
6 of TCA appear to be linear.
7 One possible explanation for these observations is that DCA and TCA may selectively
8 modify growth rates of different clones of cells. TCE and its metabolites could alter cell
9 replication and apoptosis rates in altered hepatic foci and preneoplastic foci, and provide a
10 selective growth advantage for differing clones of cells rather than a generalized proliferative
11 response. Cells resistant to the downregulation of mitosis induced by a chemical may have
12 selective growth advantage over other cells in that organ (Bull, 2000).
13 Effects of DCA and TCA on replication rates in foci and normal hepatocytes vary and
14 appear to be dose dependent in some cases. DCA at higher doses can increase replication rates of
15 tumor cells (Bull, 2000), whereas only a small stimulatory effect on replication is seen in
16 extrafocal (normal) hepatocytes. Further, as treatment progresses, an inhibitory effect is
17 observed in the extrafocal hepatocytes. This suppression of cell replication in extrafocal
18 hepatocytes is accompanied by a parallel decrease in apoptosis. These observations raise a
19 question as to which is the driver of the decreased cell turnover in extrafocal hepatocytes seen
20 with DCA. It is unknown if the decrease in cell turnover would increase the probability of
21 transformation of hepatocytes to allow clonal expansion of damaged hepatocytes that would
22 normally be eliminated (Bull, 2000). This possibility is suggested by the modeling of DCA
23 lesion growth rates by Miller et al. (2000).
24 TCA depresses replication rates in normal hepatocytes with no indication that replication
25 rate within tumors are modified by TCA treatment (Bull, 2000). TCA appears to be selecting
26 initiated cells that are more aggressive than those selected by DCA at low doses. The extent to
27 which replication rates in extrafocal hepatocytes are decreased with longer term TCE exposure
28 has not been investigated, nor those in foci or tumors.
29 Exposure to DCA has been shown to perturb a number of normal cellular functions
30 including cell signaling pathways, particularly those associated with carbohydrate handling (Bull,
31 2000). DCA has more recently been shown to alter serum insulin concentrations in mice
32 (Kato-Weinstein et al., 1998). Insulin is recognized to be mitogenic in liver (Bull, 2000) and to
33 have differing effects on apoptosis (Hermann et al., 1999). Further, insulin receptor expression
34 in extrafocal portions of the liver is suppressed by DCA (Bull, 2000). The principal intracellular
35 substrate of the insulin receptor, tyrosine kinase, is linked to enzymes regulating glycogen
36 synthesis. In fact, DCA exposure to mice inhibits the glycogen synthesis enzymes and is also
8/1/01 3-28 DRAFT-DO NOT CITE OR QUOTE
-------
1 associated with increased glycogen deposition in extrafocal hepatocytes, but not in preneoplastic
2 foci, which are glycogen poor. DCA can also inhibit pyruvate dehydrogenase kinase, although at
3 levels higher than those needed for effects on glycogen storage. The dose-response for glycogen
4 deposition in the liver is in the same range as that seen with liver tumors, raising questions of
5 whether modification of a same-cell signaling pathway may be responsible for both observations.
6 Moreover, the accumulation of glycogen in normal hepatocytes may contribute to the differential
7 effects of DCA treatment on insulin receptor expressed in extrafocal hepatocytes and those in
8 altered foci (Bull, 2000). DCA thus may alter the insulin signaling pathway and provide a
9 growth advantage to initiated cells through a signal transduction pathway necessary for tumor
10 growth (Kato-Weinstein et al., 1998). The precise nature of this pathway, however, has yet to be
11 fully elucidated.
12 Further, a number of metabolic disturbances are associated with exposure to DCA. A
13 major site of action of DCA at high doses is pyruvate dehydrogenase (PDH), the rate-limiting
14 enzyme of glucose oxidation (Stacpoole et al., 1992). DCA exposure has been shown to inhibit
15 hepatic triglyceride and cholesterol biosynthesis, whereas the stimulation of PDH results in
16 effects on gluconeogenesis (Stacpoole et al., 1992). Emerging experimental evidence, moreover,
1 7 suggests that DCA may interfere with tyrosine catabolism. DCA's further metabolism is
18 dependent on a glutathione-S-transferase isozyme, GST zeta (GSTZ) (Anderson et al., 1999),
19 also identified as maleylacetoacetate isomerase (Fernandez-Canon et al., 1998).
20 Maleylacetoacetate isomerase is one step in the catabolism of tyrosine. Chronic exposure to
21 DCA causes an irreversible inactivation of GSTZ (Anderson et al., 1999), with the rate of
22 inactivation greater for rats than for humans; the rate of inactivation in the mouse falls in
23 between those for the rat and human (Tzeng et al., 2000).
24 The interrelationship between rates of cell replication with rates of cell deaths make it
25 very difficult to determine if the effects of TCA and DCA are exerted on one process or the other
26 (Bull, 2000). Although the mechanisms underlying DCA- and TCA-induced alterations in cell
27 division and death are not yet known, the experimental observations suggest that each metabolite
28 differentially alters these processes. However, the patterns of effects on the processes of cell
29 division and death provide a reasonable framework for TCE-induced liver cancer. Moreover,
30 although it is not yet clear whether DCA-induced effects on carbohydrate handling and on other
31 metabolic processes are directly related to cancer induction, such effects are consistent with a
32 hypothesis regarding selection and can be consistent with other physiologic conditions associated
33 with hepatocarcinogenicity.
34
35 3.5.1.3. Effects on DNA
8/1/01 3-29 DRAFT-DO NOT CITE OR QUOTE
-------
1 A key question to risk assessment is whether TCE-induced tumors are mediated through
2 an induced mutational mechanism (Moore and Harrington-Brock, 2000). Examination of tumors
3 for possible mutation has been carried out for investigation of this potential mode of action. TCE
4 and its metabolites CH, DC A, and TCA are, at best, very poor genotoxicants (Moore and
5 Harrington-Brock, 2000). TCE and DCA at high concentrations have been shown to bind to
6 DNA and proteins, and to induce single-strand breaks (Moore and Harrington-Brock, 2000; Bull,
7 2000). The magnitude of response, however, is low compared to other more potent
8 genotoxicants such as MMS and DMN (Moore and Harrington-Brock, 2000). DCA is genotoxic
9 and has been shown to cause DNA strand breaks in mouse and rat liver cells following high level
10 in vivo gavage exposure. TCA is the least mutagenic of the TCE metabolites (Moore and
11 Harrington-Brock, 2000).
12 Mutation frequency and mutation spectra in the ras oncogene have been examined in
13 TCE-, TCA-, and DCA-induced liver tumors from male mice; it is interesting to note that very
14 few codon 61 h-ras mutations are seen in liver tumors from female mice treated with DCA. The
15 actual amount of mode-of-action information on induced mutational mechanisms from these data
16 is quite minimal (Moore and Harrington-Brock, 2000). However, the pattern of mutation seen
1 7 helps provide insight into the relative contribution of different metabolites to TCE-induced liver
18 cancer (Bull, 2000). Mutational frequency in DCA and TCE tumors (arising from drinking water
19 exposure) does not differ significantly from that observed in spontaneous tumors; TCA tumors
20 show both a higher mutation frequency and a pattern atypical of that found with other
21 peroxisome proliferators (Bull, 2000; Maronpot et al., 1995). Comparison of nucleotide base
22 sequences within codon 61 of h-ras shows some differences between spontaneous tumors and
23 DCA and TCE tumors, suggesting a similarity between tumors from DCA and TCE-treated
24 animals (Moore and Harrington-Brock, 2000). Route of exposure, however, appears important to
25 mutational frequency associated with TCE-induced liver tumors (Bull, 2000).
26 The similarity between DCA and TCE tumors in terms of frequency of mutation and its
27 spectra support an inference that DCA contributes an important role in the hepatocarcinogenicity
28 of TCE. The theoretical modeling of TCE, DCA, and TCA liver tumors of Chen (2000) provides
29 further support for DCA's contribution to TCE-induced liver cancer, although a role for TCA
30 cannot be excluded either. As additional supportive evidence, lower doses of DCA that
31 accompany higher doses of TCA appear to exert a large influence on tumor development.
32 Female mice initiated with N-methyl-N-nitrosourea followed by promotion with a mixture
33 containing both DCA and TCA developed tumors that appeared more similar to tumors promoted
34 by DCA alone than those by TCA alone (Pereira et al., 1997). Moreover, the number of tumors
35 produced from the mixture was greater than would be achieved by adding together tumor
36 incidences for the individual exposures.
8/1/01 3-30 DRAFT-DO NOT CITE OR QUOTE
-------
1 The interaction of TCE and its metabolites to bind directly with DNA may not be as
2 important at lower exposure levels as effects on gene expression. Specifically, effects on gene
3 expression, whether through activation of PPAR or other signal pathways, may be more
4 important to their potential carcinogenic properties than induced mutation. Moreover, effects
5 such as the mistargeting of covalent modification to the proteins that package DNA (e.g., histone
6 acetylation) or to DNA itself (e.g., cytosine methylation) may be involved in cancer induction,
7 not by altering gene sequence, but by modifying transcription of the gene (Wolffe, 1999). For
8 example, DNA methylation is believed to play a role in the control of gene transcription
9 (Goodman and Counts, 1993; Counts and Goodman, 1995).
10 TCE and its oxidative metabolites have been shown recently to decrease methylation of
11 DNA as 5-methylcytosine (5MeC) in addition to demethylation of the promoter regions of two
12 proto-oncogenes, c-jun and c-myc (Tao et al., 1998, 1999, 2000a). Moreover, the decreased
13 methylation and increased expression of c-jun and c-myc with TCA and DC A exposures also
14 occurred in the presence of increased DNA methyltransferase activity (Tao et al., 2000b). The
15 proto-oncogenes, c-jun and c-myc, participate in the control of cell proliferation and apoptosis
16 (Tao et al., 2000a). The expression of mRNA for these two genes was shown to be transiently
17 increased as well (Tao et al., 1999). This is an early observation, seen after 3 days of exposure to
18 TCE. In fact, cell proliferation is enhanced during the first few weeks of exposure to DC A and
19 TCA (Bull, 2000). Hypomethylation associated with exposure to DCA appears to be reversed
20 once exposure is terminated (Tao et al., 1998), in contrast to the extent of hypomethylation
21 associated with exposure to TCA, which remained elevated upon exposure cessation. These
22 observations support the hypothesis that DCA, and not some other process or event, is affecting
23 methylation in the tumors. Tao (2000a) recently suggested that depletion of S-
24 adenosylmethionine may be responsible for the observed hypomethylation of DNA. The
25 prevention of DCA-, TCA-, and TCE-induced DNA hypomethylation by the addition of
26 methionine supports this hypothesis.
27 The relationship between DNA methylation and cancer induction is not well elucidated,
28 and the events and their sequence, which are important to neoplastic cell transformation, are
29 lacking. Consequently, at the current time, DNA methylation may be best considered an effect of
30 exposure rather than a precursor specific to the development of tumors.
31
32 3.5.1.4. Evaluating Proposed Modes of Action for Liver Carcinogenesis
33 There are several important questions that arise from the discussion above: which of the
34 hypothesized modes of action can be judged as more plausible; is there evidence to reason any
35 one is the mode of action for TCE hepatocarcinogenicity, and if so, is it important to EPA's
36 proposed cancer guidelines (U.S. EPA, 1999); whether an event considered as a precursor to
8/1/01 3-31 DRAFT-DO NOT CITE OR QUOTE
-------
1 carcinogenesis can be identified. Hence, insight into these questions can support weight-of-
2 evidence determinations and approaches that will be used in the quantitative dose-response
3 assessment presented in Section 4. EPA's proposed cancer guidelines (U.S. EPA, 1996, 1999)
4 note the necessity to outline the sequence of events leading to cancer, to identify key events that
5 can be measured, and to weight information to determine whether there is a causal relationship
6 between events and cancer formation. The validation of a causal relationship between exposure
7 and cancer risk includes the consideration of a number of factors, many of which have been
8 borrowed from the epidemiologic literature. These criteria include (1) consideration of the
9 strength, consistency, and specificity of the association between event and tumor response; (2)
10 the presence of a dose-response relationship between the key event and tumor endpoint; (3) the
11 temporal relationship between tumor onset and occurrence of the key event (i.e., the key event
12 should precede tumor formation); (4) the biological plausibility and coherence of the mode of
13 action; and (5) consideration of other modes of action. Experimental evidence that can be used
14 to address these criteria is often derived in the species in which tumors were observed. However,
15 the ultimate question is the human relevance of the proposed mode of action. Hence, it is
16 necessary to demonstrate that the sequence of steps leading to tumorigenesis in the rodents will
1 7 also occur in humans.
18 Experimental evidence for support of mode-of-action hypotheses in the liver focuses on
19 two of TCE's oxidative metabolites, TCA and DCA. Few data are available on TCE exposure
20 directly, and only in studies involving activation of the PPAR, albeit in vitro, experiments of
21 human hepatic cells. Accordingly, it is difficult to determine which of the metabolites of TCE
22 may be responsible for its toxicity, what the key events may be, and what is their relevance to
23 humans. There is enough evidence, however, to suggest that many of the responses to TCE
24 metabolites are relevant to the ability of TCE to induce cancer of the liver in animals, as well as
25 to the extrapolation of these observations to humans.
26 TCA and DCA have been shown to induce a number of different events in livers of mice
27 (discussed in the preceding sections). Both have been shown to activate PPAR* «as evidenced by
28 peroxisomal proliferation; however, a role of PPAR activation as a possible mode of action is
29 more plausible for TCA than for DCA. TCA induces a more sustained proliferative response
30 than DCA. Furthermore, DCA induces effects on cell signaling processes and on carbohydrate
31 handling at lower concentrations than those associated with peroxisome proliferation. Besides
32 PPAR* • TCA has been shown to activate PPAR* •from mice but not humans.
33 The role of PPAR* •activation and the sequence of events following activation that are
34 important to tumorigenesis are not well defined, nor is the contribution of other forms of PPAR
35 to observed TCE carcinogenicity currently known. Furthermore, extraperoxisomal effects of
36 activation of transcription by PPAR* •may elicit a number of changes in cell signaling that may
8/1/01 3-32 DRAFT-DO NOT CITE OR QUOTE
-------
1 also be linked to increased cancer risk. Moreover, key events critical to the induction of tumors
2 and the cross-species relevance of these key events have yet to be identified. Further
3 investigations regarding activation of PPAR through the use of genetically altered experimental
4 animals may provide insight regarding the role of PPAR* *in the induction of liver tumors by
5 TCE and the contribution of TCA formation.
6 DCA has the ability to affect other cell signaling processes and to disturb carbohydrate
7 handling at lower concentrations than those associated with PPAR activation. Specifically, DCA
8 has been shown to alter cell replication, selection, and apoptosis. The observed disturbances to
9 insulin signaling and to carbohydrate handling associated with DCA exposure may also be
10 connected to these alterations. The events that lead to and result in a selective growth advantage
11 or in depressed apoptosis of altered hepatic foci have yet to be clearly defined. As with
12 activation of PPAR, critical events important to the induction of liver tumors in mice
13 administered DCA have yet to be identified.
14 Another potential mode-of-action hypothesis of DCA liver tumor induction recently put
15 forth by Cornett et al. (1999) involves disturbance of tyrosine catabolism. DCA's inhibitory
16 effects on GSTZ may also inhibit maleylacetoacetate isomerase. Inhibition of maleylacetoacetate
17 isomerase activity could consequently lead to a buildup of intermediates of tyrosine catabolism,
18 several of which have are considered alkylating agents (Cornett et al., 1999). This proposed
19 mode of action needs further investigation because direct experimental evidence regarding TCE-
20 induced effects on tyrosine intermediates such as maleylacetoacetate and maleylacetone is
21 lacking. At the present time, the hypothesis of DCA-induced effects on cell signaling is more
22 compelling.
23 Another approach for investigating the possible role of TCE metabolites in its toxicity is
24 to ascertain whether enough of the metabolite is present after administration of the parent
25 compound to induce tumors at the target of concern. A critical issue is the amount of DCA
26 produced from chronic exposure to TCE and whether this amount would be sufficient to elicit the
27 events that have been observed with experimental administration of DCA. Chronic exposure to
28 DCA has been shown to increase the elimination half-life of subsequent exposures in mice, rats,
29 and humans, suggesting that chronic exposure to DCA inhibits its own metabolism (Gonzales-
30 Leonetal., 1999). Moreover, recent reports of irreversible inactivati on of DC A by GSTZ
31 suggest larger internal doses of DCA may occur than previously assumed (Tong et al., 1998;
32 Tzeng et al., 2000). Thus, factors that may affect the clearance of DCA and the role of a possible
33 buildup of toxic metabolites are complex and are another uncertainty associated with this
34 assessment.
35 Nevertheless, Barton et al. (1999) have examined this question for mice by carrying out a
36 theoretical modeling study of both TCE and DCA pharmacokinetics. This study estimated that
8/1/01 3-33 DRAFT-DO NOT CITE OR QUOTE
-------
1 the maximum levels of DCA in the serum produced from 1,000 and 2,000 mg/kg gavage doses of
2 TCE (tumorigenic doses in the bioassay) would be equivalent to internal dose metrics that were
3 approximately half of those estimated from exposure to 0.5 mg/L DCA in drinking water.
4 Exposure of mice to 0.5 g/L DCA in drinking water produced a 48% to 63% incidence of
5 hepatocellular carcinomas and a 75% incidence of combined liver tumors (carcinomas and
6 adenomas), with an increase in the number of carcinomas per liver compared with livers from
7 control animals (Daniel et al., 1992; DeAngelo et al., 1999; Bull, 2000). The analysis of Chen
8 (2000) provides further support that DCA produced by TCE is capable of producing the
9 carcinogenic response seen in mice from gavage exposure to TCE.
10 Information needed to extrapolate the potential risk of DCA formation from TCE
11 exposure in regard to human liver cancer induction is more limited. Unfortunately, Barton et al.,
12 (1999) do not examine the amount of DCA produced in humans from TCE exposure, and only
13 one study (Fisher, 1998) is available that assayed DCA in serum from 17 human volunteers (9
14 males and 8 females) exposed via inhalation for 4 hours to 100 ppm TCE. Fisher (1998) noted
15 low but detectable levels of DCA in serum in five volunteers (three males and two females) with
16 levels ranging between 4 and 12 ppm. The lack of detection of DCA in the other 12 subjects may
17 be due to the rapid metabolism of DCA, as inhibition of DCA's metabolism would not be
18 expected to occur from a 4-hour exposure. Alternatively, but less likely, the absence of DCA in
19 these human subjects may be from lack of DCA formation from TCE. Only one subject with
20 detectable DCA levels in serum was exposed to TCE at the lower level of 50 ppm for 4 hours,
21 and DCA was not detected in this experimental protocol (Fisher, 1998, personal communication
22 with J. Fisher). Furthermore, DCA has been detected in children and adults who received chloral
23 hydrate (Henderson et al., 1997; Yan, 1999). These studies together provide support that
24 humans, like mice, have the potential to produce DCA from either TCE when administered as
25 parent or from a key TCE metabolite, chloral hydrate.
26 The technical problems with detection of DCA, however, are difficult, particularly in the
27 presence of large amounts of trichloroacetic acid, and samples are subject to the presence of
28 DCA as an artifact (Ketcha et al., 1996). Hence, it is not clear whether the reported findings of
29 DCA in humans are accurate or are a reflection of conversion of TCA to DCA under acidic
30 conditions. The more recent data of Gonzalez-Leon (1999) and analysis of Barton et al., (1999)
31 are not subject to analytical artifact; thus, it can be concluded that mice convert DCA from TCE.
32 Given the large uncertainty associated with the human data, by default, it is reasonable to assume
33 that humans would similarly produce DCA and observations in mice can be extrapolated across
34 species to humans.
35 Several lines of reasoning lead to the conclusion that no one metabolite can be identified
36 as responsible for TCE's carcinogenicity. The modeling studies of Barton et al. (1999) and Chen
8/1/01 3-34 DRAFT-DO NOT CITE OR QUOTE
-------
1 (2000) support an inference that the concentrations of DC A in mice produced from exposures
2 experienced in the TCE gavage bioassay are of sufficient quantity to induce hepatic tumors.
3 Conclusions of Chen (2000) further suggest ambiguity in identifying one metabolite as
4 explaining TCE hepatocarcinogenicity. Both metabolites may each have their separate modes of
5 action and may, in fact, work in concert to alter each other's toxicity. This is suggested from the
6 experimental evidence of joint TCA and DCA exposure by Pereira et al. (1997) and from newly
7 emerging experimental data of Bull (personal communication). Bull (personal communication)
8 has examined the frequency of TCE-induced liver tumors that are c-jun+; the frequency of
9 hepatic tumors in which c-jun has been mutated is high in DCA-induced liver tumors.
10 Preliminary results from Bull suggest a role for DCA, in that many TCE-induced tumors were
11 c-jun+ (Bull, personal communication). Also noteworthy was a finding of a similar number of
12 tumors that did not display a mutation to c-jun (Bull, personal communication); c-jun is not
13 mutated in liver tumors from TCA exposure. These findings suggest that TCA and DCA both
14 contribute to the hepatocarcinogenicity of TCE.
15
16 3.5.1.5. Consistency With Other Physiologic Conditions and Associated
17 Hepatocarcinogenicity
18 The consistency in events seen with exposure to TCE metabolites, specifically DCA, and
19 physiologic conditions associated with several human diseases can provide insight into mode-of-
20 action hypotheses and may support inferences regarding potential increased human cancer risks.
21 Many cellular events observed with exposure to DCA, including changes in carbohydrate
22 handling, are also seen with several diseases such as glycogen storage diseases and diabetes.
23 Although the pathogenesis of these diseases and what is known about liver tumor induction with
24 DCA exposure may have distinct differences, disturbances in cell metabolism and in
25 carbohydrate handling are common events. As discussed below, diabetics and individuals with
26 glycogen storage diseases are at increased risk for liver cancer. Furthermore, individuals with
27 tyrosinemia also have an increased risk of liver cancer (Tanguay et al., 1996; Laberge et al.,
28 1986).
29 Glycogen storage diseases are caused by inherited deficiencies of enzymes that regulate
30 the synthesis, transport, or degradation of glycogen. Individuals with glycogen storage diseases
31 store excess glycogen in their livers. This storage resembles that seen in the extrafocal
32 hepatocytes of DCA-treated livers. The majority of individuals with glycogen storage diseases
33 develop liver cancer early in life. In fact, multiple hepatocellular neoplasms are often seen in
34 these individuals. In a similar pattern, DCA exposure induces multiple foci in mice livers.
35 Diabetes is a disease involving insulin disturbances. Poorly controlled diabetes results in
36 excess glycogen storage in livers as well. Hepatomegaly is seen in both Type I and Type n
8/1/01 3-35 DRAFT-DO NOT CITE OR QUOTE
-------
1 diabetics (Herrman et al., 1999) and diabetics have an increased risk for liver cancer (LaVecchia
2 et al., 1994; Adami et al., 1996; Wideroff et al., 1997). It is possible that greater sensitivity of
3 tumor cells than normal cells to insulin could mimic such disturbance and contribute to DCA
4 liver tumor induction as well (Bull, 2000).
5 Moreover, observations on cellular changes that include markers in liver tumors resulting
6 from exposure to TCE, DCA, or TCA overall are consistent with a more generalized description
7 of hepatocarcinogenesis proposed by Bannasch and others (Bannasch et al., 1986, 1996, 1997; Su
8 et al., 1997). A number of findings have been reported that suggest the sequence of cellular
9 changes during hepatocarcinogenesis is in principle identical in humans and experimental
10 animals (Bannasch et al., 1986, 1996, 1997). For example, Bannasch (1986) notes a correlation
11 between changes in cell metabolism, carbohydrate handling, and neoplastic cell transformation.
12
13 3.5.2. Kidney Cancer
14 Mode-of-action studies investigating kidney tumors induced in gavage and inhalation
15 bioassays have focused on metabolites, particularly those associated with the glutathione
16 pathway, where the experimental evidence is more persuasive. There are several lines of
1 7 investigation for potential modes of action for TCE in the kidney as cited by Lash et al. (2000b):
18 peroxisome proliferation, alpha-2u globulin nephropathy, mutagenicity or other genetic toxicity,
19 formic acid formation, and cytotoxicity-induced damage. The question is which, if any, of these
20 mechanisms are operative and likely to lead to the development of cancer at doses that were
21 lower than those used in the bioassays of TCE in animals.
22
23 3.5.2.1. Genotoxicity
24 Cysteine intermediates (DCVC and DCVG) formed in the GST pathway are capable of
25 inducing point mutations, as evidenced from positive findings in the Salmonella assay (Moore
26 and Harrington-Brock, 2000). DCVC is the most potent of the TCE metabolites as a Salmonella
27 mutagen. DCVG is not as potent as DCVC, but this is consistent with the hypothesis that DCVG
28 needs to be converted to DCVC to induce point mutations. Although DCVC and DCVG have
29 not been evaluated in other screening assays, there is some indication that they can induce
30 primary DNA damage in mammalian cells in vitro and in vivo. Further, formation of covalent
31 adducts with proteins and other macromolecules in mitochondria have been documented (Lash et
32 al., 2000b). The covalent binding of DCVC is largely dependent on its metabolism by the • *
33 lyase pathway. DCVC has been shown to induce the express of two proto-oncogenes, c-fos and
34 c-myc (Lash et al., 2000b). At noted in the liver mode-of-action discussion, c-myc is involved in
35 the control of cell proliferation and apoptosis (Tao et al., 2000a).
8/1/01 3-36 DRAFT-DO NOT CITE OR QUOTE
-------
1 A key question in risk assessment, like that identified in the liver mode-of-action
2 discussion, is whether TCE-induced kidney tumors are mediated through a compound- or
3 metabolite-induced mutational mechanism. Two reports provide suggestive evidence for TCE-
4 induced mutations in the VHL gene of renal cell carcinoma patients (Briining et al., 1997b;
5 Brauch et al., 1999). Brauch et al. (1999) expanded the work of Briining et al. (1997b) and
6 observed somatic mutations in 75% of the renal cell carcinomas from TCE-exposed individuals.
7 Many (42%) of the carcinomas taken from TCE-exposed cases had multiple mutations, and
8 nucleotide base changes were found in five regions. A specific "hot spot" on gene nucleotide
9 454 was altered in 13 renal cell carcinoma patients with TCE exposure, but this mutation was not
10 detected in renal cell carcinoma patients without TCE exposure or was carried by healthy
11 subjects. Brauch et al. (1999) hypothesize that the C>T mutational changes in the VHL gene of
12 TCE-exposed renal cell carcinoma patients may be associated with alkylation and DNA adduct
13 formation of the opposite DNA strand. It is probable that these mutations play a role in the
14 development of these tumors because a function of this gene is to suppress the development of
15 kidney cancer. These data identify a profile that seems to be specific to TCE exposure. The
16 finding of the mutation in nucleotide 454 in normal kidney parenchymal tissue from four patients
1 7 adds further support that this mutation may precede tumor formation. The halothioketenes,
18 products from • *lyase cleavage of DCVC, form DNA adducts in vitro and could conceivably
19 contribute to the induction of these mutations (Miiller et al., 1994, 1998a,b). Moreover,
20 chlorothioketenes have been shown in vitro to form adducts with cytosine in an aqueous solution
21 (Volkel et al., 1998). There may, however, be alternative mechanisms such as selection of cells
22 within the mutation, rather than producing the mutation, that may account for or at least
23 contribute to these observations. Although this is not conclusive proof of a mutagenic
24 mechanism, it is prudent to consider this as an important and likely mode of action and, at the
25 minimum, a marker for TCE-induced kidney cancer.
26
27 3.5.2.2. Accumulation ofAlpha-2u Globulin
28 In the past, the production of alpha-2u globulin was considered as contributing to TCE-
29 induced kidney tumors, with the initiation of several studies to examine this hypothesis.
30 Alpha-2u globulin accumulation is unique to male rats, and when in excess is associated with
31 nephropathy, cytotoxicity, cellular proliferation, and consequent renal tumors. Goldsworthy
32 shows that TCE does not cause an accumulation of alpha-2u globulin in the male (Goldsworthy
33 et al., 1988; Lash et al., 2000b). Moreover, TCE is a renal toxicant and clearly causes kidney
34 damage in both male and female rats and in mice (Barton and Clewell, 2000). With respect to
35 kidney tumors, the analysis of Rhomberg (2000) shows that kidney tumors are found in both
36 TCE-treated male and female rats. Thus, alpha-2u globulin accumulation does not appear to be a
8/1/01 3-37 DRAFT-DO NOT CITE OR QUOTE
-------
1 mode of action of TCE-induced kidney toxicity in rats or relevant to human risk (Lash et al.,
2 2000b).
3
4 3.5.2.3. Peroxisome Proliferation
5 A potential role of peroxisome proliferation should be included in any mode-of-action
6 discussion for kidney tumor induction because renal enzymes generate chloroacetates that are
7 known to induce peroxisome proliferation, and the renal proximal tubular epithelial cells are
8 relatively rich in peroxisomes. One study specifically has addressed the issue of TCE-induced
9 peroxisome proliferation in the kidney, with other studies examining the influence of TCE
10 metabolites on peroxisome proliferation (Lash et al., 2000b). Peroxisome proliferation was
11 reported to occur in both rat and mouse kidney following exposure to TCE, with mice showing a
12 greater response, as measured by cyanide-insensitive palmitoyl-CoA oxidation, than that seen in
13 rats (Goldsworthy and Popp, 1987). TCE and its metabolites have not been shown to induce
14 kidney cancer in mice, however. Odum et al. (1988), additionally, suggest that TCA-induced
15 peroxisome proliferation does not play a role in male rat kidney carcinogenicity. A role for
16 peroxisome proliferation of TCE-induced kidney carcinogenicity is unlikely.
17
18 3.5.2.4. Nephrotoxicity and Cytotoxicity
19 TCE has been demonstrated to cause nephrotoxicity and to induce persistent changes to
20 the proximal tubules in both rats and humans. It has been proposed that kidney tumor induction
21 results from cellular necrosis and activation of repair processes that lead to cellular proliferation.
22 Most of the studies aimed at elucidating this possible mode of action have focused on DCVG or
23 DCVC, rather than the parent compound, as it is believed that flux through the GSH conjugation
24 pathway and the subsequent production of reactive species are responsible for observed
25 nephrotoxicity (Lash et al., 2000b). The in vivo formation in both animals and humans of these
26 intermediates from TCE further adds significance to these studies in an assessment of TCE. In
27 spite of the importance of • *lyase metabolism in cysteine conjugated-induced nephrotoxicity,
28 there is evidence in several other studies that suggests additional enzymatic activities may also
29 bioactivate DCVC and may be important to nephrotoxicity (Lash et al., 2000b).
30 A number of in vivo and in vitro studies support that TCE and its cysteine conjugates are
31 associated with nephrotoxicity, specifically proximal tubular damage, as measured by elevated
32 excretion of renal enzymes or by histopathologic examination. Nephrotoxicity has also been
33 shown in individuals with occupational exposure to high levels of TCE (Lash et al., 2000b;
34 Briining et al., 1999a, 1999b). Nephrotoxicity as well as tumors were observed concurrently in
35 the NTP bioassay (NTP, 1988). The incidence of nephrotoxicity in the rat bioassay was high,
36 however, and far exceeded the reported incidence of tumors (NTP, 1988). Further,
8/1/01 3-38 DRAFT-DO NOT CITE OR QUOTE
-------
1 nephrotoxicity appears to correlate with the known sex differences in metabolism rates, males
2 having a higher metabolic rate than females (Lash et al., 2000b). Exposure to TCE was
3 associated with a modest increase in lactate dehydrogenase release (a marker of cell injury) in
4 male rat kidney cells, but not in cells from female rats at similar concentrations (Lash et al.,
5 2000b).
6 Nephrotoxicity may result from oxidative stress, disturbances in calcium ion homeostasis,
7 mitochondrial dysfunction, or protein alkylation. A number of studies demonstrate
8 nephrotoxicity by these process with exposure to TCE or, more so, to DCVC. Both TCE and
9 DCVC induce an oxidative stress response that includes GSH oxidation or depletion, lipid
10 peroxidation, and oxidation or alkylation of protein sulfhydryl groups after exposure of renal
11 cells to these agents. It is likely that oxidative stress plays some role in DCVC-induced
12 nephrotoxicity, but lipid peroxidation is probably a consequence rather than a cause of cellular
13 injury (Lash et al., 2000b). DCVC-induced changes in free calcium ion concentrations include
14 inhibition of mitochondrial metabolism and function, severe mitochondrial damage, poly(ADP)-
15 ribosylation of nuclear proteins and DNA double-strand breaks, and changes in cytoskeletal
16 protein structure. The cysteine conjugates, including DCVC, have been shown to covalently bind
1 7 with proteins and other macromolecules in mitochondria. Cytosolic proteins are another target of
18 the reactive metabolites of DCVC, and protein adducts have been observed in vitro in human
19 proximal tubular cells with exposure to DCVG and DCVC and in rats exposed in vivo to DCVC.
20 The finding of higher levels of covalent binding in TCE-exposed mice compared to rats suggests
21 that other factors besides covalent adduct formation must contribute to the induction of renal
22 carcinogenesis (Lash et al., 2000b).
23 Recently, Green (Green et al., 1998) put forth another hypothesis that urinary excretion of
24 formic acid leading to decreased pH after TCE exposure may partially explain the observed renal
25 toxicity in the TCE studies. The excretion of increased levels of formic acid as observed with
26 exposure to TCE appears to be related to folate deficiency (Dow and Green, 2000). Kidney
27 toxicity has been reported in humans and rabbits with exposure to formic acid (Jacques, 1982;
28 Liesivuori and Savolainen, 1986, 1987, 1991, 1992); however, data are lacking on whether
29 formic acid induces kidney tumors. At the present time there is limited evidence that this
30 pathway is involved with renal tumorigenesis. Moreover, demonstrating a role for this
31 mechanism does not detract from a potential contribution to tumorigenesis by other pathways,
32 e.g., glutathione conjugation and processing to DCVC and subsequent reactive metabolites.
33
34 3.5.2.5. Contribution of Several Modes of Action
35 It is likely that multiple modes of action may be important in TCE-induced kidney cancer
36 (Lash et al., 2000b). As in the liver, level of exposure will be an important consideration. The
8/1/01 3-39 DRAFT-DO NOT CITE OR QUOTE
-------
1 preponderance of the evidence suggests that the cysteine conjugates and reactive metabolites
2 generated from their • *lyase metabolism are most likely responsible for kidney toxicity and
3 tumorigenicity (Lash et al., 2000b). These metabolites have not been tested in a chronic bioassay
4 and the lack of such information is an uncertainty associated with this inference. More
5 information is also needed regarding any potential role of the oxidative metabolites of TCE and
6 of formic acid.
7 Separating the contribution of genotoxicity from that of nephrotoxicity and chronic
8 cellular injury in relation to the development of renal tumors is important to a discussion of
9 potential modes of action for kidney cancer and TCE. The cysteine metabolites appear to be
10 inextricably intertwined such that the contribution of these events in tumor development cannot
11 be evaluated on the basis of data that are available today. As DCVC has been demonstrated to be
12 mutagenic in Salmonella, and mutation in the VHL gene seems to be associated with TCE-
13 related tumors in epidemiological studies, it is reasonable to regard gene mutation as a potential
14 mode of action for kidney tumors.
15
16 3.5.3. Lung Cancer
1 7 There has been much less investigation into potential modes of action for TCE-induced
18 lung cancer than for the liver and kidney. Green (2000) suggests that mice may be more sensitive
19 to the toxic effects of TCE exposure in the lung than are other species tested because of their
20 increased amounts of CYP2E1 metabolism, specifically in Clara cells, compared to humans and
21 rats. Lung tumors in mice are thought to arise through an accumulation of CH in the Clara cells
22 of the lung, causing cell damage and compensatory cell replication, which in turn leads to tumor
23 formation (Green, 2000). CH is believed to be the etiologic agent because independent studies
24 have shown that CH when administered alone causes lung lesions identical to TCE-induced
25 tumors. Neither TCA nor TCOH (metabolites of CH) causes lung lesions.
26 This hypothesis suggests that such a mechanism in mice may not be extrapolated to
27 humans, as there is little CYP2E1 activity in human lungs as a whole (Green, 2000). However,
28 activity from whole lungs may give misleading results because of the variety of cell types. High
29 activity for a few cell types may be diluted by other cells with low activity (Lash et al., 2000b).
30 While thought to be relatively scarce in the human lung, recently Boers et al. (Boers et al., 1999)
31 have characterized the number of Clara cells in the normal human airway and show that Clara
32 cells contribute substantially to cell renewal in normal conducting airway epithelium in humans.
33 Furthermore, Clara cells have been identified as important to the development of lung
34 adenocarcinoma in humans (Boers et al., 1999).
35 The hypothesis that CH is responsible for lung tumor induction requires further
36 confirmation. It would be useful to develop strong connections between metabolism of TCE, the
8/1/01 3-40 DRAFT-DO NOT CITE OR QUOTE
-------
1 actual tumor cell of origin, and the contribution of other possible mechanisms that CH might
2 utilize to increase lung tumor incidence. CH is clearly clastogenic and mutagenic at high doses
3 (Moore and Harrington-Brock, 2000), raising the possibility of genotoxic activity as well as a
4 mode of action involving simple cytotoxicity and reparative hyperplasia. Moreover, it is not yet
5 clearly demonstrated that CH is the metabolite responsible for lung tumor development.
6
7 3.5.4. Toxicity in Other Organs
8 Mode-of-action information for toxicity in other organs such as the endocrine and
9 lymphopoietic system, and in the developing fetus, is more limited than that for the liver, kidney
10 and lung. Data specific to TCE and its metabolites are few, and support for mode-of-action
11 hypotheses are based on observations of common activities with other agents. In some cases, the
12 roles of PPAR activation or cell signaling perturbations that are relevant to mode-of-action
13 discussions for the liver and kidney are applicable to discussions of toxicity observed in these
14 other organ systems. In other cases, such as lymphoma seen in the human epidemiologic studies,
15 studies using animal models are lacking or are too few to support the human observations.
16 Endocrine system effects associated with TCE exposure include the development of
1 7 testicular (Leydig cell) cancer in rats (Maltoni et al., 1988; NTP, 1988) and hormone
18 disturbances (SHBG, DHEAS, testosterone) in humans (Chia et al., 1997; Goh et al., 1998).
19 TCE and its metabolites, TCA and TCOH, have been found to partition in the male reproductive
20 organs of rats after 6 weeks of inhalation exposure (Zenick et al., 1984). A wide variety of
21 agents that affect steroid hormone levels, such as testosterone, estradiol, and luteinizing
22 hormone, will also induce Leydig cell tumors in the rat (Cook et al., 1999). Peroxisome
23 proliferating chemicals have been shown to induce Leydig cell tumors via a modulation of
24 growth factor expression by estradiol (Cook et al., 1999). Peroxisome proliferating chemicals
25 induce hepatic aromatase activity, which can increase serum and testis estradiol levels. The
26 increased interstitial fluid estradiol levels can modulate growth factors, including TGF» 'and
27 stimulate Leydig cell proliferation (Cook et al., 1999).
28 Steroid hormones such as testosterone, estradiol, and luteinizing hormone are regulated
29 through the hypothalo-pituitary-testis (HPT) axis in both rats and humans, and agents that induce
30 Leydig cell tumors in rats by disruption of the HPT axis are thought to pose a hazard to humans
31 (Cook et al., 1999). Cook et al. (1999) further suggest that Leydig cell tumors induced through
32 an aromatase activity mode-of-action such as that suggested for peroxisome proliferators are
33 considered potentially relevant to humans. Although the risk of Leydig cell cancer in humans
34 may be diminished compared to rats owing to the absence of SHBG, the observation of Leydig
35 cell tumors in rats exposed to TCE may act as a signal for disturbance of the endocrine system
36 and be indicative of potential endocrine disturbances in humans. Further, a potential target of
8/1/01 3-41 DRAFT-DO NOT CITE OR QUOTE
-------
1 testosterone disruption may be the prostate, which in humans is sensitive to testosterone and
2 other androgens (Shimada et al., 1999; Bosland et al., 1995) and is affected by Leydig cell
3 function in the rat (Yount et al., 1982). The effects of endocrine disruption in human populations
4 exposed to TCE are an area for more research.
5 TCE exposure has been associated with effects on immune system functioning and
6 includes both immune suppression and autoimmunity. Reports of both immune suppression and
7 autoimmunity may appear paradoxical at first glance; however, these observations are plausible.
8 Certain elements of the immune system may be differentially affected by a chemical, affecting
9 the ability to mount well-regulated immune responses to both foreign and self-antigens and
10 leading to conditions such as autoimmunity via an alteration in the balance between specific
11 immune components. Immune suppression, furthermore, may play a role in the induction of
12 cancer; many immune-suppressive agents are human carcinogens (Tomatis et al., 1989).
13 Additionally, a common event or events may exist in mode-of-action pathways for tumorigenesis
14 and for autoimmunity (e.g., dysregulation of apoptosis). The observation of an increased risk of
15 multiple myeloma in patients with autoimmune diseases provides some support for this
16 hypothesis (Cooper et al., 1999).
17 Organic solvent exposure in general is associated with autoimmune disease such as
18 scleroderma (systemic sclerosis) (Nietert et al., 1998). Mode-of-action hypotheses for
19 autoimmunity resulting from environmental agents have been grouped into three broad
20 categories: (1) a change in the hormonal milieu to favor estrogenic stimulation of the immune
21 system; (2) suppression of one section of the immune system, which disrupts normal immune
22 surveillance; and (3) chemical binding to a self-antigen forming a neoantigen, thus breaking
23 tolerance by inducing immunity to the unmodified native molecule as well as to the modified
24 antigen (Mayes, 1999). These mechanisms are currently speculative at best, but are supported by
25 animal models (Mayes, 1999). TCE exposure, specifically, has been shown to induce
26 autoantibody formation against dichloroacetyl chloride in genetically susceptible autoimmune-
27 prone mice (Griffin et al., 2000a,b, Khan et al., 1995). Moreover, T-cells were shown to secrete
28 more interferon-* «and less interleukin-4 after TCE exposure (Griffin et al., 2000a). Griffin et al.,
29 (2000a) note that the pattern of response was consistent with a T-helper Type 1 immune or
30 inflammatory response. Furthermore, these effects have been shown to contribute to the
31 development of autoimmune disease in the liver (autoimmune hepatitis) (Griffin et al., 2000c).
32 Oxidative metabolism is necessary for inducing the inflammatory response because blocking the
33 CYP450 pathway was shown to primarily inhibit the response (Griffin et al., 2000b). One
34 epidemiologic study also reported an strong association between TCE exposure and systemic
35 sclerosis patients with autoantibodies (Nietert et al., 1998). The mode(s) of action for these
36 observations is not known at this time.
8/1/01 3-42 DRAFT-DO NOT CITE OR QUOTE
-------
1 The Kupffer cell may also play an important role in autoimmune response. Kupffer cells
2 have been discussed previously in the liver mode-of-action section. Activated Kupffer cells
3 release a number of chemicals, including hydrogen peroxide, superoxide ion, and hydrolytic
4 enzymes, which aid in antigen destruction, and also cytokines, which have immunoregulatory
5 and inflammatory activity (e.g., interleukins, TGF» 'and TGF» ). Several of these chemicals
6 have also been associated with apoptosis (see liver cancer discussion) (Laskin, 1996). Treatment
7 of mice with TCE for 30 days has been shown to affect Kupffer cells and the production of
8 inflammatory cells (Kjellstrand et al., 1983). Moreover, peroxisome-proliferating chemicals in
9 general are also reported to affect Kupffer cell function (Youseff and Badr, 1998), and PPAR
10 activation may play a role in autoimmune disease.
11 Immune suppression and TCE exposure are areas needing more research. Immune
12 suppression may play a significant role in the development of non-Hodgkin's lymphoma, as seen
13 in the epidemiologic studies (Wartenberg et al., 2000) and in the mouse bioassay (Henschler et
14 al., 1980). Non-Hodgkin's lymphoma is found with a higher frequency among individuals with
15 compromised immune systems resulting from either viruses or past chemotherapy (Smith, 1996).
16 Although this is not well studied, one report of TCE exposure to female mice via the drinking
1 7 water shows exposure-related inhibition of humoral immunity with effects on bone marrow
18 function, cell-mediated immune responses, and macrophage function (Sanders et al., 1982).
19 Mode-of-action hypotheses for observed developmental effects seen with TCE, TCA, and
20 DCA exposure are not well developed, and data specific to TCE exposure are few.
21 Developmental effects that have been associated with TCE or TCE metabolite exposure include
22 micropthalmia/anopthalmia (eye anomalies) in rats, cardiac defects in rats and humans and,
23 more vaguely, neural tube defects in humans. Micropthalmia has been reported in human
24 offspring with maternal alcohol and retinoic acid exposures. Both retinoic acid and ethanol have
25 in common peroxisome receptor activity. PPAR* •activation may be important to the
26 development of eye anomalies, although no data currently support this hypothesis. In fact,
27 another agent that activates PPAR, DEHP, induces pronounced anopthalmia and micropthalmia
28 in rats (Narotsky et al., 1995a,b).
29 Researchers at the University of Arizona are examining the cardiac teratogenicity of
30 trichloroethylene. Research is focused on evaluating whether gene expression is affected during
31 cardiogenesis by altering several molecules which are hypothesized to be critical for normal heart
32 development. Recently Boyer et al. (2000) have reported that TCE treatment produces a dose-
33 dependent inhibition of mesenchymal cell transformation (a critical event in development of the
34 heart) in progenitors of the valves and septa in the heart in vitro. Boyer et al. (2000) also note a
35 concurrent study that shows alteration of gene expression in rat embryo hearts with maternal
36 exposure in drinking water of 110 ppm TCE. Although questions may surround the experimental
8/1/01 3-43 DRAFT-DO NOT CITE OR QUOTE
-------
1 evidence of whether cardiac anomalies observed in the developmental assays are related to TCE,
2 Boyer et al. (2000) report that TCE can affect events important to the development of the heart,
3 events which are consistent with an induction of cardiac anomalies. Moreover, the TCE
4 metabolites, TCA and DCA, both produce cardiac anomalies in rats (Smith et al., 1989, 1992;
5 Johnson et al., 1998a,b; Epstein et al., 1993). DCA also concentrates in rat myocardial
6 mitochondria (Kerbey et al., 1976), freely crosses the placenta (Smith et al., 1992), and has
7 known toxicity to tissues dependent on glycolysis as an energy source (e.g., the testes, lens, and
8 nervous system in humans and dogs) (Yount et al., 1982; Stacpoole et al., 1979; Katz et al.,
9 1981; Cicmanec et al., 1991). Mammalian embryos also rely on glycolysis for energy, therefore
10 suggesting a common vulnerability for toxicity (Smith et al., 1992). However, there is no direct
11 evidence for DCA to affect glycolysis in the embryo. More research into TCE and its
12 metabolites is needed to more fully elucidate possible modes of action for the effects observed in
13 standard developmental protocols.
14
15 3.6. HAZARD CHARACTERIZATION
16 3.6.1. Consideration of Causality in the Epidemiologic Evidence
17 Overall, the epidemiologic studies suggest an association between TCE exposure and
18 excess risks of kidney cancer, liver cancer, lympho-hematopoietic cancer, cervical cancer, and
19 prostate cancer (Wartenberg et al., 2000). Determining whether an association is one of cause
20 and effect involves considering several aspects that are characteristic of cause-and-effect
21 relationships (Hill, 1965; Rothman, 1986).45 No aspect is either necessary or sufficient; rather,
22 they must be weighed to determine whether the most likely explanation of an observed
23 association is cause and effect.
24 (1) Strength of the observed association. The finding of large, precise risks increases
25 confidence that the association is not likely due to chance, bias, or other factors. A
26 modest risk, however, does not preclude a causal association and may reflect a lower
27 level of exposure or an agent of lower potency. For TCE, the observed risks are generally
28 modest (twofold or less) for cancers of the kidney, liver, and lympho-hematopoietic
29 system. An exception is the large odds ratio for kidney cancer (OR=10.8, 95% CI=
30 3.4-34.8) reported by Vamvakas et al. (1998). Differences between cohort and case-
45EPA's proposed cancer guidelines (U.S. EPA, 1999) use this framework to determine whether a cause-and-effect
interpretation of the epidemiologic evidence is credible. The proposed guidelines may extend this framework to the
separate question of determining whether the mode-of-action information is sufficient to establish cause and effect.
For this question, the key aspect of causality is whether experimental evidence in the laboratory demonstrates that a
specific sequence of key events leads to the observed tumors and that intervening to prevent a key event will prevent
tumor formation.
8/1/01 3-44 DRAFT-DO NOT CITE OR QUOTE
-------
1 control studies, in part, may reflect exposure differences. The observed risks are not
2 thought to be attributable to smoking (important in liver cancer) or socioeconomic status
3 (important in cervical cancer).
4 (2) Consistency of the observed association. An inference of causality is strengthened
5 when a pattern of elevated risks is observed across several independent studies. Some
6 degree of consistency was observed for associations of TCE exposure with cancers of the
7 kidney, liver, lympho-hematopoietic system, cervix, and prostate (Wartenberg et al.,
8 2000). Consistency is strongest for kidney cancer, supported by tier-I, tier-II, and tier-in
9 mortality (but not tier-Hi incidence) studies and case-control studies (see Section 3.4.2).
10 For liver cancer, consistent results were observed in tier-I and tier-II but not tier-in
11 studies (see Section 3.4.1). For lympho-hematopoietic cancer, associations observed in
12 the tier-I studies were supported by the ecologic studies of residential drinking water
13 exposure, though not by the tier-II or tier-Hi studies (see Section 3.4.4). Cervical cancer,
14 though sparsely reported, is elevated in tier-I incidence studies and tier-Hi mortality
15 studies (see Section 3.4.5). Prostate cancer is elevated in the tier-I studies (see Section
16 3.4.5).
17 (3) Specificity of the observed association. Traditionally, specificity has been defined in
18 terms of one cause, one disease (Hill, 1965). TCE causes cancer at several sites in rats
19 and mice; hence, there is no expectation that TCE would be associated with only one
20 human cancer. Many agents cause cancer at multiple sites, and many cancers have
21 multiple causes. Evidence of specificity may come from the recent observation of
22 multiple mutations of the VHL tumor suppressor gene, primarily C>T changes including
23 nucleotide 454, in renal cell carcinoma patients with high, prolonged TCE exposure
24 (Brauch et al., 1999; Briining et al., 1997b) (see Section 3.5.2). Key research related to
25 specificity would include investigation of VHL gene mutations in other cohorts exposed
26 to TCE (see Section 5).
27 (4) Temporal relationship of the observed association. A causal interpretation is
28 strengthened when exposure is known to precede development of the disease.
29 Associations between TCE exposure and several forms of cancer are supported primarily
30 by cohort and case-control studies, in which the temporal relationship is well described.
31 All drinking water studies, except the case-control analysis by MA-DOH (1997), are
32 ecologic or prevalence studies, in which knowledge of the temporal relationship is
33 lacking. For this reason, the conclusions place greater weight on the cohort and case-
34 control studies.
8/1/01 3-45 DRAFT-DO NOT CITE OR QUOTE
-------
1 (5) Biologic gradient (exposure-response relationship). A clear exposure-response
2 relationship strongly suggests cause and effect. For TCE, biologic gradients are only
3 sporadically observed, though the number of site-specific cancers in a study is often too
4 small to identify biologic gradients. For kidney cancer, only the studies by Morgan et al.
5 (1998) and Vamvakas et al. (1998) observed exposure-response relationships.
6 Both the urinary biomarker studies and the Blair et al. (1998) study observed liver cancer
7 incidence to increase with exposure. Anttila et al. (1995) observed incidence to increase with
8 time since first exposure, with the largest risk among those with the longest time since first
9 exposure (SIR=6.1, 95% CI= 2.8-17.7). Similar findings were not observed for liver cancer
10 mortality; however, liver cancer incidence is considered more definitive because liver cancer
11 mortality is extensively miscoded on death certificates.
12 Exposure-response analyses presented in the drinking water studies collectively suggest
13 that greater exposure to drinking water contaminated with TCE and other chlorinated solvents is
14 associated with lymphatic cancer, particularly leukemia. Studies by Cohn et al. (1994), Fagliano
15 et al. (1990), and MA-DOH (1997) reported exposure-response relationships between drinking
16 water exposure to TCE and other chlorinated solvents and the risk of lympho-hematopoietic
1 7 cancer, particularly childhood leukemia and non-Hodgkin's lymphoma.
18 For cervical cancer, Anttila et al. (1995) and Blair et al. (1998) observed incidence to
19 increase with exposure, with a significantly elevated risk among those with the highest urinary
20 TCA levels in the Anttila study (SIR=4.4, 95% CI= 1.4-10.1). Comparison with internal controls
21 in the Blair study suggests that TCE may be the etiologic agent rather than a confounder such as
22 socioeconomic status.
23 (6) Biologic plausibility. Section 3.5 discusses the many recent mechanistic studies
24 investigating TCE's carcinogenic effects in rats and mice and their relevance to humans. The
25 mechanistic results to date indicate that TCE-induced carcinogenesis is complex, involving
26 multiple carcinogenic metabolites acting through multiple mechanisms, all of which have
27 relevance to humans (see Section 3.5).
28 (7) Coherence. A cause-and-effect interpretation of the human data does not conflict
29 with other lines of evidence. The strongest associations between TCE exposure and human
30 cancer are for the kidney, liver, and lympho-hematopoietic system, sites where TCE causes
31 cancer in either rats or mice. In both humans and laboratory animals, TCE can cause adverse
32 effects other than cancer in the kidney, liver, and immune system, and it is biologically plausible
33 that some of these effects may play a role in the development of cancer of the kidney, liver, and
34 lympho-hematopoietic system, respectively (see Sections 3.5.2, 3.5.1, 3.5.4). The associations
35 between human cervical and prostate cancer have no suitable animal models to compare.
8/1/01 3-46 DRAFT-DO NOT CITE OR QUOTE
-------
1 (8) Experimental evidence (from human populations). Experimental evidence is
2 seldom available from human populations (Rothman, 1986) and exists only when conditions of
3 human exposure are altered to create a "natural experiment" at different levels of exposure. For
4 example, leukemia cases became more evenly distributed throughout the town of Woburn after
5 closure of two wells contaminated with TCE and other solvents (MA-DOH, 1997).
6 (9) Analogy. The pattern of effects associated with TCE, particularly in the liver, has
7 similarities to those of several other chlorinated solvents and to mixed-solvent exposures.
8
9 3.6.1.1. Synthesis of the Epidemiologic Evidence
10 Together, these nine aspects of causality suggest that a cause-and-effect association
11 between TCE exposure and cancer is credible. Among the different tumor sites, support is
12 strongest for kidney cancer, followed by liver and lympho-hematopoietic cancer and then cervical
13 and prostate cancer.
14 The strength of this conclusion is tempered by some still-open questions about TCE's
15 cancer potential. The statistical significance of the results of the joint epidemiologic analysis
16 could change with omission of one study or another. Most studies lack quantitative exposure
1 7 information. More research is needed on mutations of the VHL gene, particularly in renal cell
18 carcinoma patients not exposed to TCE. In addition, a nested case-control study could help
19 resolve questions about causality, as this study design has the potential to examine the effects of
20 lifestyle factors and exposures to mixtures of solvents that may share common metabolites and
21 modes of action (see Section 5).
22
23 3.6.2. Weight of Evidence Under the Current and Proposed Cancer Guidelines
24 TCE has been extensively tested in animals, with mice developing liver tumors, lung
25 tumors, and lymphomas, and rats developing kidney tumors and testicular tumors.
26 Epidemiologic studies, considered as a whole, have associated TCE exposure with excess risks
27 of kidney cancer, liver cancer, lympho-hematopoietic cancer, cervical cancer, and prostate
28 cancer. Weight-of-evidence characterizations have had to address issues concerning the strength
29 of the human evidence and the relevance of the animal tumors to humans.
30 With this assessment, the weight of the epidemiologic evidence of TCE's potential
31 carcinogenicity has become stronger than before (see Section 3.4). To bridge the opposing views
32 of the past, this assessment commissioned a joint analysis of the epidemiologic studies
33 (Wartenberg et al., 2000). This joint analysis used a statistically based weight-of-evidence
34 approach that stratified the available studies into tiers according to how well each study's results
8/1/01 3-47 DRAFT-DO NOT CITE OR QUOTE
-------
1 can be associated with TCE exposure specifically.46 Within each tier, study results were
2 weighted according to inverse variance.47 The joint analysis found that the most consistent and
3 compelling results are statistically significant increased incidences of kidney cancer and liver
4 cancer in workers exposed to TCE. These are followed by lympho-hematopoietic cancer,
5 cervical cancer, and prostate cancer. There are few studies of highly exposed populations,
6 consequently, the magnitude of response is generally less than a twofold increase. The joint
7 analysis leads to a conclusion that these increases are statistically significant, thus, TCE may be
8 capable of causing cancer in humans at multiple sites.
9 The epidemiologic studies have recently been augmented by molecular information, in
10 which multiple mutations of the VHL tumor suppressor gene,48 primarily C>T changes including
11 nucleotide 454, were found in renal cell carcinoma patients with high, prolonged TCE exposure
12 (Briining et al., 1997b) (see Section 3.5.2). A followup study found mutations of the VHL gene
13 in 75% of TCE-exposed workers with renal cell carcinoma, with an association between number
14 of mutations and TCE exposure level (Brauch et al., 1999). Adding specificity to this
15 association, a C>T change at nucleotide 454 was found in 13 of 33 renal cell carcinoma patients
16 with TCE exposure; this specific mutation was not found in those without identifiable TCE
17 exposure.
18 The mechanistic research into the mode-of-action for each animal tumor site has begun to
19 link TCE with disturbances in cell signaling and carbohydrate metabolism, which can lead to
20 human cancer and other diseases (see Section 3.5). Subject to dose-response adjustments for
21 relative human-to-animal sensitivity (see Sections 4.5.2, 4.5.3), this research makes it plausible
22 that TCE acts through mechanisms that can cause cancer in humans.
23 EPA's proposed cancer guidelines provide that molecular information can be used in
24 choosing a weight-of-evidence descriptor. TCE could be described as "carcinogenic to humans,"
25 supported by (1) association of TCE exposure with increased risk of human kidney cancer, liver
46Tier I studies are those in which TCE exposure has been inferred for individual study subjects and in which it is
best characterized. Tier II studies are those in which there is putative TCE exposure, but individuals are not
identified as uniquely exposed to TCE. Tier III studies are the studies of dry cleaning and laundry workers in which
subjects are exposed to a variety of solvents including TCE (Wartenberg et al., 2000).
41 Inverse variance gives greater weight to larger studies that yield more precise estimates. Study results are
averaged, with each study result weighted (i.e., multiplied) by a factor that is inversely proportional to the study
result's variance.
48The VHL gene normally suppresses renal cell carcinomas. Mutations to this gene have been noted in kidney
cancers and may be an important risk factor and mode-of-action for chemically induced renal cell cancer (Lash et al.,
2000b). To illustrate the difference between "mode" of action and "mechanism" of action, knowledge that VHL
gene mutations are involved may be enough to identify a plausible "mode" of action, whereas knowing how such
modifications induce subsequent events leading to kidney cancer would be needed to identify the "mechanism" of
action.
8/1/01 3-48 DRAFT-DO NOT CITE OR QUOTE
-------
1 cancer, lympho-hematopoietic cancer, cervical cancer, and prostate cancer; (2) induction of
2 kidney tumors and testicular tumors in rats and liver tumors, lung tumors, and lymphomas in
3 mice; and (3) induction of multiple VHL gene mutations, primarily OT changes including
4 nucleotide 454, in a human population. At this time, however, fuller identification of the mode-
5 of-action and associated key events is lacking (see Section 5). It is not known whether the OT
6 change at nucleotide 454 is a marker of exposure, a key event leading to kidney cancer, or
7 whether other factors selectively favor the growth of cells with this change. Functional alteration
8 of the VHL gene can be a mode of kidney cancer induction; whether this is how TCE induces
9 kidney cancer has not been established.
10 Alternatively, TCE could be described as "likely to be carcinogenic to humans,"
11 supported by evidence associating TCE exposure with human cancer and strong evidence of
12 carcinogenicity in animals involving processes that have relevance for humans. Among "likely"
13 carcinogens, the evidence for TCE is strong and is drawn from several types of information (see
14 Sections 3.2-3.5).
15 At this time, a strong characterization as "highly likely to be carcinogenic to humans" is
16 more appropriate, given the still-open questions about TCE's cancer potential. The statistical
1 7 significance of the results of the joint epidemiologic analysis could change with omission of one
18 study or another. This does not decrease confidence in the results of the joint analysis, but
19 emphasizes that the TCE studies collectively have only recently accumulated enough power to
20 begin to detect associations between moderate TCE exposures and some relatively common
21 cancers. On the other hand, the new evidence associating TCE exposure with a transformation at
22 nucleotide 454 is striking evidence specific to TCE exposure, putting a "genetic fingerprint" that
23 associates such kidney tumors with TCE exposure. Replication of this result in another cohort
24 showing kidney tumors could warrant a future description of TCE as "carcinogenic to humans"
25 (see Section 5).
26 Under the current (1986) cancer guidelines, TCE would be classified as a "probable
27 human carcinogen" (group Bl), with "limited" human evidence and "sufficient" animal evidence
28 of carcinogenicity. The principal changes from the controversy of the late 1980s are the stronger
29 weight of epidemiologic evidence (see Section 3.4) and the new mechanistic information
30 suggesting that TCE's modes of action may be relevant to humans (see Section 3.5).
8/1/01 3-49 DRAFT-DO NOT CITE OR QUOTE
-------
Table 3-1. Liver tumors in mice exposed to TCE
Hepatocellular carcinomas and adenomas after 103 weeks gavage exposure, beginning at 8
weeks of ?ge (NTP, 1990)
Administered dose (mg/kg-d) Vehicle ctl 1,000
Male B6C3F1 mice ** 14/48 (29%) 39/50 (78%)
Female B6C3F1 mice** 6/48 (13%) 22/49 (45%)
Hepatocellular carcinomas" after 90 weeks gavage exposure, beginning at 5 weeks of age
(NCI, 1976)
Dose group Vehicle ctl Low doseb High dose0
Male B6C3F1 mice ** 1/20 (5%) 26/50 (52%) 31/48 (65%)
Female B6C3F1 mice** 0/20(0%) 4/50(8%) 11/47(23%)
Hepatomas after 78 weeks inhalation exposure, beginning at 11-12 weeks of age (Maltoni et
al., 1986)
Administered daily concentration
(mg/m3)" Control 112.5 337.5 675
Male Swiss mice6 ** 4/66(6%) 2/53(4%) 8/59(14%) 13/61(21%)
Female Swiss mice6 0/84(0%) 0/89(0%) 0/86(0%) 1/86(1%)
MaleB6C3Flmicef** 1/59(2%) 1/31(3%) 3/38(8%) 6/37(16%)
Female B6C3F1 micef** 3/88(3%) 4/89(4%) 4/88(5%) 9/85(11%)
MaleB6C3Flmiceg 17/77 (22%) 19/47 (40%) 27/67 (40%) 21/63 (33%)
**Statistically significant by Cochran-Armitage trend test (p<0.05).
aHepatocellular adenomas were not reported.
bLow dose is 1,200 mg/kg-d for male mice, 900 mg/kg-d for female mice (5 d/wk).
cHigh dose is 2,400 mg/kg-d for male mice, 1,800 mg/kg-d for female mice (5 d/wk).
Equivalent to 100, 300, 600 ppm (100 ppm=540 mg/m3), adjusted for 7 hr/d, 5 d/wk exposure.
6Mice alive at week 43, when first hepatoma was observed.
^ice alive at week 33, when first hepatoma was observed (expt BT306).
8Mice alive at week 68, when first hepatoma was observed (expt BT306 bis).
Sources: NTP (1990) tables 8, 9; NCI (1976) table VIE; Maltoni et al (1986) IV/VI table 14,
IV/VII table 14, IV/Vin table 14.
8/1/01 3-50 DRAFT-DO NOT CITE OR QUOTE
-------
Table 3-2. Kidney tumors in rats, adjusted for reduced survival"
Tubular cell adenomas and adenocarcinomas after 103 weeks gavage exposure, beginning
at 6.5-8 weeks of age (NTP, 1988,1990)
Administered dose (mg/kg-d)
Male ACT rats
Male August rats
Male Marshall rats
Male Osborne-Mendel rats **
Male F344/N rats**
Pooled males **
Female ACT rats
Female August rats
Female Marshall rats
Female Osborne-Mendel rats
Female F344/N rats
Pooled females **
Untreated ctl
0/48 (0%)
0/47 (0%)
2/43 (4%)
0/46 (0%)
0/48 (0%)
2/232 (1%)
0/46 (0%)
0/47 (0%)
1/48 (2%)
1/47 (2%)
0/46 (0%)
2/234 (1%)
Vehicle ctl
0/46 (0%)
0/47 (0%)
0/44 (0%)
0/47 (0%)
0/46 (0%)
0/230 (0%)
0/43 (0%)
1/47 (2%)
1/49 (2%)
0/43 (0%)
0/46 (0%)
2/228 (1%)
500
1/29 (3%)
2/35 (6%)
1/25 (4%)
6/44 (14%)
2/46 (4%)
12/179 (7%)
3/36 (6%)
4/42 (10%)
2/44 (5%)
0/44 (0%)
0/45 (0%)
9/211 (4%)
1,000
0/22 (0%)
1/33 (3%)
1/28 (4%)
2/33 (6%)
3/33 (9%)
7/149 (5%)
1/31 (3%)
0/34 (0%)
i/oo /"?0/\
i/j2 \J /o)
1/45 (2%)
1/44 (2%)
4/186(2%)
Renal tubuli adenocarcinomas after 104 weeks inhalation exposure, beginning at 12 weeks
of age (Maltoni et al., 1986)
Administered daily
concentration (mg/m3)b
M Sprague-Dawley rats **
F Sprague-Dawley rats
Control
0/120 (0%)
0/139 (0%)
112.5
0/118(0%)
0/128 (0%)
337.5
0/116(0%)
0/127 (0%)
675
4/122 (3%)
1/127 (1%)
**Statistically significant by Cochran-Armitage trend test (p<0.05).
aACI, August, Marshall, Osborne-Mendel, and F344/N rats alive at week 57, Sprague-Dawley
rats at week 47 (expt BT304) or week 62 (expt BT304 bis).
Equivalent to 100, 300, 600 ppm (100 ppm=540 mg/m3), adjusted for 7 hr/d, 5 d/wk exposure.
Sources: NTP (1988) Tables A2, C2, E2, G2; NTP (1990) table A3; Maltoni et al. (1986) IV/IV
Table 19, IV/V Table 19.
8/1/01
3-51
DRAFT-DO NOT CITE OR QUOTE
-------
Table 3-3. Malignant lymphomas in mice exposed to TCE
Malignant lymphomas after 103 weeks gavage exposure, beginning at 8 weeks of age (NTP,
1990)
Administered dose (mg/kg-d)
Vehicle ctl
1,000
MaleB6C3Fl mice
Female B6C3F1 mice
11/50(22%) 13/50(26%)
7/48(15%) 13/49(27%)
Lymphosarcomas and reticulum cell sarcomas after 90 weeks gavage exposure, beginning
at 5 weeks of age (NCI, 1976)
Dose group
Vehicle ctl Low dosea
High dose"
MaleB6C3Fl mice
Female B6C3F1 mice
1/20 (5%) 4/50 (8%) 2/48 (4%)
1/20 (5%) 5/50 (10%) 5/47 (11%)
Malignant lymphomas after 78 weeks inhalation exposure (Henschler et al., 1980)
Administered daily concentration
(mg/m3)'
Control
96
480
Male Han:NMRI mice
Female Han:NMRI mice **
7/30 (23%) 7/29 (24%) 6/30 (20%)
9/29(31%) 17/30(57%) 18/28(64%)
**Statistically significant by Cochran-Armitage trend test (p<0.05).
aLow dose is 1,200 mg/kg-d for male mice, 900 mg/kg-d for female mice (5 d/wk).
bHigh dose is 2,400 mg/kg-d for male mice, 1800 mg/kg-d for female mice (5 d/wk).
Equivalent to 100 and 500 ppm (100 ppm=540 mg/m3), adjusted for 6 hr/d, 5 d/wk exposure.
Sources: NTP (1990) tables 8, 9; NCI (1976) table XXXa; Henschler et al (1980) table 3a.
8/1/01
3-52
DRAFT-DO NOT CITE OR QUOTE
-------
Table 3-4. Testicular tumors in male rats exposed to TCE, adjusted for reduced survival3
Interstitial cell tumors after 103 weeks gavage exposure, beginning at 6.5-8 weeks of age
(NTP, 1988,1990)
Administered dose (mg/kg-d)
Male ACT rats
Male August rats
Male Marshall rats **
Male Osborne-Mendel rats
Male F344/N rats
Untreated ctl
38/45 (84%)
36/46 (78%)
16/46 (35%)
1/30 (3%)
44/47 (940/0)
Leydig cell tumors after 104 weeks inhalation
(Maltoni et al., 1986)
Administered daily
concentration (mg/m3)b
M Sprague-Dawley rats **
Control
6/114(5%)
Vehicle ctl
36/44 (82%)
34/46 (74%)
17/46 (37%)
0/28 (0%)
47/48 (98%)
500
23/26 (88%)
30/34 (88%)
21/33 (64%)
0/25 (0%)
47/48 (98%)
1,000
17/19 (89%)
26/30 (87%)
32/39 (82%)
1/19 (5%)
32/44 (73%)
exposure, beginning at 12 weeks of age
112.5
16/105 (15%)
337.5
30/107 (28%)
675
31/113(27%)
**Statistically significant by Cochran-Armitage trend test (p<0.05).
aACI rats alive at week 70, August rats at week 65, Marshall rats at week 32, Osborne-Mendel
rats at week 97, F344/N rats at week 32, Sprague-Dawley rats at week 81 (expt BT304) or week
62(exptBT304bis).
Equivalent to 100, 300, 600 ppm (100 ppm=540 mg/m3), adjusted for 7 hr/d, 5 d/wk exposure.
Sources: NTP (1988) Tables A2, C2, E2, G2; NTP (1990) Table A3; Maltoni et al. (1986) IV/IV
Table 21, IV/V Table 21.
8/1/01
3-53
DRAFT-DO NOT CITE OR QUOTE
-------
Table 3-5. Lung tumors in mice exposed to TCE
Pulmonary adenomas and adenocarcinomas after 78 weeks inhalation exposure, beginning
at 11-12 weeks of age (Maltoni et al., 1986)
Administered daily concentration
(mg/m3) Control 112.5 337.5 675
Male Swiss mice** 10/90(11%) 11/90(12%) 23/90(26%) 27/90(30%)
Female Swiss mice 15/90(17%) 15/90(17%) 13/90(14%) 20/90(22%)
Male B6C3F1 mice 2/90(2%) 2/90(2%) 3/90(3%) 1/90(1%)
Female B6C3F1 mice** 4/90(4%) 6/90(7%) 7/90(8%) 15/90(17%)
Pulmonary adenomas and adenocarcinomas by inhalation (Fukuda et al., 1983)
Administered daily concentration
(mg/m3) Control 56 168 504
Female ICR mice** 6/49(12%) 5/50(10%) 13/50(26%) 11/46(24%)
Carcinomas, adenocarcinomas, and adenomas of the lung or alveoli after 90 weeks gavage
exposure, beginning at 5 weeks of age (NCI, 1976)
Dose group Vehicle ctl Low dosea High doseb
Male B6C3F1 mice 0/20 (0%) 5/50 (10%) 2/48 (4%)
Female B6C3F1 mice 1/20 (5%) 4/50 (8%) 7/47 (15%)
**Statistically significant by Cochran-Armitage trend test (p<0.05).
aLow dose is 1,200 mg/kg-d for male mice, 900 mg/kg-d for female mice (5 d/wk).
bHigh dose is 2,400 mg/kg-d for male mice, 1,800 mg/kg-d for female mice (5 d/wk).
Sources: Maltoni et al., 1986, Tables 48, 51; Fukuda et al. (1983); NCI (1976) Table XXV.
8/1/01 3-54 DRAFT-DO NOT CITE OR QUOTE
-------
H ' Cl
TCE
GST pathway
\Gxidative
metabolism
P450
CH a,
ALDH
TCA
CLC —C,
i
T
OH
0
I I
C C,
ADH
(Cytochrome P450?)
TCOH
CLCCH..OH
DCVG
0 -
DCVC
Beta-Lyase / "\\ NAT
Activation
Detoxification
' OH
DCA
glucuronide
Figure 3-1. Metabolism of Trichloroethylene (TCE).
8/1/01
3-55 DRAFT-DO NOT CITE OR QUOTE
-------
4. DOSE-RESPONSE ASSESSMENT AND CHARACTERIZATION
1 4.1. GENERAL CONSIDERATIONS AND APPROACH
2 This dose-response assessment updates the cancer slope factor for TCE and develops an
3 oral RfD and an inhalation RfC for effects other than cancer. These effects include
4 neurotoxicity, immunotoxicity, developmental toxicity, liver toxicity, kidney toxicity, endocrine
5 effects, and several forms of cancer (see Section 3.4). The overall approach is to develop
6 alternative estimates using different studies and modeling approaches, then to choose values that
7 are supported by several sources of data and lines of reasoning. This provides a measure of
8 confidence that the risk estimates are robust and not likely to be substantially changed by a new
9 study or analysis.
10 This dose-response assessment draws from the state-of-the-science papers that appear in
11 volume 1. Pharmacokinetic models49 were developed to provide dose metrics50 for the RfD, RfC,
12 and cancer assessments (Fisher, 2000; Clewell et al., 2000). Statistical analyses improved these
13 models by calibrating them to fit more data sets, then quantifying the uncertainty in each dose
14 metric (Bois, 2000a,b). Analyses supporting the RfD and RfC applied pharmacokinetic models
15 and empirical dose-response models51 whenever suitable data were available (Barton and
16 Clewell, 2000). The cancer assessment pursued several lines of analysis and considered both
17 linear and nonlinear approaches, based on knowledge about mode-of-action at each tumor site
18 (see Section 3.5). Empirical dose-response models were fitted to data on liver tumors and lung
19 tumors in mice and kidney tumors and testicular tumors in rats (Rhomberg, 2000). A
20 mechanism-based dose-response model52 was developed for testing hypotheses about the mouse
21 liver tumor response (Chen, 2000). Slope factors for kidney tumors, liver tumors, and
22 non-Hodgkin's lymphoma were also developed from the epidemiologic data.
49Pharmacokinetic models simulate the relationship between external exposure levels and the biologically effective
dose at a target tissue. Pharmacokinetic models take into account absorption, distribution, metabolism, and
elimination of the administered chemical and its metabolites. Pharmacodynamic models simulate the relationship
between a biologically effective dose and the occurrence of a disease response.
50Dose metrics are alternative ways of describing dose. They involve what is being measured (parent compound,
metabolite, or biomarker), where it is measured (whole body, blood, or specific tissue), and how it is measured
(cumulative, average, or peak exposure). For example, total administered TCE and average TCA in liver are dose
metrics.
51An empirical dose-response model is one based on fitting a curve to data. (They are sometimes called "curve-
fitting" models or "benchmark dose" models.) In contrast to mechanism-based dose-response models, an empirical
model is not based on specific knowledge about the biological mechanisms leading to disease.
52A mechanism-based dose-response model (also called a biologically based dose-response model) is a
pharmacodynamic model whose mathematical structure reflects the ascertained mode-of-action and whose
parameters are measured experimentally.
8/1/01 4-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 The assessment for effects other than cancer follows EPA's methods for developing RfDs
2 (Barnes and Dourson, 1988) and RfCs (U.S. EPA, 1994). Despite the existence of numerous
3 studies, however, the database for developing an RfD or RfC is problematic. ATSDR did not
4 derive analogous chronic-duration levels53 for TCE, viewing the chronic studies as limited by
5 inadequate characterization of exposure, inadequate quantification of results, or lack of endpoints
6 suitable for deriving chronic levels (ATSDR, 1997). This assessment, through modeling and
7 comparing results for different adverse effects, species, exposure durations, and exposure routes,
8 has overcome some data gaps and developed an RfD and RfC that are supported by multiple
9 lines of reasoning.
10 The cancer assessment is consistent with the approach of EPA's proposed cancer
11 guidelines (U.S. EPA, 1996). These guidelines encourage use of models to incorporate a wider
12 range of experimental data and use of mode-of-action information to guide modeling and
13 extrapolation approaches. This assessment develops alternative risk values from many sources
14 before comparing and reconciling them. Because information about sensitive populations,
15 children, and cumulative risks indicates the potential for vast human variation in risk, a range of
16 cancer estimates is described, with guidance on how to choose a particular estimate based on risk
17 factors in the exposed population and exposure scenario.
18
19 4.2. DOSIMETRY MODELING
20 This dose-response assessment begins by expressing doses in common terms to facilitate
21 better comparisons across studies, health effects, species, exposure durations, and exposure
22 routes. Reflecting the importance of metabolism to TCE-induced toxicity, doses are estimated
23 from pharmacokinetic models whenever data are suitable. Otherwise, doses are scaled from
24 animals to humans based on equivalence of mg/kg3/4-d (U.S. EPA, 1992). Pharmacokinetic
25 models can account for high- to low-dose nonlinearity and animal-to-human differences in
26 metabolism. The use of pharmacokinetic modeling and dose scaling in developing RfDs is
27 relatively new, representing a move toward harmonizing RfD methods with those already in use
28 for RfCs and cancer assessments.
29
53ATSDR calls these minimal risk levels (MRLs), defined as "an estimate of the daily human exposure to a hazardous
substance that is likely to be without appreciable risk of adverse noncancer health effects over a specified duration of
exposure" (ATDSR, 1997).
8/1/01 4-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 4.2.1. Pharmacokinetic Modeling
2 A pharmacokinetic model for ingested TCE in male mice is described in a state-of-the-
3 science paper by Fisher (2000). The model predicts area-under-the-curve54 in the liver for TCA
4 and DCA. Compartments55 include TCE in liver, lung, kidney, fat, gut, rapidly perfused tissues,
5 and slowly perfused tissues. Metabolite submodels include CH, TCA, TCOH, TCOG, and DCA
6 in liver, lung, kidney, and the rest of the body; DCVC and DCVG were not modeled. The model
7 simplifies metabolism by modeling it only in the liver. Model parameters were taken from
8 published literature, measured experimentally, or fitted by running the model with data at 1,200
9 mg/kg. The model was validated against data at 300, 600, and 2,000 mg/kg, suggesting the
10 ability of the model to make predictions below the dose where it was developed. Parameters
11 were scaled according to body weight. This mouse model was then adapted for inhaled TCE in
12 human males and females. The human model does not include CH and DCA, because CH was
13 not detected in experimental subjects and DCA was detected only intermittently.
14 Another pharmacokinetic model for ingested and inhaled TCE in male mice, rats, and
15 humans is described in a state-of-the-science paper by Clewell et al. (2000). This model predicts
16 metabolite areas-under-the-curve in plasma. Compartments include TCE in liver, lung,
17 tracheobronchial region, fat, gut, rapidly perfused tissues, and slowly perfused tissues. One-
18 compartment metabolite submodels include TCA, TCOH, TCOG, and DCA. The model also
19 includes descriptions of GST metabolism in the kidney, CH metabolism and clearance in the
20 lung, GST metabolism of TCE to DCVC in the liver, activation of • *lyase and clearance by NAT
21 in the kidney, biliary excretion of TCOG, and enterohepatic recirculation of TCOH. Model
22 parameters were taken from published literature or fitted by running the model with selected data
23 sets. Parameters were scaled allometrically.56
24 Statistical analyses of these pharmacokinetic models are discussed in two additional state-
25 of-the-science papers (Bois, 2000a,b). These analyses use a Bayesian57 statistical framework and
S4Area-under-the-curve refers to the tissue concentration of a chemical integrated over time. It is a measure of the
long-term average concentration of a chemical in a specific tissue.
55Pharmacokinetic models specify compartments to represent different tissues or groups of tissues, for example,
liver, fat, and the rest of the body.
56Allometric scaling is a method for calculating equivalent doses across species. It scales volumes according to body
weight, blood flow rates and metabolic capacities according to the 3/4 power of body weight, and rate constants
according to the inverse 1/4 power of body weight.
51 Bayesian statistical methods regard a model's parameters as random variables with a probability distribution. New
data can alter what is thought to be the best distribution for describing each parameter. The distribution based only
on prior information and assumptions is called the prior distribution. Analysis of new data yields a posterior
distribution that reconciles the prior information and assumptions with the new data.
8/1/01 4-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 Markov-chain Monte-Carlo58 simulation to refine the models by using more data sets to estimate
2 each model's parameters. The result is a set of calibrated models that better fits a wider range of
3 experimental data. In some cases the calibrated parameters are quite different from the originals,
4 thus, substantial information has been gained by fitting the models to additional experimental
5 data sets. Dose estimates from the calibrated models were used in subsequent state-of-the-
6 science papers (Chen, 2000; Rhomberg, 2000) supporting the cancer assessment, whereas dose
7 estimates from the Clewell et al. (2000) pharmacokinetic model were used in the state-of-the-
8 science paper supporting the RfD and RfC (Barton and Clewell, 2000).
9 EPA's National Exposure Research Laboratory is developing a Dose Estimating
10 Exposure Model, a model architecture that simulates internal doses for different chemicals and
11 species. As part of this assessment, this model has been configured for TCE, initially following
12 the structure of the Fisher and Clewell models. This new TCE model is being calibrated by
13 comparing its predictions to several experimental data sets. A notable feature of this model is its
14 capacity to simulate simultaneous exposure to TCE, TCA, and DCA. This will be particularly
15 useful in future site-specific applications (for example, at Superfund sites or in assessing the
16 safety of a drinking water supply) where people are exposed to TCE along with TCA and DCA
1 7 from other solvents or other sources.
18
19 4.2.2. Uncertainty Analyses
20 Another product of the statistical analyses is a quantitative description of uncertainty in
21 the dose estimates (Bois, 2000b). Table 4-1 summarizes distributions of potential dose metrics
22 for modeling liver, lung, and kidney tumors from the calibrated model (Bois, 2000b) on the basis
23 of the Clewell et al. (2000) pharmacokinetic model, the only model to calculate estimates outside
24 the liver. From Table 4-1 it is apparent that lung and kidney dose metrics are quite uncertain,
25 with 95% confidence intervals spanning more than a 5,000-fold range. This may be due to the
26 difficulty in measuring and modeling a transient metabolite such as CH, the possibility that
27 plasma CH may not be a good surrogate for CH accumulated in Clara cells of the lung, or the
28 poorer database for estimating GST metabolites in the kidney. Whatever the reason, this is more
29 uncertainty than is generally present in risk analyses supporting public health decisions. One
30 approach to addressing this uncertainty would be to take lung and kidney dose estimates from the
31 lower end, perhaps the first percentile, of their distributions; this would effectively increase risk
^Markov-chain Monte-Carlo simulation is an iterative process that can be used within a Bayesian statistical
framework to (1) sample each model's parameters from their prior distributions, (2) fit that model with the sampled
parameters to several additional experimental data sets, and (3) compare the model's predictions with the
experimental results to obtain posterior distributions for the model's parameters that improve the model's fit. These
steps are repeated thousands of times until each parameter's posterior distribution converges to a more robust
distribution that reflects a wider database.
8/1/01 4-4 DRAFT-DO NOT CITE OR QUOTE
-------
1 estimates several hundred-fold above the median. Rather than make such a large adjustment, this
2 assessment uses default RfC dosimetry models (U.S. EPA, 1994) when modeling lung tumors by
3 inhalation and the default 3/4-power scaling factor when modeling kidney tumors by ingestion.
4 Liver dose metrics, in contrast, are less uncertain (see Table 4-1); consequently, the cancer
5 assessment uses the calibrated model's median dose estimates for the liver. In this way,
6 uncertainty analysis is used to distinguish between uncertain applications (lung and kidney) and
7 more robust applications (liver), so that pharmacokinetic modeling is used when the results are
8 robust and other methods are considered when there is too much uncertainty.
9 The statistical analyses also reveal several differences between males and females.
10 Females have a significantly lower alveolar ventilation rate (beyond that explained by allometric
11 scaling), higher TCOH body-to-blood partition coefficient, lower TCA body-to-blood partition
12 coefficient, higher Vmax/Km ratio for TCOH glucuronidation, higher conversion of TCOH to
13 TCA, and higher urinary excretion of TCA (Bois, 2000a). These insights are possible because
14 Fisher (2000) collected and modeled data from both males and females. The number of variables
15 with significant differences shows the complexity of overall human variation.
16
1 7 4.2.3. Route Extrapolation
18 Pharmacokinetic models can be used to investigate questions about extrapolation from
19 one exposure route to another. This can be done by determining oral and inhalation exposures
20 that yield equal internal doses of the active agent.
21 For this assessment, the pharmacokinetic model developed at EPA's National Exposure
22 Research Laboratory calculated that chronic oral exposure to TCE at 1 mg/kg-d yields similar
23 TCA area-under-the-curve as chronic inhalation exposure to TCE at 75 mg/m3. This relationship
24 was investigated at several dose levels near 1 mg/kg-d and found to be approximately linear at
25 this dose and below. The relationship is different, however, when matching DCA area-under-
26 the-curve as the internal dose. Chronic oral exposure to TCE at 1 mg/kg-d yields similar DCA
27 area-under-the-curve as chronic inhalation exposure to TCE at 3 mg/m3 (instead of 75 mg/m3).
28 This 25-fold uncertainty highlights the importance of further research to identify the appropriate
29 internal dose metric for each toxic effect (see Section 5).
30
31 4.3. ORAL REFERENCE DOSE FOR EFFECTS OTHER THAN CANCER
32 For effects other than cancer, risk assessments have not attempted to describe dose-
33 response curves in the range of environmental exposures. Rather, the focus has been to estimate
34 an exposure level where there is little concern for adverse effects. This RfD is derived through a
35 process of (1) considering all studies and selecting the critical effects that occur at the lowest
8/1/01 4-5 DRAFT-DO NOT CITE OR QUOTE
-------
1 dose, (2) selecting a dose (or point of departure59) at which the critical effect either is not
2 observed or would occur at a relatively low incidence (for example, 10%), and (3) reducing this
3 dose by uncertainty factors to reflect differences between study conditions and conditions of
4 human environmental exposure.
5
6 4.3.1. Critical Effects
7 Many different toxic effects are associated with oral TCE exposure, as TCE can disrupt
8 fundamental cellular processes through multiple metabolites and mechanisms (see Section 3.4).
9 At higher doses (above approximately 100 mg/kg-d), targets of oral TCE toxicity include the
10 liver, kidney, nervous system, reproductive system, and developing fetus (ATSDR, 1997).
11 Effects have been observed at these sites in acute studies (2 weeks or less), intermediate studies,
12 or chronic studies (1 year or more). At lower doses (below approximately 100 mg/kg-d), there
13 are fewer studies, but effects continue to be observed in several systems, often in subchronic
14 studies. At the lowest doses tested (approximately 1-10 mg/kg-d), effects are observed in the
15 liver, kidney, and developing fetus. These are considered to be the critical effects of oral TCE
16 exposure.
17
18 4.3.2. Point of Departure
19 Table 4-2 summarizes the results of oral studies conducted at the lower doses. To
20 compare studies, doses are expressed in human-equivalent terms. For liver effects,
21 pharmacokinetic modeling (Clewell et al., 2000) was used to estimate plasma TCA as the dose
22 metric (Barton and Clewell, 2000). For kidney effects, statistical analyses (Bois, 2000b) revealed
23 substantial parameter uncertainty in the pharmacokinetic modeling (see Table 4-1), consequently,
24 human-equivalent doses were based on equivalence of mg/kg3/4-d (U.S. EPA, 1992). Response
25 levels are presented as either a NOAEL,60 a LOAEL if the study did not identify a NOAEL, or a
26 modeled LED10 if the study results were suitable for modeling (Barton and Clewell, 2000).
27 Adverse effects have been observed in several studies at a human-equivalent dose range
28 of 1-10 mg/kg-d. For the rat kidney there is a chronic NOAEL at 10 mg/kg-d (Maltoni et al.,
59Point of departure denotes a dose at the lower end of the observed dose-response curve where extrapolation to
lower doses begins. For effects other than cancer, the point of departure is either a NOAEL (no-observed-adverse-
effect level), a LOAEL (lowest-observed-adverse-effect level) if no NOAEL can be identified, or a modeled point
(for example, an LED10 or LED01) if the data are suitable for curve-fitting. For cancer, the point of departure is an
LED10 (or lower point if one can be reliably estimated) for tumors, or for a key tumor precursor when there is
information to describe the mode-of-action.
60A NOAEL (no-observed-adverse-effect level) is the highest experimental dose without a statistically or biologically
significant effect. Note that at this dose there may be effects that are not biologically significant (/'. e., not judged
adverse), or there may be biologically significant effects that are not statistically significant. Statistical significance
is affected by study design and sample size, consequently, a NOAEL in one study may show a statistically significant
adverse effect in another study with a different design or larger sample.
8/1/01 4-6 DRAFT-DO NOT CITE OR QUOTE
-------
1 1986). For the liver there is a subchronic NOAEL at 1 mg/kg-d in mice (Tucker et al., 1982), a
2 subchronic LOAEL at 1 mg/kg-d in mice (Buben and O'Flaherty, 1985), and a subchronic LED10
3 of 0.6 mg/kg-d in rats (Berman et al., 1995). Among these liver values, 1 mg/kg-d is chosen as
4 the point of departure, supported by these three studies as a dose where liver toxicity can begin to
5 be observed in two species after subchronic dosing. For cardiac anomalies in developing rats,
6 there is a LOAEL at 34 mg/kg-d and a NOAEL at 0.05 mg/kg-d (Dawson et al., 1993). The 700-
7 fold difference between this LOAEL and NOAEL reflects the wide dose spacing in this study,
8 which makes use of this endpoint as a point of departure highly uncertain.
9
10 4.3.3. Uncertainty Factors
11 To address differences between study conditions and conditions of lifetime human
12 environmental exposure, the point of departure is reduced by uncertainty factors61 that reflect
13 different areas of uncertainty.
14 Several studies have identified similar points of departure for adverse liver effects,
15 indicating some degree of confidence in using this effect as a basis for calculating an RfD. There
16 are several sources of uncertainty in using these studies.
1 7 (a) Human variation: The NOAELs, LOAELs, and LED10s for adverse liver effects (see
18 Section 4.3.2, Table 4-2) were estimated using a pharmacokinetic model (Barton and
19 Clewell, 2000; Clewell et al., 2000). The parameter uncertainty in these modeled dose
20 estimates (estimated between the 50th and 99th percentiles, see Section 4.2.2, Table 4-1)
21 is 15-fold if plasma TCA is used as the dose metric and 20-fold if plasma DCA is used
61RfDs apply to lifetime human environmental exposure, including sensitive subgroups. Differences between study
conditions and conditions of human environmental exposure may make a dose that appears safe in an experiment not
be safe in the environment. Uncertainty factors account for differences between study conditions and conditions of
human environmental exposure. These include:
(a) Variation from average humans to sensitive humans: RfDs apply to the human population, including
sensitive subgroups, but studies rarely target sensitive humans. Sensitive humans could be adversely
affected at lower doses than a general study population; consequently, general-population NOAELs are
reduced to cover sensitive humans.
(b) Uncertainty in extrapolating from animals to humans: If an RfD is developed from animal studies, the
animal NOAEL is reduced to reflect pharmacokinetic and pharmacodynamic factors that may make humans
more sensitive than animals.
(c) Uncertainty in extrapolating from subchronic NOAELs to chronic NOAELs: RfDs apply to lifetime
exposure, but sometimes the best data come from shorter studies. Lifetime exposure can have effects that
do not appear in a shorter study; consequently, a safe dose for lifetime exposure can be less than the safe
dose for a shorter period. If an RfD is developed from less-than-lifetime studies, the less-than-lifetime
NOAEL is adjusted to estimate a lifetime NOAEL.
(d) Uncertainty in extrapolating from LOAELs to NOAELs: RfDs estimate a dose without appreciable risks,
but sometimes adverse effects are observed at all study doses. If an RfD is developed from a dose where
there are adverse effects, that dose is adjusted to estimate a NOAEL.
(e) Other factors to reflect professional assessment of scientific uncertainties not explicitly treated above,
including completeness of the overall database, minimal sample size, or poor exposure characterization.
8/1/01 4-7 DRAFT-DO NOT CITE OR QUOTE
-------
1 (Bois, 2000b, Rhomberg, 2000). Not known is whether either of these dose metrics is an
2 appropriate indicator of liver toxicity. This pharmacokinetic variation is complemented
3 by pharmacodynamic variation (Renwick, 1998). In the absence of data to quantify
4 human pharmacodynamic variation, a default factor of 101/2 62 has been suggested
5 (Renwick, 1998). Multiplying 15-to 20-fold pharmacokinetic variation by 101/2-fold
6 pharmacodynamic variation indicates overall human variation of about 50-fold (rounded
7 to one significant digit).
8 (b) Animal-to-human uncertainty: This aspect of uncertainty can also be viewed in terms of
9 pharmacokinetic and pharmacodynamic components. Animal-to-human pharmacokinetic
10 uncertainty is reflected in the 15- to 20-fold factor just discussed. Animal-to-human
11 pharmacodynamic uncertainty is covered by a default factor of 101/2, similar to the
12 practice used for deriving RfCs (U.S. EPA, 1994).
13 (c) Subchronic-to-chronic uncertainty: The NOAEL, LOAEL, and LED10 for adverse liver
14 effects (see Section 4.3.2, Table 4-2) were based on subchronic exposures. When
15 subchronic studies are used to derive an RfD for lifetime exposure, an uncertainty factor
16 of up to 10-fold is generally applied. Although duration-response trends are not evident
17 in the animal studies, recent human studies have found statistically significant duration-
18 response trends for decreased testosterone, decreased FSH, decreased SHBG, and
19 increased DHEA (Chia et al., 1997; Goh et al., 1998). Duration-response trends have
20 also been reported for central nervous system toxicity (Rasmussen et al., 1993; Ruitjen et
21 al., 1991). These studies suggest that prolonged exposure to TCE can increase the
22 severity of effects, prompting the use of a partial factor of 101/2 until duration-response
23 relationships are better characterized in humans.
24 (d) LOAEL-to-NOAEL uncertainty: The 1 mg/kg-d point of departure is a subchronic
25 LOAEL in one study, a NOAEL in another, and an LED10 in a third. When adverse
26 effects are observed at the point of departure, an uncertainty factor of up to 10-fold is
27 generally applied. In this case, a reduced factor of 101/2 is appropriate because 1 mg/kg-d
28 appears to be at the boundary where effects can begin to be observed.
29 (e) Other factors: The general population is routinely exposed to many agents that induce
30 CYP2E1, for example, ethanol, acetaminophen, and many ubiquitous environmental
31 contaminants. Moreover, several metabolites of TCE are major environmental
32 contaminants in their own right: for example, TCA and DC A are two principal toxic
33 byproducts of drinking water disinfection. Thus, humans generally have higher
34 background exposures to TCE and its metabolites compared to background exposures in
62101/2 denotes the square root of 10. On a logarithmic scale, this represents half of a factor of 10. Numerically, 101/2
is approximately equal to 3.1, which is usually rounded to 3 in a final calculation.
8/1/01 4-8 DRAFT-DO NOT CITE OR QUOTE
-------
1 test animals; in other words, humans start higher on the dose-response curve than the test
2 animals. To account for this difference between humans and laboratory animals, a
3 modifying factor of 101/2 will be applied. This factor appears reasonable in view of the
4 large numbers of people who are at increased risk because of certain medications (for
5 example, acetaminophen and barbiturates) or diseases (for example, diabetes and
6 alcoholism) that affect metabolism and may exacerbate the effects of TCE exposure
7 (Pastino et al., 2000).
8 This modifying factor to reflect background exposures to TCE and its metabolites is new.
9 It is meant to address the issue of cumulative risks, thus responding to calls by the risk
10 assessment and public health communities that risk assessments would be more realistic and
11 relevant if they address how exposure to other chemicals and stressors can alter a chemical's
12 toxicity. More than for most other chemicals, the general population is exposed to TCE and its
13 metabolites from multiple sources (see Section 3). Consideration of a preexisting background
14 exposure to toxic metabolites is important when these chemicals are assumed to have a sublinear
15 dose-response curve (see Section 1.8, footnote 26). This modifying factor applies only to
16 populations that have background exposure to the metabolites that are involved in TCE's
17 toxicity.
18
19 4.3.4. Calculation and Characterization of the Oral Reference Dose
20 For TCE, an RfD can be based on critical effects in the liver, kidney, and developing
21 fetus. The point of departure is 1 mg/kg-d, a dose where adverse liver effects can begin to be
22 observed in two species after subchronic dosing. A composite uncertainty factor of 5,000 is
23 obtained by multiplying factors of 50 for average-to-sensitive human variation, 101/2 for animal-
24 to-human uncertainty, 101/2 for using subchronic instead of lifetime studies, 101/2 for using a point
25 of departure where adverse effects have been observed, and a modifying factor of 101/2 to reflect
26 background exposures to TCE and its metabolites, thus beginning to address the issue of
27 cumulative risks involving TCE. Dividing the 1 mg/kg-d point of departure by a composite
28 uncertainty factor of 5,000 would yield an RfD of 2x 10~4 mg/kg-d.
29 When RfDs are calculated using conventional 10-fold uncertainty factors, EPA limits the
30 composite factor to 3,000 when human-equivalent doses are used (U.S. EPA, 1994). With this
31 limitation, the RfD would be 3 x 10~4. The TCE uncertainty factors are not, for the most part,
32 conventional 10-fold factors: the 50-fold factor for human variation is based on an uncertainty
33 analysis, and the other factors are reduced below 10 based on the available data. Each factor is
34 appropriate at some level, as the RfD is otherwise derived from a dose where adverse effects are
35 observed in subchronic animal studies, but the size of each factor is an open question.
8/1/01 4-9 DRAFT-DO NOT CITE OR QUOTE
-------
1 Another perspective on the size of the uncertainty factors is provided by research in
2 progress at ATSDR (El Masri, 2000), in which a data-derived factor covering human variation
3 and animal-to-human uncertainty for TCE was 625 (compared with 50x 101/2=150 used in this
4 assessment). ATSDR's analysis showed more uncertainty for TCE than for the three other
5 chlorinated solvents that were modeled. This research suggests that an RfD based on a
6 composite uncertainty factor of 3,000 would leave only a five-fold margin (3,000/625=5) to
7 cover subchronic-to-chronic uncertainty, LOAEL-to-NOAEL uncertainty, and higher cumulative
8 background exposures in humans. This further indicates the high level of uncertainty inherent in
9 any RfD currently derivable for TCE.
10 Barton and Clewell (2000) suggest an RfD of 0.06-0.12 mg/kg-d based on subchronic
11 liver effects in rats and mice.63 Their RfD is based on increased liver-weight-to-body-weight
12 ratio, one of the critical effects used here. Their point of departure is equal to the 1 mg/kg-d used
13 here. The difference is their use of an uncertainty factor only for average-to-sensitive human
14 variation. This indicates the potential for discussion of alternative RfDs using fewer uncertainty
15 factors.
16 The argument for omitting the animal-to-human factor would rest on an assertion that
1 7 humans are no more sensitive than rats and mice in terms of pharmacodynamic response to TCE.
18 Although some mechanisms of liver toxicity may be quantitatively less in humans (for example,
19 PPAR activation), others might not (for example, cell signaling with DC A). TCE's modes of
20 action are not known, nor is the potential for interaction between modes of action. In addition,
21 critical effects in the kidney and developing fetus occur at doses only slightly above those
22 causing liver toxicity. It has not been established that humans are more sensitive than rats and
23 mice to these other effects.
24 The argument for omitting the subchronic-to-chronic factor would rest on an assertion
25 that TCE's subchronic effects would not increase over longer durations (the principal studies
26 supporting liver effects involved exposure for only 14 and 30 days). Such short durations usually
27 indicate application of a 10-fold factor. In addition, recent human and animal studies have found
28 statistically significant duration-response trends for several hormone levels, providing further
29 support for using a subchronic-to-chronic factor.
30 The argument for omitting the LOAEL-to-NOAEL factor would rest on an assertion that
31 the LOAELs for increased liver-weight-to-body-weight ratio represent a minimal effect level.
32 This assertion receives some support from the use of "benchmark dose" modeling, which
"Barton and Clewell (2000) suggest an oral RfD in the range of 0.06-0.12 mg/kg-d based on liver effects (increased
liver-weight-to-body-weight ratio) in rats (Herman et al, 1995) and mice (Buben and O'Flaherty, 1985), obtaining
Weibull-model ED10s of 308-515 mg/kg-d using modeled TCA area-under-the-curve (Clewell et al., 2000) as the
dose metric. The corresponding human-equivalent doses were divided by a composite uncertainty factor of 10 (for
human variation).
8/1/01 4-10 DRAFT-DO NOT CITE OR QUOTE
-------
1 indicates that the LOAELs are comparable to LED10s. Benchmark dose modeling, however, does
2 not mandate rote treatment of LED10s as if they were NOAELs. In this case, the observation of
3 adverse effects at the lowest doses tested after only 14 and 30 days exposure warrants some
4 concern for effects at still lower doses.
5 This assessment uses partial uncertainty factors of 101/2 (approximately threefold) in each
6 of these areas. This is commensurate with the limited nature of the principal studies, in which
7 rats and mice dosed for 14 or 30 days, respectively, showed adverse effects at the lowest tested
8 doses. Further, ATSDR's decision that the database is not strong enough to support a chronic
9 value indicates the presence of much uncertainty in these areas.
10 Limitations of the database are many. As noted by ATSDR (1997), the chronic-duration
11 studies are problematic. Epidemiologic studies of TCE-contaminated drinking water are limited
12 in their ability to quantify dose-response relationships because of difficulties in estimating past
13 contamination levels and the presence of other solvents with similar metabolic profiles. Most
14 animal studies focus on liver and kidney, organs where TCE induces tumors; studies assessing
15 alteration of organ function or enzyme levels are less common. Finally, neither the occupational
16 studies nor the animal studies provide information about sensitive individuals. In view of these
17 limitations, the use of several uncertainty factors appears reasonable.
18
19 4.4. INHALATION REFERENCE CONCENTRATION FOR EFFECTS OTHER
20 THAN CANCER
21 4.4.1. Critical Effects
22 For acute inhalation exposure to high concentrations (above approximately 1,000 ppm, or
23 5,400 mg/m3), targets of TCE toxicity include the central nervous system, cardiovascular system,
24 kidney, and developing fetus (ATSDR, 1997). At lower concentrations (approximately
25 100-1,000 ppm, or 540-5,400 mg/m3), acute exposure still affects these systems, and there are
26 also intermediate and chronic studies (1 year or more) that show effects on the central nervous
27 system, kidney, liver, and endocrine system (see Section 3.4). At the lowest concentrations,
28 effects are observed in the central nervous system, liver, and endocrine system. These are
29 considered to be the critical effects of inhaled TCE.
30
31 4.4.2. Point of Departure
32 Table 4-3 summarizes the results of inhalation studies of lower concentrations.
33 Concentrations in occupational studies were converted to continuous (24 hr/d, 7 d/wk)
34 concentrations. Concentrations in animal studies were taken from the analysis of Barton and
35 Clewell (2000). If pharmacokinetic-adjusted doses were not available, the RfC methodology
36 (U.S. EPA, 1994) for systemic effects for a water-soluble, perfusion-limited gas was used
8/1/01 4-11 DRAFT-DO NOT CITE OR QUOTE
-------
1 instead. Response levels are presented as either a NOAEL, a LOAEL if the study did not identify
2 a NOAEL, or a modeled LEC10 if the study results were suitable for modeling.
3 Adverse effects have been observed in several studies at a concentration range of 5-16
4 ppm (27-86 mg/m3). These include LOAELs of 7-16 ppm (38-86 mg/m3) for signs of central
5 nervous system toxicity in several occupational studies (Rasmussen et al., 1993; Ruitjen et al.,
6 1991; Vandervort et al., 1973; Okawa and Bodner, 1973). Also, an LEC10 of 5 ppm (27 mg/m3)
7 was identified for heart rate and electroencephalographic changes in rats (Arito et al., 1994). For
8 endocrine effects, a LOAEL of 11 ppm (59 mg/m3) was identified for decreased mean levels of
9 testosterone, decreased FSH, decreased SHBG, and increased DHEA in an occupational study
10 (Chia et al., 1997; Goh et al., 1998). A pharmacokinetic-model-derived LEC10 of 5 ppm (27
11 mg/m3) was identified for increased liver-weight-to-body-weight ratio in mice (Kjellstrand et al.,
12 1983).
13 Together, these studies indicate 5-16 ppm (27-86 mg/m3) as the lower end of the range of
14 concentrations at which adverse effects have been observed. From this range, 7 ppm (38 mg/m3)
15 is chosen as the point of departure. This concentration was identified as a subchronic LOAEL
16 for central nervous system effects in two occupational studies and is supported by central nervous
17 system effects in rats at 5 ppm (27 mg/m3). This latter dose is also associated with a 10%
18 increased incidence of liver effects in mice.
19
20 4.4.3. Uncertainty Factors
21 To address differences between study conditions and conditions of human environmental
22 exposure, the point of departure is reduced by several uncertainty factors.
23 (a) Human variation: The occupational studies observed healthy adult workers, who do not
24 reflect the potential for effects in children, the elderly, or those with disease or other
25 conditions that can increase susceptibility. This supports use of the standard 10-fold
26 factor for human variation. If human variation for this effect were as large as that for
27 liver effects (see Section 4.3.3), then this factor would be an underestimate.
28 (b) Animal-to-human uncertainty: Because the point of departure is supported by human
29 studies, this factor is not needed.
30 (c) Subchronic-to-chronic uncertainty: The point of departure is based on subchronic
31 studies. Duration-response trends have been reported for central nervous system toxicity
32 (Rasmussen et al., 1993; Ruitjen et al., 1991). These observations suggest that continuing
33 exposure to TCE can increase the severity of effects, supporting a 10-fold factor to
34 address the potential for more severe toxicity from lifetime exposure to TCE.
35 Statistically significant duration-response trends have also been observed for decreased
36 mean levels of testosterone, decreased follicle-stimulating hormone, decreased SFffiG,
8/1/01 4-12 DRAFT-DO NOT CITE OR QUOTE
-------
1 and increased DHEA in humans (Chia et al., 1997; Goh et al., 1998), plus decreased
2 testosterone in rats (Kumar et al., 2000).
3 (d) LOAEL-to-NOAEL uncertainty: The central nervous system effects and the endocrine
4 effects observed in occupational studies are LOAELs, supporting a 10-fold factor to
5 approach the range where a NOAEL could be expected.
6 (e) Other factors: These are not being used.
7
8 4.4.4. Calculation and Characterization of the Inhalation Reference Concentration
9 For TCE, an RfC can be based on critical effects in the central nervous system, liver, and
10 endocrine system. The point of departure is 7 ppm (38 mg/m3). This concentration was
11 identified as a subchronic LOAEL for central nervous system effects in two occupational studies
12 and is further supported by central nervous system effects in rats and liver effects in mice at 5
13 ppm (27 mg/m3). A composite uncertainty factor of 1,000 is obtained by multiplying factors of
14 10 for average-to-sensitive human variation, 10 for starting from subchronic instead of lifetime
15 studies, and 10 for starting from effect levels instead of NOAELs. Dividing the 38 mg/m3 point
16 of departure by a composite uncertainty factor of 1,000 yields an RfC of 4><10~2 mg/m3.
1 7 Barton and Clewell (2000) suggest an RfC of 0.4-1 ppm (2-5 mg/m3) based on
18 subchronic neurological effects in rats.64 Their RfC is based on one of the studies supporting the
19 point of departure. The difference is their use of an uncertainty factor for only average-to-
20 sensitive human variation. This indicates the potential for discussion of alternative RfCs that do
21 not incorporate uncertainty factors for using a subchronic instead of a lifetime study or for using
22 an effect level instead of a NOAEL.
23
24 4.4.5. Comparison of Reference Dose and Reference Concentration
25 The similarity of effects by oral or inhalation exposure to TCE suggests using route
26 extrapolation (see Section 4.2.3):
27 (a) to compare effect levels for oral and inhalation exposure, and
28 (b) to compare the RfD and RfC.
29 The results will depend on the equivalence between oral dose and inhaled concentration
30 that yields similar internal levels of the active agent causing a particular toxic effect. If TCA
31 area-under-the-curve is the appropriate dose metric, then ingesting TCE at 1 mg/kg-d yields
32 similar internal TCA as inhaling TCE at 75 mg/m3. On the other hand, if DCA area-under-the-
64Barton and Clewell (2000) suggest an inhalation RfC in the range of 0.4-1 ppm (2-5 mg/m3) based on neurologic
effects (changes in sleep and wakefulness) in rats (Arito et al., 1994), obtaining an ED10 of 0.57 ppm using modeled
TCOH in blood (Clewell et al., 2000) as the dose metric. The corresponding human-equivalent concentration was
divided by a composite uncertainty factor of 10 (for human variation).
8/1/01 4-13 DRAFT-DO NOT CITE OR QUOTE
-------
1 curve is the appropriate dose metric, then ingesting TCE at 1 mg/kg-d yields similar internal
2 DCA as inhaling TCE at 3 mg/m3 (see Section 4.2.3).
3 The oral point of departure is an equivalent human dose of 1 mg/kg-d (see Section 4.3.2).
4 This corresponds to an inhaled concentration of 75 mg/m3 if TCA is the appropriate dose metric
5 for route extrapolation, and 3 mg/m3 if DCA is the appropriate dose metric. The 25-fold range
6 between these two concentrations contains the inhalation point of departure of 38 mg/m3 (see
7 Section 4.4.2). This suggests a convergence of effect levels for these two exposure routes.
8 Similarly, the oral RfD is 3 x 1CT4 mg/kg-d (see Section 4.4.4). This corresponds to an
9 inhaled concentration of 2>
-------
1 risk decreases proportionately with dose.67 The slope of the dose-response curve at low doses is
2 called a slope factor.
3 EPA's proposed cancer guideline revisions (U.S. EPA, 1996) describe a two-step process
4 of modeling the observed data followed by extrapolation to lower doses. Extrapolation to lower
5 doses follows either a linear or a nonlinear approach, depending on what is known about mode-
6 of-action. Extrapolation begins from apoint of departure near the lowest doses. When linear
7 extrapolation is used, the slope factor is determined by the line from the point of departure
8 toward zero. When the mode-of-action is understood well enough to support nonlinear
9 extrapolation, a dose-response curve is not estimated below the point of departure. Instead, a
10 discussion provides information about the distance from the point of departure to a dose where
11 there would be little concern for cancer.
12 The use of pharmacokinetic and dosimetry models carries an implicit assumption of
13 tumor site concordance across species; therefore, it is important to assess and consider each
14 tumor site that may be relevant to human environmental exposure. In this regard, it is important
15 to note that lack of suitable dose-response data has precluded development of slope factors for
16 some sites where there might be a cancer risk, notably for cervical cancer and prostate cancer.
17
18 4.5.1. Risk Estimates From Cancer Epidemiology Studies
19 Quantitative risk methods for human studies must be tailored to the type of data available
20 in each particular case. Of utmost importance is the availability of quantitative exposure data,
21 which can be the principal determinant of the quality of the risk estimates.
22
23 4.5.1.1. Finnish Cohort
24 One set of risk estimates was derived from a cohort of 2,050 male and 1,924 female
25 Finnish workers exposed to TCE and other solvents (Anttila et al., 1995). This cohort had
26 statistically significant increases in non-Hodgkin's lymphoma and, among workers exposed for
27 more than 20 years, liver cancer. Although kidney cancer was not statistically significantly
28 increased in this cohort, the observed incidence was used to bound the potential risk identified
29 qualitatively from the epidemiologic studies as a whole (Wartenberg et al., 2000). For this
67Low-dose-linear models are appropriate for extrapolation to lower doses when a carcinogen acts in concert with
other exposures and processes that cause a background incidence of cancer (Crump et al., 1976; Lutz, 1990).
Further, even when the mode-of-action indicates a nonlinear dose-response curve in homogeneous animal
populations, the presence of genetic and lifestyle factors in a heterogeneous human population tends to make the
dose-response curve more linear (Lutz, 1990). This is because genetic and lifestyle factors contribute to a wider
spread of human sensitivity, which extends and straightens the dose-response curve over a wider range. Although
these considerations provide a reasonable argument for a model that is linear at low doses, the relation of the low-
dose slope to one from the experimental range is uncertain; this uncertainty increases with the distance from the
experimental range.
8/1/01 4-15 DRAFT-DO NOT CITE OR QUOTE
-------
1 assessment, Anttila made available to EPA urinary TCA measurements and the cancer status for
2 each worker.68 A separate study of 51 male workers (Ikeda, 1972) was used to derive a
3 relationship between urinary TCA and air concentrations of TCE:
4 Urinary TCA (mg/L) = 2.956 x Ambient TCE (ppm, 8 hr/d, 6 d/wk)
5 Occupational exposure duration, which was not available, was assumed to be 15 years,
6 and each worker's mean urinary TCA measurement was taken as representative for this duration.
7 The resulting inhalation unit risk estimates, adjusted for lifetime exposure to an average of 1
8 ug/m3 of TCE in air, are:
9 Liver cancer: 1 x 1CT7; upper bound, 9x 1CT7
10 Kidney cancer: 2x 1CT5; upper bound, 3 x 1CT5
11 Non-Hodgkin'slymphoma: SxlCT5; upper bound, 9xlCT5
12 Overall: 7x 1(T5; upper bound, >9x 1(T5
13 These estimates are based on a small number of cancer cases. As with other occupational
14 studies, they are based on healthy adult workers and may not be representative of children or
15 other groups. Moreover, the exposure duration is uncertain. In addition, the workers were
16 exposed to other solvents (albeit solvents with similar metabolic pathways, potential modes of
1 7 action, and toxic effects), so that attributing all risk to TCE can overestimate the risk from TCE
18 (though not, perhaps, for the combined solvent exposure). On the other hand, the exposure
19 metric is based on biological measurements of a major metabolite of TCE.
20
21 4.5.1.2. Hill Air Force Base Cohort
22 Another analysis considered the cohort of Hill Air Force Base workers exposed to TCE
23 and other chemicals (Blair et al., 1998). This cohort showed nonsignificant increases in mortality
24 from liver cancer, kidney cancer, and non-Hodgkin's lymphoma, cancers that are elevated in the
25 epidemiologic studies overall (Wartenberg et al., 2000). Blair made available to EPA a
26 qualitative exposure index and the cancer mortality status for each worker. The exposure index
27 did not quantify exposure, instead, it classified each job as either high or low peak exposure.
28 Difficulties in converting this exposure index into a credible quantitative estimate include
29 differences between peak exposure and cumulative exposure and the diversity of jobs and
30 exposure levels in the cohort; indeed, Blair (1998) cautioned that the exposure index was not
68The data for each cancer case are as follows:
Liver cancer Kidney cancer Non-Hodgkin'slymphoma
Average urinary TCA (• mol/L) 15257 17112 92 23 77 363 2,217 3 3 48207312422 13 44 21
Age at first TCA measurement 64 50 29 15 52 48 41 40 32 54 54 36 63 34 33 33 18 21
Age at last TCA measurement 65 52 30 25 54 49 42 41 33 55 55 37 64 35 34 34 19 27
Ageatendoffollowup 78 79 56 41 68 62 62 63 49 80 74 61 81 58 62 57 39 43
8/1/01 4-16 DRAFT-DO NOT CITE OR QUOTE
-------
1 convertible into ppm. Without a quantitative estimate of exposure, it is not possible to derive
2 risk estimates.
3
4 4.5.1.3. New Jersey Drinking Water Study
5 Risk estimates were derived from a study of drinking water contamination in a 75-town
6 area of New Jersey (Cohn et al., 1994). Female residents had increased incidences of leukemia
7 (RR=1.4, 95% CI=1.1-1.9, N=56) and non-Hodgkin's lymphoma (RR=1.4, 95% CI=1.1-1.7,
8 N=87). For homes exceeding the Safe Drinking Water Act maximum contaminant level of
9 5 |ig/L, the average concentration was 23 * §/L and the highest concentration was 67 |ig/L.
10 Using this average from the most-exposed homes, a drinking water unit risk estimate for non-
11 Hodgkin's lymphoma from exposure to 1 * §/L of TCE in drinking water is:
12 (RR - 1) x Background risk / Average concentration =
13 (1.4-l)x(6xl(T4)/(23 • g/L)=lxl(T5per» g/L
14 where 6x 1CT4 is the background risk of non-Hodgkin's lymphoma in the United States. This unit
15 risk can be converted to a slope factor by multiplying by 35,000 (based on a 70-kg adult drinking
16 2 L/d). The resulting slope factor is 3.5x 10"1 per mg/kg-d average lifetime exposure to TCE.
1 7 The residents were exposed to other drinking water contaminants, so that attributing all risk to
18 TCE can overestimate the risk from TCE. On the other hand, this risk estimate is derived from
19 human environmental exposure, avoiding the uncertainties of animal-to-human extrapolation and
20 high- to low-dose extrapolation. Moreover, using the average concentration from the most-
21 exposed homes would tend to overestimate exposure and, hence, underestimate unit risks.
22
23 4.5.1.4. German Cohort
24 The California Environmental Protection Agency (Cal/U.S. EPA, 1999) derived a slope
25 factor from the increased incidence of kidney cancer in German cardboard workers exposed to
26 TCE (Henschler et al., 1995). In the absence of exposure data, the German threshold limit value
27 of 50 ppm (270 mg/m3) was used as a surrogate for the average workplace concentration. This
28 was converted to a lifetime average daily dose, assuming 50% absorption of inhaled TCE:
29 50 ppm x (8 hr/24 hr) x (5 d/7 d) x (15.2 yr/70 yr) x 0.50 x 5.37 (mg/m3)/ppm
30 x 20 m3/d / 70 kg = 1.98 mg/kg-d
31 From the reported kidney cancer incidence (7/169, SIR=13.53 based on East German
32 background rates), a slope factor was obtained by the following calculation, based on an additive
33 risk model that is linear in dose:
34 (7/169) x (1-1/13.53) / 1.98 mg/kg-d = 1.9 xlQ'2 per mg/kg-d
8/1/01 4-17 DRAFT-DO NOT CITE OR QUOTE
-------
1 By this calculation, Cal/EPA converted the air concentration into mg/kg-d by figuring
2 gross amount inhaled by a 70-kg adult breathing 20 m3/d. Therefore, to convert the slope factor
3 to an inhalation unit risk, 20 m3/d and 70 kg would be removed from the preceding calculations.
4 The resulting inhalation unit risk is 5x 10~6 per * gVm3.
5 To the extent that the workers were exposed to mixed solvents and other risk factors, this
6 slope factor derived from the Henschler study would tend to overestimate risks from TCE,
7 because it attributes all kidney cancers to TCE and not to the other risk factors. On the other
8 hand, to the extent that the threshold limit value of 50 ppm overestimates average workplace
9 concentration, this slope factor would tend to underestimate risks. Further uncertainty arises
10 from using an unadjusted incidence (7/169) instead of one based on life-table methods, from
11 comparing this incidence with East German background rates (the plant was in West Germany,
12 but West German rates were unavailable), and from treating exposure concentration and
13 exposure duration as constant across all workers.
14 Variation in human susceptibility was explored by further analyses based on observed
15 differences in kidney cancer risk attributed to GST polymorphisms (Briining et al., 1997). The
16 odds ratio for the combined genotype GSTM/GSTT (found in 40% of Caucasians) was 3.75.
1 7 This corresponds to a higher slope factor (3.5x 10~2 per mg/kg-d) for those with the combined
18 genotype and a lower slope factor (9x10~3 per mg/kg-d) for the rest of the population (Cal/U. S.
19 EPA, 1999).
20
21 4.5.2. Risk Estimates From Liver Tumors in Mice
22 TCE causes hepatocellular carcinomas and adenomas in male and female mice exposed
23 by ingestion or inhalation (see Table 3-1). TCE is also associated with liver cancer in humans
24 (Wartenberg et al., 2000). Several approaches are explored for estimating these risks.
25
26 4.5.2.1. Mechanism-Based Modeling
27 Chen (2000) developed a mechanism-based dose-response model for mouse liver tumors,
28 based on mechanistic hypotheses discussed by Bull (2000). The model is a stochastic form of the
29 two-stage model, in which normal cells can become initiated (the first stage); initiated cells either
30 die, proliferate into a clone of initiated cells, or are converted to malignancy (the second stage);
31 and malignant cells either die or progress to a detectable tumor. Dose metrics were area-under-
32 the-curve of TCA or DCA in the liver, taken from Bois's calibration of Fisher's pharmacokinetic
33 model (Bois, 2000a). Chen's model assumes that TCE induces liver tumors through TCA and
34 DCA acting by clonal expansion of preexisting initiated cells. This assumption of action on only
8/1/01 4-18 DRAFT-DO NOT CITE OR QUOTE
-------
1 preexisting initiated cells implies that TCE-induced liver tumor incidences are proportional to
2 background liver tumor incidences.
3 In a sequence of applications of the mechanism-based model, (1) a dose-response curve
4 for DCA-induced liver tumors was developed from bioassays of DC A in mice, (2) this curve was
5 found to be consistent with bioassays of TCA in mice, considering the rate at which some TCA is
6 metabolized to DC A, and (3) these curves were found to be consistent with the bioassays of TCE
7 in mice, considering the rate at which some TCE is metabolized to TCA and DCA (Chen, 2000).
8 This suggests that TCE produces enough DCA to theoretically explain the mouse liver tumors
9 induced in the TCE bioassays, though it cannot rule out a role for TCA or other metabolites.
10 Thus, the relative importance of DCA or TCA in TCE-induced mouse liver tumors is not
11 resolved by mechanism-based modeling.
12 The mechanism-based model also explored extrapolation to lower doses. Following the
13 model below the observed range yields an (upper bound) unit risk of 2.4* 10~8 for oral exposure
14 to 1 ug/L of TCE in drinking water (Chen, 2000).
15
16 4.5.2.2. Empirical Modeling
17 Rhomberg (2000) fitted empirical dose-response models to the liver tumor data. Dose
18 metrics were area-under-the-curve of TCA or DCA in the liver or plasma, taken from Bois's
19 calibrations of Fisher's and Clewell's pharmacokinetic models (Bois, 2000a,b). Table 4-4
20 presents the modeled LED10s that provide a point of departure for low-dose extrapolation. These
21 LED10s reveal substantial model uncertainty between Fisher's and Clewell's models. They also
22 reveal that Bois's calibrations bring some convergence: harmonic means69 of LED10s from the
23 four experiments span 0.16-25.82 mg/kg-d when dose estimates come from Fisher's and
24 Clewell's original models, but narrow to 0.5-3.1 mg/kg-d when dose estimates come from Bois's
25 calibrations of those models. (Bois did not calibrate Fisher's model with DCA as the dose
26 metric; in this case, only Fisher's original model is considered.) This latter range (0.5-3.1
27 mg/kg-d) is used as the point of departure for low-dose extrapolation. Although both endpoints
28 of this range use TCA as the dose metric, this does not imply a preference for TCA over DCA.
29 The DCA range (1.7-2.5 mg/kg-d) simply falls within the TCA range, moreover, Section 4.5.1
30 discusses why neither is established as the appropriate dose metric.
31 It is tempting to note that the point-of-departure range for liver tumors (0.5-3.1 mg/kg-d)
32 includes the point of departure for liver toxicity (a subchronic LOAEL of 1 mg/kg-d for increased
33 liver-weight-to-body-weight ratio; see Section 4.3.2). This concordance, however, is likely to be
69The harmonic mean is computed as M(Ł 1/JQ. For this application, this is equivalent to averaging slope factors
(which are inversely proportional to LED10s), then taking the reciprocal to convert the average slope factor back to a
dose.
8/1/01 4-19 DRAFT-DO NOT CITE OR QUOTE
-------
1 only coincidental, as liver weight change is a nonspecific effect observed in both rats and mice,
2 but rats have not developed liver tumors from TCE exposure. The liver weight changes,
3 therefore, are not considered a key event in liver tumor development.
4
5 4.5.2.3. Extending the Observed Range Using Subchronic-dosing Studies
6 A key determinant of risk is the shape of the dose-response curve below the point of
7 departure. One source of information about lower doses is the comprehensive study by Maltoni
8 et al. (1986), who had the foresight to investigate different exposure durations in parallel
9 experiments. Groups of Swiss mice inhaled TCE at 100-600 ppm (540-3,240 mg/m3), 7 hours
10 daily, 5 days weekly, for either 8 weeks or 78 weeks; then the experiments continued past 2 years
11 to ensure that the latent period for tumor development had passed.
12 Table 4-5 and Figure 4-1 show the incidence of hepatomas in male mice (females did not
13 develop hepatomas). If response is considered as a function of cumulative exposure
14 (concentration times duration), the 8-week exposure groups can be regarded as extending the
15 observed range to lower cumulative exposures. Overall, response decreases as cumulative
16 exposure decreases, but the curve does not appear to be sublinear anywhere in this extended
17 range. This suggests that exposures must be far below the observable range before the risk drops
18 appreciably.
19 This finding may be pertinent for risk assessments of less-than-lifetime exposure
20 durations. If the lifetime average dose is low because the exposure duration is relatively short, it
21 may be important to consider that risk may not fall proportionately as exposure duration
22 decreases. That is, cumulative exposure (concentration times duration, or area-under-the-curve)
23 may not be a good dose metric to use to predict liver tumor risk.
24
25 4.5.2.4. Extending the Observed Range Using DCA and TCA Studies
26 Another source of information at lower doses are the studies of DCA- or TCA-induced
27 liver cancer. Bioassays of these TCE metabolites cover a 100-fold dose range, wider than the
28 dose range of the TCE bioassays. Chen (2000) analyzed these studies jointly, using area-under-
29 the-curve of DCA in the liver as a common dose metric.
30 Table 4-6 and Figure 4-2 bring together the dose-response information from the DCA,
31 TCA, and TCE bioassays. The DCA and TCA studies extend the dose range as much as 100-fold
32 below the TCE studies. Overall, the responses appear to decrease as dose decreases, but the
33 dose-response curve does not appear to be sublinear anywhere in this extended range, again
34 suggesting that exposures must be far below an LED10 before the risk drops appreciably.
35 Looking at the multiple-dose TCA and DCA bioassays individually, the dose-response
36 curve for TCA appears linear, whereas the curve for DCA appears sublinear. If we consider
8/1/01 4-20 DRAFT-DO NOT CITE OR QUOTE
-------
1 tumor multiplicity along with tumor incidence, however, the DCA dose-response curve, too,
2 becomes linear (DeAngelo et al., 1999). Tumor multiplicity may more closely reflect rates of
3 underlying carcinogenic processes.
4
5 4.5.2.5. Extrapolation to Lower Doses
6 Under EPA's proposed cancer guideline revisions (U.S. EPA, 1996), choice of linear or
7 nonlinear extrapolation to lower doses is based on what is known about the mode-of-action. The
8 evidence indicates that TCE-induced liver tumors arise through multiple metabolites and
9 multiple modes of action (see Section 3.5.1). The preponderance of evidence suggests that the
10 CYP450 metabolites are sufficient to explain the liver tumors (Bull, 2000; Chen, 2000). CH and
11 TCA cause liver tumors in mice, and DCA causes liver tumors in mice and rats. A plausible
12 mode-of-action is that TCE induces liver tumors through TCA and DCA modifying cell signaling
13 systems that control rates of cell division and cell death (Bull, 2000). Tumors from the TCA and
14 DCA bioassays arise from different mechanisms and can be distinguished: TCA-induced tumors
15 are associated with peroxisome proliferation, while DCA-induced tumors are associated with
16 selectively inhibiting growth of normal hepatocytes and suppressing apoptosis (Bull, 2000).
17 DCA may also have different modes of action at different doses, possibly involving differential
18 accumulation of glycogen in normal and initiated liver cells (Bull, 2000). Moreover, DCA is
19 also weakly mutagenic (Moore and Harrington-Brock, 2000), although this property is not
20 needed to explain the bioassay results. Characterizing the tumors from the TCE bioassays is a
21 critical research need that presently limits understanding of the respective roles of TCA and DCA
22 in TCE-induced tumors (see Section 5). At present, however, the extensive mode-of-action
23 information still lacks identification of the sequence of key events and a quantitative description
24 of the doses at which those key events begin to occur. In such cases, EPA's proposed cancer
25 guideline revisions (U.S. EPA, 1996) support consideration of both linear and nonlinear
26 extrapolation to lower doses.
27 Linear extrapolation follows a straight line toward the origin, the slope of this line (or
28 slope factor) is 0.10/LED10. The LED10 range of 0.5-3.1 mg/kg-d yields slope factors of 3><10"2
29 to 2x 10"1 per mg/kg-d. Because the pharmacokinetic models allow for a nonlinear relationship
30 between (external) exposure and (internal) dose, this "straight-line" extrapolation can reflect
31 nonlinearity in the exposure-dose-response data.
32 Nonlinear extrapolation under the proposed cancer guideline revisions (U.S. EPA, 1996)
33 does not attempt to describe dose-response curves in the range of environmental exposures,
34 instead, it examines whether an exposure is small enough to pose little risk of cancer. The goal is
35 that exposures be small enough that "the key events in tumor development would not occur
36 among sensitive individuals in a heterogeneous human population, thus representing an actual
8/1/01 4-21 DRAFT-DO NOT CITE OR QUOTE
-------
1 'no effect level'" (U.S. EPA, 1999). A nonlinear analysis considers several factors70 where the
2 available data may fall short of this ideal.
3 (a) Nature of the response: The point of departure is ideally based on key precursor events.
4 If tumors are used instead, then the dose is adjusted to bring it into the range where the
5 key precursor events would not occur (U.S. EPA, 1999). For TCE-induced liver cancer,
6 further research is needed to identify the key events and the doses at which the key events
7 begin to occur (see Section 5). These doses will be below the LED10s where tumors are
8 clearly observable, how far below cannot now be determined.
9 (b) Shape of the observed dose-response curve: The shape of the dose-response curve can
10 indicate how fast risk decreases below the point of departure. For TCE-induced liver
11 cancer, dose-response curves are ambiguous, consistent with either low-dose linearity or
12 nonlinearity (Rhomberg, 2000). Extending the dose-response curve using information
13 from subchronic-dosing studies or from DCA and TCA bioassays suggests that the dose-
14 response curve may remain linear quite far below, perhaps two orders of magnitude
15 below, the LED10s. At still lower doses, the mode-of-action may ultimately be nonlinear;
16 nonetheless, current data do not identify how far below the LED10s sublinearity begins to
1 7 prevail. Further research is needed to identify the key precursor events and describe their
18 dose-response curves (see Section 5).
19 (c) Human sensitivity compared with experimental animals: This factor has been viewed as
20 having pharmacokinetic and pharmacodynamic components (U.S. EPA, 1994). For TCE,
21 the pharmacokinetic component is addressed by the pharmacokinetic models, which
22 predict higher internal doses of TCA and DCA in humans than in mice. Little
23 information is typically available on the pharmacodynamic component, and the proposed
24 cancer guideline revisions (U.S. EPA, 1996) recommend that humans be considered
70A nonlinear analysis considers the following factors (U.S. EPA, 1999):
(a) Nature of the response: The point of departure is ideally based on key precursor events, not tumors, just as
RfDs and RfCs are based on critical (precursor) effects occurring at lower doses than those associated with
frank toxicity. Precursor events are likely to occur at doses below those where tumors are observable. If a
point of departure is based on tumors, then the point of departure is adjusted to reflect the dose at which key
precursor events would begin to occur.
(b) Shape of the observed dose-response curve: If the dose-response curve is steep at the point of departure,
then the occurrence of key events declines rapidly in this range. On the other hand, if the dose-response
curve is relatively shallow, then the point where the key event no longer occurs may lie far below this range.
(c) Human sensitivity compared with experimental animals: If the point of departure is developed from animal
studies, this dose is adjusted to reflect pharmacokinetic and pharmacodynamic factors that may make
humans and animals differ in sensitivity.
(d) Nature and extent of human variability in sensitivity: Pertinent information would come from human
studies, because animal studies, particularly those using homogeneous animal strains, do not provide
information about human variation.
(e) Human exposure: Cancer response can depend on the expected pattern of human exposure, including the
magnitude, frequency, and duration of exposure.
8/1/01 4-22 DRAFT-DO NOT CITE OR QUOTE
-------
1 potentially more sensitive by a default factor of 101/2. For TCE, however, the prevailing
2 view is that humans would be quantitatively less sensitive than mice to the PPAR activity
3 and other processes leading to the mouse liver tumors (Bull, 2000). In this case, an
4 inverse factor of 10~1/2 could be used to reflect the potential lesser sensitivity of humans
5 (U.S. EPA, 1996).
6 (d) Nature and extent of human variability in sensitivity: For TCE, this factor can be based
7 on the statistical uncertainty analyses (Bois, 2000b). A factor of 15-20 reflects the
8 pharmacokinetic uncertainty in the liver between the 50th and 99th percentiles (see Table
9 4-1). Coupled with a factor of 101/2 to cover pharmacodynamic differences between
10 humans, a factor of 50-60 would appear appropriate to reflect human variation in
11 sensitivity.
12 This discussion of data-derived factors indicates several large uncertainties: 50-60-fold
13 to reflect human variation, approximately 100-fold to reach a range where the dose-response
14 curve does not appear linear in some bioassay, an as-yet-undeterminable factor to reflect the
15 difference in dose between tumors and key precursor events, and a factor of 10~1/2 to reflect the
16 potential for humans to be less sensitive than mice. Overall, then, the composite factor would
17 appear comparable to the 3000-10,000 that is the maximum for developing a credible RfC or
18 RfD (U.S. EPA, 1994; Barnes and Dourson, 1988).
19
20 4.5.2.6. Comparison of Empirical and Mechanism-Based Modeling Results
21 Mechanism-based modeling yields a unit risk of 2.4x 10~8 for oral exposure to 1 * g/L in
22 water (Chen, 2000). This unit risk can be converted to a slope factor by multiplying by 35,000
23 (based on a 70-kg adult drinking 2 L/d). The resulting slope factor is 8.4* 10~4 per mg/kg-d. This
24 result lies below the empirically based slope factor range of 3 x 10~2 to 2* KT1.
25 The mechanism-based slope factor includes a 100-fold reduction attributable to an
26 assumption that chemically induced tumor rates scale across species in proportion to background
27 tumor rates, the rate in mice being 100 times that of humans. This difference is not based on any
28 data for TCE, rather, it is a corollary of the model's assumptions that TCE acts by clonal
29 expansion of preexisting initiated cells and that there is strict tumor site concordance across
30 species. If these assumptions, which generally are not used in cross-species extrapolations for
31 other chemicals, were not used here, then the mechanism-based slope factor would be 100-fold
32 higher, or 8* 10~2 per mg/kg-d. This latter result lies within the slope factor range based on
33 empirical modeling.
34
35 4.5.3. Risk Estimates From Lung Tumors in Mice
8/1/01 4-23 DRAFT-DO NOT CITE OR QUOTE
-------
1 4.5.3.1. Modeling
2 TCE causes lung adenomas and adenocarcinomas in mice exposed by inhalation (see
3 Table 3-5). Pharmacokinetic models (Clewell et al., 2000; Bois, 2000b) were used to estimate
4 area-under-the-curve and peak concentration of CH in plasma, plasma CH being considered as a
5 surrogate for CH accumulation in Clara cells of the lung. Bois's uncertainty analysis (Bois,
6 2000b), however, revealed the presence of substantial parameter uncertainty in the lung dose
7 estimates (see Table 4-1, Section 4.2.2). Consequently, dosimetry models for deriving inhalation
8 RfCs were used instead (U.S. EPA, 1994). Rhomberg (Rhomberg, 2000) fitted empirical dose-
9 response models to the lung data (see Table 4-7). The resulting LEC10s are quite similar, ranging
10 from 1.3xl05to3.6xl05»g/m3.
11
12 4.5.3.2. Extrapolation to Lower Doses
13 The proposed mode-of-action for TCE-induced lung tumors is that rapid CYP2E1
14 metabolism of TCE in the Clara cells of the mouse lung leads to an accumulation of CH and
15 causes cell damage and compensatory cell replication that leads to tumor formation (Green,
16 2000). Because humans have fewer Clara cells and little CYP2E1 activity in the lung as a whole,
1 7 humans may be less sensitive to this particular mode-of-action. (This argument, however,
18 pertains more to the relative sensitivity of mice and humans than to nonlinearity in either mice or
19 humans.) Although this hypothesis is consistent with the higher level of CYP2E1 in mouse lung,
20 it has not yet been determined whether accumulation of CH is a key event causing tumors or a
21 coincidental event unrelated to tumors. Moreover, several questions about this hypothesis
22 remain unresolved (see Section 3.5.3). In addition, CH is clearly clastogenic and mutagenic at
23 high doses (Moore and Harrington-Brock, 2000), raising the possibility of multiple modes of
24 action. For these reasons, this assessment will pursue both linear and nonlinear extrapolation to
25 lower doses.
26 Linear extrapolation takes the form of a straight line to the origin; the slope of this line
27 (or inhalation unit risk) is 0.10/LEC10. The LEC10 range of 1.3xl05 to 3.6xl05 • g/m3 yields unit
28 risksof3xlO~7to8xlO~7per»g/m3.
29 Nonlinear extrapolation under the proposed cancer guidelines (U.S. EPA, 1996) considers
30 several factors in identifying a dose where there would be little concern for cancer.
31 (a) Nature of the response: The proposed key event, CH accumulation in Clara cells, has
32 been studied only at high doses (450 ppm and above, equivalent to 2,430 • g/m3) (Green,
33 2000) relative to those where lung tumors have been observed (150-300 ppm, or
34 810-1,620 mg/m3). This makes determination of cause-and-effect premature, and it does
35 not identify doses at which CH accumulation would not occur. Further research is needed
8/1/01 4-24 DRAFT-DO NOT CITE OR QUOTE
-------
1 to establish the quantitative relationship between CH accumulation and lung tumors so
2 that a point of departure can be based on the key precursor event.
3 (b) Shape of the observed dose-response curve: Dose-response curves for TCE-induced lung
4 cancer are ambiguous, consistent with either linearity or nonlinearity at low doses
5 (Rhomberg, 2000). Current data do not identify how far below the LEC10s sublinearity
6 begins to prevail. Further research is needed to link CH accumulation to lung tumors and
7 describe the dose-response curve for CH accumulation.
8 (c) Human sensitivity compared with experimental animals: The pharmacokinetic
9 component is addressed by inhalation dosimetry models (U.S. EPA, 1994). For the
10 pharmacodynamic component, if rapid CYP2E1 metabolism leading to CH accumulation
11 in Clara cells is the key event, humans may be quantitatively less sensitive than mice to
12 this effect (Green, 2000). In this case, an inverse factor of 10~1/2 could be used to reflect
13 the potential lesser sensitivity of humans (U.S. EPA, 1996). Nonetheless, humans may
14 not necessarily be less sensitive if other CYP450 enzymes play a role or if other lung cell
15 types are involved.
16 (d) Nature and extent of human variability in sensitivity: For TCE in the lung, this factor is
1 7 problematic in light of the substantial variation revealed by the uncertainty analysis of the
18 pharmacokinetic modeling (Bois, 2000b). A factor of more than 300 would be needed to
19 reflect just the pharmacokinetic uncertainty in the lung between the 50th and 99th
20 percentiles (see Table 4-1). A further factor would be needed to reflect
21 pharmacodynamic differences. Clearly, more research is needed in the area of human
22 variation.
23 In light of these research needs (see Section 5), it is difficult to proceed with nonlinear
24 extrapolation to establish a dose at which lung tumors would not be of concern.
25
26 4.5.4. Risk Estimates From Kidney Tumors in Rats
27 4.5.4.1. Modeling
28 TCE causes kidney adenocarcinomas and adenomas in male and female rats by oral
29 gavage and in male rats by inhalation (see Table 3-2). TCE is also associated with kidney cancer
30 in humans (Wartenberg et al., 2000). Pharmacokinetic models (Clewell et al., 2000; Bois,
31 2000b) were used to estimate metabolites in the kidney. Bois's uncertainty analysis (Bois,
32 2000b), however, revealed the presence of substantial parameter uncertainty in the kidney dose
33 estimates (see Table 4-1, Section 4.2.2). Consequently, dose estimates for the kidney were based
34 on equivalence of mg/kg3/4-d instead (U.S. EPA, 1992). Rhomberg (2000) fitted empirical dose-
35 response models to the kidney data (see Table 4-8). Following NTP guidance to combine kidney
36 adenocarcinomas and adenomas (McConnell et al., 1986), the resulting human-equivalent LED01
8/1/01 4-25 DRAFT-DO NOT CITE OR QUOTE
-------
1 for kidney tumors is 33 mg/kg-d. LED01s were calculated instead of LED10s because kidney
2 tumor incidences can be observed below the 10% level.
3
4 4.5.4.2. Extrapolation to Lower Doses
5 TCE-induced kidney tumors may arise through multiple modes of action (see Section
6 3.5.2). The preponderance of evidence suggests that TCE's GST metabolites are responsible for
7 kidney toxicity and kidney tumors (Lash et al., 2000b). Although these metabolites have not
8 been tested in cancer bioassays, the GST metabolite DCVC is mutagenic (Moore and
9 Harrington-Brock, 2000). Involvement of a mutagenic metabolite would indicate use of linear
10 extrapolation to lower doses (U.S. EPA, 1996). Cytotoxicity may also be involved, as
11 cytotoxicity is observed in both rats and humans exposed to high levels of TCE.
12 Linear extrapolation from the LED01 of 33 mg/kg-d yields a slope factor of 3 x 10~4 per
13 mg/kg-d (0.01/33 = 3xlO~4).
14
15 4.5.5. Risk Estimates From Testicular Tumors in Rats
16 4.5.5.1. Modeling
17 Independent studies provide some evidence that TCE causes testicular tumors in rats by
18 oral gavage and by inhalation (see Table 3-4). Health Canada derived a cancer potency estimate
19 (a TD05, the dose associated with an increased tumor risk of 5%) from the increased incidence of
20 testicular interstitial cell tumors in Marshall rats exposed to TCE (see Table 3-4). Health
21 Canada's dosimetry assumed cross-species equivalence of mg/kg-d. Their resulting TD05 was
22 200 mg/kg-d.
23 Using EPA's methodology with the same study yields an LED10 of 25 mg/kg-d. The
24 eightfold ratio of Health Canada's TD05 to EPA's LED10 is attributable to a factor of 3.8 for
25 EPA's use of 3/4-power scaling and a factor of 2.2 between the lower-bound dose at 10%
26 response and the central-estimate dose at 5% response.
27
28 4.5.5.2. Extrapolation to Lower Doses
29 TCE-induced testicular tumors would likely result from endocrine disturbances (see
30 Section 3.5.4). There is evidence of endocrine disturbance in both rats (Kumar et al., 2000) and
31 humans (Chia et al., 1997; Goh et al., 1998). Endocrine disturbance in humans is one of the
32 critical effects used in developing the RfC, where a subchronic LOAEL of 11 ppm (59 mg/m3)
33 was observed for decreased testosterone, decreased follicle-stimulating hormone, decreased
34 SHBG, and increased DHEA (see Section 4.4.2). Route extrapolation yields equivalent oral
35 doses of 0.8 mg/kg-d to 20 mg/kg-d, depending on whether TCA or DCA is used as the dose
8/1/01 4-26 DRAFT-DO NOT CITE OR QUOTE
-------
1 metric (see Section 4.2.3). As a check for consistency with the tumor data, these precursor doses
2 lie below the LED10 for tumors of 25 mg/kg-d.
3 EPA's proposed cancer guidelines (U.S. EPA, 1996) prefer to extrapolate from precursor
4 data instead of tumor data, using an RfD or RfC for the precursor if the tumors were
5 demonstrated to occur only subsequently to the precursor. If this is the case, then an RfD for
6 endocrine-related tumors would be calculated by dividing the point of departure for endocrine
7 disturbance (0.8 mg/kg-d) by the composite uncertainty factor (1,000, comprising 10 for human
8 variation, 10 for a subchronic study, 10 for the LOAEL) previously applied to endocrine
9 disturbance as a critical effect (see Section 4.4.3). The resulting RfD is 8* 10~4 mg/kg-d.
10
11 4.5.6. Discussion of Alternative Cancer Risk Estimates
12 4.5.6.1. Slope Factors
13 Table 4-9 and Figure 4-3 compile the alternative points of departure and cancer risk
14 estimates for TCE. Inhalation-to-oral extrapolation of estimates from the Anttila study was
15 based on internal TCA (see Section 4.2.3), because that study quantified exposure by measuring
16 urinary TCA. Route extrapolation is also supported by the general similarity of tumor sites from
17 oral or inhalation exposure (see Section 3.4). Nonetheless, the best approach for route
18 extrapolation remains a topic for further research (see Section 5).
19 Two sets of estimates appear to lie apart from all others. On the low end, rats appear to
20 be less sensitive than mice or humans. On the high end, estimates from the Anttila study are
21 rather uncertain, based on a small number of cancer cases and an assumed uniform exposure
22 duration of 15 years. Setting aside these lowest and highest estimates, there appears to be
23 convergence of the other estimates, even though they are derived from different sources. The
24 remaining slope factors, per mg/kg-d, are 2x 10~2 (derived from occupational inhalation
25 exposure), 3 x 10~2 to 2x 10"1 (derived from mice), and 4x 10"1 (derived from residential drinking
26 water exposure). Because they are supported by diverse studies and do not reflect the highest
27 estimates (from the Anttila study) or the lowest estimates (from the rat studies), these remaining
28 estimates constitute a middle range of risk estimates where confidence is greatest. This middle
29 range is robust in the sense that it is not likely to be dramatically altered by a new study or by
30 minor changes in the analysis of existing studies.
31 This new slope factor range, 2x 10"2 to 4x 10"1 per mg/kg-d, lies just above EPA's previous
32 slope factor for TCE, 1.1 x 10"2 per mg/kg-d. The increase can be attributed principally to two
33 factors: (1) the pharmacokinetic models predict that humans would experience higher internal
34 doses of TCA and DC A than would mice from similar exposures (Rhomberg, 2000), perhaps due
35 to more extensive enterohepatic recirculation (Barton and Clewell, 2000); and (2) the generation
36 of TCA and DC A would be more efficient at the lower doses expected from environmental
8/1/01 4-27 DRAFT-DO NOT CITE OR QUOTE
-------
1 exposure than from doses used in the bioassays (Rhomberg, 2000). The new slope factor range
2 also lies just above the recent slope factor used by Cal/EPA, 1.3 xlO"2 per mg/kg-d (Cal/U.S.
3 EPA, 1999).71
4
5 4.5.6.2. Nonlinear Analyses
6 Nonlinear extrapolations were attempted for liver tumors, lung tumors, and testicular
7 tumors. For liver tumors, the point of departure was in the range of 0.5-3.1 mg/kg-d, with a dose
8 reduction of 3,000-10,000 needed to reach a dose where there may be little concern for cancer.
9 For lung tumors, it was problematic to determine the reduction needed to reach such a dose. For
10 testicular tumors, an RfD of 8 x 10~4 mg/kg-d was developed.
11
12 4.5.6.3. Joint Consideration of Slope Factors and Nonlinear Analyses
13 Joint consideration of results from linear and nonlinear extrapolations is an area for which
14 there has been no guidance. This may prove to be a complex question, as the purpose of linear
15 extrapolation is to estimate a bound on risk whereas nonlinear extrapolation attempts to make a
16 qualitative assurance of safety. It would not be appropriate to settle on a single result by
1 7 choosing one analysis and ignoring the other.
18 In the meantime, linear and nonlinear analyses can be harmonized somewhat by being
19 expressed in common units. For example, slope factors can be inverted to obtain risk-specific
20 doses12 For an increased risk of 1 in 1 million, the slope factor range (2x 10~2 to 4x 10"1 per
21 mg/kg-d) yields 2.5* 10~6 to 5* 10~5 mg/kg-d as a range of risk-specific doses. Similarly, for a
22 risk of 1 in 10,000, the range is 2.5x 10~4 to 5x 10~3 mg/kg-d. Nonlinear extrapolation can be
23 expressed as a dose where there may be little concern for cancer. For example, for a point of
24 departure of 1.8 mg/kg-d (the midpoint of 0.5-3.1 mg/kg-d, the point-of-departure range for liver
25 tumors) and a dose reduction of 6500 (the midpoint of 3000-10,000, the dose-reduction range
26 developed for mouse liver tumors), a dose where there may be little concern for liver cancer is
27 roughly 3x 10~4 mg/kg-d (1.8/6500=3x 10~4). Higher or lower results could be obtained by using
28 other choices from these two ranges, but the midpoints serve to illustrate the comparison between
29 linear and nonlinear health benchmarks.
30 Figure 4-3 shows that for TCE the health benchmarks under nonlinear extrapolation are
31 approximately 10-100 times the doses associated with an increased cancer risk of 1 in 1 million
71Cal/EPA's cancer estimates are based on liver tumors in mice (NCI, 1976; Maltoni et al, 1986), modeled to an
LED10 of 7.9 mg/kg-d using total amount of TCE metabolized as the dose metric. Using linear extrapolation to
lower doses, their slope factor is 0.10/LED10 = 1.3 x 1(T2. Cal/EPA considered nonlinear extrapolation to be
insufficiently supported by the data.
72A risk-specific dose is calculated by dividing a risk level by a slope factor. For example, if the slope factor is 2 per
mg/kg-d, then the risk-specific dose associated with a risk of 1 in 1 million is 10^7(2 per mg/kg-d) = 5 x 1CT7 mg/kg-d.
8/1/01 4-28 DRAFT-DO NOT CITE OR QUOTE
-------
1 under linear extrapolation. In other words, nonlinear extrapolation leads to doses that would be
2 associated with increased cancer risks of 1 in 100,000 to 1 in 10,000 under linear extrapolation.
3 This suggests that if confidence increases in the appropriateness of nonlinear extrapolation for
4 TCE, then levels that we now associate with risks approaching 1 in 10,000 may be considered
5 virtually safe in the future.
6
7 4.5.6.4. Sensitive Populations and Cumulative Risks
8 Information is beginning to emerge about differences in metabolism, disease, and other
9 factors that make humans vary in their response to TCE, as well as the potential for other
10 chemicals to alter TCE's metabolism and toxicity. This information indicates that a single risk
11 value is not appropriate to describe the differential effects of TCE. For this reason, the
12 alternative slope factors have not been consolidated into a single estimate. Rather, a range of
13 slope factors has been described. Depending on the characteristics of the exposed population and
14 the exposure scenario, each risk assessment should select an appropriate slope factor from this
15 range. Risk assessments involving the presence of risk factors such as diabetes or alcohol
16 consumption, or high background exposure to TCE or its metabolites, would more appropriately
17 choose a higher slope factor. Conversely, lower slope factors may be appropriate when these risk
18 factors and background exposures are not involved.
19 It will be a challenge for future research to quantify the differential risk indicated by each
20 risk factor or exposure scenario (see Section 5). In this way, nonlinear cancer assessments will
21 stop resembling single-point RfDs and will move toward risk ranges with the flexibility to
22 quantitatively accommodate population-specific and site-specific considerations.
23
24 4.5.7. Sources of Uncertainty in the Dose-Response Assessment
25 All risk assessments involve uncertainty, as study data are extrapolated to make
26 inferences about potential effects from human environmental exposure. One type of uncertainty
27 stems from different interpretations of the available data (known as model uncertainty), another
28 from estimation errors due to incomplete or imprecise data (parameter uncertainty). An
29 understanding of uncertainty can be used:
30 »• To make more informed choices during the conduct of a risk assessment, and
31 *• To better characterize the range of plausible risk estimates for different populations.
32
33 4.5.7.1. Model Uncertainty
34 This assessment's exploration of multiple modeling approaches affords an opportunity to
35 understand and address different sources of model uncertainty. The full extent of model
8/1/01 4-29 DRAFT-DO NOT CITE OR QUOTE
-------
1 uncertainty cannot be quantified, only the models that have been analyzed. In this assessment,
2 several sources of model uncertainty have been discussed:
3 *• Two pharmacokinetic models initially led to risk estimates that differed by 15-fold (see
4 Table 4-4). To reduce this uncertainty, the models were fitted to additional data sets
5 (Bois, 2000a,b). This calibration improved the models and made the models' results
6 more compatible, reducing this source of model uncertainty (see Section 4.2.1).
7 *• A mechanism-based model assumption that TCE-induced liver tumor incidences are
8 proportional to a species' background liver tumor incidence results in 100-fold lower
9 human risk estimates (see Section 4.5.2). Because current mechanistic understanding
10 suggests that cross-species extrapolation is far more complex (see Section 3.5.1), this
11 assessment chose estimates not based on this assumption.
12 + Different exposure routes yield different proportions of metabolite formation. Depending
13 on whether TCA formation or DCA formation is used as the basis of comparison, route
14 extrapolations can differ by 25-fold (see Section 4.2.3). To convey an appreciation for
15 this unresolved uncertainty, this assessment has developed estimates using both
16 approaches.
17 *- Different exposure patterns73 may yield different risks, but this has not been well
18 investigated. In the absence of TCE-specific data, this assessment has followed the
19 common, neutral practice of assuming that equal risks result from equal cumulative
20 doses. There are some data, however, to suggest that this approach may underestimate
21 risks from short-duration exposures (see Section 4.5.2).
22 > The choice between linear and nonlinear extrapolation to environmental doses creates
23 two distinct classes of estimates (see Sections 4.5.2, 4.5.3). This dichotomous uncertainty
24 stems from different interpretations of the information on TCE's mode-of-action (see
25 Section 3.5). Some toxicologists maintain that cross-species differences would make
26 humans less sensitive to TCE according to the modes of action proposed for mice and
27 rats, suggesting less concern for human cancer. Other scientists maintain that such
28 conclusions are premature without more specific definition and experimental
29 confirmation of the active agents and sequences of key events leading to cancer. In
30 addition, some of TCE's effects on cell signaling may augment ongoing processes and
31 conditions in some groups of humans, leading to an increased cancer risk for some.
32 EPA's proposed cancer guidelines take the latter view, that the active agents and key
33 events be identified in order to establish the mode-of-action. Consequently, this
^Exposure patterns refer to magnitude, frequency, duration, and timing of exposure. For example, exposure to 30
ppm for 10 hr at age 30 and exposure to 300 ppm for 1 hr at age 1 are two different exposure patterns with a total
exposure of 300 ppm-hr.
8/1/01 4-30 DRAFT-DO NOT CITE OR QUOTE
-------
1 assessment has prepared estimates using both linear and nonlinear approaches, the former
2 to bound the risks and the latter to demonstrate the degree of uncertainty and value of
3 further mechanistic research. In addition, this assessment has made efforts to develop
4 risk estimates from human studies as an alternative to extrapolating from animal models.
5
6 4.5.7.2. Parameter Uncertainty
7 Uncertainty analyses and confidence intervals were developed for some of the key
8 parameters used in this assessment. Each description of parameter uncertainty assumes that the
9 underlying model is valid.
10 > Uncertainty in the pharmacokinetic data is reflected by 95% confidence intervals
11 spanning approximately 100-fold for the liver, 14,000-fold for the lung, and 5,000-fold
12 for the kidney (see Table 4-1, Section 4.2.2). Because of this high uncertainty for lung
13 and kidney, this assessment reverted to default animal-to-human dose-scaling models for
14 these organs. In all cases, median dose estimates were used in subsequent calculations.
15 > Uncertainty in the animal dose-response data is reflected by the ratio of ED 10s to
16 LED 10s. These generally do not exceed a factor of 2 (Rhomberg, 2000).
8/1/01 4-31 DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-1. Distributions of potential dose metrics from the uncertainty analysis of the
pharmacokinetic models
Ratio of dose estimate at each percentile to the
mediana
Span of
Dose metric GSD 1 2.5 10 50 90 97.5 99 95% CIb
Mouse liver TCA AUC 3.2 1/15 1/10 1/4 1 4 10 15 100
Mouse liver DCA AUC 3.6 1/20 1/12 1/5 1 5 12 20 144
Mouse lung CH AUC 11.7 1/300 1/120 1/23 1 23 120 300 14,400
Mouse lung CH max 12.5 1/360 1/140 1/25 1 25 140 360 19,600
Rat kidney thiol 9.0 1/170 1/74 1/17 1 17 74 170 5,476
"Assuming lognormally distributed errors.
bRatio of estimate at 97.5 percentile to estimate at 2.5 percentile.
Source: Adapted from Bois, 2000b, Table 5-7; Rhomberg, 2000, Table 15.
8/1/01 4-32 DRAFT-DO NOT CITE OR QUOTE
-------
oo
o
Table 4-2. Oral studies supporting development of an RfD
Study
Tucker etal. (1982)
and
Sanders etal. (1982)
Buben and O'Flaherty
(1985)
Herman etal. (1995)
and
Moser etal. (1995)
Maltoni etal. (1986)
Dawsonetal. (1993)
Narotsky etal. (1995)
Griffin et al. (2000c)
Species
Mouse
Mouse
Rat
Rat
Rat
Rat
Mouse
Duration;
exposure route
Subchronic
(6 months)
Oral - drinking
water
Subchronic
(6 weeks)
Oral - gavage
14 days
Oral - gavage
Chronic bioassay
(52 weeks
exposure, followed
until natural death)
Oral - gavage
Developmental
Oral - drinking
water
Developmental
Oral - gavage
Subchronic
(4 weeks and 32
weeks)
Oral - drinking
water
Exoerimental doses
(mg/kg/day)1'2
0, 184, 2175, 393, 660
(m)
0, 184, 1935, 437, 793
(f)
0, 100, 200, 400, 800,
1600, 2400, 3200
0, 506, 1507. 500, 1500
0, 50, 250
0, L58, HOOppm
-0.18, 129mg/ks/dav
0, 10, 32, 101, 320,
475, 633, 844, 1,125
0, 21, 100, 400
Organ;
effect
Liver:
liver weight-
body weight ratio
Immune system:
antibody (plaque)-
forming cell assay
Liver:
liver weight-
body weight ratio
Liver:
liver weight-
body weight ratio
Neurobehavioral
Kidney:
renal
megalonucleocytosis
Developmental:
cardiac anomalies
Developmental:
anopthalmia/
micropthalmia
Immune system/liver:
serum antinuclear
antibodies (ANA),
histopathologcial
changes, serumalanine
aminotransferase
Human-equivalent dose2'3
NOAEL,PK-adj. 1 mg/kg-d
NOAEL 29 mg/kg-d
LOAEL,PK-adj. 1 mg/kg-d
LOAEL,PK-adj. 0.3 mg/kg-d
BMD10, PK-adj.
(Weibull) 0.6 mg/kg-d
(Power) 0.5 mg/kg-d
NOAEL 40 mg/kg-d
NOAEL 10 mg/kg-d
LOAEL 34 mg/kg-d
NOAEL 0.05 mg/kg-d
NOAEL 9 mg/kg-d
BMD10 133 mg/kg-d
LOAEL 4 mg/kg-d
H
W
O
c
o
H
W
1 NOAELs or LOAELs are underlined.
2 From Barton and Clewell (1999).
3
Moser etal. (1995).
! NOAEL.
Human equivalent doses calculated using RfD methodology (Barnes and
Dourson, 1988) and scaling to the ratio of body weights to the 0.75 power.
4 Tucker etal. (1982).
5 Saunder et al. (1982).
6 Berman et al. (1995).
-------
Table 4-3. Inhalation studies supporting development of an RfC
Study
Kjellstrandetal. (1983)
Chiaetal. (1997)
Gohetal. (1998)
Chiaetal. (1996)
Maltonietal. (1986)
Ruijtenetal. (1991)
Rasmussen et al. (1993)
Vandervort and
Pelakoff(1973)
Okawa and Bodner
(1973)
Aritoetal. (1994)
Kumar et al. (2000)
Species
Mouse
Human
Rat
Human
Human
Human
Human
Rat
Rat
Duration;
exposure route
Subchronic
(30 days)
inhalation
Subchronic
(mean =5 years)
inhalation
Chronic bioassay
(104 weeks)
inhalation
Subchronic
(mean =16 years)
inhalation
Subchronic
(mean = 7 years)
inhalation
Subchronic
(8 years)
inhalation
Subchronic
inhalation
Subchronic
(6 weeks)
inhalation
Subchronic
(12 weeks and 24
weeks)
inhalation
Concentration1
0, 37, 75, 150, 300 ppm
30 ppm (mean)
0, 100. 300, 600 ppm
44 ppm (calculated
from mean cumulative
exposure)
40 - 60 mg/1 urinary
TCA
- - 20 ppm5
32 ppm
71 mg/1 urinary TCA
(average)
- -20-30 ppm5
0, 50, 100, 600 ppm
0, 376 ppm
Organ;
effect
Liver:
liver weight-
body weight ratio
Endocrine/Liver:
FSH, testosterone, DHEAS4,
FAI4, SHBG4, insulin
Kidney:
megalonucleocytosis
CNS:
massiter reflex latency
CNS:
motor dyscoordination
CNS:
drowiness, fatigue,
dizziness, eye irritation
CNS:
nausea, headache, dizziness,
fatigue, upper respiratory
irritation
CNS:
heart rate,
electroencephalographic
changes
Endocrine/Reproductive :
testosterone, testicular
cholesterol, 17 B-hydroky
steroid dehydrogenase,
glucose 6-P- dehydrogenase,
sperm number and motility
Human-equivalent
concentration2'3
LOAEL,PK-adj. 12 ppm
BMD10, PK - adj . (f) 5 ppm
(m) 9,11 ppm
LOAEL 11 ppm
NOAEL,PK-adj. 21 ppm
LOAEL 16 ppm
LOAEL 7 ppm
LOAEL 11 ppm
LOAEL 7- 11 ppm
LOAEL,PK-adj. 9 ppm
BMD10, PK-adj. 5 ppm
(wakefulness)
LOAEL 41 ppm
1 NOAELs or LOAELs are underlined.
2 Human equivalent doses calculated using RfC methodology (U.S. EPA,
1994) for a category 3 gas, extrathoracic effects.
NOAEL* [HEC] (mg/m3) = NOAEL* [ADJ] (mg/m3) x [(H b/g) A / (H b/g) H].
The value of 1.0 is used if (H b/g) A > (H b/g) H. (H b/g) A= 15.9, mouse
(Fisher, 2000); 14, mouse, 18.5, rat (Clewell et al., 2000). (H b/g) H = 9.13,
male, 11.15, female (Fisher, 2000); 9.2 (Clewell et al., 2000).
3 HECs for mice and rats are from Barton and Clewell (1999).
4 DHEAS - dihydroepiandrosterone sulphate, FAI - free androgen index,
SHBG - sex hormone binding globin.
5 U-TCA interpolated to atmospheric concentration assuming 50 mg/1 U-
TCA is equivalent to 20 ppm (Axelson et al., 1994).
-------
Table 4-4. Human-equivalent LED10s (mg/kg-d) derived from liver tumor incidences in
four mouse studies (NTP males or females, NCI males or females), using two dose metrics
(TCA area-under-the-curve, DCA area-under-the-curve) and four pharmacokinetic models
(Fisher 2000; Bois, 2000a; Clewell et al., 2000; Bois, 2000b)
Based on TCA area-under-the-curve Based on DCA area-under-the-curve
Bois/ Bois/ Bois/
Fisher Fisher Clewell Clewell Fisher Fisher Clewell Bois/Clewell
Male mice (NTP) 0.19 1.47 0.077 0.23 0.77 NA 12.05 1.15
Female mice
(NTP) 0.33 3.03 0.15 0.45 1.39 NA 24.39 2.33
Male mice (NCI) 0.83 4.76 0.24 0.83 3.45 NA 40.00 4.35
Female mice
(NCI) 2.63 18.52 1.01 3.57 10.75 NA 169.49 18.18
Mean1of4exp'ts 0.40 3.14 0.16 0.50 1.67 NA 25.82 2.52
NA = not available.
Harmonic mean = N/(Ł 1/^Q; equivalent to averaging slope factors (which are inversely
proportional to LED10s), then taking the reciprocal to convert back to a dose.
Source: Adapted from Rhomberg (2000) (LED10s are computed as 0.10 divided by slope factors
in Rhomberg, 2000, Table 4).
8/1/01 4-35 DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-5. Extending the observed range with a study of less-than-lifetime dosing of TCE
Concentration
(ppm)
0
100
600
0
100
300
600
Duration
(wk)
8
8
8
78
78
78
78
Exposure index
(cone x dur)
0
800
4,800
0
7,800
23,400
46,800
Hepatoma incidence
1/41 (2%)
3/36 (8%)
4/23 (17%)
4/66 (6%)
2/53 (4%)
8/59 (14%)
13/61 (21%)
Source: Adapted from Maltoni et al., 1986.
8/1/01 4-36 DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-6. Extending the observed range with studies of TCE's metabolites
Compound
DCA
TCA
TCE
Dose
(DCA AUC)
1.38
14.01
63.66
88.4
0.07
7.12
40.8
67
71
67
Liver tumor incidence
3/21 (14%)
1/18 (5%)
8/12 (67%)
25/30 (83%)
2/9 (22%)
8/21 (38%)
21/24 (88%)
(61%)
(86%)
(65%)
Source: Adapted from Chen, 2000.
8/1/01
4-37
DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-7. Human-equivalent LEC10s (ug/m3) derived from
lung tumor incidences in four mouse studies, using RfC
dosimetry
LEC
10
Female ICR mice (Fukuda) 1 AX 10s
Male Swiss mice (Maltoni) 1.3 x 105
Female Swiss mice (Maltoni) 3.6x 10s
Male B6C3F1 mice (Maltoni) 2.8x 10s
Mean of 4 experiments 1.9x 10s
Source: Adapted from Rhomberg, 2000. (LEC10s are computed as
0.10 divided by unit risks in Rhomberg, 2000, Table 9).
8/1/01 4-38 DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-8. Human-equivalent LED01s (mg/kg-d) derived from kidney tumor incidences in
rat studies, using 3/4-power scaling
Based on cross-species scaling
factor of 3/4-power of relative
body weight
Male and female rats, 5 strains pooled (NTP), gavage,
adenocarcinomas only 6.7X101
Male and female rats, 5 strains pooled (NTP), gavage,
adenocarcinomas and adenomas1 3.3x 101
Male Sprague-Dawley rats (Maltoni), inhalation,
adenocarcinomas only N/A
JNTP guidance (McConnell et al., 1986) recommends combining adenocarcinomas and
adenomas; Rhomberg (2000) Table 12 states that including adenomas yields values about half
those for adenocarcinomas alone.
Source: Adapted from Rhomberg, 2000. (LED01s are computed as 0.01 divided by slope factors
in Rhomberg, 2000, Table 13.)
8/1/01 4-39 DRAFT-DO NOT CITE OR QUOTE
-------
Table 4-9. Compilation of cancer estimates
Point of departure
(mg/kg-d)
Slope factor
(mg/kg-d)-1
Risk-specific dosea
(mg/kg-d)
Cancer estimates based on human studies
Liver cancer
Finnish cohort13
1.4C
7x10
,-2
1.4x10
,-5
Kidney cancer
Finnish cohort13
German cohort
Non-Hodgkin's lymphoma
Finnish cohort13
New Jersey cohort
0.05C
5C
0.014C
0.25C
2>
2>
7>
4>
ao°
a o-2
ao°
ao-1
5
5
1
2
xlO~7
xlO-5
,4xlO~7
.5xlO~6
Cancer estimates based on mouse studies
Liver cancer
Mechanism-based modeld
Mechanism-based model6
Linear extrapolation
Nonlinear extrapolation
Lung cancer8
Not applicable
Not applicable
0.5-3.1
0.5-3.1
1.7-4.8
SxlO^xlO-1
Not applicable
Not applicable
1.25xlO~3
1.25xlO~5
0.5-3. lxlO~5
(3xlO~4/
(Not calculable/
Cancer estimates based on rat studies
Kidney cancer
Testicular cancer
33h
25
,-4
3x10
Not indicated
3.3x10
,-3
aDose associated with an upper-bound increased cancer risk of 1 in 1 million, calculated as 10 /slope factor. When
nonlinear low-dose extrapolation is used, this is a dose considered to pose minimal cancer risks, calculated as the
point of departure divided by the composite risk reduction factor from a nonlinear dose-response analysis.
bFor the Finnish cohort, the slope factors are based on route extrapolation from air unit risks, using a correspondence
between inhaled and oral doses giving equal TCA area-under-the-curve (75 mg/m3 ~ 1 mg/kg-d, see Section 5.2.3).
Air unit risks for liver cancer, kidney cancer, and non-Hodgkin's lymphoma are 9 x 1CT7, 3 x 1CT5, and 9 x 1CT5 per
ug/m3, respectively.
The point of departure (anLED10) is calculated from the slope factor (LED10=0.10/slope factor).
dFrom a mechanism-based model, assuming that tumor response is proportional to background liver tumor incidence
(Chen, 2000).
Trom a mechanism-based model (see note m), without assuming that liver tumor response is proportional to
background liver tumor incidence.
This is a dose considered to pose minimal cancer risks, calculated as the point of departure divided by the composite
risk reduction factor from a nonlinear dose-response analysis. This is not a dose associated with a 1-in-l million
cancer risk.
The point of departure is based on route extrapolation from the point of departure for an inhalation study in mice
using a correspondence between inhaled and oral doses giving equal TCA area-under-the-curve [75 mg/m3 ~ 1
mg/kg-d, see Section 5.2.3]. The point of departure for mouse lung tumors (anLEC10) spans a range of 1.3*105 to
3.6xl05ug/m3.
The point of departure is an LED01 rather than an LED10 (see Section 5.5.3).
8/1/01
4-40
DRAFT-DO NOT CITE OR QUOTE
-------
Figure 4-1. Extending the observed range with less-than-lifetime dosing
50%
40%
30%
20%
10%
0%
8 wks -•- 78 wks
10 20 30 40
Exposure index(concentration x duration)
50
8/1/01
4-41 DRAFT-DO NOT CITE OR QUOTE
-------
Figure 4-2. Extending the observed range with studies of TCE's metabolite
s.
o
100%
80%
60%
40%
20%
0%
20
DCA -*- TCA n TCE
n
40 60
AUC DCA
80
100
8/1/01
4-42 DRAFT-DO NOT CITE OR QUOTE
-------
Figure 4-3. TCE health benchmarks
11
••• Risk- specific doste Point of departure
o
o
O
•— i
0
C 0
. — .
}
o
o
o
Mid- range
o RfD
o RfC
» Kidne)
* Non-Hod^
•
4—
«—
<
of cancer
<
»
* Liver cance
cancer (Finnisl
«— > Kid
kin's lymphom;
Non-Hodgkiri:
—* Liver c;
—+ Liver c
—* Lung cance
o
0
risk- spec
>
r (Finnish worki
i workers)
ney cancer (pol
(Finnish works
5 lymphoma (N
mcer (range in i
ancer (range in
r (range in mice
Sidney cancer (
festicular canes
ific doses
jrs)
iTnorphism in G
rs)
Jersey drinking
nice, linear extr
mice, nonlinear
, no extrapolatic
rats, linear extn
r (rats, nonlinec
(excludii
erman workers
water exposure
apolation)
extrapolation)
m)
polation)
r extrapolation)
igr highest
:-8 1E-6 1E-4 1E-2 1E+0 1E+2 1E+4 1E+6 1E+8
TCE dose (mg/kg-d)
8/1/01
4-43
DRAFT-DO NOT CITE OR QUOTE
-------
5. RESEARCH NEEDS
1 Although the database for TCE is rather extensive and much has been learned in recent
2 years, this is a time when new research could significantly increase understanding and lead to
3 later refinement of this health risk assessment. Many research needs were discussed in the state-
4 of-the-science papers that comprise volume 1 of this assessment, and this research promises to
5 increase understanding of TCE's adverse health effects, the active agents and modes of action for
6 each effect, and their relevance to human environmental exposure. This section focuses on those
7 research needs that left this assessment with open questions that were filled by default
8 assumptions and methods. Further research into these areas could, therefore, have direct bearing
9 on specific sections of this assessment. The practical importance of describing these research
10 needs is not to decrease confidence in this assessment, as all assessments have areas that can be
11 improved by further research, but to be explicit about what specific research could help resolve
12 the major open questions in this assessment.
13
14 5.1. RESEARCH NEEDS FOR HAZARD ASSESSMENT
15 (a) TCE's effects other than cancer have not been adequately identified or studied. There is a
16 lack of definitive chronic studies of effects other than cancer (see Section 3.4).
17 Investigation into the modes of action for effects other than cancer lags behind what has
18 been developed for TCE-induced cancer (see Section 3.5).
19 (b) With EPA's proposed cancer guidelines becoming more specific (U.S. EPA, 1999), it is
20 now apparent that what is needed is a description of the sequence of key events for each
21 tumor type and the dose levels at which those events begin to occur. Although there has
22 been much progress in developing and investigating several plausible modes of action,
23 the active agents and key events have not yet been identified (see Section 3.5).
24 (c) For TCE-induced liver cancer, further research is needed to identify the key events and
25 the doses at which they begin to occur. Characterizing biological markers of the liver
26 tumors from the TCE bioassays would promote understanding of the respective roles and
27 interaction of TCA and DC A in the TCE-induced tumors (see Section 3.5.1).
28 (d) Fuller identification of the mode-of-action and associated key events for kidney tumors,
29 especially with respect to the potential role of VHL mutations in human kidney cancer,
30 could affect the description of TCE as either "carcinogenic to humans" or "likely to be
31 carcinogenic to humans." Replication of recent findings in another cohort showing
32 kidney tumors could warrant a future description of TCE as "carcinogenic to humans"
33 (see Section 1.3).
34
8/1/01 5-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 5.2. RESEARCH NEEDS FOR DOSE-RESPONSE ASSESSMENT
2 (a) Route extrapolations can differ by 25-fold, depending on whether internal TCA or DCA
3 is used as the dose metric. Further research could identify the appropriate internal dose
4 metric for each toxic effect (see Section 4.2.3).
5 (b) There is a need for consensus on a preferred empirical model to be used for each type of
6 data set. This assessment's experience with benchmark dose modeling revealed
7 differences of more than an order of magnitude depending on which empirical model was
8 used. This invites "model shopping" and introduces too much uncertainty for use in risk
9 assessments that support public health decisions (see Section 4.3.2, Table 4-2).
10 (c) The RfD includes a data-derived factor of 50 for human variation and default factors of
11 100 overall for uncertainty in extrapolating from animals to humans, from subchronic
12 studies to lifetime exposure, from effect levels to NOAELs, and from single-chemical
13 toxicity tests to complex exposures involving multiple chemicals. This latter 100-fold
14 factor indicates the potential for future research to reduce uncertainty and improve the
15 assessment's accuracy (see Section 4.3.3. 4.3.4).
16 (d) There is considerable pharmacokinetic and pharmacodynamic uncertainty for children.
17 An approach is needed for evaluating the need for and size of an additional factor to
18 ensure that the RfD and RfC are protective of children, while also providing incentives
19 for research to quantify the potential difference in sensitivity between children and adults.
20 (e) At low doses, the mode-of-action for liver tumors may be nonlinear; nonetheless, current
21 data do not identify how far below the LED10s sublinearity begins to prevail. Further
22 research is needed to identify the key precursor events and describe their dose-response
23 curves (see Section 4.5.2).
24 (f) The mode-of-action for lung tumors is undeveloped with respect to developing dose-
25 response curves for the key events. In addition, there is substantial parameter uncertainty
26 in current pharmacokinetic estimates of a dose metric in the lung (see Section 4.5.3).
27 (g) Similarly, there is substantial parameter uncertainty in current pharmacokinetic estimates
28 of a dose metric in the kidney (see Section 4.5.4).
29 (h) This assessment identifies several risk factors that would make a population more
30 sensitive. Further research is needed to estimate the magnitude of the increased risk.
31 Understanding the functional relationship between a risk factor and the associated disease
32 is a critical research need that presently prevents estimating the differential risk faced by
33 sensitive populations (see Sections 2.4, 2.5, 2.6, 4.5.6).
34 (i) It may be important to understand the joint effect of several risk factors together. If the
35 effects of several factors are multiplicative, the combined effect could be quite large (see
36 Section 4.5.6).
8/1/01 5-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 (j) Risk assessment guidelines and methods will need to explicitly consider the effect of
2 background exposures as risk assessments consider the cumulative effect of multiple
3 chemicals and stressors on an exposed population. Finding a safe dose in an otherwise
4 unexposed population does not mean that that dose is safe when background exposures
5 are considered (see Section 1.6).
6
7 5.3. RESEARCH NEEDS FOR EXPOSURE ASSESSMENT
8 (a) Limited data indicate that TCE has been detected in all of eight samples of human milk
9 from four U. S. urban areas. Quantification of levels of TCE and its metabolites in milk,
10 in both highly exposed and relatively unexposed populations, is a critical research need
11 that presently prevents comparing levels in milk with levels allowed in drinking water
12 (see Section 1.5).
13 (b) As TCE's metabolites have a large role in TCE-induced toxicity, there is a need to
14 estimate cumulative exposure to these metabolites from all parent compounds, sources,
15 and pathways (see Section 1.6).
8/1/01 5-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 REFERENCES
2
3 Adami, HO; Chow, WH; Nyrene, O; et al. (1996) Excess risk of primary liver cancer in patients with diabetes
4 mellitus. J Natl Cancer Inst 88:1472-1477.
5
6 Anderson, PM; Schultze, MO. (1965) Cleavage of s-(l,2-dichlorovinyl)-L-cysteine by an enzyme of bovine origin.
7 Arch Biochem Biophys 111:593-602.
8
9 Anderson, WB; Board, PG; Gargano, B; etal. (1999) Inactivation of glutathione transferase zeta by dichloroacetic
10 acid and other fluorine-lacking • »haloalkanoic acids. Chem Res Toxicol 12:1144-1149.
11
12 Anttila, A; Pukkala, E; Sallmen, M; et al. (1995) Cancer incidence among Finnish workers exposed to halogenated
13 hydrocarbons. J Environ Occup Med 37:797-806.
14
15 Aranyi, C; O'Shea, WJ; Graham, JA; etal. (1986) The effects of inhalation of organic chemical air contaminants on
16 murine lung host defenses. Fundam Appl Toxicol 6:713-720.
17
18 Arito, H; Takahashi, M; Ishikawa, T. (1994) Effect of subchronic inhalation exposure to low-level trichloroethylene
19 on heart rate and wakefulness-sleep in freely moving rats. Jpn J Ind Health 36:1-8.
20
21 AT SDR (Agency for Toxic Substances and Disease Registry). (1997) Toxicological Profile for Trichloroethylene.
22 Atlanta: Agency for Toxic Substances and Disease Registry.
23
24 Bannasch, P. (1996) Pathogenesis of hepatocellular carcinoma: sequential cellular, molecular, and metabolic
25 changes. Prog Liver Dis 16:161-197.
26
27 Bannasch, P; Hacker, HJ; Klimec, F; et al. (1984) Hepatocellular glycogenosis and related pattern of enzymatic
28 changes during hepatocarcinogenesis. Adv Enzyme Regul 22:97-121.
29
30 Bannasch, P; Hacker, HJ; Tsuda, H; et al. (1986) Aberrant regulation of carbohydrate metabolism and
31 metamorphosis during renal carcinogenesis. Adv Enzyme Regul 25:279-296.
32
33 Bannasch, P; Klimek, F; Mayer, D. (1997) Early bioenergetic changes in hepatocarcinogenesis: preneoplastic
34 phenotypes mimic responses to insulin and thyroid hormone. J Bioenerget Biomembranes 29(4):303-313.
35
36 Bardodej, Z; Vyskocil, J. (1956) The problem of trichloroethylene in occupational medicine: trichlororethylene
37 metabolism and its effect on the nervous system evaluated as a means of hygienic control. AMA Arch Ind Health
38 56:581-592.
39
40 Barnes, D; Dourson, M. (1988) Reference dose (RID): description and use of health risk assessments. Regul
41 Toxicol Pharmacol 8:471-486.
42
43 Barton, HA; Clewell, HJ III. (2000) Evaluating noncancer effects of trichloroethylene: dosimetry, mode-of-action,
44 and risk assessment. Environ Health Perspect 108(suppl 2)323-334.
45
46 Barton, HA; Das, S. (1996) Alternatives for a risk assessment on the chronic noncancer effects from oral exposure
47 to trichloroethylene. Regul Toxicol Pharmacol 24:269-285.
48
49 Barton, HA; Bull, R; Schultz, I; et al. (1999) Dichloroacetate (DCA) dosimetry: interpreting DCA-induced liver
50 cancer dose response and the potential for DCA to contribute to trichloroethylene-induced liver cancer. Toxicol Lett
51 106:9-21.
52
53 Berman, E; Schlicht, M; Moser, VC; et al. (1995) A multidisciplinary approach to lexicological screening: I.
54 Systemic toxicity. J Toxicol Environ Health 45:127-143.
55
8/1/01 R-1 DRAFT-DO NOT CITE OR QUOTE
-------
1 Bhat, HK; Kanz, MF; Campbell, GA; et al. (1991) Ninety day toxicity study of chloroacetic acid in rats. Fundam
2 ApplToxicol 17:240-253.
3
4 Bhattacharya, RK; Schultze, MO. (1967) Enzyme from bovine and turkey kidneys which cleaves
5 S-(l,2-dichlorovinyl)L-cysteine. Comp Biochem Physiol 22:723-735.
6
7 Blair, A; Hartge, P; Stewart, PA; et al. (1998) Mortality and cancer incidence of aircraft maintenance workers
8 exposed to trichloroethylene and other organic solvents and chemicals: extended follow up. Occup Environ Med
9 55:161-171.
10
11 Boers, JE; Ambergen, AW; Thunnissen, FBJ. (1999) Number and proliferation of clara cells in normal human
12 airway epithelium. Am J Respir Crit Care Med 159:1585-1591.
13
14 Bois, FY. (2000a) Statistical analysis of Fisher et al. PBPK model of trichloroethylene kinetics. Environ Health
15 Perspect 108(suppl 2)275-282.
16
17 Bois FY. (2000b) Statistical analysis of Clewell et al. PBPK model of trichloroethylene kinetics. Environ Health
18 Perspect 108(suppl 2)307-316.
19
20 Bosland, MC; Ford, H; Horton, L. (1995) Induction at high incidence of ductal prostate adenocarcinomas in
21 NBC/CR and Sprague-Dawley HSD:SD rats treated with a combination of testosterone and estradiol, 17-beta or
22 diethylstilbestrol. Carcinogenesis 16:1311-1317.
23
24 Bove, FJ; Fulcomer, MC; Klotz, J; et al. (1992) Public drinking water contamination and birth outcomes. Am J
25 Epidemiol 141:850-862.
26
27 Bovenzi, M; Barbone, F; Betta, A; et al. (1995) Scleroderma and occupational exposure. Scand J Work Environ
28 Health 21:289-292.
29
30 Boyer, AS; Finch, WT; Runyan, RB. (2000) Trichloroethylene inhibits development of embryonic heart valve
31 precursors in vitro. Toxicol Sci 53:109-117.
32
33 Boyes, WK; Bushnell, PJ; Crofton, KM; et al. (2000) Neurotoxic and pharmacokinetic responses to
34 trichloroethylene as a function of exposure scenario. Environ Health Perspect 108(suppl 2)317-322.
35
36 Brauch, H; Weirich, G; Hornauer, M; et al. (1999) Trichloroethylene exposure and specific somatic mutations in
3 7 patients with renal cell carcinoma. J Natl Cancer Inst 91:854-861.
38
39 Briining, T; Lammert, M; Kempkes, M; et al. (1997a) Influence of polymorphisms of GSTM1 and GSTT1 for risk
40 of renal cell cancer in workers with long-term high occupational exposure to trichloroethene. Arch Toxicol
41 71:596-599.
42
43 Briining, T; Weirich, G; Hornauer, MA; et al. (1997b) Renal cell carcinomas in trichloroethene (TRI) exposed
44 persons are associated with somatic mutations in the VHL (VHL) tumour suppressor gene. Arch Toxicol
45 71:332-335.
46
47 Briining, T; Mann, H; Melzer, H; et al. (1999a) Pathological excretion patterns of urinary proteins in renal cell
48 cancer patients exposed to trichloroethylene. Occup Med (Lond) 49:299-305.
49
50 Briining, T; Sundberg, AG; Birner, G; et al. (1999b) Glutathione transferase alpha as a marker for tubular damage
51 after trichloroethylene exposure. Arch Toxicol 73:246-254.
52
53 Buben, JA; O'Flaherty, EJ. (1985) Delineation of the role of metabolism in the hepatotoxicity of trichloroethylene
54 and perchlorethylene: a dose-effect study. Toxicol Appl Pharmacol 78:105-122.
55
8/1/01 R-2 DRAFT-DO NOT CITE OR QUOTE
-------
1 Bull, RJ. (2000) Mode-of-action of liver tumor induction by trichloroethylene and its metabolites, trichloroacetate
2 and dichloroacetate. Environ Health Perspect 108(suppl 2):241-259.
3
4 Burg, JA; Gist, GL. (1997) The potential impact on women from environmental exposures. J Womens Health
5 6:159-161.
6
7 Bushnell, PJ. (1997) Concentration-time relationships for the effects of inhaled trichloroethylene on signal
8 detection behaviour in rats. Fundam Appl Toxicol 36:30-38.
9
10 Byers, VS; Levin, AS; Ozonoff, DM; et al. (1988) Association between clinical symptoms and lymphocyte
11 abnormalities in a population with chronic domestic exposure to industrial solvent-contaminated domestic water
12 supply and a high incidence of leukemia. Cancer Immunol Immunother 27:77-81.
13
14 Cai, H; Guengerich, P. (2000) Acylation of protein lysines by trichloroethylene oxide. Chem Res Toxicol
15 13:327-335.
16
17 Cal/EPA. (February 1999) Public Health Goal for Trichloroethylene in Drinking Water. Office of Environmental
18 Health Hazard Assessment, California Environmental Protection Agency.
19
20 Cal/EPA (California Environmental Protection Agency). (June 1999) Air Toxics Hot Spots Program Risk
21 Assessment Guidelines Part III: Technical Support Document for the Determination of Noncancer Chronic
22 Reference Exposure Levels and OEHHA's Responses to Public Comments. Chronic Toxicity Summary for
23 Trichloroethylene. http://www.oehha.ca.gov/hotspots/.
24
25 Caldwell-Kenkel, JC; Currin, RT; Tanaka, Y; et al. (1991) Kupffer cell activation and endothelial cell damage after
26 storage of rat livers: effects of reperfusion. Hepatology 13(l):83-95.
27
28 Calle, EE; Murphy, TK; Rodriguez, C; et al. (1998) Diabetes mellitus and pancreatic cancer mortality in a
29 prospective cohort of United States adults. Cancer Causes Control 9:403-410.
30
31 Cattley, RC; DeLuca, J; Elcombe, C; et al. (1998) Do peroxisome proliferating compounds pose a
32 hepatocarcinogenic hazard to humans? Regul Toxicol Pharmacol 27:47-60.
33
34 Chen, CW. (2000) Biologically based dose-response model for liver tumors induced by trichloroethylene. Environ
35 Health Perspect 108(suppl 2):335-342.
36
37 Chevalier, S; Roberts, RA. (1998) Perturbation of rodent hepatocyte growth control by nongenotoxic
38 hepatocarcinogens: mechanisms and lack of relevance for human health (review). Oncol Reports 5:1319-1327.
39
40 Chia, SE; Ong, C-N; Tsakok, MF; et al. (1996) Semen parameters in workers exposed to trichloroethylene. Reprod
41 Toxicol 10:295-299.
42
43 Chia, SE; Goh, VHH; Ong, CN. (1997) Endocrine profiles of male workers with exposure to trichloroethylene. Am
44 JIndMed 32:217-222.
45
46 Chiu, KM; Schmidt, MJ; Havig Hurst, TC; et al. (1999) Correlation of serum L-carnitine and
47 dehydroepiandrosterone sulphate levels with age and sex in healthy adults. Age, Aging 28:89-91.
48
49 Cicmanec, JL; Condie, LW; Olson, GR; et al. (1991) 90-Day toxicity study of dichloroacetate in dogs. Fundam
50 Appl Toxicol 17:376-389.
51
52 Clewell, HJ III; Gentry, PR; Covington, TR; et al. (2000) Development of a physiologically based pharmacokinetic
53 model of trichloroethylene and its metabolites for use in risk assessment. Environ Health Perspect 108(suppl
54 2):283-305.
55
8/1/01 R-3 DRAFT-DO NOT CITE OR QUOTE
-------
1 Cohn, P; Klotz, J; Bove, F; et al. (1994) Drinking water contamination and the incidence of leukemia and non-
2 Hodgkin's lymphoma. Environ Health Perspect 102:556-561.
3
4 Cook, JC; Klinefelter, GR; Hardisty, JF; et al. (1999) Rodent Leydig cell tumorogenesis: a review of the
5 physiology, pathology, mechanisms and relevance to humans. Crit Rev Toxicol 29:169-261.
6
7 Cooper, GS; Miller, FW; Pandey, JP. (1999) The role of genetic factors in autoimmune disease: implication for
8 environmental research. Environ Health Perspect 107:693-700.
9
10 Cornett, R; James, MO; Henderson, GN; et al. (1999) Inhibition of glutathione s-transferase zeta and tyrosine
11 metabolism by dichloroacetate: a potential unifying mechanism for its altered biotransformation and toxicity.
12 Biochem Biophys Res Commun 262:752-756.
13
14 Counts, JL; Goodman, JI. (1995) Hypomethylationof DNA: a nongenotoxic mechanism involved in tumor
15 promotion. Toxicol Lett 82-83:663-672.
16
17 Crump, KS; Hoel, DG; Langley, CH; et al. (1976) Fundamental carcinogenic processes and their implications for
18 low dose risk assessment. Cancer Res 36:2973-2979.
19
20 Daniel, FB; DeAngelo, AB; Stober, JA; etal. (1992). Hepatocarcinogenicity of chloral hydrate,
21 2-chloroacetaldehyde, and dichloroacetic acid in male B6C3F1 mouse. Fundam Appl Toxicol 19:159-168.
22
23 Danielsson, BR. (1990) Effects on Reproduction of Trichloroethylene. Kem I report. Nordic Chemicals Control
24 Group, 13-31.
25
26 Dawson, BV; Johnson, PD; Goldberg, SJ; et al. (1993) Cardiac teratogenesis of halogenated hydrocarbon-
27 contaminated drinking water. J Am Coll Cardiol 21:1466-1472.
28
29 D'Cruz, D. (2000) Autoimmune diseases associated with drugs, chemicals and environmental factors. Toxicol Lett
30 112-113:421-432.
31
32 DeAngelo, AB; Daniel, FB; McMillan, L; etal. (1989) Species and strain sensitivity to the induction of peroxisome
33 proliferation by chloroacetic acids. Toxicol Appl Pharmacol 101:285-298.
34
35 DeAngelo, AB; Daniel, FB; Most, BM; et al. (1996) The carcinogenicity of dichloroacetic acid in the male Fischer
36 344 rat. Toxicology 114:207-221.
37
38 DeAngelo, AB; George, MH; House, DE. (1999) Hepatocarcinogenicity in the male B6C3F1 mouse following a
39 lifetime exposure to dichloroacetic acid in the drinking water: dose-response determination and modes of action. J
40 Toxicol Environ Health 58:485-507.
41
42 Dosemeci, M; Cocco, P; Chow, WH. (1999) Gender differences in risk of renal cell carcinoma and occupational
43 exposures to chlorinated aliphatic hydrocarbons. Am J Ind Med 36:54-59.
44
45 Dow, JL; Green, T. (2000) Trichloroethylene induced vitamin B(12) and folate deficiency leads to increased formic
46 acid excretion in the rat. Toxicology 146:123-136.
47
48 Driscoll, TR; Hamdan, HH; Wang, G; et al. (1992) Concentration of individual serum or plasma bile acids in
49 workers exposed to chlorinated aliphatic hydrocarbons. Br J Ind Med 49:700-705.
50
51 Eaton, DL; Bammler, TK (1999) Concise review of the glutathione s-transferases and their significance to
52 toxicology. Toxicol Sci 49:156-164.
53
54 Elcombe, CR. (1985) Species differences in carcinogenicity and peroxisome proliferation due to trichloroethylene:
55 a biochemical human hazard assessment. Arch Toxicol (suppl 8):6-17.
56
8/1/01 R-4 DRAFT-DO NOT CITE OR QUOTE
-------
1 Elcombe, CR; Rose, MS; Pratt, IS. (1985) Biochemical, histological, and ultrastructural changes in rat and mouse
2 liver following the administration of trichloroethylene: possible relevance to species differences in
3 hepatocarcinogenicity. Toxicol Appl Pharmacol 79:365-376.
4
5 El-Masri, HA; Mumtaz, MM. (2000) A test for reasonableness of default values for environmental chemicals using
6 Monte-Carlo simulations and physiologically based pharmacokinetic (PBPK) models. Toxicol Sci 54(1): 183.
7
8 Executive Order on Protection of Children from Environmental Health Risks and Safety Risks.
9 http://www.epa.gov/children/executiv.htm. April 21, 1997.
10
11 Epstein, DL; Nolen, G; Randall, JL; et al. (1993) Cardiopathic effects of dichloroacetate in the Long-Evans rat
12 fetus. Teratology 47:529-529.
13
14 Fagliano, J; Berry, M; Bove, F; et al. (1990) Drinking water contamination and the incidence of leukemia: an
15 ecologic study. Am J Public Health 80:1209-1212.
16
17 Falk, H. (1982) Liver. In: Cancer Epidemiology and Prevention. Schottenfeld, D; Fraumeni, JF Jr, eds.
18 Philadelphia: W.B. Saunders, pp. 668-682.
19
20 Farland, WH. (1992) Similarities and differences between children and adults: future directions and research needs.
21 In: Similarities and differences between children and adults: implications for risk assessment. Guzelian, PS; Henry,
22 CJ; Olin, SS, eds. Washington: ILSI Press, pp. 273-280.
23
24 Ferencz, C; Loffredo, CA; Correa-Villasenor, A; et al. (1997) Genetic and Environmental Risk Factors of Major
25 Cardiovascular Malformation. The Baltimore-Washington Infant Study 1981-1989. Vol 5: Perspective in Pediatric
26 Cardiology. New York: Futura Publishing.
27
28 Fernandez-Canon, JM; Panalva, MA. (1998) Characterization of a fungal maleylacetoacetate isomerase gene and
29 identification of its human homologue. J Biol Chem 273:329-337.
30
31 Fisher, JW. (2000) Physiologically based pharmacokinetic models for trichloroethylene and its oxidative
32 metabolites. Environ Health Perspect 108(suppl 2):265-273.
33
34 Fisher, JW; Whittaker, TA; Taylor, DH. (1989) Physiologically based pharmacokinetic modeling of the pregnant
35 rat: a multiroute exposure model for trichloroethylene and its metabolite, trichloroacetic acid. Toxicol Appl
36 Pharmacol 99:395-414.
37
38 Fisher, JW; Mahle, D; Abbas, R. (1998) A human physiologically based pharmacokinetic model for
39 trichloroethylene and its metabolites, trichloroacetic acid and free trichloroethanol. Toxicol Appl Pharmacol
40 152:339-359.
41
42 Fredriksson, A; Danielsson, BRG; Eriksson, P. (1993) Altered behaviour in adult mice orally exposed to tri-and
43 tetrachloroethylene as neonates. Toxicol Lett 66:13-19.
44
45 Fukuda, K; Takemoto, K; Tsuruta, H. (1983) Inhalation carcinogenicity of trichloroethylene in mice and rats. Ind
46 Health 21:243-254.
47
48 Ghantous, H; Danielson, BR; Denecker, L; et al. (1986) TCA accumulates in murine amniotic fluid after TCE
49 inhalation. Acta Pharmacol Toxicol 58:105-114.
50
51 Gilbert, KM; Griffin, JM; Pumford, NR. (1999) Trichloroethylene activates CD4+ T cells: potential role in an
52 autoimmune response. DrugMetab Rev 31:901-916.
53
54 Goh, VH-H; Chia, S-E; Ong, C-N. (1998) Effects of chronic exposure to low doses of trichloroethylene on steroid
55 hormone and insulin levels in normal men. Environ Health Perspect 106:41-44.
56
8/1/01 R-5 DRAFT-DO NOT CITE OR QUOTE
-------
1 Goldberg, SJ; Lebowitz, MD; Graver, EJ; et al. (1990) An association of human congenital cardiac malformations
2 and drinking water contaminants. J Am Coll Cardiol 16:155-164.
3
4 Goldman, J. (1996) Connective tissue disease in people exposed to organic chemical solvents: systemic sclerosis
5 (scleroderma) in dry cleaning plant and aircraft industry workers. J Clin Rheumatol 2:185-190.
6
7 Goldsworthy, TL; Popp, JA. (1987) Chlorinated hydrocarbon-induced peroxisomal enzyme activity in relation to
8 species and organ carcinogenicity. Toxicol Appl Pharmacol 88:225-233.
9
10 Goldsworthy, TL; Lyght, O; Burnette, VL; et al. (1988) Potential role of • »2» 'globulin protein droplet
11 accumulation and cell replication in the renal carcinogenicity of rats exposed to trichloroethylene, perchlorethylene,
12 and pentachloroethane. Toxicol Appl Pharmacol 96:367-379.
13
14 Gonzales-Leon, A; Merdink, JL; Bull, RJ; et al. (1999) Effect of pre-treatment with dichloroacetic or
15 trichloroacetic acid in drinking water on the pharmacokinetics of a subsequent challenge dose in B6C3F1 mice.
16 ChemBiol Interact 123:239-53.
17
18 Goodman, JI; Counts, JL. (1993) Hypomethylation of DNA: a possible nongenotoxic mechanism underlying the
19 role of cell proliferation in carcinogenesis. Environ Health Perspect 101(suppl 5): 169-72.
20
21 Graeter, LJ; Channel, SR; Eggers, JS; et al. (2000) Evaluating the role of gavage treatment of trichloroethylene,
22 trichloroacetic acid, and dichloroacetic acid on the developing Sprague-Dawley rat fetus. Toxicologist 54:289
23 (Abstract No. 1356).
24
25 Grandjean, E; Muchinger, R; Turrian V; et al. (1955) Investigations into the effect of exposure to trichloroethylene
26 in mechanical engineering. Br J Ind Med 12:131-142.
27
28 Green, T. (2000) Pulmonary toxicity and carcinogenicity of trichloroethylene: species differences and modes of
29 action. Environ Health Perspect 108(suppl 2):261-264.
30
31 Green, T; Dow, J; Foster, JR; et al. (1998) Formic acid excretion in rats exposed to trichloroethylene: a possible
32 explanation for renal toxicity in long-term studies. Toxicology 127:39-47.
33
34 Griffin, JM; Blossom, SJ; Jackson, SK; et al. (2000a) Trichloroethylene accelerates an autoimmune response by
35 Thj T cell activation in MRL +/+ mice. Immunopharmacology 46:123-137.
36
37 Griffin, JM; Gilbert, KM; Pumford, NR. (2000b) Inhibiton of CYP2E1 reverses CD4+ T-cell alterations in
38 trichloroethylene-treated MRL +/+ mice. Toxicol Sci 54:384-389.
39
40 Griffin, JM; Gilbert, KM; Lamps, LW; et al. (2000c) CD4+ T-cell activation and induction of autoimmune hepatitis
41 following trichloroethylene treatment in MRL +/+ mice. Toxicol Sci 57:345-352.
42
43 Hamdan, H; Stacey, NH. (1993) Mechanism of trichloroethylene-induced elevation of individual serum bile acids.
44 I. Correlation of trichloroethylene concentrations to bile acids in rat serum. Toxicol Appl Pharmacol 121:291-295.
45
46 Hayes, JD; Pulford, DJ. (1995) The glutathione S-transferase supergene family: regulation of GST and the
47 contribution of the isoenzymes to cancer chemoprotection and drug resistance. Crit Rev Biochem Mol Biol
48 30:445-600.
49
50 Henderson, GN; Yan, Z; James, MO; et al. (1997) Kinetics and metabolism of chloral hydrate in children:
51 identification of dichloroacetate as a metabolite. Biochem Biophys Res Commun 236:695-698.
52
53 Henschler, D; Romen, W; Elsasser, HM; et al. (1980) Carcinogenicity study of trichloroethylene by long-term
54 inhalation in three animal species. Arch Toxicol 43:237-248.
55
8/1/01 R-6 DRAFT-DO NOT CITE OR QUOTE
-------
1 Henschler, D; Elsasser, H; Romen, W; et al. (1984) Carcinogenicity study of trichloroethylene, with and without
2 epoxide stabilizers, in mice. J Cancer Res Clin Oncol 107:149-156.
3
4 Henschler, D; Vamvakas, S; Lammert, M; et al. (1995) Increased incidence of renal cell tumors in a cohort of
5 cardboard workers exposed to trichloroethylene. Arch Toxicol 69:291-299.
6
7 Herrmann, CE; Sanders, RA; Klaunig, JE; et al. (1999) Decreased apoptosis as a mechanism for hepatomegaly in
8 streptozotocin-induced diabetic rats. Toxicol Sci 50:146-151.
9
10 Hill, AB. (1965) The environment and disease: association or causation? Proc R Soc Med 58:295-300.
11
12 Hobara, T; Kobayashi, H; Higashihara, E; et al. (1984) Acute effects of 1,1,1-trichloroethane, trichloroethylene,
13 and toluene on the hematologic parameters in dogs. Arch Environ Contam Toxicol 13:589-593.
14
15 Holmberg, PC; Hernberg, S; Kurppa, K; et al. (1980a) Oral clefts and organic solvent exposure during pregnancy.
16 Int Arch Occup Environ Health 50:371-376.
17
18 Holmberg, PC; Nurminen, M. (1980b) Congenital defects of the central nervous system and occupation factors
19 during pregnancy: a case-referent study. Am J Ind Med 1:167-176.
20
21 Holmberg, PC; Hernberg, S; Kurppa, K; et al. (1982) Oral clefts and organic solvent exposure during pregnancy.
22 Int Arch Occup Environ Health 50:371-376.
23
24 Hytten, FE; Chamberlain, G. (1980) Clinical Physiology in Obstetrics (Blackwell, Dosmecki C, eds).
25 Oxford:Oxford University Press.
26
27 IARC (International Agency for Research on Cancer). (1995) Dry Cleaning, Some Chlorinated Solvents and Other
28 Industrial Chemicals. Lyon: International Agency for Research on Cancer, World Health Organization.
29
30 Ikeda, M; Otsuji, H; Imamura, T; et al. (1972) Urinary excretion of total trichloro-compounds, trichloroethanol, and
31 trichloroacetic acid as a measure of exposure to trichloroethylene and tetrachloroethylene. Br J Ind Med 29:328-333.
32
33 Isaacson, LG; Spohler, SA; Taylor, DH. (1990) Trichloroethylene affects learning and decreases myelin in the rat
34 hippocampus. Neurotoxicol Teratol 12:377-381.
35
36 Jacques, S. (1982) Acute renal failure following ingestion of formic acid. Nursing Times 7:1312-1315.
37
38 Jaffe, DR; Gandolfi, AJ; Nagle, RB. (1984) Chronic toxicity of S-(trans-l,2-dichlorovinyl)-L-cysteine in mice. J
39 Appl Toxicol 4:315-319.
40
41 Jaffe, DR; Hassall, CD; Gandolfi, AJ; et al. (1985) Production of DNA single strand breaks in rabbit renal tissue
42 after exposure to 1,2-dichlorovinylcysteine. Toxicology 35:25-33.
43
44 Johnson, PD; Dawson, BV; Goldberg, SJ. (1998a) A review: trichloroethylene metabolites: potential cardiac
45 teratogens. Environ Health Perspect 106 (suppl 4):995-999.
46
47 Johnson, PD; Dawson, BV; Goldberg, SJ. (1998b) Cardiac teratogenicity of trichloroethylene metabolites. J Am
48 Coll Cardiol 32:540-545.
49
50 Kato-Weinstein, J; Lingohr, MK; Orner, GA; et al. (1998) Effects of dichloroacetate on glycogen metabolism in
51 B6C3F1 mice. Toxicology 130:141-154.
52
53 Katz, R; Tai, CN; Deiner, RM; et al. (1981) Dichloroacetate, sodium: 3-month oral toxicity studies in rats and dogs.
54 Toxicol Appl Pharmacol 57:273-287.
55
8/1/01 R-7 DRAFT-DO NOT CITE OR QUOTE
-------
1 Kedderis, GL. (1996) Hepatic blood flow limitation of the bioactivation of hazardous chemical air pollutants.
2 Fundam Appl Toxicol Suppl "The Toxicologist" 30:34-35.
3
4 Kedderis, GL. (1997) Extrapolation of in vitro enzyme induction data to humans in vivo. Chem Biol Interact
5 107:109-121.
6
7 Kerbey, AL; Randle, PJ; Cooper, RH; et al. (1976) Regulation of pyruvate dehydrogenase in rat heart. Biochem J
8 154:327-348.
9
10 Ketcha, MM; Stevens, DK; Warren, DA; et al. (1996) Conversion of trichloroacetic acid to dichloroacetic acid in
11 biological samples. J Anal Toxicol 20:236-241.
12
13 Khan, MF; Kaphalia, BS; Prabhakar, BS; et al. (1995) Trichloroethene-induced autoimmune response in female
14 MRL +/+ mice. Toxicol Appl Pharmacol 134:155-160.
15
16 Kilburn, KH; Warshaw, RH. (1992) Prevalence of symptoms of systemic lupus erythematosus (SLE) and of
17 fluorescent antinuclear antibodies associated with chronic exposure to trichloroethylene and other chemicals in well
18 water. Environ Res 57:1-9.
19
20 Kjellstrand, P; Kanje, M; Mansson, L; et al. (1981) Trichloroethylene: effects on body and organ weights in mice,
21 rats and gerbils. Toxicology 21:105-115.
22
23 Kjellstrand, P; Holmquist, B; Aim, P; etal. (1983) Trichloroethylene: further studies of the effects on body and
24 organ weights and plasma butyrylcholinesterase activity in mice. Acta Pharmacol Toxicol 53:375-384.
25
26 Klaussen, K. (1980) Chapter 3, Absorption, distribution, and excretion of toxicants. In: Casarett and Doull's
27 Toxicology, The Basic Science of Poisons. 2nd Edition. Doull, J; Klassen, CD; Amdur, MO, eds. New York:
28 Macmillan Publishing Co., Inc.
29
30 Knecht, KT; Adachi, Y; Bradford, BU; et al. (1995) Free radical adducts in the bile of rats treated chronically with
31 intragastric alcohol: inhibition by destruction of Kupffer cells. Mol Pharmacol 45:1028-1034.
32
33 Kumar, P; Prasad, AK; Dutta, KK. (2000) Steroidogenic alterations in testes and sera of rats exposed to
3 4 trichloroethylene (TCE) by inhalation. Hum Exp Toxicol 19:117-21.
35
36 Laberge, C; Lescault, A; Tanguay, RM. (1986) Hereditary tyrosinemias (type I): a new vista on tyrosine toxicity
3 7 and cancer. Adv Exp Med Biol 206:209-221.
38
39 Lash, LH; Fisher, JW; Lipscomb, JC; et al. (2000a) Metabolism of trichloroethylene. Environ Health Perspect
40 108(suppl 2): 177-200.
41
42 Lash, LH; Parker, JC; Scott, CS. (2000b) Modes of action of trichloroethylene for kidney tumorigenesis. Environ
43 Health Perspect 108(suppl 2):225-240.
44
45 Laskin, DL. (1996) Sinusoidal lining cells and hepatotoxicity. Toxicol Pathol 24:112-118.
46
47 Laskin, DL; Pilaro, AM; Ji, S. (1986) Potential role of activated macrophages in acetaminophen hepatotoxicity. II.
48 Mechanisms of macrophage accumulation and activation. Toxicol Appl Pharmacol 2:216-226.
49
50 Laskin, DL; Robertson, FM; Pilaro, AM. et al. (1988) Activation of liver macrophages following phenobarbital
51 treatment of rats. Hepatology 8:1051-1055.
52
53 Laskin, DL; Gardener, CR; Price, VF. et al. (1995) Modulation of macrophage functioning abrogates the acute
54 hepatotoxicity of acetaminophen. Hepatology 4:1045-1050.
55
8/1/01 R-8 DRAFT-DO NOT CITE OR QUOTE
-------
1 LaVecchia, C; Negri, E; Franceschi, S; et al. (1994) A case-control study of diabetes mellitus and cancer risk. Br J
2 Cancer 70:950-953.
3
4 Lieber, CS. (1997) Cytochrome P-4502E1: its physiological and pathological role. Physiol Rev 77:517-543.
5
6 Liesivuori, J; Savolainen, H. (1986) Urinary excretion of formic acid in rabbits. Acta Pharmacol Toxicol
7 58:161-162.
8
9 Liesivuori, J; Kosma, V-M; Naukkarinen, A. (1987) Kinetics and toxic effects of repeated intravenous dosage of
10 formic acid in rabbits. Br J Exp Pathol 68:853-861.
11
12 Liesivuori, J; Savolainen, H. (1991) Methanol and formic acid toxicity: biochemical mechnaisms. Pharmacol
13 Toxicol 69:157-163.
14
15 Liesivuori, J; Laitinen, J; Savolainen, H. (1992) Kinetics and renal effects of formic acid in occupationally exposed
16 farmers. Arch Toxicol 66:522-524.
17
18 Lindblad, P; Chow, WH; Chan, J; et al. (1999) The role of diabetes mellitus in the aetiology of renal cell cancer.
19 Diabetologia42:107-112.
20
21 Lipscomb, JC; Garret, C; Snawder, JE. (1997) Cytochrome P450-dependent metabolism of trichloroethylene:
22 interindividual differences in humans. Toxicol Appl Pharmacol 142:311-318.
23
24 Liu, YT; Jin, C; Chen, Z; et al. (1988) Increased subjective symptom prevalence among workers exposed to
25 trichloroethylene at sub-OEL levels. Tohoku J Exp Med 155:183-195.
26
27 Lock, EA; Sani, Y; Moore, RB; et al. (1996) Bone marrow and renal injury associated with haloalkene cysteine
28 conjugates in calves. Arch Toxicol 70:607-619.
29
30 Lockey, JE; Kelly, CR; Cannon, GW; et al. (1987) Progressive systemic sclerosis associated with exposure to
31 trichloroethylene. Occup Med 29:493-496.
32
33 Lutz, WK. (1990) Dose-response relationship and low dose extrapolation in chemical carcinogenesis.
34 Carcinogenesis 11(8): 1243-1247.
35
36 MA, DOH. (Massachusetts Department of Health). (1997) Woburn Childhood Leukemia Follow-Up Study. 2 vols.
3 7 Boston: Department of Public Health, Bureau of Environmental Health.
38
39 Maloney, EK; Waxman, DJ. (1999) Trans-Activation of PPAR* "and PPAR* "by structurally diverse environmental
40 chemicals. Toxicol Appl Pharmacol 161:209-218.
41
42 Maltoni, C; Lefemine, G; Cotti, G. (1986) Experimental research on trichloroethylene carcinogenesis. In: Archives
43 of Research on Industrial Carcinogenesis. Maltoni, C; Mehlman, MA, eds. Princeton: Princeton Scientific
44 Publishing Co., 1-393.
45
46 Maltoni, C; Lefemine, G; Cotti, G; et al. (1988) Long-term carcinogenic bioassays on trichloroethylene
47 administered by inhalation to Sprague-Dawley rats and Swiss andB6C3Fl mice. AnnN Y Acad Sci 534:316-351.
48
49 Maronpot, R; Fox, T; Malarkey, DE; et al. (1995) Mutations in the ras proto-oncogene: clues to etiology and
50 molecularpathogenesis of mouse liver tumors. Toxicology 101:125-126.
51
52 Mayer, D; Metzger, C; Leonetti, P; et al. (1998) Differential expression of key enzymes of energy metabolism in
53 preneoplastic and neoplastic rat liver lesions induced by nitrosomorpholine and dehydroepiandrosterone. Int J
54 Cancer 79:232-240.
55
8/1/01 R-9 DRAFT-DO NOT CITE OR QUOTE
-------
1 Mayer, D; Buniatian, G; Metzger, C; et al. (1999) Dehydroepiandrosterone increases in zone of glutamine
2 synthetase-positive hepatocytes in female rat liver: a putative androgenic effect. Histochem Cell Biol
3 111(5):375-380.
4
5 Mayes, MD. (1999) Epidemiologic studies of environmental agents and systemic autoimmune diseases. Environ
6 Health Perspect 107:743-748.
7
8 McConnell, EE. (1992) Comparative responses in carcinogenesis bioassays as a fuction of age at first exposure. In:
9 Similarities and differences between children and adults: implications for risk assessment. Guzelian, PS; Henry, CJ;
10 Olin, SS, eds, Washington: ILSI Press, 66-78.
11
12 McConnell, EE; Solleveld, HA; Swenberg, JA; et al. (1986) Guidelines for combining neoplasms for evaluation of
13 rodent carcinogenesis studies. J Natl Cancer Inst 76:283-289.
14
15 Mellemgaard, A; Engholm, G; McLaughlin, JK; et al. (1994) Occupational risk factors for renal-cell carcinoma in
16 Denmark. Scand J Work Environ Health 20:160-165.
17
18 Melnick, RL; Jameson, CW; Goehl, TJ; et al. (1987) Application of microencapsulation for toxicology studies. II.
19 Toxicity of microencapsulated trichloroethylene in Fischer 344 rats. Fundam Appl Toxicol 8:432-442.
20
21 Metzger, C; Bannasch, P; Mayer, D. (1997) Enhancement and phenotypic modulation of N-nitrosomorpholine-
22 induced hepatocarcinogenesis by dehydroepiandrosterone. Cancer Lett 121:125-131.
23
24 Miller, JH; Minard, KM; Wind, RA; et al. (2000) In vivo MRI measurements of tumor growth induced by
25 dichloroacetate: implications for mode-of-action. Toxicology 145:115-125.
26
27 Moore, GW; Swift, LL; Rabinowitz, D; et al. (1979) Reduction of serum cholesterol in two patients with
28 homozygous familial hypercholesterolemiaby dichloroacetate. Atherosclerosis 33:285-293.
29
30 Moore, MM; Harrington-Brock, K. (2000) Mutagenicity of trichloroethylene and its metabolites: implications for
31 the risk assessment of trichloroethylene. Environ Health Perspect 108(suppl 2):215-223.
32
33 Moser, VC; Phillips, PM; McDaniel, KM; et al. (1999) Behavioral evaluation and the neurotoxicity produced by
34 dichloroacetic acid in rats. Neurotoxicol Teratol 21:719-731.
35
36 Mtiller, M; Birner, G; Henschler, D; et al. (1994) Structure of DNA adducts formed by halothioketenes and
37 haloketenes: possible adducts formed from trichloro- and perchloroethene via glutathione conjugation. In: DNA
38 Adducts: Identification and Biological Significance. Hemminki, K; Dipple, A; Shuker, DEG; et al., eds. IARC
39 Scientific Publications No. 125. Lyon: International Agency for Research on Cancer, 423-428.
40
41 Mtiller, M; Birner, G; Dekant, W. (1998a) Reactivity of haloketenes and halothioketenes with nucleobases:
4 2 chemical characterization of reaction products. Chem Res Toxicol 11:454-463.
43
44 Mtiller, M; Birner, G; Sander, M; et al. (1998b) Reactivity of haloketenes and halothioketenes with nucleobases:
45 reactions in vitro with DNA. Chem Res Toxicol 11:464-470.
46
47 Nagaya, T; Ishikawa, N; Hata, H; et al. (1993) Subclinical and reversible hepatic effects of occupational exposure
48 to trichloroethylene. Int Arch Occup Environ Health 64:561-563.
49
50 Nakajima, T; Kamijo, R; Usuda, N; et al. (2000) Sec-dependent regulation of hepatic peroxisome proliferation in
51 mice by trichloroethylene via peroxisome proliferator-activated receptor • "(PPAR» ). Carcinogenesis 21:677-682.
52
53 Narotsky, MG; Kavlock, RJ. (1995a) A multidisciplinary approach to lexicological screening: II. Developmental
54 toxicity. J Toxicol Environ Health 45:145-171.
55
8/1/01 R-10 DRAFT-DO NOT CITE OR QUOTE
-------
1 Narotsky, MG; Weller, EA; Chinchilli, VM; et al. (1995b) Nonadditive developmental toxicity in mixtures of
2 trichloroethylene, di-(2-ethylhexyl) phthalate, and heptachlor in a 5 x 5 x 5 design. Fundam Appl Toxicol
3 27:203-216.
4
5 NCI (National Cancer Institute). (1976) Carcinogenesis Bioassay of Trichloroethylene. National Cancer Institute
6 Carcinogenesis Technical Report Series No. 2. HEW Publication No. (NIH) 76-802.
7
8 Neghab, M; Qu, S; Bai, CL; et al. (1997) Raised concentration of serum bile acids following occupational exposure
9 to halogenated solvents, 1, l,2-trichloro-l,2,2-trifluoroethane and trichloroethylene. Int Arch Occup Environ Health
10 70:187-194.
11
12 Nietert, PJ; Sutherland, SE; Silver, RM; et al. (1998) Is occupational organic solvent exposure a risk factor for
13 scleroderma? ArthritRheum41:llll-1118.
14
15 NRC (National Research Council). (1994) Science and Judgment in Risk Assessment. Committee on Risk
16 Assessment of Hazardous Air Pollutants, National Research Council. Washington: National Academy Press.
17
18 NTP. (National Toxicology Program). (1988) Toxicology and Carcinogenesis Studies of Trichloroethylene (CAS
19 No. 79-01-6) in Four Strains of Rat (ACI, August, Marshall, and Osborne-Mendel) (Gavage Studies). NTP-TR-273.
20
21 NTP. (1990) Carcinogenesis Studies of Trichloroethylene (Without Epichlorohydrin) (CAS No. 79-01-6) inF344/N
22 Rats and B6C3F1 Mice (Gavage Study). NTP TR 243. Research Triangle Park, NC: U.S. Department of Health and
23 Human Services.
24
25 NTP. (2000a). 9th Report on Carcinogens. Research Triangle Park, NC: U.S. Department of Health and Human
26 services.
27
28 NTP. (2000b) Call for Public Comments on Substances, Mixtures and Exposure Circustances Proposed for Listing
29 in the Report on Carcinogens, Tenth Edition. Federal Register 65:17889-17891.
30
31 Odum, J; Green, T; Foster, JR; et al. (1988) The role of trichloracetic acid and peroxisome proliferation in the
32 differences in carcinogenicity of perchloroethylene in the mouse and rat. Toxicol Appl Pharmacol 92:103-112.
33
34 Odum, J; Foster, JR; Green, T. (1992) A mechanism for the development of Clara cell lesions in the mouse lung
35 after exposure to trichloroethylene. ChemBiol Interact 83:135-153.
36
37 Okawa, MT; Bodner, A. (1973) Health Hazard Evaluation of Western Electric Company, Inc. HHE 72-74-51.
38 National Institute for Occupational Safety and Health.
39
40 Parks, PA; Gilmour, MI; Selgrade, MK. (1993) Pulmonary defenses to streptococcal infection following acute
41 exposure to trichloroethylene (TCE) or chloroform (CHC13). Toxicologist 13:107.
42
43 Pastino, GM; Yap, WY; Carroquino, M. (2000) Human variability and susceptibility to trichloroethylene. Environ
44 Health Perspect 108(suppl 2):201-214.
45
46 Pattengale, PK. (1994) Tumours of the lymphohaematopoietic system. IARC SciPubl (111):651-670.
47
48 Pattengale, PK; Firth, CH. (1986) Contributions of recent research to the classification of spontaneous lymphoid
49 cell neoplasms in mice. CRC Crit Rev Toxicol 16:185-212.
50
51 Pereira, MA; Phelps, JB. (1996) Promotion by dichloroacetic acid and trichloroacetic acid or
52 N-methyl-N-nitrosourea-initiated cancer in the liver of female B6C3F1 mice. Cancer Lett 102:133-141.
53
54 Pereira, MA; Li, K; Kramer, P. (1997) Promotion by mixtures of dichloroacetic acid and trichloroacetic acid of
55 N-methyl-N-nitrosourea-initiated cancer in the liver of female B6C3F1 mice. Cancer Lett 115:15-23.
8/1/01 R-11 DRAFT-DO NOT CITE OR QUOTE
-------
1 Peters, R; Bocker, R; Beaune, PH; et al. (1990) Hydroxylation of chlorzoxazone as a specific probe for human liver
2 cytochrome P450IM. ChemRes Toxicol 3:566-573.
3
4 Rasmussen, K; Arlien-Seborg, P; Sabroe, S. (1993) Clinical neurological findings among metal degreasers exposed
5 to chorinated solvents. Acta Neurol Scand 87:200-204.
6
7 Raucy, JL; Kraner, JC; Lasher, JM. (1993) Bioactivation of halogenated hydrocarbons by cytochrome P4502E1.
8 Crit Revs Toxicol 23:1-20.
9
10 Raucy, J. (1995) Risk assessment: toxicity from chemical exposure resulting from enhanced expression of CYP2E1.
11 Toxicology 105:217-223.
12
13 Renwick, AG. (1998) Toxicokinetics in infants and children in relation to the ADI and TDI. Food Add Contam
14 15(suppl): 17-35.
15
16 Rhomberg, L. (2000) Dose-response analyses of the carcinogenic effects of trichloroethylene in experimental
17 animals. Environ Health Perspect 108(suppl 2):343-358.
18
19 Ritz, B. (1999) Cancer mortality among workers exposed to chemicals during uranium processing. J Occup Environ
20 Med 41:556-566.
21
22 Roberts, RA; Soames, AR; Gill, JH; et al. (1995) Non-genotoxic hepatocarcinogens stimulate DNA synthesis and
23 their withdrawal induces apoptosis, but in different hepatocyte populations. Carcinogenesis 16:1693-1698.
24
25 Rose, ML; Germolec, D; Schoonhoven, R; et al. (1997) Kuppfer cells are causally responsible for the mitogenic
26 effect of peroxisome proliferators. Carcinogenesis 18:1453-1456.
27
28 Rothman, KJ. (1986) Modern Epidemiology. Boston: Little, Brown and Company.
29
30 Ruitjen, MWMM; Verberk, MM; Salle, HJA. (1991) Nerve function in workers with long term exposure to
31 trichloroethene. Br J Ind Med 48:87-92.
32
33 Runyan, RB. (2000) Letter to editor, reply. Toxciol Sci 56:437-438.
34
35 Saihian, EM; Burton, JL; Heaton, KW. (1978) A new syndrome with pigmentation, scleroderma, gynaecomastia,
36 Raynaud's phenomenon and peripheral neuropathy. Br J Dermatol 99:437-440.
37
38 Sanders, VM; Tucker, AN; White, KL; et al. (1982) Humoral and cell-mediated immune status in mice exposed to
39 trichloroethylene in drinking water. Toxicol Appl Pharmacol 62:358-368.
40
41 Schaeverbeke, T; Halle, O; Moride, Y; et al. (1995) Systemic sclerosis and prolonged exposure to toxics [abstract].
42 Arthritis Rheum 38(suppl 9):S332.
43
44 Schlehofer, B; Heuer, C; Blettner, M; et al. (1995) Occupation, smoking and demographic factors, and renal cell
45 carcinoma in Germany. Int J Epidemiol 24:51-57.
46
47 Schraml, P; Zhaou, M; Richter, J; et al. (1999) Analysis of kidney tumors in trichloroethylene exposed workers by
48 comparative genomic hybridization and DNA sequence analysis. Verh Dtsch Ges Pathol 83:218-224.
49
50 Schultz, R; Yan, W; Toppari, J; et al. (1999) Expression of peroxisome proliferator-activated receptor alpha
51 messenger ribonucleic acid and protein in human and rat testis. Endocrinology 149:2968-2975.
52
53 Selgrade, MJ. (1995) Immunotoxicity and risk assessment: effect of temporal factors. Inhal Toxicol 7:891-901.
54
55 Shimada, H; Achanzar, W; Waalkes, MP; et al. (1999) Testosterone pretreament enhances cadmium-induced
56 increases in c-myc and c-jun transcript levels in L6 myoblasts. Toxic Subst Median 18:129-137.
8/1/01 R-12 DRAFT-DO NOT CITE OR QUOTE
-------
1 Smith, MK; Randall, JL; Read, EJ; et al. (1989) Teratogenic activity of trichloroacetic acid in the rat. Teratology
2 40:445-451.
3
4 Smith, MK; Randall, JL; Read, EJ; et al. (1992) Developmental toxicity of dichloroacetate in the rat. Teratology
5 46:217-223.
6
7 Smith, MR. (1996) Non-Hodgkin's lymphoma. Curr Prob Cancer XX: 1-8.
8
9 Stacpoole, PW; Moore, GW; Kornhauser, DM. (1979) Toxicity of chronic dichloroacetate. NEnglJMed 300:372.
10
11 Stacpoole, P; Wright, EC; Baumgartner, TG; et al. (1992) A controlled clinical trial of dichloroacetate treatment in
12 patients with lactic acidosis. The dichloroacetate-lactic acidosis study group. N Engl J Med 327:1564-1569.
13
14 Stevens, EL; Taylor, JA; Kaplan, N; et al. (1994) Ethnic variations in the CYP2E1 gene: polymorphism analysis of
15 695 African-Americans, European-Americans and Taiwanese. Pharmacogenesis 4:185-192.
16
17 Stewart, RD; Dodd, HC; Gay, HH; et al. (1970) Experimental human exposure to trichloroethylene. Arch Environ
18 Health 20:64-71.
19
20 Su, Q; Benner, A; Hofmann, WJ; et al. (1997) Human hepatic preneoplasia: phenotypes and proliferation kinetics
21 of foci and nodules of altered hepatocytes and their relationship to liver cell dysplasia. Virchows Arch 431:391-406.
22
23 Tanguay, RM; Jorquera, R; Poudrier, J; et al. (1996) Tyrosine and its catabolites: from disease to cancer. Acta
24 Biochim Pol 43:209-216.
25
26 Tao, L; Kramer, PM; Ge, R; et al. (1998) Effect of dichloroacetic acid on DNA methylation in liver and tumors of
27 female B6C3F1 mice. Toxicol Sci 43:139-144.
28
29 Tao, L; Ge, R; Xie, M; et al. (1999) Effectof trichloroethylene on DNA methylation and expression of early-
30 intermediate protooncogenes in the liver of B6C3F1 mice. BiochemMol Toxicol 13:231-237.
31
32 Tao, L; Yang, S; Zie, M; et al. (2000a) Effect of trichloethylene and its metabolites, dichloroacetic acid and
33 trichloroacetic acid, on the methylation and expression of c-jun and c-myc protooncogenes in mouse liver;
34 prevention by methionine. Toxicol Sci 54:399-407.
35
36 Tao, L; Yang, S; Xie, M; et al. (2000b) Hypomethylation and overexpression of c-jun and c-myc protooncogenes
37 and increased DNA methyltransferase activity in dichloroacetic and trichloroacetic acid-promoted mouse liver
38 tumors. Cancer Lett 158:185-193.
39
40 Taylor, DH; Lagory, KE; Zaccaro, J; et al. (1985) Effectof trichloroethylene on the exploratory activity of rats
41 exposed during development. Sci Total Environ 47:415-420.
42
43 Tomatis, L; Aitio, A; Wilbourn, J; et al. (1989) Human carcinogens so far identified. Jpn J Cancer Res 80:795-807.
44
45 Tong, Z; Board, PG; Anders, MW. (1998) Glutathione transferase zeta catalyses the oxygenation of the carcinogen
46 dichloroacetic acid to glyoxylic acid. Biochem J 33 l(Pt 2):371-374.
47
48 Tucker, AN; Sanders, VM; Barnes, DW; et al. (1982) Toxicology of trichloroethylene in the mouse. Toxicol Appl
49 Pharmacol 62:351-357.
50
51 Tugwood, JD; Aldridge, TC; Lambe, KG; et al. (1998) Peroxisome proliferator-activated receptor-alpha and the
52 pleiotropic response to peroxisome proliferators. Arch Toxicol Suppl 20:377-386.
53
54 Tzeng, HF; Blackburn, AC; Board, PG; et al. (2000) Polymorphism-and species-dependent inactivation of
55 glutathione transferase zeta by dichloroacetate. Chem Res Toxicol 13:231-236.
56
8/1/01 R-13 DRAFT-DO NOT CITE OR QUOTE
-------
1 U.S. EPA (Environmental Protection Agency). (1985) Health Assessment Document for Trichloroethylene. Final
2 Report. EPA/600/8-82/006F. Washington: U.S. Environmental Protection Agency, Office of Health and
3 Environmental Assessment.
4
5 U.S. EPA. (1986) Guidelines for Carcinogen Risk Assessment. Federal Register 51(185):33992-34003.
6
7 U.S. EPA. (1987) Addendum to the Health Assessment Document for Trichloroethylene: Updated Carcinogenicity
8 Assessment for Trichloroethylene. External Review Draft EPA/600/8-82/006FA. Washington: U.S. Environmental
9 Protection Agency, Office of Health and Environmental Assessment.
10
11 U.S. EPA. (1992) Draft Report: A Cross-Species Scaling Factor for Carcinogen Risk Assessment Based on
12 Equivalence of mg/kg^/day. Federal Register 57:24152-24173.
13
14 U.S. EPA. (1994) Methods for Derivation of Inhalation Reference Concentrations and Application of Inhalation
15 Dosimetry. EPA/600/8-90/066F. Washington: U.S. Environmental Protection Agency, Office of Research and
16 Development.
17
18 U.S. EPA. (1996) Proposed Guidelines for Carcinogen Risk Assessment; Notice. Federal Register
19 61(79):17960-18011.
20
21 U.S. EPA. (1997) Cumulative Risk Assessment Guidance—Phase I Planning and Scoping. Memorandum from
22 Carol M. Browner, Administrator, and Fred Hansen, Deputy Administrator.
23
24 U.S. EPA. (1999) Guidelines for Carcinogen Risk Assessment. SAB Review Draft. NCEA-F-0644. Washington:
25 U.S. Environmental Protection Agency, Office of Research and Development.
26
27 Vamvakas, S; Bruning, T; Thomasson, B; etal. (1998) Renal cell cancer correlated with occupational exposure to
28 trichloroethylene. J Cancer Res Clin Oncol 124:374-382.
29
30 Vandervort, R; Polakoff, PL; Flesch, JP. (1973) Health Hazard Evaluation of Dunham-Bush, Inc. HHE 72-84-31.
31 National Institute for Occupational Safety and Health.
32
33 Vartiainen, T; Pukkala, E; Strandman, T; et al. (1993) Population exposure to tri- and tetrachloroethylene and
34 cancer risk: two cases of drinking water pollution. Chemosphere 27:1171-1181.
35
36 Villaschi, S; Giovanetti, A; Lombardi, CC; et al. (1991) Damage and repair of mouse bronchial epithelium
37 following acute inhalation of trichloroethylene. Exp Lung Res 17:601-614.
38
39 Volkel, W; Friedewald, M; Lederer, E; et al. (1998) Biotransformation of perchloroethene: dose-dependent
40 excretion of trichloroacetic acid, dichloroacetic acid, and N-acetyl-S-(trichlorovinyl)-L-cysteine in rats and humans
41 after inhalation. Toxicol Appl Pharmacol 153:20-27.
42
43 Waller, PA; Clauw, D; Cupps, T; et al. (1994) Fasciitis (not scleroderma) following prolonged exposure to an
44 organic solvent (trichloroethylene). J Rheumatol 21:1567-1570.
45
46 Wang, G; Stacey, NH. (1990) Elevation of individual serum bile acids on exposure to trichloroethylene or
47 • •naphthylisothiocyanate. Toxicol Appl Pharmacol 105:209-215.
48
49 Wang, RS; Nakajima, T; Honma, T. (1999) Trichloroethylene inhibits aldehyde dehydrogenase only for aliphatic
50 aldehydes of short chains in rats. Toxicology 132:9-18.
51
52 Wang, T; Shankar, K; Ronis, MJ; et al. (2000) Potentiation of thioacetamide liver injury in diabetic rats is due to
53 induced CYP2E1. J Pharmacol Exp Ther 294:473-479.
54
55 Wartenberg, D; Reyner, D; Scott, CS. (2000) Trichloroethylene and cancer: epidemiologic evidence. Environ
56 Health Perspect 108(suppl 2)161-176.
8/1/01 R-14 DRAFT-DO NOT CITE OR QUOTE
-------
1 White, RF; Feldman, RG; Eviator, II; et al. (1997) Hazardous waste and neurobehavioral effects: a developmental
2 perspective. Environ Res 73:113-124.
3
4 Wideroff, L; Gridley, G; Mellemkjaer, L; et al. (1997) Cancer incidence in a population-based cohort of patients
5 hospitalized with diabetes mellitus in Denmark. J Natl Cancer Inst 89:1360-1365.
6
7 Wilson, PD; Loffredo, CA; Correa-Villasenor, A; et al. (1998) Attributable fraction for cardiac malformations. Am
8 JEpidemiol 148:414-423.
9
10 Wolffe, AP. (1999) The cancer-chromatin connection. Science and Medicine July/August:28-37.
11
12 Wu, C; Schaum, J. (2000) Exposure assessment of trichloroethylene. Environ Health Perspect 108(suppl
13 2):359-363.
14
15 Wu, C; Schaum, J. (2001) Sources, emission and exposure for trichloroethylene (TCE) and related chemicals.
16 Washington: U.S. Environmental Protection Agency, EPA/600/R-00/099.
17
18 Yan, Z; Henderson, GN; James, MO; et al. (1999) Determination of chloral hydrate metabolites in human plasma
19 by gas chromatography-mass spectrometry. J Pharm Biomed Anal 19(3-4):309-318.
20
21 Yoo, JS; Guengerich, FP; Yang, CS. (1988) Metabolism of N-nitrosodialkyl-amines in human liver microsomes.
22 Cancer Res 88:1499-1505.
23
24 Yount, EA; Felten, SY; O'Connor, BL; et al. (1982) Comparison of the metabolic and toxic effects of
25 2-chloropropionate and dichloroacetate. J Pharmacol Exp Ther 222:501-508.
26
27 Youseff, J; Badr, M. (1998) Extraperoxisomal targets of peroxisome proliferators: mitochondrial, microsomal, and
28 cytosolic effects. Implications for health and disease. Crit Revs Toxicol 28:1-33.
29
30 Youseff, J; Badr, M. (1999) Biology of senescent liver peroxisomes: role in hepatocellular aging and disease.
31 Environ Health Perspect 107:791-798.
32
33 Zenick, H; Blackburn, K; Hope, E; et al. (1984) Effects of trichloroethylene exposure on male reproductive
34 function in rats. Toxicol 31:237-250.
35
36 Zielinski, A. (1973) General health state of women professionally exposed to trichloroethylene vapours. Med Pr
37 24:263-271.
38
39 Zhou, YC; Waxman, DJ. (1998) Activation of peroxisome proliferator activated receptors by chlorinated
40 hydrocarbons and endogenous steroids. Environ Health Perspect 106:983-987.
8/1/01 R-15 DRAFT-DO NOT CITE OR QUOTE
-------
United States
Environmental Protection Agency/ORD
National Center for
Environmental Assessment
Washington, DC 20460
Official Business
Penalty for Private use
$300
Please make all necessary changes on the below label,
detach or copy, and return to the address in the upper left-
hand corner.
If you do not wish to receive these reports CHECK HERE G;
detach, or copy this cover, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
EPA/600/R-01/055A
------- |