UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON D.C. 20460
OFFICE OF THE ADMINISTRATOR
SCIENCE ADVISORY BOARD
January 11,2011
EPA-SAB-11-002
The Honorable Lisa P. Jackson
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Subject: Review of EPA's Draft Assessment entitled "lexicological Review of
Trichloroethylene " (October 2009)
Dear Administrator Jackson:
EPA's Office of Research and Development (ORD) requested the Science Advisory
Board (SAB) to conduct a peer review of EPA's draft Integrated Risk Information System (IRIS)
assessment entitled, "lexicological Review of Trichloroethylene" (October 2009). This draft
document responded to the National Academy of Sciences (NAS) 2006 recommendations
published in a report entitled, "Assessing the Human Health Risks of Trichloroethylene: Key
Scientific Issues" (National Research Council, 2006). In response to ORD's request, the SAB
convened an expert panel to conduct this review. The SAB Panel was asked to comment on the
scientific soundness of the hazard and dose-response assessments of trichloroethylene (TCE)-
induced cancer and non-cancer health effects. Specifically, the SAB was asked to comment on
the use of a physiologically-based pharmacokinetic (PBPK) model for dose and route of
exposure extrapolation within species and across species; TCE metabolism and mode of action;
the derivation of an oral reference dose (RfD) and inhalation reference concentration (RfC) for
non-cancer toxicity; the weight of evidence of potential human carcinogenicity; and the
estimated cancer oral slope factor and inhalation unit risk for TCE.
The SAB commends EPA for its comprehensive approach and responsiveness to the NAS
recommendations. Overall, the SAB Panel supported EPA's scientific approaches to the risk
assessment and found these to appropriately adhere to EPA's risk assessment guidelines. The
SAB Panel made a number of recommendations aimed at enhancing the transparency of the draft
assessment and strengthening the scientific basis for the conclusions presented. The SAB
responses to the EPA's charge questions are detailed in the report. SAB major comments and
recommendations are provided below:
-------
• EPA has made significant changes that improve the existing PBPK model for TCE. The
Panel supported the use of this updated PBPK model for dose- response assessment for
the extrapolation of doses within species, across species and route-to-route extrapolation.
The Panel also supported the use of the Bayesian framework for estimation and
characterization of the PBPK model parameter uncertainties. The Panel made a number
of suggestions for better documentation of the model.
• The Panel found that the draft document adequately synthesizes the available scientific
information to support a conclusion that TCE poses a potential human health hazard for
non-cancer toxicity, including effects on the central nervous system, the kidney, the liver,
the immune system, the male reproductive system, and the developing fetus.
• The Panel supported the selection of an RfC and an RfD based on multiple candidate
reference values that fell within a narrow range rather than reliance on a single most
sensitive critical endpoint. Although recognizing the kidney hazards of TCE, the Panel
was concerned about the use of three candidate RfD/RfCs based on kidney effects as the
primary basis for the RfD and RfC because of uncertainties regarding the relative rate of
formation of toxic metabolites in humans vs. animals. The Panel recommends that EPA
derive RfD/RfC values based on immunological endpoints and cardiac malformations.
• The Panel found that the EPA's meta-analyses for kidney cancer, lymphoma, and liver
cancer were well-conducted, with results that bolster the weight of evidence for potential
human carcinogenicity from TCE exposure. Accordingly, the Panel agreed with EPA's
conclusion that TCE is considered to be "Carcinogenic to Humans" by all routes of
exposure, based on convincing epidemiological evidence of a causal association between
TCE exposure and kidney cancer, compelling evidence for lymphoma, and limited
evidence for liver cancer. This conclusion is further supported by consistent evidence
from animal studies and pharmacokinetic and metabolism information.
• EPA concluded that a mutagenic mode of action (MOA) was operative in TCE-induced
kidney tumorigenesis. However, the Panel concluded that the available evidence also
supports MO As involving cell death and compensatory cell proliferation. The Panel
agreed with EPA's conclusion that there is inadequate evidence for an MO A mediated by
activation of peroxisome proliferator receptor-alpha for TCE-induced liver cancer in
humans.
• Finally, the Panel supported EPA's approaches for deriving cancer inhalation unit risk
and oral slope factors, including the use of default age-dependent adjustment factors to
address susceptible populations. The Panel supported the use of the French occupational
study (Charbotel et al., 2006) as the basis for estimating cancer unit risks, and the use of a
default linear extrapolation from the point of departure for cancer dose-response
assessment. The Panel, however, recommended inclusion of a more detailed discussion
of assumptions used in the analysis to support the calculation of the unit risks.
11
-------
The SAB appreciates the opportunity to provide EPA with advice on this important subject.
The SAB urges EPA to move expeditiously to finalize the IRIS document for trichloroethylene.
We look forward to receiving the Agency's response.
Sincerely,
/signed/ /signed/
Dr. Deborah L. Swackhamer, Chair Dr. Deborah Cory-Slechta, Chair
EPA Science Advisory Board SAB Trichloroethylene Review Panel
in
-------
NOTICE
This report has been written as part of the activities of the EPA Science Advisory Board,
a public advisory committee providing extramural scientific information and advice to the
Administrator and other officials of the Environmental Protection Agency. The Board is
structured to provide balanced, expert assessment of scientific matters related to problems facing
the Agency. This report has not been reviewed for approval by the Agency and, hence, the
contents of this report do not necessarily represent the views and policies of the Environmental
Protection Agency, nor of other agencies in the Executive Branch of the Federal government, nor
does mention of trade names or commercial products constitute a recommendation for use.
Reports of the EPA Science Advisory Board are posted on the EPA Web site at:
http ://www. epa.gov/sab
IV
-------
U.S. Environmental Protection Agency
Science Advisory Board
Trichloroethylene Review Panel
CHAIR
Dr. Deborah Cory-Slechta, Professor, Department of Environmental Medicine, School of
Medicine and Dentistry, University of Rochester, Rochester, NY
(Member of the Board 2003 - 2010)
MEMBERS
Dr. Scott Bartell, Assistant Professor, Program in Public Health, University of California -
Irvine, Irvine, CA
Dr. Aaron Blair, Scientist Emeritus, National Cancer Institute, National Institutes of Health,
Rockville, MD
Dr. Anneclaire De Roos, Associate Professor, Department of Epidemiology, University of
Washington and Associate Member, Epidemiology Program, Fred Hutchinson Cancer Research
Center, Seattle, WA
Dr. Rodney Dietert, Professor, Department of Microbiology and Immunology, College of
Veterinary Medicine, Cornell University, Ithaca, NY
Dr. Claude Emond, Adjunct Clinical Professor , Department of Environmental and
Occupational Health , Faculty of Medicine, University of Montreal, Montreal, QC, Canada
Dr. Montserrat Fuentes, Professor, Department of Statistics, North Carolina State University,
Raleigh, NC
Dr. David G. Hoel, Distinguished University Professor, Department of Biometry and
Epidemiology, Medical University of South Carolina, Charleston, SC
Dr. Gunnar Johanson, Professor and Deputy Director, Institute of Environmental Medicine,
Karolinska Institutet, Stockholm, Sweden
Dr. Deborah Keil, Professor, Medical Laboratory Sciences, Department of Pathology,
University of Utah School of Medicine, Salt Lake City, UT
Dr. Jose Manautou, Associate Professor & Marlene L. Cohen and Jerome H. Fleisch Scholar,
Department of Pharmaceutical Sciences, School of Pharmacy, University of Connecticut, Storrs,
CT
Dr David McMillan, Associate Professor, Department of Pharmacology and Experimental
Neuroscience, College of Medicine, University of Nebraska Medical Center, Omaha, NE
-------
Dr. Michael Pennell, Assistant Professor, Division of Biostatistics, College of Public Health,
The Ohio State University, Columbus, OH
Dr. Kenneth M. Portier, Director of Statistics, Department of Statistics and Evaluation,
American Cancer Society, National Home Office, Atlanta, GA
Dr. Gloria Post, Research Scientist, Office of Science, New Jersey Department of
Environmental Protection, Trenton, NJ
Dr Gary Rankin, Professor and Chair of Pharmacology, Physiology and Toxicology,
Pharmacology, Physiology and Toxicology, Joan C. Edwards School of medicine, Marshall
University, Huntington, WV
Dr. Ivan Rusyn, Associate Professor, Environmental Sciences and Engineering, Gillings School
of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC
Dr. Ornella Selmin, Associate Research Scientist, Nutritional Sciences, Shantz Building 38,
Room 309, University of Arizona, Tucson, AZ
Dr. Brian Thrall, Technical Group Leader, Cell Biology Group , Pacific Northwest National
Laboratories, Richland, WA
Dr. John Vena, Professor and Department Head, Department of Epidemiology and
Biostatistics, College of Public Health, University of Georgia, Athens, GA
Dr. Virginia Weaver, Associate Professor, Departments of Environmental Health Sciences &
Medicine, Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD
SCIENCE ADVISORY BOARD STAFF
Dr. Marc Rigas, EPA Science Advisory Board, Science Advisory Board Staff Office,
Washington, DC (2009 - 2010)
Dr. Holly Stallworth, EPA Science Advisory Board, Science Advisory Board Staff Office,
Washington, DC
Dr. Diana Wong, EPA Science Advisory Board, Science Advisory Board Staff Office,
Washington, DC
VI
-------
U.S. Environmental Protection Agency
Science Advisory Board (2011)
CHAIR
Dr. Deborah L. Swackhamer, Professor and Charles M. Denny, Jr., Chair in Science,
Technology and Public Policy and Co-Director of the Water Resources Center, Hubert H.
Humphrey Institute of Public Affairs, University of Minnesota, St. Paul, MN
SAB MEMBERS
Dr. David T. Allen, Professor, Department of Chemical Engineering, University of Texas,
Austin, TX
Dr. Claudia Benitez-Nelson, Full Professor and Director of the Marine Science Program,
Department of Earth and Ocean Sciences , University of South Carolina, Columbia, SC
Dr. Timothy Buckley, Associate Professor and Chair, Division of Environmental Health
Sciences, College of Public Health, The Ohio State University, Columbus, OH
Dr. Patricia Buffler, Professor of Epidemiology and Dean Emerita, Department of
Epidemiology, School of Public Health, University of California, Berkeley, CA
Dr. Ingrid Burke, Director, Haub School and Ruckelshaus Institute of Environment and Natural
Resources, University of Wyoming , Laramie, WY
Dr. Thomas Burke, Professor, Department of Health Policy and Management, Johns Hopkins
Bloomberg School of Public Health, Johns Hopkins University, Baltimore, MD
Dr. Terry Daniel, Professor of Psychology and Natural Resources, Department of Psychology,
School of Natural Resources, University of Arizona, Tucson, AZ
Dr. George Daston, Victor Mills Society Research Fellow, Product Safety and Regulatory
Affairs, Procter & Gamble, Cincinnati, OH
Dr. Costel Denson, Managing Member, Costech Technologies, LLC, Newark, DE
Dr. Otto C. Doering III, Professor, Department of Agricultural Economics, Purdue University,
W. Lafayette, IN
Dr. David A. Dzombak, Walter J. Blenko Sr. Professor of Environmental Engineering ,
Department of Civil and Environmental Engineering, College of Engineering, Carnegie Mellon
University, Pittsburgh, PA
Dr. T. Taylor Eighmy, Vice President for Research, Office of the Vice President for Research,
Texas Tech University, Lubbock, TX
vn
-------
Dr. Elaine Faustman, Professor, Department of Environmental and Occupational Health
Sciences, School of Public Health and Community Medicine, University of Washington, Seattle,
WA
Dr. John P. Giesy, Professor and Canada Research Chair, Veterinary Biomedical Sciences and
Toxicology Centre, University of Saskatchewan, Saskatoon, Saskatchewan, Canada
Dr. Jeffrey Griffiths, Associate Professor, Department of Public Health and Community
Medicine, School of Medicine, Tufts University, Boston, MA
Dr. James K. Hammitt, Professor, Center for Risk Analysis, Harvard University, Boston, MA
Dr. Bernd Kahn, Professor Emeritus and Associate Director, Environmental Radiation Center,
Georgia Institute of Technology, Atlanta, GA
Dr. Agnes Kane, Professor and Chair, Department of Pathology and Laboratory Medicine,
Brown University, Providence, RI
Dr. Madhu Khanna, Professor, Department of Agricultural and Consumer Economics,
University of Illinois at Urbana-Champaign, Urbana, IL
Dr. Nancy K. Kim, Senior Executive, Health Research, Inc., Troy, NY
Dr. Catherine Kling, Professor, Department of Economics, Iowa State University, Ames, IA
Dr. Kai Lee, Program Officer, Conservation and Science Program, David & Lucile Packard
Foundation, Los Altos, CA
Dr. Cecil Lue-Hing, President, Cecil Lue-Hing & Assoc. Inc., Burr Ridge, IL
Dr. Floyd Malveaux, Executive Director, Merck Childhood Asthma Network, Inc., Washington,
DC
Dr. Lee D. McMullen, Water Resources Practice Leader, Snyder & Associates, Inc., Ankeny,
IA
Dr. Judith L. Meyer, Professor Emeritus, Odum School of Ecology, University of Georgia,
Lopez Island, WA
Dr. James R. Mihelcic, Professor, Civil and Environmental Engineering, State of Florida 21st
Century World Class Scholar, University of South Florida, Tampa, FL
Dr. Jana Milford, Professor, Department of Mechanical Engineering, University of Colorado,
Boulder, CO
Vlll
-------
Dr. Christine Moe, Eugene J. Gangarosa Professor, Hubert Department of Global Health,
Rollins School of Public Health, Emory University, Atlanta, GA
Dr. Horace Moo-Young, Dean and Professor, College of Engineering, Computer Science, and
Technology, California State University, Los Angeles, CA
Dr. Eileen Murphy, Grants Facilitator, Ernest Mario School of Pharmacy, Rutgers University,
Piscataway, NJ
Dr. Duncan Patten, Research Professor, Hydroecology Research Program , Department of Land
Resources and Environmental Sciences, Montana State University, Bozeman, MT
Dr. Stephen Polasky, Fesler-Lampert Professor of Ecological/Environmental Economics,
Department of Applied Economics, University of Minnesota, St. Paul, MN
Dr. Arden Pope, Professor, Department of Economics, Brigham Young University , Provo, UT
Dr. Stephen M. Roberts, Professor, Department of Physiological Sciences, Director, Center for
Environmental and Human Toxicology, University of Florida, Gainesville, FL
Dr. Amanda Rodewald, Professor of Wildlife Ecology, School of Environment and Natural
Resources, The Ohio State University, Columbus, OH
Dr. Jonathan M. Samet, Professor and Flora L. Thornton Chair, Department of Preventive
Medicine, University of Southern California, Los Angeles, CA
Dr. James Sanders, Director and Professor, Skidaway Institute of Oceanography, Savannah,
GA
Dr. Jerald Schnoor, Allen S. Henry Chair Professor, Department of Civil and Environmental
Engineering, Co-Director, Center for Global and Regional Environmental Research, University
of Iowa, Iowa City, IA
Dr. Kathleen Segerson, Philip E. Austin Professor of Economics , Department of Economics,
University of Connecticut, Storrs, CT
Dr. Herman Taylor, Director, Principal Investigator, Jackson Heart Study, University of
Mississippi Medical Center, Jackson, MS
Dr. Barton H. (Buzz) Thompson, Jr., Robert E. Paradise Professor of Natural Resources Law
at the Stanford Law School and Perry L. McCarty Director, Woods Institute for the
Environment, Stanford University, Stanford, CA
Dr. Paige Tolbert, Professor and Chair, Department of Environmental Health, Rollins School of
Public Health, Emory University, Atlanta, GA
IX
-------
Dr. John Vena, Professor and Department Head, Department of Epidemiology and
Biostatistics, College of Public Health, University of Georgia, Athens, GA
Dr. Thomas S. Wallsten, Professor and Chair, Department of Psychology, University of
Maryland, College Park, MD
Dr. Robert Watts, Professor of Mechanical Engineering Emeritus, Tulane University,
Annapolis, MD
Dr. R. Thomas Zoeller, Professor, Department of Biology, University of Massachusetts,
Amherst, MA
SCIENCE ADVISORY BOARD STAFF
Dr. Angela Nugent, Designated Federal Officer, U.S. Environmental Protection Agency,
Science Advisory Board Staff Office, Washington, DC
Dr, Thomas Armitage, Designated Federal Officer, U.S. Environmental Protection Agency,
Washington, DC
-------
TABLE OF CONTENTS
ABBREVIATIONS AND ACRONYMS xii
EXECUTIVE SUMMARY 1
RESPONSES TO EPA'S CHARGE QUESTIONS 6
1. PBPK Modeling 6
2. Meta-analysis of cancer epidemiology 10
3. Non-Cancer Hazard Assessment 14
4. Cancer Hazard Assessment 18
5. Role of Metabolism on TCE Toxicity 22
6. Mode of Action 27
7. Susceptible Populations 30
8. Non-Cancer Dose-Response Assessment 32
9. Cancer Dose-Response Assessment 42
10. Age-Dependent Adjustment Factors 45
11. Additional key studies 47
12. Research Needs 48
REFERENCES 51
Appendix A: Editorial Comments 55
XI
-------
ABBREVIATIONS AND ACRONYMS
AIC
ADAF
BMD
BMDL
BW
CI
cRfCs
cRfDs
DCA
DCVC
DCVG
DEHP
EPA
ESRD
GC-MS
GSH
HEC
HED
HPLC-UV
idPOD
IQR
IRIS
LOAEL
MCMC
MOA
NAG
NCI
NHL
NOAEL
NRC
NTP
OR
ORD
PBPD
PBPK
p-cRfC
p-cRfD
PERC
POD
PPARa
RCC
RfC
RfD
Akaike Information Criteria
age-dependent adjustment factor
benchmark dose
benchmark dose lower bound
body weight
confidence interval
candidate RfCs
candidate RfDs
dichloroacetic acid
dichlorovinyl cysteine
S-dichlorovinyl glutathione
di(2-ethylhexyl) phthalate
Environmental Protection Agency
end stage renal disease
gas chromatography-mass spectrometry
gluthione
human equivalent concentration
human equivalent dose
high performance liquid chromatography-ultraviolet
internal dose points of departure
interquartile range
Integrated Risk Information System
Lowest Adverse Effect Level
Markov Chain Monte Carlo
mode of action
N-acetyl-p-D-glucosaminidase
National Cancer Institute
non-Hodgkin's lymphoma
No Adverse Effect Level
National Research Council
National Toxicology Program
odds ratio
Office of Research and Development
physiologically-based pharmaodynamic
physiologically-based pharmacokinetic
PBPK model-based candidate RfCs
PBPK model-based candidate RfDs
perchloroethylene
point of departure
peroxisome proliferator activated receptor alpha
renal cell carcinoma
reference concentration
reference dose
xn
-------
RR relative risk
SIR standardized incidence ratio
SMR standardized mortality ratio
TCA trichloroacetic acid
TCE trichloroethylene
TCOH trichloroethanol
UF uncertainty factor
VSD ventricular defects
Xlll
-------
EXECUTIVE SUMMARY
This report was prepared by the Science Advisory Board (SAB) Trichloroethylene
Review Panel (the "Panel") in response to a request by EPA's Office of Research and
Development (ORD) to review the Draft IRIS Toxicological Review of Trichloroethylene (TCE)
(hereafter referred to as the draft document). The Panel deliberated on the charge questions (see
Appendix A) during a May 10 - 12, 2010 face-to-face meeting and subsequent conference calls
on June 24, 2010 and September 13, 2010. The Panel's draft report was considered and
approved by the Chartered SAB in a public teleconference on December 15, 2010. There were
12 charge questions that focused on: hazard assessment of non-cancer and cancer health effects,
the use of a PBPK model for TCE and its metabolites for the derivation of a proposed oral
reference dose (RfD), an inhalation reference concentration (RfC) for non-cancer endpoints,
cancer weight of evidence classification, mode of action of TCE carcinogenicity, as well as
inhalation and oral unit risks for TCE. This Executive Summary highlights the Panel's major
findings and recommendations.
PBPK Modeling
The Panel commended the updated physiologically-based pharmacokinetic (PBPK)
model (Evans et al., 2009; Chiu et al., 2009) for dose-response assessment. The Panel found that
while the PBPK model was generally well presented, its description was incomplete in that
mass-balance equations were not presented. The Panel provided suggestions to improve model
documentation and clarity, including clearer descriptions of the strategy behind the model
structure and the biological relevance of each model equation. Model assumptions need to be
more clearly described and the consequences of potential violations of these assumptions should
be discussed. In addition, a more detailed justification was needed for the handling of between-
animal variability in the model. The Panel agreed that use of the Bayesian framework for
estimation and characterization of the PBPK model parameter uncertainties was appropriate.
However, a more thorough description was needed for the choice of prior distributions, the
Bayesian fitting methodology, and the fit of the posterior distribution for each model parameter.
The Panel also generally endorsed the hierarchical calibration approach that uses the posterior
results in mice to establish the rat priors, and the rat posterior results to set the human priors.
The Panel also recommended performance of a local sensitivity analysis to identify key model
parameters that drive changes in modeling results.
Meta-Analyses of Cancer Epidemiology
The Panel agreed that EPA's updated meta-analyses for kidney cancer, lymphoma and
liver cancer followed the National Research Council (NRC, 2006) recommendations. The Panel
agreed with EPA's conclusions that TCE increased the risk for the three cancers studied, based
on appropriate inclusion criteria for studies, the methods of conducting the meta-analysis that
included consideration of bias and confounding, and the robustness of the findings based on the
tests for heterogeneity and sensitivity. The Panel also suggested performing a meta-analysis for
lung cancer to further support the absence of smoking as a possible confounder.
-------
Non-Cancer Hazard Assessment
EPA has provided a comprehensive synthesis of the available evidence regarding the
effects of TCE and its major metabolites on the central nervous system, the kidney, the liver, the
immune system, the male reproductive system, and the developing fetus. One issue of concern
was the inconsistencies between reported levels of glutathione conjugation pathway metabolites.
The Panel recommended that the impact of these divergent levels be more transparently
presented. The Panel recommended inclusion of the potential for TCE-induced immune
dysfunctions (i.e., immunosuppression, autoimmunity, inappropriate and/or excessive
inflammation) to mechanistically underlie other adverse health endpoints.
Carcinogenic Weight of Evidence
The Panel agreed with EPA's conclusion that TCE is "Carcinogenic to Humans" by all
routes of exposure. This is based on convincing evidence of a causal association between TCE
exposure and kidney cancer, compelling evidence for lymphoma, and more limited evidence for
liver cancer as presented in the draft document. The epidemiologic data, in the aggregate, were
quite strong. The summary risk estimates from the meta-analyses provided a clear indication of
a cancer hazard from TCE. In addition, both animal data and toxicokinetic information provide
biological plausibility and support the epidemiologic data.
Role of Metabolism
The Panel agreed with EPA's conclusion that oxidative metabolites of TCE were likely
responsible for mediating the liver effects. The Panel recommended that EPA examine studies
that provided quantitative assessment of trichloroacetic acid (TCA) and dichloroacetic acid
(DCA) formation after TCE exposure. Dose-response modeling, similar to that performed for
tetrachloroethylene, may be considered by EPA to provide scientifically-based information on
relative contribution, or lack thereof, of TCA and/or DCA to the liver carcinogenesis effect of
TCE.
EPA has provided a clear and comprehensive summary of the available evidence that
metabolites derived from gluthione (GSH) conjugation of TCE mediate kidney effects. The
Panel noted that uncertainties exist with regard to the extent of formation of the dichlorovinyl
metabolites of TCE between humans and rodents. The issue of quantitative assessment of the
metabolic flux of TCE through the GSH pathway vs. the oxidative metabolism pathway needs to
be considered carefully. A more complete discussion of the strengths and limitations of the
analytical methodologies used should be provided to address the large discrepancies in estimates
of S-dichlorovinyl glutathione (DCVG) formation.
Mode of Action (MOA)
The Panel agreed that the weight of evidence supports a mutagenic MOA for TCE-
induced kidney tumors. However, the Panel concluded that the weight of evidence also
-------
supported an MOA involving cytotoxicity and compensatory cell proliferation and including
these may more accurately reflect kidney tumor formation than does a mutagenic mechanism
alone. The combination of cytotoxicity, proliferation and DNA damage together may be a much
stronger MOA than any individual components.
The Panel agreed that the data are inadequate to conclude that any of the TCE-induced
cancer and non-cancer effects in rodents are not relevant to humans.
The Panel agreed that there is inadequate support for peroxisome proliferator activated
receptor alpha (PPARa) agonism and its sequellae being key events in TCE-induced human liver
carcinogenesis. Recent data from animal models (Yang et al., 2007) suggest that activation of
PPARa is an important but not limiting factor for the development of mouse liver tumors, and
additional molecular events may be involved. The Panel viewed the mode of action (MOA) for
liver carcinogenicity in rodents as complex rather than unknown. It is likely that key events from
several pathways may operate leading to acute, subchronic and chronic liver toxicity of TCE.
Susceptible Populations
The Panel found EPA's hazard assessment provided a good review of potentially
susceptible populations, and identified factors (genetics, lifestage, background, co-exposures and
pre-existing conditions) that may modulate susceptibility to TCE carcinogenicity and non-cancer
effects. However, the Panel disagreed with EPA's conclusion that toxicokinetic variability can
be adequately quantified using existing data. The Panel recommended that exposure to solvent
mixtures should be considered for potential co-exposures, since exposure to more than one
chemical with the same target organ likely increases risk.
Selection of Critical Studies and Effects
The Panel supported the selection of a RfC and RfD based on multiple candidate
reference values that lie within a narrow range at the low end of the full range of candidate
reference values developed, rather than basing these values on the single most sensitive critical
endpoint. The Panel expressed concerns about the use of several candidate critical studies and
effects, specifically National Toxicology Program (NTP, 1988) [toxic nephropathy], National
Cancer Institute (NCI, 1976) [toxic nephrosis], and Woolhiser et al. (2006) [increased kidney
weights]. However, the Panel noted that uncertainties about the quantitative risk assessment
based on kidney effects in NTP (1988), NCI (1976) and Woolhiser et al. (2006) did not indicate
that there was uncertainty that TCE caused renal toxicity. As discussed previously, the three
PBPK model-based candidate RfCs/RfDs (p-cRfCs/RfDs) for renal endpoints were based on an
uncertain dose metric, especially in regard to the relative rate of formation of the toxic metabolite
in humans and animals. Additional issues related to choice of toxic nephropathy in female
Marshall rats from NTP (1988) included excessive mortality due to dosing errors and possibly
other causes, and a high level of uncertainty in the extrapolation to the benchmark dose (BMD)
due to the use of very high doses and a high incidence (>60%) of toxic nephropathy at both dose
levels used. With respect to toxic nephrosis in mice from NCI (1976), the BMD analysis was not
supported because the effect occurred in nearly 100% of animals in both dose groups, and
because a high level of uncertainty is associated with extrapolation from the Lowest Adverse
-------
Effect Level (LOAEL) at which nearly 100% animals were affected. Renal cytomegaly and
toxic nephropathy, which were not selected as critical effects, occurred at high frequency in all
treated groups.
The Panel recommended that the two endpoints for immune effects from Keil et al.
(2009) and the cardiac malformations from Johnson et al. (2003) be considered the principal
studies supporting the RfC. The Panel also recommended that the endpoints for immune effects
from Keil et al. (2009) and Peden-Adams et al. (2009) and the cardiac malformations from
Johnson et al. (2003) be considered as the principal studies supporting the RfD.
Derivation of RfD and RfC
The screening, evaluation, and selection of candidate critical studies and effects used for
the development of the RfC and RfD were sound. The derivation of the points of departure
(PODs) was generally appropriate. However, the BMD modeling results were uncertain for
some datasets. For example, the log-logistic BMD analysis for toxic nephropathy in female
Marshall rats in NTP (1988), shown in Figure F-10 in Appendix F, may greatly overestimate the
risks at low doses. As discussed above, this modeling involved extrapolation from a high
LOAEL at which a high percentage of the animals were affected.
EPA used PBPK-based dose metrics for interspecies, intraspecies, and route-to-route
extrapolation. The Panel supported this approach for development of the RfC and RfD. The
Panel noted that the candidate RfDs /RfCs for kidney endpoints were highly sensitive to the rate
of renal bioactivation of the cysteine conjugate, S-dichlorovinyl glutathione (DCVC), in humans
relative to rodents. Candidate RfDs/RfCs developed using this dose-metric were several
hundred-fold lower than RfD/RfCs for the same endpoints based on applied dose with standard
uncertainty factors. The Panel noted that the uncertainties about the in vitro and in vivo data
used to estimate the rate of renal bioactivation of DCVC were much greater than for other dose
metrics [e.g. there are large discrepancies in the rates of human glutathione conjugation reported
by Lash et al. (1999a) and Green et al. (1997a)]. These uncertainties should be clarified and
should be the basis of a sensitivity analysis in the next update of the TCE draft risk assessment.
The Panel also recommended that the rationale for scaling the dose metric to body weight3 4, in
conjunction with the interspecies extrapolation based on PBPK modeling, should be presented in
a clearer and more transparent way.
Uncertainty Factors
The Panel agreed that, in general, the selection of uncertainty factors was clearly and
transparently described and appropriate. EPA developed equivalent doses and concentrations for
sensitive humans to replace standard uncertainty factors for inter- and intra-species
toxicokinetics. The Panel concluded that the approach used, including the selections of PODs
and the extrapolations from rodent to human, followed by consideration of the 99th percentile
human estimates, was acceptable to address the sensitive population. In future work, the
variability and uncertainty could be better characterized by considering other quantiles of the
distribution.
-------
Inhalation Unit Risk and Oral Unit Risk
In this assessment, EPA developed an inhalation unit risk and oral unit risk for the
carcinogenic potency of TCE in accordance with the approach outlined in the U.S. EPA Cancer
Guidelines (U.S. EPA, 2005a, b). The unit risks for renal cell carcinoma were based on a case
control study published by Charbotel et al. (2006). The Panel found that the analysis of the
Charbotel et al. (2006) data was well described and that the selection of this study to estimate
unit risks was appropriate. However, more discussion is needed on whether or not it is necessary
to adjust for exposure to cutting oils when computing an odds ratio or relative risk relating TCE
exposure to kidney cancer. The Panel recommended that EPA take a closer look at the literature
to determine if there are other studies which suggest that exposure to cutting oils is a risk factor
for kidney cancer. EPA should also provide a more detailed discussion on the implication of
assumptions made in their analysis. In addition, background kidney cancer rates in the United
States were used in constructing the life table, although the Charbotel et al. (2006) data was
based on a French cohort. A comparison of background cancer rates in France and the United
States would be helpful in supporting their conclusions. The Panel supported the adjustment of
the renal cell carcinoma unit risks to account for the added risk of other cancers, using the meta-
analysis results and Raaschou-Nielsen et al. (2003).
The Panel agreed that human data, when available, should be preferred over rodent data
when estimating unit risk since within species uncertainty is easier to address than between
species uncertainty. The Panel supported the use of linear extrapolation from the POD for cancer
dose-response assessment of TCE as a default approach. The Panel agreed that characterization
of uncertainty and variability was appropriate, and was exceptionally strong in the PBPK
models.
Age-Dependent Adjustment Factors (ADAFs)
The Panel agreed that application of age-dependent adjustment factors (ADAFs) in the
TCE analysis consistently followed recommendations in the U.S. EPA Cancer Guidelines (U.S.
EPA, 2005a). All of the steps were clearly presented for inhalation exposure. However, the
discussion for the oral exposure route was shortened and referred back to the inhalation section,
making understanding of the example difficult to follow. Currently, EPA's IRIS assessment
provides lifetime cancer risk drinking water concentrations for adults only. The Panel
recommended that drinking water concentrations for specified cancer risk levels should also be
derived for various age groups.
-------
RESPONSES TO EPA'S CHARGE QUESTIONS
1. PBPK Modeling
Is EPA's updated PBPK model for TCE and its metabolites (also reported in Evans et al.,
2009, and Chiu et al., 2009) clearly and transparently described and technically and
scientifically adequate for supporting EPA's hazard characterization and dose-response
assessment? Specifically, please address the PBPK model structure; Bayesian statistical
approach; parameter calibration; model predictions of the available in vivo data; and
characterization of PBPK model dose metric predictions, including those for the GSH
conjugation pathway.
Response
la PBPK model structure
According to the TCE Review Document (page 3-64), the version of the PBPK model
published by Hack et al. (2006) consisted of many parameter values that differed by study,
particularly in the case of metabolism. In addition, according to the authors, DCA metabolism in
the lung compartment remained highly uncertain. Subsequently, the EPA made efforts to
improve the 2006 model using an extensive analysis with different datasets to produce the PBPK
model used in this risk assessment. The Panel found that this PBPK model expansion seemed to
accurately predict the internal dose in the target tissue. The Panel agreed that using a PBPK
model improved the quality of the predictions for risk assessment and anticipated that the current
model will reduce uncertainties that resulted from the use of previous PBPK models.
The Panel found that, for the most part, the PBPK model was well presented in the TCE
Review Document but also noted that improvement was still possible. For example, the
conceptual representation of the PBPK model given in Figure 3-7 (page 3-69) was useful in
understanding the changes made to the Hack (2006) model, but did not facilitate a full
understanding of the model structure. Figure 3-7 could be expanded to also include the symbols
used for the model parameters (e.g. blood flow and metabolic parameters along with the
appropriate arrows and volumes in the compartments).
The Panel agreed that the details provided in Appendix A fully explain how the
population model was structured. However, the description of the PBPK model was incomplete
in that the mass-balance equations are not presented. In parallel to presenting these equations,
references should be given to Figure 3-7 (PBPK model structure) and Table A-4 (PBPK model
parameters). A better description would facilitate a complete understanding of both the
conceptual and mathematical structure of the model. The Panel suggested the following
additions: 1) a more detailed explanation of how interspecies extrapolation was performed,
especially the use of scaling equations, 2) graphical comparisons of prior vs. posterior
distributions for all key parameters, and 3) fits and the graphs of the concentration-time profiles
and the predictions of critical dose metrics. These additions can be made to either the master
-------
document or incorporated into Appendix A. Many of the desired graphics could be found in the
"linked documents" but these were overlooked by many reviewers because they were not part of
the formal documentation. Placing many of these graphics alongside the model descriptions will
improve both clarity and transparency.
On the issue of PBPK model structure, the Panel had some difficulty in fully
understanding its structure, and also noted deficiencies in the mathematical descriptions for each
compartment. With enough work and persistence, the structure was understandable, but these
deficiencies will be a bigger issue for users who are not experts in PBPK modeling. The Panel
made recommendations regarding improvements to the documentation of the PBPK model.
The Panel believed that the model documentation should also highlight any questionable
assumptions and discuss the potential implications of these assumptions being wrong. The Panel
observed that there remained a significant amount of variability between animals that did not
seem to be accounted for in the final model. Because the raw data sets were not available to the
Panel, it was difficult to determine if this was indeed the case. In addition, some analyses
discussed by the Panel would appear to be computationally unfeasible. The Panel initially
discussed extensions of the model which would avoid some of these problems (e.g., inclusion of
animal-specific parameters), but decided that these extensions are computationally unfeasible
given current resources.
Recommendations:
• Provide a better description of the final model structure and, in particular, provide a revised
model structure diagram that identifies model parameters with model states and pathways
(flows).
• Clarify the strategy behind the model structure and describe the biological relevance of each
model equation.
• Document model assumptions and discuss the consequences of potential violations of these
assumptions, e.g. impacts on bias and accuracy.
• Provide a more detailed justification for how between animal variability is accounted for in
the model.
Ib Bayesian statistical approach
The Panel agreed with the EPA that use of the Bayesian framework for estimation and
characterization of the PBPK model parameter uncertainties was appropriate. The general
description of the Bayesian approach presented in the TCE review document was acceptable.
The description of how uncertainty and variability are characterized was confusing mainly due to
the inconsistent use of the terms "population" and "group." The description of the Bayesian
model fit suffered from a lack of sufficient detail to provide complete transparency. Several
model parameters entered the Bayesian estimation method with wide and uniform prior
distributions. The large number of such parameters made the Markov Chain Monte Carlo
(MCMC) chains longer, resulting in long time to convergence and wide posterior distributions.
The Panel noted high variability in the posterior distributions of many model inputs and the
stated parameters. However the posterior distributions for many internal dose stated parameters
were much less variable.
-------
The Panel would have liked to see the extent to which posterior parameter distributions
are correlated. If rodent parameters were correlated as might be expected, how this correlation
was accounted for in human-specific model parameter estimates should be discussed.
Recommendations:
• Present better descriptions and/or details on the choice of prior distributions, the Bayesian
fitting methodology and fit of the posterior distribution for each model parameter.
• Provide some information on correlations around posterior medians for species-specific
parameters.
• Supply more information on the model ordinary differential equations and on the likelihood
function used in the Bayesian estimation.
Ic Parameter Calibration
Parameter calibration as described in the draft Document was accomplished via a
hierarchical fitting approach that used the posterior results in mice to establish the rat priors and
the rat posterior results to set the human priors. The Panel generally endorsed this hierarchical
fitting approach.
Recommendation:
• Improve the quality and the description of the assumptions underlying the use of the
hierarchical approach to parameter calibration. Help the reader to understand the extent to
which these assumptions are used consistently throughout the parameter calibration process.
Id Model Fit Assessment and Dose Metric Projections
There were a very large number of parameters in the PBPK model which made critical
review of the whole model and in particular identifying the key issues around model fit a
significant challenge.
A review of Figures 3-9, 3-10, A-3 and A-4, suggested that the updated model has
adequate fit. Table 3-45 was particularly useful, as were the graphs in the linked documents that
provided detailed descriptions of how well the model fit for the individual in vivo studies. When
evaluating the quality of each prior, the draft document focused on agreement of the interquartile
ranges. In Figure 3.9 (page 3-107), the vertical axes changed from the Hack model fit to the
updated model fit. This added a challenge to assessing model fit since the models were
predicting two slightly different quantities [N-Ac(l,2-DCVC) excreted (ug) for the Hack model
and N-Ac(l,2 or 2,2 -DCVC) excreted (ug) for the updated model].
As a measure of model goodness of fit, the draft document presented the residual error
geometric standard deviations (Table 3-41, page 3-98). The Panel was not certain how to use this
statistic. For example, what does it say about model fit when the residual error is GSD 2.7 for
venous blood TCE? Does this indicate a good fit or poor fit? For people who are not familiar
with the design of the PBPK model, it is hard to critically interpret the values in this table.
-------
The Panel pointed out other issues related to the evaluation of the posterior distributions.
Some of the posteriors were flatter than their priors, which was an unexpected result. In
addition, in Table 3-36, (section 3.5.6.2), pages 3-88 to 3-89, the Panel observed that prior and
posterior distributions of model parameters were almost identical and only in a few cases were
the distributions different.
The Panel noted that a large number of studies were available to EPA for this review.
Some of the rat studies were not used for parameter calibration and hence were used to assess the
validity of the model; that is, to determine whether the fitted model was adequate to predict data
from situations not specifically covered in the parameter estimation exercise. The Panel
approved of this approach, finding that even a limited validation analysis improved the
confidence of users in the final PBPK model and helped point to areas where the model may still
be inadequate.
Recommendations:
• Move some graphical presentations from the linked graphics documents into the body of the
report or into Appendix A.
• Incorporate more discussion on model fit and in particular indicate areas where the model fits
well and areas where it did not fit well. Tie this discussion somehow to Table 3-41.
• Include graphs that show predicted versus observed values for all data points used in the
analysis (one graph per endpoint).
• To help readers identify which parameters are better specified than others, provide a table of
model parameters listed in reverse order by the width of their posterior variability (width of
the IQR or width of 95% CI).
• Identify those parameters with very different prior and posterior distributions and discuss
why this might be a reasonable result of the parameter calibration process. An alternative
would be to provide a table where parameters are ranked based on the percent change of the
posterior from the prior.
• Clarify which parameters are related to variability and which address parameter uncertainty.
Separate the discussion of the two types of parameters.
le Lack of an adequate sensitivity analysis
The charge to the Panel did not specifically address parameter sensitivity but the Panel
did discuss the lack of and need for some form of sensitivity analysis. A common feature of
PBPK models is that the output is highly sensitive to a few parameters (key parameters) and far
less sensitive to the remaining parameters.
Recommendation:
• Perform a local sensitivity analysis, starting from the final fitted PBPK model, to assess how
small changes in model parameter estimates impact predictions. Provide graphical
presentations of the sensitivity of the model to changes in key model parameters in the final
documentation.
-------
2. Meta-analysis of cancer epidemiology
NRC (2006) recommended that EPA develop updated meta-analyses of the epidemiologic
data on TCE exposure and cancer, and provided advice as to how EPA should conduct
such analyses. Is EPA's updated meta-analysis of the epidemiologic data on TCE exposure
and kidney cancer, lymphoma, and liver cancer clearly and transparently described and
technically and scientifically adequate for supporting EPA's hazard characterization and
dose-response assessment? Specifically, please address the standards of epidemiologic
study design and analysis as they were applied to select studies for inclusion in the meta-
analysis; the rationales for study relative risk estimate selections; the meta-analysis
methods; and the characterization of the conclusions of the meta-analyses. [Note: The
scope of this charge question only includes the meta-analysis methods and results and not
the overall weight of evidence for TCE carcinogenicity, which is addressed as part of a
subsequent charge question.]
Response
NRC recommended that EPA conduct a new meta-analysis and to (1) pay attention to
essential design features; (2) include only studies where exposure is documented; (3) classify
studies on objective characteristics; (4) assess study power for each; (5) combine cohort and
case-control studies unless it introduces substantial heterogeneity; (6) test for heterogeneity; and
(7) perform sensitive analyses.
The Panel agreed that EPA followed these principles in their meta-analyses for
lymphoma, and cancers of the kidney and liver. The EPA approach was clearly and transparently
described and technically and scientific appropriate for supporting EPA's hazard characterization
and dose-response assessment. The Panel found EPA performed a thorough literature review and
clearly developed a comprehensive listing of candidate studies for the meta-analyses. The
strengths and weaknesses of each study were characterized and clearly presented in the draft
document. Procedures for selection of studies for the meta-analyses were clearly described.
Studies selected for inclusion had clear indications of TCE exposure and included
exposure assessments for each study participant. Exposure levels differed considerably among
and within the studies, which was an advantage. Candidate studies were also evaluated based on
study design, endpoints evaluated, TCE exposure assessment, follow-up procedures for cohort
studies, interview type (for case-control studies), use of proxy respondents (for case-control
studies), sample size, and statistical analysis. Information on these factors was clearly presented
for each candidate study. Appropriate criteria for including and excluding studies from the meta-
analysis were developed and carefully applied. Reasons for excluding studies were clearly
stated. Studies included had cohort or case-control designs, appropriate evaluation of cancer
incidence or mortality, adequate selection of study subjects, characterization of individual TCE
exposure for each subject, and relative risk estimates for lymphoma or cancers of the kidney or
liver adjusted for at least age, sex, and race. For example, studies where individual exposure to
TCE could not be reasonably determined were excluded, even though some exposure to
individuals in the group was a reasonable assumption. Although excluded studies likely included
some individuals who had exposure to TCE, exclusion was appropriate because inclusion would
10
-------
likely result in classification of some unexposed individuals as exposed, which would increase
exposure misclassification and bias estimates of relative risk downward. The Panel found EPA
carefully considered and described overlap between different studies (because of slightly
overlapping study populations and extended follow-up of individual cohorts) and made
appropriate selection of the results to include in the meta-analyses. The strengths and weaknesses
of the meta-analyses were appropriately considered in the evaluation and interpretation of the
results in relation to hazard characterization.
The Panel found that EPA discussed possible misclassification of exposure and disease
for the studies included in the meta-analyses. EPA appropriately noted that most exposure
assessment limitations would diminish relative risks and mute exposure-response gradients.
EPA indicated that in only one study were the interviewers blinded with regard to
case/control status. Although it is desirable to attempt blinding for case-control studies, it is
usually not possible to fully accomplish this because subject responses during the interview
provide clues as to subject status. The Panel thought this was not a serious limitation.
The Panel found that EPA clearly described the statistical techniques used in the meta-
analyses. Both random and fixed-effect models were used in the meta-analyses. This was useful
to assess the accuracy of the underlying assumptions regarding study variation. The Panel
agreed with EPA's reliance upon the random effects models for interpretation. Use of several
approaches to evaluate heterogeneity provided a fuller characterization than would be available
from any single technique. The potential for publication bias was appropriately evaluated. The
robustness of the findings was highlighted based on the tests for heterogeneity and sensitivity.
Results from the meta-analyses were fully and clearly presented in tables and figures.
Meta-analyses were performed only for lymphoma, and cancers of the kidney and liver.
The text did not make clear why only these three were selected for the meta-analysis approach,
although it was assumed this was because prior reviews of the literature had identified these
cancers as possibly associated with TCE exposure. The Panel found it might be useful to have
information on other cancers to provide evidence regarding possible confounding. For example,
kidney cancer was associated with smoking. Most cohort studies lacked information on tobacco
use. However, if there was confounding by smoking, there would have to be an excess of lung
cancer and other tobacco-related diseases in the cohorts. Absence of an excess of lung cancer
was very strong evidence that workers exposed to TCE did not smoke more than the unexposed,
or comparison population. Although no studies had excess of lung cancer, a meta- analysis of
lung cancer showing no association with TCE would document this conclusion regarding
possible confounding. Smoking could not cause excesses of kidney cancer, liver cancer or
lymphoma without also causing an excess of lung cancer. The lack of effect of TCE for lung
cancer in individual studies provided convincing evidence that confounding by smoking is
unlikely.
The Panel agreed that EPA carefully evaluated the data from the studies included in their
review and results from the meta-analyses against standard epidemiologic criteria for causality,
i.e., consistency, strength of the association, specificity of the association, temporal relationship,
exposure-response gradient, biologic plausibility, coherence, experimental evidence, and
analogy. The document provided a full discussion of these issues.
11
-------
Bias and confounding are concerns in epidemiologic studies. The Panel agreed that the
draft document had a strong discussion on potential confounding. Age, gender and race were
appropriate potential confounders to include in the meta-analyses and the meta-analyses included
effect estimates that were adjusted. The potential for confounding was evaluated in a number of
ways. Several of the case-control studies could directly adjust for potential confounding from
important risk factors and provide directly adjusted relative risks. EPA also pointed out that
many potential confounders, e.g., obesity, diabetes, tobacco, and hypertension in kidney cancer,
were unlikely to be associated with the level of TCE exposure and, thus, were unlikely to
confound. If these factors did confound, other cancers would be affected. Other occupational
exposures were mentioned as possible confounders, e.g., other organic solvents, cutting fluids,
and hydrazine. The link between most of these and the cancers of concern relative to TCE was
weak or non-existent, so they were not strong candidates for confounding. Biases are also a
concern in observational studies. In case-control studies, case-response bias and case or control
selection bias are a concern, while in cohort studies biases associated with follow-up and
exposure are a concern. No obvious bias that would occur across studies of different designs, in
different countries, and with different exposure metrics falsely produced an association with
TCE. The Panel did not think confounding or bias were likely explanations for the findings from
the epidemiologic studies and meta-analyses.
The Panel agreed that the findings of several community studies although intriguing,
were appropriately omitted from the meta-analyses due to large misclassification errors and lack
of control for confounding, which would tend to bias estimates from the meta-analysis.
The Panel found that EPA appropriately discussed the changing classification of
hematopoietic and lymphatic system tumors and selected lymphoma (predominately non-
Hodgkin's lymphoma (NHL) as an outcome for meta-analysis. EPA specifically wanted to
select studies with the best outcome definitions, rather than pick at studies where the
hematopoietic cancers were grouped, (e.g. myeloid and lymphoid neoplasms together). EPA
selected studies representing various groupings of NHLs (with some studies that included
chronic lymphocytic leukemia) or focused on specific subtypes of NHL (including one study that
focused on hairy cell leukemia), but did not include studies of Hodgkin lymphoma (if any such
studies existed). Given that the EPA's intent was to conduct a meta-analysis with NHL as the
outcome, the Panel felt that the terminology should be changed to 'non-Hodgkin lymphoma'
instead of 'lymphoma', throughout the document. The term 'NHL' more accurately describes
the intent of the analysis as well as the overwhelming majority of cases in the analysis, despite
changing classification schemes. The focus of the meta-analysis on NHL and any indication in
the meta-analysis where cases definition may diverge from classical NHL (as in studies that
included chronic lymphocytic leukemia) should be clearly explained in both Appendix C and in
the Hazard Characterization section (section 4.6.1.2.2) in the main document.
The Panel agreed that appropriate approaches were used in the meta-analysis. Effect size
(the relative risks or odds ratios) included in the meta-analyses were selected appropriately using
the most appropriate selection criteria. However the Panel had a few questions of clarification
about the meta-analysis for kidney cancer.
12
-------
There are a number of technical points that should be mentioned as footnotes to the meta
analysis plots. First, the exact confidence intervals given in the original publications have been
replaced with approximations. The Panel suggests that the explanation in Appendix C be
reiterated in the main document. For reference, Appendix C, Table C-6 (pages C-26 to C-27)
shows the actual SE(logRR) used to calculate the weights. In addition, Appendix C, page C-3,
lines 14-20 explains the discordant confidence intervals in the figures. A second example is that
a 20 year lag was used for the Zhao study while lags were either not given or not used in the
other studies. Clarify the rationale for selecting the "20 yr lag" result from Zhao et al. (2005) and
not selecting the "20 yr lag" result from Raaschou- Nielsen et al. (2003).
The Panel agreed with EPA's conclusions from the meta-analyses that TCE increased the
risk for the three cancers studied. The Panel's agreement with EPA's conclusion was based on
the strict and appropriate inclusion criteria, the methods of conducting the meta-analyses,
including consideration of bias and confounding, and the robustness of the findings based on the
tests for heterogeneity and sensitivity.
Recommendations:
• Provide a rationale for the three cancer sites selected for the meta-analysis. The rationale
could be nicely summarized in a table.
• Consider including meta-analysis for lung cancer for confounding purposes or other sites
for comparison for which some association with TCE exposure has been reported in
epidemiologic studies, such as childhood leukemia and cervical cancer. It might also be
possible to provide this information without a formal meta-analysis.
• Provide measures of heterogeneity such as the I2 statistic for each meta-analysis.
Although this information was provided and accurately explained in Appendix C, it was
mischaracterized at several points in the primary document. For example, the summary
of the kidney cancer meta-analysis on p. 4-167 of the primary document states that "there
was no observable heterogeneity across the studies for any of the meta-analyses," but
Appendix C indicates "the I value of 38% suggested the extent of the heterogeneity was
low-to-moderate." Non-significant heterogeneity is indeed observed heterogeneity.
• Evaluate the likely impact of converting odds ratios to relative risk estimates (i.e., using
the method of Greenland (2004) or Zhang and Yu (1998), and decide if necessary to
perform these conversions for the meta-analysis.
• Change the terminology regarding the meta-analysis results for 'lymphoma' to 'non-
Hodgkin lymphoma' throughout the document.
13
-------
3. Non-Cancer Hazard Assessment
Does EPA's hazard assessment of non-cancer human health effects of TCE logically,
accurately, clearly, and objectively represent and synthesize the available scientific
evidence to support its conclusions that TCE poses a potential human health hazard for
non-cancer toxicity to the central nervous system; the kidney; the liver; the immune
system; the male reproductive system; and the developing fetus, including the role of TCE
in inducing fetal cardiac defects?
Response:
The Panel agreed that the EPA's TCE hazard assessment has clearly, accurately, logically
and objectively represented and synthesized the available scientific evidence to support its
conclusions that TCE poses a potential human health hazard for non-cancer toxicity.
Specifically, the EPA has provided a comprehensive and thorough synthesis of the available
evidence regarding the effects of TCE and its major metabolites in each of the tissues addressed
in the charge question. This includes human epidemiological studies, animal studies, in vitro
studies using renal cell cultures, and in vivo and in vitro metabolism studies.
3a Central Nervous System
TCE-associated auditory impairment was discussed in this section (4.3.2.3.). It is noted
that auditory impairment is commonly seen with various autoimmune conditions and
inflammation-based diseases and these were among the immune dysfunctions observed with
TCE exposure.
3b The Kidney
In regard to the effects of TCE in the kidney, EPA had provided a thorough and clear
description of these effects. One issue of concern here was the quantitative aspect of the GSH
pathway metabolites. Dr. Wolfgang Dekant, in his public comment, suggested that data obtained
using the "Reed method" overestimated the amount of DCVG produced. This HPLC method is
characterized by variability and overall decline in retention times over the life of the HPLC
column due to derivatization of amine groups on the column (Lash etal., 1999b). Although data
are limited, GSH pathway metabolite levels reported by methods that utilize 14C TCE and
radiochemical detection followed by mass spectrometry identification of the metabolites (Green
et al, 1997a) are lower than those from reports using the "Reed method". In addition, studies
using HPLC-MS/MS techniques with stable isotope-labeled DCVG and DCVC standards have
also been used to detect GSH pathway metabolite levels (Kim et al, 2009). Based on the in vitro
work presented in Table 3-23 (page 3-44 of the draft EPA document) determining DCVG
formation by the "Reed method" in human, rat and mouse liver, one would expect mouse serum
DCVG levels to be -4-6 times lower than humans. However, using the HPLC-MS/MS technique
of Kim et al., the peak DCVG serum levels are -1,000 times lower in mouse serum than
determined by Lash et al. (1999a) in human serum. Although differences in exposure routes,
exposure doses, etc. should be considered, this much larger than expected difference also
suggests that the "Reed method" provides an overestimation of DCVG levels in humans. This
could occur if the "Reed method" identifies non-specific derivatives as DCVG or other GSH
14
-------
pathway metabolites. Thus, interpretation of DCVG levels from the Lash et al. (1999a) paper
should be made with caution.
It is noted that the focus on animal data in the EPA report is appropriate because human
data on non-cancer kidney effects from TCE are limited by two factors. The first is outcome
assessment. Due to the insensitivity of the clinical kidney outcomes such as glomerular filtration
rate and end stage disease, human nephrotoxicant work often uses kidney early biological effect
markers. Unfortunately, research to accurately determine the prognostic value of these
biomarkers is fairly limited and data analysis in many of these studies is quite rudimentary often
involving only a comparison of unadjusted mean values between an exposed and a control group.
A range of biomarkers are used and results are frequently not entirely consistent as noted in
Section 4.4. The second challenge is that human exposure often involves a mixture of solvents
making determination of the impact of an individual solvent difficult. For example, the GN-
PROGRESS retrospective cohort study in Paris, France, which examined the impact of solvents
on risk of end stage renal disease (ESRD) and progression of glomerulonephritis, included
patients with a wide range of solvent exposures. Solvent exposure was assessed by industrial
hygienists from lifetime occupational histories collected by interview and a list of the 30 most
common solvents. These authors noted an elevated risk for progression of glomerulonephritis to
ESRD from TCE although numbers were small and did not achieve statistical significance
(adjusted hazard ratio [95% CI] 2.5 [0.9 to 6.5]) (Jacob et al, 2007). These authors also did not
discuss how they addressed exposure to solvent mixtures as they attempted to focus on specific
agents.
3c The Liver
The only criticism noted for this section was the (perhaps unavoidable) repetitive nature
of their coverage, as these issues appeared elsewhere in the document. Less repetition and better
integration of these sections would improve the readability of the document.
3d The Immune System
It is noted that the children's exposure data and adverse outcomes are consistent with the
immunotoxicity reported in the animal developmental models. It is noted that while TCE
exposure can produce a range of immune dysfunctions, including immunosuppression, elevated
risk of autoimmunity and dysregulation of inflammation, it is possible that the doses of TCE
producing each category of adverse immune outcomes may differ. For example, most studies
reporting autoimmune dysregulation used higher doses of exposure compared with at least some
studies where immunosuppression was observed.
3e The Male Reproductive System
It is noted that male potency/sterility issues can be associated with inflammatory
dysfunction in the testes produced by some environmental pollutants (usually associated
testicular macrophage dysfunction) (see Pace et al., 2005). Since inflammatory dysfunction is
associated with TCE exposure, this is an additional possible mechanism that may be associated
with adverse outcome for male potency. For in utero exposure studies in rodents using lower
doses of TCE and metabolites, where effects (carcinogenic and non-carcinogenic) can be
15
-------
observed transgenerationally, attention should be directed to epigenetic changes as possible
MOA for TCE-mediated effects on the reproductive systems.
3f The Developing Fetus, Including the Role ofTCE in Inducing Fetal Cardiac Defects
It is noted that the type of cytokine dysregulation seen with TCE exposure (e.g.,
involving IL-6) can play a role in cardiac dysfunction. The report explains logically why the
Johnson et al. (2003) study was used to derive some reference points. Some recent publications
confirm and reinforce the results obtained in the Johnson et al. (2003) study and could be cited
to make a stronger argument. They are listed as follows:
• TCE effects on the cardiac system were specific for a narrow window of development
corresponding to myocardial expansion and endocardial cushion formation, consistent
with previous findings from Drake et al,2006a and b; Mishima 2006; Boyer et al. 2000,
and consistent with the definition of a teratogen.
• The types of defects and morphological changes (e.g cardiac hypertrophy and hypoplasia)
were consistent with a mechanism of action involving disruption of calcium handling and
cardiac contractility, observed by Caldwell et al, 2008 in rat cardiomyocytes. Numerous
literature data (reviewed in Lehnart et al., 2008; Lebeche et al, 2008; Yano et al., 2008;
Gyorke et al., 2008) confirm the notion that alteration of calcium homeostasis is
sufficient to induce alteration of contractility and in turn heart defects.
• A non-monotonic dose-response relationship was found that confirms several other
studies (Caldwell et al., 2008; Drake et al., 2006) suggesting the presence of more than
one MOA due to presence of metabolites, enzymatic sensitivity, etc.
Recommendations
• If additional endpoints of renal dysfunction (e.g. diuresis, increased glucose excretion)
were present in the reported studies, they should be included in the report. Often only
one or two parameters of renal function and histopathology were presented. A better
overall description of renal dysfunction should be presented if available (especially for
animal studies).
• There should be a better description of the location of the renal lesion, including nephron
segment, if known. For example, TCE and DCVC appeared to affect the proximal tubule
at the level of the outer stripe of the medulla (S3 segment of proximal tubule). Is this the
site of lesions seen with other TCE metabolites? Explaining the role (or lack of a role) of
any other TCE metabolites in TCE nephrotoxicity could be strengthened by comparing
the sites of the renal lesion.
• On page 4-338, please clarify the use of the phrase, "subpopulation levels", on lines 31
and 33.
16
-------
• A statement should be added that the spectrum of TCE-induced immune dysfunctions
(immunosuppression, autoimmunity, inappropriate and/or excessive inflammation)
included in this EPA draft report has the potential to produce adverse effects that are seen
well beyond lymphoid organs and involving several other physiological tissues and
systems. The types of immune-inflammatory dysfunctions described in this report have
been observed to affect function and risk of disease in the nervous system (e.g., loss of
hearing), the skin, the respiratory system, the liver, the kidney, the reproductive system
(e.g., male sterility), and the cardiovascular system (e.g., heart disease, atherosclerosis).
• A statement should be added to emphasize the cell-mediated immune effects of TCE as
some of this has been supported by the human epidemiology data and the issue is
pertinent to risk of cancer.
17
-------
4. Cancer Hazard Assessment
Using the approach outlined in the U.S. EPA Cancer Guidelines (U.S. EPA, 2005a), does
EPA's hazard assessment of carcinogenicity logically, accurately, clearly, and objectively
represent and synthesize the available scientific evidence to support its conclusions that
TCE is carcinogenic to humans by all routes of exposure? Specifically, please address the
epidemiologic evidence for associations between TCE and kidney cancer, lymphoma, and
liver and biliary tract cancer; the extent to which the results of the meta-analyses
contribute to the overall weight of evidence for TCE carcinogenicity; the laboratory animal
data for rat kidney tumors, mouse liver tumors, and lymphatic cancers in rats and mice;
and the toxicokinetic and other data supporting TCE carcinogenicity by all routes of
exposure.
Response:
The Panel agreed that cancer hazard characterization hinges on the synthesis of the
accumulated scientific evidence, especially the epidemiologic evidence supporting the
carcinogenicity of TCE. Assessment of the causal association and weight of evidence supported
the conclusion that TCE is carcinogenic to humans by all routes of exposure as outlined in the
US EPA cancer guidelines. Results from animal bioassays and toxico-kinetic data provide
further support to the EPA conclusion. The report logically, accurately, clearly, and objectively
presented the methodological review of the epidemiologic evidence, highlighted the criteria for
study inclusion in meta-analyses and the meta-analysis methods (as noted in charge question 2)
and appropriately assessed the weight of the evidence to conclude that TCE is causally related to
lymphoma, and kidney and liver cancer.
Epidemiological Data
The report appropriately highlighted the causal criteria in support of the conclusion. The
consistency of the findings was notable given the rarity of the cancers, differences in latency and
potential for exposure misclassification as described in the study assessments highlighted in the
hazard characterization. Multiple explanations would be needed to account for the associations
between TCE and several cancers from studies with differing designs, strengths and weaknesses.
The summary risk estimates from the meta-analyses provided a clear indication of a
cancer hazard from TCE. The pooled risk estimates from the meta-analyses for kidney cancer
and liver cancer, although modest, were robust with no indication of publication bias or
heterogeneity. Meta-analyses for both kidney cancer and lymphoma found higher increases in
the risk estimates associated with higher TCE exposure than for any TCE exposure and no
evidence of strong confounding, which further supported a causal association.
EPA concluded TCE is carcinogenic to humans by all routes of exposure. This
conclusion was based on convincing evidence of a causal association between TCE exposure and
kidney cancer, compelling evidence for lymphoma, and more limited evidence for liver cancer.
The epidemiologic data, in the aggregate, were quite strong. In addition, the epidemiologic data
were supported by bioassays and toxicokinetic data. Although issues of concern could be raised
18
-------
about individual studies, the overall pattern and the results from the meta-analyses were quite
compelling. Potential confounding from established risk factors for these cancers of concern
could be directly assessed in some studies and indirectly evaluated by reviewing cancer excesses
that did not occur in TCE exposed populations, e.g., the absence of an excess for lung cancer
indicates confounding from smoking is not likely.
Some studies had low power to evaluate the TCE-cancer relationship, but the meta-
analysis provides a tool to combine underpowered studies and assess the overall effect.
Exposure assessment in epidemiologic studies is difficult in the best of circumstances. EPA
appropriately focused on studies with the stronger exposure assessment efforts to minimize the
effects of exposure misclassification. However, misclassification of exposure undoubtedly
occurred. In the cohort studies the effect of exposure misclassification on estimates of relative
risk will be largely non-differential because factors used in exposure assessment were recorded
before occurrence of the disease. Thus, it will tend to depress estimates of relative risk and mute
exposure-response gradients and is not an explanation for any observed excesses. Non-
differential exposure misclassification would also occur in case-control studies. Differential
misclassification is more of a concern in case-control studies. Differential misclassification can
bias relative risks upward or downward, although the upward bias is usually raised in positive
studies. However, no evidence is available to suggest that differential exposure bias occurs
across all the case-control studies. The summary estimates from the meta-analysis provided a
clear indication of a cancer hazard from TCE. EPA concluded the association between TCE and
lymphoma and liver cancer were more limited than that for kidney cancer. These conclusions
about the epidemiologic data were supported by the statistically significant excesses for these
tumors in the meta-analyses, no statistically significant heterogeneity, and consistency of
findings after exclusion of individual studies in sensitivity analyses. The consistency of the
findings was remarkable given the rarity of the cancers, differences in latency and potential for
exposure misclassification, as described in the study assessments highlighted in the hazard
characterization.
EPA concluded that the epidemiology data were convincing for a causal association
between TCE and kidney cancer, compelling for lymphoma, and positive but more limited for
liver cancer. The Panel did not have strong disagreement with this statement, although some felt
that the data for liver cancer were as compelling as that for lymphoma. Liver cancer has a much
lower incidence than kidney cancer or lymphoma in Western countries (where most of the
epidemiologic studies were conducted) and this requires more reliance on the meta-analysis for a
summary effect estimate with adequate power. The meta-analysis found that the association of
TCE exposure with liver cancer was elevated and statistically significant. Further grouping liver
cancer cases by the level of exposure resulted in numbers that were too small to adequately
evaluate risks among persons with higher exposures. Nevertheless, we considered these results
for liver cancer to be strong because there was no evidence of heterogeneity or publication bias
in the meta-analysis, and because the epidemiologic findings were supported by observations of
liver cancer in animal models. Although potential confounding by other risk factors for liver
cancer is possible, strong risk factors such as hepatitis are very rare in Western countries (where
most of these studies were conducted), so this is unlikely to have caused such a degree of
confounding. There were no studies to evaluate whether hepatitis might be a confounder in
TCE studies, although this seemed unlikely.
19
-------
The meta-analysis results were impressive for lymphoma, showing a significantly
elevated relative risk for ever-exposure to TCE and an even higher effect estimate for high TCE
exposure. However, it is important to note that there was weak evidence of publication bias in
the lymphoma meta-analysis results, which means that studies showing no TCE effect or inverse
associations may not have been published. In addition, there was significant heterogeneity in the
meta-analysis results for lymphoma for ever-exposure to TCE, indicating that there is an
unexplained factor causing heterogeneity that indicates it may be inappropriate to combine the
estimates in a meta-analysis. This heterogeneity may reflect the complicated and changing
definitions for lymphoma across studies and over time. It is also possible that effects from TCE
may differ by type of lymphoma. The association with lymphoma was further supported by the
larger relative risk in meta-analyses for the higher exposure categories compared to the overall
relative risk. This was evidence for an exposure response gradient, even though no individual
studies showed much evidence of this.
Animal Data and Toxicokinetics
The Panel agreed that human data, when available, should be preferred over rodent data
when estimating unit risk since within species uncertainty is easier to address than between
species uncertainty. The Panel believed that the animal and toxicokinetic data were thoroughly
reviewed and the biologic plausibility and coherence of the epidemiologic findings were
supported by the laboratory animal data and the toxicokinetic data.
Recommendations:
• The immune effects as highlighted in the hazard assessment should be referred to in the
conclusion especially in the criteria of biological plausibility and coherence because of the
relationship between immune system dysfunction and cancer risk.
• Although the summary evaluation focused on the scientific evidence and meta-analysis for
kidney, lymphoma and liver cancers, there is also some suggestive evidence for TCE as a
risk factor for cancer at other sites including bladder, esophagus, prostate, cervix, breast and
childhood leukemia. This evidence that also supports the conclusion should be mentioned in
the summary evaluation (section 4.11.2.1).
• Add a paragraph describing the definition of lymphoma as used in IRIS. Change the
terminology regarding the meta-analysis to 'non-Hodgkin lymphoma' instead of
'lymphoma', throughout the document. The term 'NHL' more accurately describes the
intent of the analysis as well as the overwhelming majority of cases in the analysis, despite
changing classification schemes. The focus of the meta-analysis on NHL and the exact
classifications the meta-analysis includes where it may diverge from classical NHL (as in
studies that included chronic lymphocytic leukemia) should be clearly explained in both
Appendix C and in the Hazard Characterization document (section 4.6.1.2.2).
• To assist the reader, please include references in the summary section (section 4.11.2). For
example, "The other 13 high-quality studies [note: besides Hardell and Hansen] reported
20
-------
elevated Relative Risk estimates with overall TCE exposure that were not statistically
significant." References for statements like this would be helpful. The Panel counted
fewer than 13 studies in the meta-analysis after subtracting out Hardell and Hansen, and not
all of these showed elevated risk estimates, so it would be helpful for the reader to know
which 13 studies this statement refers to.
21
-------
5. Role of Metabolism on TCE Toxicity
Does EPA's hazard assessment logically, accurately, clearly, and objectively represent and
synthesize the available scientific evidence to support its conclusions regarding the role of
metabolism in TCE carcinogenicity and non-cancer effects? Specifically, please address
EPA's conclusions that the liver effects induced by TCE are predominantly mediated by
oxidative metabolism, but not adequately accounted for by the metabolite trichloroacetic
acid (TCA) alone and that the kidney effects induced by TCE are predominantly mediated
by metabolites formed from the GSH-conjugation pathway.
Response
The Panel agreed that EPA's hazard assessment in the draft document has produced a
systematic, thorough, objective and clear summary of information on the role of metabolism in
TCE-induced toxicity with regards to both cancer and non-cancer health effects. The Panel also
found that EPA has presented a comprehensive review of metabolite formation in animals and
humans, and has provided a clear, logical assessment of the role these metabolites play in
mediating its carcinogenic and non-cancer effects.
5a Mediation of TCE-induced Liver Effects by Oxidative Metabolism
The Panel found that EPA's conclusion that oxidative metabolites of TCE are responsible for
mediating the liver effects is sound and based on a wealth of supportive studies.
The document was a thorough review of the extensive literature on the role of oxidative
metabolism in TCE toxicity to the liver. Direct evidence that oxidative metabolism was required
for liver toxicity, such as studies which modulated TCE toxicity by modulating P450 activity,
was somewhat limited. One noted exception is the study by Ramdhan et al. (2008), that reported
CYP2E1-deficient mice produced considerably less oxidative metabolites and showed reduced
hepatoxicity, although due to a small number of animals studied, effects were significant only at
the highest TCE dose. Nonetheless, the collective evidence, especially from studies with two
major oxidative metabolites of TCE - TCA and DC A, was very strong that in rodents, at doses
where metabolism is not saturated, the majority of TCE was metabolized and that metabolites
from the oxidative pathway predominated over those of the glutathione conjugation pathway.
Mice are the most susceptible species with respect to TCE-induced liver effects and the majority
of studies support the conclusion the oxidative metabolites are playing the major role.
5b Contribution of TCA to Adverse effects on the Liver
The Panel found the conclusion that "the adverse effects on the liver of one of the TCE
metabolites, trichloroacetic acid, cannot adequately account for the liver effects of TCE" is
sound and supported by several lines of experimental evidence.
TCA is the predominant oxidative metabolite of TCE and its effects are well known to be
associated with liver toxicity and carcinogenicity. However, oxidative metabolism of TCE
generates a number of molecules and the confidence in the ability to identify TCE's oxidative
metabolite(s) that may be responsible for hepatotoxicity and/or liver cancer in rodents or humans
is much less than that for the overall role of oxidative metabolism. This uncertainty is due in
22
-------
part to the problems with quantitative assessment of DCA formation after TCE administration.
There is sufficient evidence to implicate DCA in mediating carcinogenic effects of TCE that are
not related to those produced by TCA. The EPA correctly stated that DCA was a minor
metabolite of TCE in vivo, at least in rodents, and that some of the earlier reports on DCA
dosimetry may have been erroneous due to the issues with the analytical methods. There are,
however, several studies (Delinsky et al., 2005; Kim et al., 2009) which provide information on
the blood levels of DCA after oral exposure to TCE in rats and mice. Such data, together with a
large body of literature on TCA formation after treatment with TCE, should be carefully
evaluated with regards to the estimation of the internal dose (or relative amounts) of each of
these key metabolites.
The Panel found that EPA has taken several approaches to determine whether liver
tumors induced by TCE can be accounted for by TCA formation alone. The first approach was to
compare dose-response profiles for non-cancer liver toxicity endpoints from TCE and TCA
based on TCA dose equivalents, an internal dose metric. In contrast to DCA, the quantitative
data available for TCA and TCOH, together with PBPK models relying on their measurements,
are among the most consistent and allow for the assessment of the oxidative metabolite flux from
TCE. Analysis of liver weight changes (Fig 4-7, 4-8) suggested that while total TCE oxidative
metabolism was strongly correlated with liver weight changes (R2 = 0.89), the amount of TCA
formed underestimated the degree of liver hypertrophy observed. The dose-response
relationships for liver hypertrophy observed between TCE and TCA, based on TCA daily dose
equivalents, were strikingly different in both slope of the dose-response and overall magnitude,
suggesting that the mechanisms of hypertrophy, and/or the metabolites involved, were different.
This analysis was compelling because TCA daily liver dose equivalents were used for
comparison. The internal dose metrics, if accurately applied, should account for potential
differences due to bioavailability and exposure route issues that have been previously raised for
TCE and TCA. The Panel notes that the bioavailability of TCE, DCA and TCA in oral gavage
studies was dependent, among many factors, on the type of the vehicle and the magnitude of the
administered dose. It has been suggested [Sweeney et al., 2009; NRC review of the IRIS
assessment of Tetrachloroethylene (Appendix B)] that the bioavailability of TCA (when
administered directly) was highly non-linear with an increasing dose. Thus, the internal dose of
each metabolite of interest, either through metabolism from TCE or following direct
administration, was key for the comparison of health effects between the parent and its
metabolites.
The second approach used in the draft document to support the conclusion that multiple
metabolites were involved in liver tumors induced by TCE included comparisons of liver
phenotypic markers (glycogen staining, c-jun staining) and tumor-derived genetic markers
(incidence of H-ras mutations). This analysis was interesting, yet qualitative in nature. The use
of phenotypic markers such as H&E staining, glycogen staining, antibody reactivity, tumor
tincture, etc., must be interpreted with caution since the underlying biochemistry/molecular
biology of these descriptive attributes is often not well understood and may be highly dependent
on the state of progression of the tumors The criteria used in each study for phenotypic
classification (i.e., staining intensity, background staining) is not always clearly outlined in the
original literature reports. The EPA has included adequate discussion noting the technical
limitations for each of the studies, which increased the confidence that such evidence from a
23
-------
single study was not overly weighted in drawing conclusions about the role of TCA. While
individual studies comparing phenotype/genotype of TCE-, TCA- and DCA- induced tumors
have important limitations, the collective group of studies was consistent with the interpretation
that TCE tumors displayed phenotypic and genotypic heterogeneity that was different than that
of tumors induced by TCA alone. This was in agreement with the EPA conclusion that these
data also did not support the hypothesis that TCA was a sole acting liver metabolite of TCE.
However, since factors such as interactions among metabolites and tumor progression state may
have unknown influences in the phenotype/genotypes observed, this type of qualitative evidence
was not sufficient to invoke specific roles for other contributing metabolites, or to discount
potential contributing roles of other metabolites.
The draft included little in terms of the comparative quantitative evaluation of the
hepatocarcinogenic potency of TCE, TCA and DCA even though extensive information was
available, especially in mice. A recent draft of the IRIS assessment of a highly related chemical,
tetrachloroethylene (PERC), provided the evaluation of the consistencies between PERC and
TCA with regards to the liver cancer endpoint (Appendix 4A of PERC IRIS draft document).
TCA is a major metabolite of both TCE and PERC and it is debatable whether TCA toxicity can
account for the majority (if not all) of the adverse liver effects of PERC.
Given the controversy of DCA as a contributing metabolite in liver effects induced by
TCE and the importance of this issue as it relates to understanding TCA's role, it is somewhat
surprising that there was relatively little analysis of the literature related to the use of DCA as a
therapeutic agent in humans as an integrated part of this section of the review. Although these
studies obviously involved high doses, they are relevant to the potential spectrum of effects
observed in humans.
Recommendations:
• The EPA should examine studies that provide quantitative assessment of TCA and DCA
formation after TCE exposure in vivo and draw conclusions with regards to the relative
amount and kinetics of the oxidative metabolites of interest for liver toxicity.
• A careful evaluation of the concentration-time kinetics is needed to achieve certainty in the
comparisons of liver effects and the conclusions drawn by the EPA which suggest that
TCA-induced adverse liver effects do not explain those observed with TCE. Equally
important is to fully consider the bioavailability of TCE itself with regards to the vehicle
effects between studies.
• The body of the document could be further strengthened by reporting EPA's evaluation on
the strength of the specific criteria used for phenotypic classification described in each study
discussed, and noting where specific criteria were not reported. While most of this
information was included in the appendix, the EPA may consider constructing a summary
table for Section 4.5.6.
• Dose-response modeling, similar to that performed for PERC, may be considered by the
EPA to provide science-based information on relative contribution, or lack thereof, of TCA
and/or DCA to the apical liver carcinogenesis effect of TCE. While data gaps exist and there
are limitations in the comparisons between independent cancer bioassays, the document
should clearly state what the limitations are should such analysis be deemed futile.
24
-------
• The draft assessment may be strengthened by including information from human use of
DCA in clinical practice.
5c Role of GSH-Conjugation Pathway on TCE-InducedKidney Effects
The Panel concluded that EPA has provided a clear and comprehensive summary of the
available evidence that metabolites derived from GSH conjugation of TCE are responsible for
mediating kidney effects.
The Panel found the integration of the data from human epidemiological, animal and in
vitro mechanistic studies produced a clear and transparent weight-of-evidence assessment
supportive of TCE GSH conjugation metabolites' role in kidney toxicity and cancer. Whereas
sufficient amounts of oxidative metabolites of TCE (i.e., TCOH) may be formed which could
contribute to kidney effects, potentially through formic acid, the literature indicated the
pathological effects on the kidney induced by oxidative metabolites were not consistent with
those observed with TCE. In contrast, the pathological effects on the kidney induced by
DCVC/DCVG were similar to TCE. Thus, a reasonable conclusion was that the glutathione
conjugation pathway played a more important role in driving these effects. The primary
challenge was to determine the true flux through the glutathione conjugation pathway.
Many uncertainties exist in PBPK model estimates for the GSH pathway. This issue is
critical, since these uncertainties can result in orders of magnitude differences in flux between
rodents and humans. The argument that mercapturates of the glutathione conjugates, as
detoxication pathway products, are not quantitative markers of flux through the GSH pathway is
rational and supported by in vivo human and rodent data. The level of urinary mercapturates, as
deactivation products, is evidence that the pathway operates in humans, but does not necessarily
reflect the amount of DCVC formed. Direct data on DCVG/DCVC formation, or its reactive
metabolites, are the more appropriate measures of flux for this pathway. This was clearly and
adequately discussed in the review.
The quantitative analysis of the species differences in GSH metabolism was somewhat
narrow. Specifically, the issue of vast differences in human vs rodent metabolism of TCE to
GSH conjugates hinged on the very limited experimental evidence. Only one human in vivo
study was available that directly quantified DCVG in urine in a few subjects (Lash et al. 1998).
The rodent in vivo data (Kim et al. 2009) was limited to only one isogenic (hybrid) mouse strain.
Other important differences between these studies were that they utilized different exposure
routes, doses, and analytical methods. The uncertainties associated with the potential several
orders of magnitude difference in TCE metabolism through GSH pathway between species
should be considered more carefully.
In addition, multiple in vitro studies have been published in the peer reviewed literature. For
example, in vitro GSH conjugation data were used to develop prior distributions for GSH
conjugation rates, something which was not done for previous PBPK models of TCE. Ample
discussion was given to the data generated by the Lash laboratory, which was clearly the most
extensive set of data relative to DCVG and DCVC levels in humans. These data indicated
DCVG may be formed at levels similar to that of oxidative metabolites in humans. Based on
25
-------
these data, the conclusion that the GSH conjugation pathway plays an important role in kidney
tumors/toxicity in both rodents and likely in humans is logical.
However the discussion of additional published in vitro studies that show disparately lower
results for DCVG formation (beyond mercapturates) was not given a comparable level of
attention. For example, the documents pointed out discrepancies between in vitro studies of
DCVG formation conducted by the Green and Lash laboratories that report results differing by
orders of magnitude. The studies from these labs reported very similar assay conditions using
the same strain of rats, but differed in the analytical techniques used (HPLC-UV versus GC-MS).
The analysis of these disparate results provided in the review was limited to nondescript
statements that the differences may be "related to the different analytical methods employed such
as detection of radiolabeled substrate vs. derivatized analytes" (section 3.3.2.7). Unfortunately,
the authors of the original studies do not really provide technical explanations for the disparities
either. Given such disparate results, the EPA has chosen to use the geometric mean of these
two studies in estimating DCVG formation. This decision process and its impacts on the final
rates for DCVG formation need to be more clearly spelled out in the discussion of these studies.
The discrepancies in estimates of DCVG formation are among the most contentious issues
associated with TCE risk analysis. Given the difficult task of drawing conclusions from such
different results, the conservative approach the EPA has taken is defensible from a public safety
policy perspective. From a strictly scientific perspective however, at a minimum, such large
literature disparities call for a more complete discussion of the strengths and limitations of the
analytical methodologies used than what is described in the review.
Recommendations:
• The issue of quantitative assessment of the metabolic flux of TCE through the GSH pathway
vs. the oxidative metabolism pathway should be considered carefully since uncertainties
exist with regard to the extent of formation of the dichlorovinyl metabolites of TCE between
humans and rodents. EPA may need to provide appropriate reservations to the conclusions
based on the limited data for GSH metabolites.
• The discussion of how each of the in vitro and in vivo data sets were used to estimate
DCVG formation parameters for the PBPK model should be more transparent indicating
strengths and weaknesses in the database.
26
-------
6. Mode of Action
Using the approach outlined in the U.S. EPA Cancer Guidelines (U.S. EPA, 2005a), does
EPA's hazard assessment logically, accurately, clearly, and objectively represent and
synthesize the available scientific evidence to support its conclusions regarding the mode(s)
of action [MOA(s)] of TCE carcinogenicity and non-cancer effects? Specifically, please
address the conclusions that the weight of evidence supports a mutagenic MOA for TCE-
induced kidney tumors; that a MOA for TCE-induced kidney tumors involving cytotoxicity
and compensatory cell proliferation, possibly in combination with a mutagenic MOA, is
inadequately supported by available data; that there is inadequate support for PPARa
agonism and its sequellae being key events in TCE-induced liver carcinogenesis; that there
are inadequate data to specify the key events and MOAs involved in other TCE-induced
cancer and non-cancer effects; and that the available data are inadequate to conclude that
any of the TCE-induced cancer and non-cancer effects in rodents are not relevant to
humans.
Response
6a Hazard Assessment and Mode of Action
The Panel agreed that the IRIS TCE hazard assessment logically, accurately, clearly, and
objectively represented and synthesized the available scientific evidence to support its
conclusions regarding the mode(s) of action [MOA(s)] of TCE carcinogenicity and non-cancer
effects. For each end point, the hazard assessment described the possible MOA and underlying
mechanisms. In general, the assessment provided explanations for inconsistent data or lack of
results. For example, Section 4.8.3.3.2 provided a comprehensive, detailed, and very useful
discussion of potential reasons for inconsistencies in the body of literature on TCE exposure in
utero and heart defects.
The Panel agreed that the MOA for TCE nephrotoxicity involves conversion of TCE to
GSH derived metabolites followed by conversion of the glutathione conjugate (DCVG) to the
cysteine conjugate (DCVC) and activation by p-lyase in the kidney to the ultimate nephrotoxic
species. Thus, the EPA's hazard assessment logically, accurately, clearly, and objectively
represents and synthesizes the available scientific evidence to support the conclusion regarding
the MOA for TCE kidney non-cancer toxicity. However, as discussed in the response to charge
question 3, the Panel noted that uncertainties remain with regards to quantity of metabolites
formed in humans and rodents. The panel concluded that the narrative presentation of the data,
along with the evaluation of the strengths and weaknesses of each study, was appropriate with
supplemental information.
Recommendations:
• The impact of the inconsistencies in data on the quantity of GSH pathway metabolites
formed in humans and rodents should be presented more transparently.
• In the body of the document, MOA information should be systematized and broken down
into key events for each proposed MOA. The EPA may consider using a tabular format to
facilitate the ease of evaluation. Information on supporting/refuting (if any) evidence (with
27
-------
appropriate references indicated), human relevance (if available), and "strength" of each line
of evidence/study should be included.
• EPA should consider tabular summaries by specific metabolites when studies used
metabolite exposure rather than the parent compound.
• Data gaps should be clearly identified to help guide future research.
• Key conclusions supporting/refuting each key event should be presented in bullet form
indicating where in the document a more detailed narrative/tables can be found.
6b MO A for TCE-Induced Kidney Tumors
The Panel agreed that the weight of evidence supported a mutagenic MOA for TCE-
induced kidney tumors. However, the Panel concluded that the weight of evidence did not
exclude the MOA for TCE-induced kidney tumors involving cytotoxicity and compensatory cell
proliferation and including this MOA may more accurately reflect kidney tumor formation than a
mutagenic mechanism alone. Furthermore, the combination of cytotoxicity, proliferation and
DNA damage together may be a much stronger MOA than the individual components.
Recommendations:
• Modify the relevant text to reflect that the available data do, in fact, provide support for
TCE-induced kidney tumors involving cytotoxicity and compensatory cell proliferation,
possibly in combination with a mutagenic MOA, although not to the extent that support for
a mutagenic MOA was provided.
6c Inadequate Support for PPARa agonism and its sequellae being key events in TCE-
induced liver carcinogenesis
The Panel agreed that there was inadequate support for PPARa agonism and its sequellae
being key events in TCE-induced human liver carcinogenesis. The Panel noted that PPARa
agonists do not elicit peroxisomal proliferation in humans, a pathological change which is a
hallmark effect of TCE and other peroxisome proliferators in rodents.
The Panel noted that a number of studies important for consideration of the relevance of
PPARa mode of action to human liver carcinogenesis have been completed recently. These
include, but are not limited to, studies in PPARa-null mice (Ito et al. 2007; Takashima et al.
2008; Eveillard et al. 2009), PPARa humanized transgenic mice (Morimura et al. 2006), and
hepatocyte-specific constitutively-activated PPARa transgenic mice (Yang et al. 2007). The data
from these animal models suggest that activation of PPARa is an important but not limiting
factor for the development of mouse liver tumors and that additional molecular events may be
involved.
The Panel noted the quantitative differences in the affinity of the various isoforms of
PPARs to TCA, DCA and other model peroxisome proliferators are well established. Likewise,
the quantitative differences in affinity between species are also known.
28
-------
Recommendations:
• Graphical or tabular presentation of these data to strengthen the comparative analysis
between metabolites and chemicals.
• Including some of the analyses which compare the receptor transactivation potency and the
carcinogenic potential of TCA, DCA and other model peroxisome proliferators from Guyton
et al (2009) to strengthen the arguments.
6d Inadequate Data to specify Key Events and MO As involved in other TCE-Induced Cancer
and Non-Cancer Effects
The Panel agreed that the data are inadequate to specify the key events and MO As
involved in other TCE-induced cancer (lung, lymphoma) and non-cancer effects (central nervous
system, immune system, respiratory tract toxicity, reproductive effects, developmental effects).
6e Human Relevance of TCE-Induced Cancer and Non-Cancer Effects in Rodents
The Panel agreed that the data are inadequate to conclude that any of the TCE-induced
cancer and non-cancer effects in rodents are not relevant to humans.
Recommendations:
• The impact of potential overestimation of the extent of the GSH pathway in humans in
Section 4.4.7 (Kidney) must be transparent
• The MOA for carcinogenicity should be described as complex rather than unknown in
Section 4.5.7.4. Mode of Action (MOA). With respect to conclusions regarding the liver,
while the complete MOA in animals may not be clear at this time, complex is a more
appropriate descriptor since it is likely that key events from several pathways may operate
leading to acute, sub-chronic and chronic liver toxicity of TCE.
29
-------
7. Susceptible Populations
Does EPA's hazard assessment logically, accurately, clearly, and objectively represent
and synthesize the available scientific evidence to support its conclusions that the
factors that could modulate susceptibility to TCE carcinogenicity and non-cancer
effects include genetics, lifestage, background and co-exposures, and pre-existing
conditions, but that only toxicokinetic variability in adults can be quantified given the
available data?
Response
The Panel agreed that Section 4.10 of the Hazard Assessment provided a good review of
potentially susceptible populations, and that the identified factors (genetics, lifestage,
background, co-exposures and pre-existing conditions) may modulate susceptibility to TCE
carcinogenicity and non-cancer effects. The review included adequate data to support factors
that modulate exposure and pharmacokinetics in both animals and humans, but few data to
demonstrate differing susceptibility to health effects from TCE exposure in either animals or
humans. The Panel agreed with the conclusion that the existing data are inadequate to form a
conclusion about whether the potentially modulating factors do or do not impact risk
estimates for TCE and human health effects. The Panel agreed with the use of standard age-
dependent adjustment factors in the protection of children.
Recommendations:
• The Panel disagreed with the statement that "toxicokinetic variability in adults can be
quantified given the existing data," as the main study characterizing toxicokinetic
variability in adults was small (n<100) and was composed of subjects selected non-
randomly. The Hazard Assessment document should note the limitations of the adult
data for toxicokinetic modeling in terms of uncertainty and possible bias in section
4.10.3, and elsewhere in the document where these data are used for hazard
characterization modeling.
• Section 4.10 of the Hazard Assessment should discuss explicitly the lack of data
demonstrating modulation of health effects from TCE by the identified factors (genetics,
lifestage, background, co-exposures, and pre-existing conditions), and the need for such
data in risk assessment.
• EPA should make specific recommendations for studies that would fill the data gap for
susceptible groups. For example, epidemiologic studies in which TCE exposure is well-
characterized and in which internal comparisons can be made to determine whether there
is effect modification, and animal studies comparing subgroups (e.g., based on genetics,
obesity, multiple solvent exposures).
• Modulation of TCE exposure-related hypersensitivity dermatitis by genetic variation may
be relevant for future study, given results of the study of hypersensitivity dermatitis in
Asian workers reported in Li et al. (2007) and increasing industrial chemical exposures in
China.
30
-------
The wording in Section 4.10 was often not clear about whether it was describing results
for a study that looked at effect modification of the TCE effect or not, as opposed to
direct effects of age, gender, etc. Also, the draft document needs to state explicitly where
effects of TCE within one subgroup were stated, whether the other subgroup was also
examined in the same study.
The Panel recommended that exposure to solvent mixtures should be added as a potential
susceptibility factor (co-exposures) to Section 4.10, since exposure to more than one
chemical to the same target organ likely increases risk.
Section 4.10.2.4.1 (page 4-585) should be more accurately titled 'Obesity', rather than
'Obesity and metabolic syndrome'. As presently written, Section 4.10.2.4.1 gives no
clear message as to how obesity affected the kinetics of TCE, and the section should be
revised to provide clarification.
-------
8. Non-Cancer Dose-Response Assessment
EPA's dose-response assessment includes the development of a chronic inhalation
Reference Concentration (RfC) and chronic oral Reference Dose (RfD) for non-cancer
effects. Please address the following methods and results from EPA's non-cancer dose-
response assessment in terms of the extent to which they are clearly and transparently
described and technically/scientifically adequate to support EPA's draft RfC and RfD:
a. The screening, evaluation, and selection of candidate critical studies and
effects;
b. The points of departure, including those derived from benchmark dose
modeling (e.g., selection of dose-response models, benchmark response
levels);
c. The selected PBPK-based dose metrics for inter-species, intra-species, and
route-to-route extrapolation, including the use of body weight to the 3/4 power
scaling for some dose metrics;
d. The selected uncertainty factors;
e. The equivalent doses and concentrations for sensitive humans developed
from PBPK modeling to replace standard uncertainty factors for inter- and
intra-species toxicokinetics, including selection of the 99th percentile for
overall uncertainty and variability to represent the toxicokinetically-sensitive
individual;
f. The qualitative and quantitative characterization of uncertainty and
variability;
g. The selection of NTP (1988) [toxic nephropathy], NCI (1976) [toxic
nephrosis], Woolhiser et al. (2006) [increased kidney weights], Keil et al.
(2009) [decreased thymus weights and increased anti-dsDNA and anti-ssDNA
antibodies], Peden-Adams et al. (2006 [developmental immunotoxicity], and
Johnson et al. (2003) [fetal heart malformations] as the critical studies and
effects for non-cancer dose-response assessment;
h. The selection of the draft RfC and RfD on the basis of multiple critical effects
for which candidate reference values are in a narrow range at the low end of
the full range of candidate critical effects, rather than on the basis of the
single most sensitive critical effect.
Response
8a The screening, evaluation, and selection of candidate critical studies and effects
The Panel agreed that the screening, evaluation, and selection of candidate critical studies
and effects were generally adequate to support EPA's draft RfC and RfD. The Panel noted that a
very large number of studies were considered and included in the tables, and agreed that it was
appropriate to evaluate all studies showing dose-response for neurological, kidney, liver,
immunologic, respiratory system, reproductive, and developmental effects, and body weight
32
-------
change. The Panel's comments on sub-question (a) related primarily to making the information
presented in the document more clear and transparent to the reader, rather than to the screening,
evaluation, and selection process itself.
The Panel believed that it was important that the reader easily be able to find the details
of the studies included in the Chapter 5 tables.
For instance, four different studies with different durations were cited as "Crofton and
Zhao (1997)" in Table 4-23, and it was not clear which duration was the basis for the cRfD in
Table 5-1. In other cases, it was not stated whether the cRfD or cRfC was based on males or
females when both were included in the study, or which strain was the basis when multiple
strains were used. For example, from Table 5-2 and the text on p. 5-15 to 5-16, it was not clear
which strain, gender, or exposure duration was used for the RfC for increased liver weight based
on Kjellstrand et al. (1983b) (discussed in Chapter 4 and Appendix E). Another example for
which cross-referencing the different sections of the document would be helpful is the
information on the doses in the drinking water study of Keil et al. (2009). In the description of
the study on p. 4-395, the doses were given as drinking water concentrations (ppb), but in Table
5-3, the LOAELs for this study were given in mg/kg/day, and the conversion from ppb in
drinking water to mg/kg/day is found in Appendix E (p. E-34). A final example of where cross-
referencing would be helpful relates to the studies of Carney et al. (2006) and Schwetz et al.
(1975). These studies were listed in Table 5-4 (Reproductive Toxicity) because the key effect,
decreased maternal body weight gain in a developmental study, was considered a "reproductive"
effect. However, these studies were discussed under developmental toxicity in Chapter 4,
making it difficult to locate them while reading the section on reproductive toxicity in Chapter 5.
Finally, it was stated on p. 5-1, point (1) that studies with "quantitative dose-response
data" were considered. Some of the studies which were considered as the basis for RfCs and
RfDs used only one dose of TCE and a control group (for example, Barrett et al., 1992). If a
control group and a single treated group were considered adequate "quantitative dose-response
data," this should be stated.
Recommendations:
• Chapter 5 should include a list of all non-cancer health effects and studies discussed in
Chapter 4, noting those which were considered candidate critical effects and studies.
• Tables 5.1-5.5 should provide cross-references to the table or page in Chapter 4 and/or to the
Appendices (such as Appendix E for hepatic studies) where the listed study was discussed,
and should include more details (e.g. gender, strain, duration) of the studies selected as the
basis for cRfDs and cRfCs when these details were needed to prevent ambiguity.
• Consistent dose units should be used in discussing the same study in different places in the
document.
33
-------
8b The points of departure, including those derived from benchmark dose modeling (e.g.,
selection of dose-response models, benchmark response levels)
The Panel agreed that the derivation of the points of departure (PODs) was generally
technically/scientifically adequate to support EPA's draft RfC and RfD. The Panel noted that the
graphics in Appendix F provided a good presentation of the BMD analyses.
The Panel noted that, although BMD modeling was generally an appropriate approach for
POD determination, the results of BMD modeling were very uncertain with some datasets. For
example the log-logistic BMD analysis for toxic nephropathy in female Marshall rats in the NTP
(1988) study, shown in Figure F-10, may greatly overestimate the risks at low doses. This
modeling involved extrapolation from a high LOAEL at which a high percentage of the animals
were affected.
Recommendation:
• Chapter 5 should include the information on POD derivation from Table F-13 of Appendix
F, including approach, selection criterion and decision points.
8c The selected PBPK-based dose metrics for inter-species, intra-species, and route-to-route
extrapolation, including the use of body weight to the % power scaling for some dose metrics
The Panel agreed that the use of PBPK-based dose metrics for inter-species, intra-species,
and route-to-route extrapolation modeling were, for the most part, technically and scientifically
adequate to support EPA's draft RfC and RfD.
However, it was noted by the Panel that the RfDs and RfCs for kidney endpoints were
highly sensitive to the rate of renal bioactivation of DCVC (ABioactDCVCBW34) in human
versus rodents. Specifically, it was noted that p-cRfDs/RfCs based on this dose-metric were
several hundred-fold lower than RfDs/RfCs for the same endpoints based on applied dose with
standard uncertainty factors, while p-cRfDs/RfCs for endpoints based on other dose metrics were
much closer to RfDs/RfCs based on applied dose and standard uncertainty factors.
In addition to the strong dependence of the p-cRfDs and p-cRfCs on the rate of renal
bioactivation of DCVC, the Panel noted that the uncertainties about the in vitro and in vivo data
used to estimate this dose metric were much greater than for other dose metrics. For example,
there were very large discrepancies in the rates of human glutathione conjugation reported by
Lash et al. (1999a) and Green et al. (1997a).
The Panel understood that the rationale for scaling the dose metric to body weight3 4, in
conjunction with the interspecies extrapolation, is that the PBPK model predicted the dose rate to
the target tissue rather than the internal concentration of TCE. However, this distinction and the
associated rationale would likely not be readily apparent to most readers of the document as
currently written. Confusion might arise because, for other contaminants, PBPK models were
used to estimate serum levels or other metrics of internal concentration, rather than delivered
doses, and in such case, scaling of body weighty4 would not be used.
34
-------
The discussion of "empirical dosimetry" vs. "concentration equivalence dosimetry"as
presented in the draft document would likely not be readily understandable to many readers.
Furthermore, since body weight374 scaling was used for all of the dose metrics discussed in
sections 5.1.3.1.1-5.1.3.1.5, it may not be necessary to include the extensive discussion of the
two dosimetry approaches in each of these sections.
Recommendations:
• The uncertainty about the rate of human glutathione conjugation found in Lash et al. (1999a)
versus Green et al. (1997a) should be highlighted in the current assessment.
• The basis for the renal bioactivation dose metric should be more clearly and transparently
presented and discussed in Chapter 3 and other appropriate sections. If this dose metric was
derived indirectly from data on other metabolic pathways leading to and/or competing with
bioactivation, this should be more clearly discussed.
• The rationale for scaling the dose metric to body weight3 4, in conjunction with the
interspecies extrapolation based on PBPK modeling, should be presented in a clearer and
more transparent way (e.g. on pp. 5-33 - 5-36).
• The discussion of "empirical dosimetry" vs. "concentration equivalence dosimetry" should
be made clearer and more transparent (pp. 5-33 - 5-36).
8d Uncertainty factors
The Panel agreed that, in general, the selection of uncertainty factors was clearly and
transparently described and technically/scientifically adequate to support EPA's draft RfC and
RfD. The uncertainty factors were consistently applied in Tables 5-8 to 5-13.
However it was noted that the uncertainty factors were appropriately applied only if the BMD-
PBPK 99th percentile (HECgg and HEDgg) dose metrics were correctly derived.
The Panel recognized that EPA guidance defines the duration of subchronic rodent
studies as 4 weeks to 90 days, and chronic rodent studies as 90 days to 2 years, and noted that
some of the subchronic studies considered as the basis for risk assessment were of duration as
short as 4 weeks (e.g. Isaacson, 1990). Also, some studies of duration only slightly greater than
90 days (e.g. 18 weeks for Kulig et al., 1987) were classified as chronic, as appropriate under the
EPA definition of chronic as longer than 90 days. However, exposures for 18 weeks may not
always accurately predict effects for lifetime duration, since 18 weeks is only a small percentage
of a two year (104 week) rodent lifespan (less than 18%).
Recommendations:
• The definitions of chronic and subchronic studies should be provided in the document and a
citation given.
• The discussion of the subchronic to chronic uncertainty factor on p. 5-6 should be clarified as
far as durations of studies considered suitable as the basis of a chronic risk assessment.
• The draft document should include discussion of whether studies in the lower end of the
range defined as subchronic (e.g. 4 weeks) are of sufficient duration to be used as the basis
for a chronic (lifetime) risk assessment.
35
-------
Studies only slightly longer than the minimum needed to be considered chronic should be
noted as such, and the use of an uncertainty factor to account for less than lifetime exposure
(of less than the full uncertainty factor of 10) could be considered for studies of such
durations, especially for endpoints thought to progress in incidence or severity with time.
8e The equivalent doses and concentrations for sensitive humans developed from PBPK
modeling to replace standard uncertainty factors for inter- and intra-species toxicokinetics,
including selection of the 99* percentile for overall uncertainty and variability to represent the
toxicokinetically-sensitive individual
The Panel generally agreed that this information is clearly and transparently described
and technically/scientifically adequate to support EPA's draft RfC and RfD. It was noted that
the 99th percentile estimates may be very sensitive to modeling assumptions, such as the choice
of prior distribution and the shape of the distribution for population variability in the
toxicokinetic parameters. The Panel concluded that the approach used, including the selections
of idPODs and the extrapolations from rodent to human followed by consideration of the 99th
percentile human estimates, was acceptable to address the sensitive population. It was also
concluded that the approach used to simulate a large range of exposure doses in order to obtain
the distribution for the relationship between human exposure and internal dose (page 5-68) was
appropriate.
Recommendations:
• The Panel noted variability/uncertainty for the toxicokinetically-sensitive individual could be
quantified in future work by considering distributions in addition to the distribution of the
99th percentile, such as the 95th percentile.
• A quantile regression looking simultaneously at several quantiles could be developed in the
future and presented in future refinements of this assessment.
Additional issue related to sub-questions (c), (d), and (e) discussed by the Panel:
The question arose as to whether the general approach used in the draft document to
develop p-RfDs and p-RfCs was appropriately protective, as opposed to being overly
conservative. Specifically, the Panel noted that the PODs identified through BMD analysis were
based on most sensitive species, strain, and sex, and that the idPODs based on lower bound
estimates of the 1% or 5% response in animals were used as a central dose estimate in humans.
It was also noted that uncertainty factors for interspecies and intra-human pharmacodynamic
variability were applied to the 99th percentile estimates (i.e. the doses for the 1% most
pharmacokinetically sensitive humans) of the internal dose (HECgg and
The Panel endorsed the use of BMD modeling instead of an approach based on an
uncertainty factor for LOAEL-to-NOAEL extrapolation, and the use of PBPK modeling instead
of default uncertainty factors for inter- and intra-species pharmacokinetic differences, when these
approaches were supported by the data. The Panel recognized that these approaches were not
36
-------
intended to introduce greater conservatism, but rather to incorporate data to replace default
assumptions when appropriate.
There was consensus among the Panel members that the general approach described
above was consistent with accepted EPA methodology for RfD/RfC development. It was
specifically noted that the uncertainty factors for interspecies and intra-human pharmacodynamic
variability were intended to account for variability as well as uncertainty, and that some p-
RfDs/p-RfCs based on PBPK modeling were higher than RfDs/RfCs for the same endpoints
based on the default methodology. The Panel recommended that HECso and HEDso values be
included in Tables 5-8 to 5-13 for informational purposes.
Finally, as discussed further under sub-question (h), the Panel concluded that the
consistency of RfDs and RfCs, and that selected endpoints utilized relatively certain dose
metrics, gave confidence in the PBPK approach. Dose metrics used for selected endpoinds and
their levels of certainty are summarized as follows:
Dose Metric
DCVC activation
Total metabolism
Total oxidative metabolism
Applied Dose (dose metric
based on PBPK modeling not
used
Level of Certainty
uncertain
Relatively certain
Relatively certain
Dose Metric Use
Renal endpoints
Decreased thymus weight,
anti-ss and ds DNA antibodies
Cardiac malformations
Developmental
immunotoxicity
8f The qualitative and quantitative characterization of uncertainty and variability;
The Panel generally agreed that the uncertainties related to the RfC and RfD were clearly
and transparently described and technically/scientifically adequate to support EPA's draft RfC
and RfD.
It was noted that in the PBPK model, the uncertainty and variability were quantified with
the posterior distributions, as appropriate for any Bayesian framework, while in the more general
dose-response framework, the uncertainty is characterized with uncertainty factors which
account for the main sources of variability and uncertainty. One Panel member commented that
it was inconsistent to use a Bayesian approach in the PBPK modeling but not in the dose-
response analysis, which uses numeric uncertainty factors, rather than distributions, which
represent variability and uncertainty as a fixed effect.
The Panel recognized that the use of uncertainty factors in the TCE assessment followed
the currently accepted EPA approach.
37
-------
Recommendations:
• The quantitative uncertainty analysis of PBPK model-based dose metrics for LOAEL or
NOAEL based PODs (Section 5.1.4.2) should be revised to clarify the objective of this 2-D
type analysis, as well as the methodology used.
• In future work, EPA could develop an approach using distribution to characterize uncertainty
in a Bayesian framework.
8g The selection ofNTP (1988) [toxic nephropathy], NCI (1976) [toxic nephrosis], Woolhiser
et al. (2006) [increased kidney weights], Keil et al. (2009) [decreased thymus weights and
increased anti-dsDNA and anti-ssDNA antibodies], Peden-Adams et al. (2006 [developmental
immunotoxicity], and Johnson et al. (2003) [fetal heart malformations] as the critical studies
and effects for non-cancer dose-response assessment
The Panel concluded that the choices of Keil et al. (2009) [decreased thymus weights and
increased anti-dsDNA and anti-ssDNA antibodies], Peden-Adams et al. (2006) [developmental
immunotoxicity], and Johnson et al. (2003) [fetal heart malformations] as critical studies and
effects were technically/scientifically adequate to support EPA's draft RfC and RfD. The Panel
noted that questions related to the use of cardiac malformations from Johnson et al. (2003) as a
critical endpoint were adequately addressed in the response to Charge Question 3. It was noted
that BMD modeling for the data from Johnson et al. (2003) was highly sensitive to model choice.
It was also noted that, although a tremendous amount of information was available on liver
toxicity, hepatic effects were not a critical endpoint because they were less sensitive than other
endpoints.
The Panel expressed concerns about the use of NTP (1988) [toxic nephropathy], NCI
(1976) [toxic nephrosis], and Woolhiser et al. (2006) [increased kidney weights] as critical
studies and effects. For all three of these studies, uncertainties exist for the PBPK modeling
based on renal bioactivation of DCVC, as discussed in sub-question (c) above.
Additional issues related to choice of toxic nephropathy in female Marshall rats from
NTP (1988) as a critical effect and study include excessive mortality due to dosing errors and
possibly other causes, and a high level of uncertainty in the extrapolation to the BMD due to the
use of very high doses and a high incidence (>60%) of toxic nephropathy at both dose levels
used. It was also noted that the incidence of this effect was lower in this study in other strains of
rats and in male Marshall rats, suggesting that the sensitivity for this effect was highest in the
female Marshall rats.
It should be noted that the uncertainties noted by the Panel about the quantitative risk
assessment based on toxic nephropathy in NTP (1988) did not indicate that there was uncertainty
that TCE caused renal toxicity in this study. The Panel noted that renal cytomegaly, which was
not selected as a critical effect, occurred at a very high frequency in both sexes of all four strains
used in this study, with 90-100% incidence in almost all dosed groups, and toxic nephropathy
also occurred in all treated groups. In contrast, neither renal cytomegaly nor toxic nephropathy
was seen in any of 396 control animals in study, which included groups of 50 males and females
of the four different rat strains.
38
-------
Additional issues related to the choice of toxic nephrosis in mice from NCI (1976) were
that BMD analysis was not supported because the effect occurred in nearly 100% of animals in
both dose groups, and that a high level of uncertainty was associated with extrapolation from the
LOAEL at which nearly 100% animals were affected. It was noted by the Panel that toxic
nephrosis did not occur in any control animals of either sex in this study.
Thus, although the numerical values for the RfD and RfC based on the renal endpoints
were highly uncertain, TCE could clearly cause renal toxicity in both sexes of the four strains of
rats tested, as well as in both sexes of mice, when administered in sufficient doses.
8h The selection of the draft RfC and RfD on the basis of multiple critical effects for which
candidate reference values are in a narrow range at the low end of the full range of candidate
critical effects, rather than on the basis of the single most sensitive critical effect.
The Panel supported the selection of a draft RfC and a draft RfD based on multiple
candidate reference values in a narrow range which was at the low end of the full range of
candidate reference values developed, rather than basing these values on the single most
sensitive critical endpoint. This approach was supported by the Panel because it was a very
robust approach that increases confidence in the final RfC and RfD.
Reference Concentration
As noted in the draft assessment, the proposed RfC, 0.001 ppm (5 ug/m3), was within a
factor of 3 of the p-cRfCs for the six critical endpoints selected. The Panel agreed with the use of
PBPK modeling for route-to-route extrapolation for the five p-cRfCs which were based on oral
studies.
EPA stated in the draft document (p. 5-83) that there was high confidence in the three p-
cRfCs based on renal endpoints [increased kidney weight (Woolhiser et al., 2006), toxic
nephrosis (NCI, 1976), and toxic nephropathy, (NTP,1988)] because of the clearly adverse
nature of the effects, the fact that two of them were based on chronic studies, and high
confidence in its estimate of the dose metric which was clearly related to toxicity, while there
was somewhat less confidence in the three p-cRfCs based on other endpoints [decreased thymus
weight and anti-DNA antibodies (Keil et al., 2009) and cardiac malformation (Johnson et al.,
2003)]. As stated in the response to (g), TCE can clearly cause significant renal toxicity when
administered in sufficient doses. Thus, the Panel agreed that kidney toxicity was indisputably a
key effect of TCE from a hazard identification perspective. However, as discussed above, the
Panel concluded that the three p-cRfCs for renal endpoints were based on an uncertain dose
metric, especially in regard to the relative rate of formation of the toxic metabolite in humans
and animals. Although there was somewhat less confidence in the immune and cardiac
malformation endpoints from a hazard identification perspective, for reasons discussed
extensively in other sections of this response, there was sufficient confidence in them to consider
them critical endpoints to support the RfC. While the confidence in these three endpoints was
less than for the kidney endpoints as far as hazard identification, the three p-cRfCs for these
endpoints were based on relatively certain dose metrics.
39
-------
Although there was much greater pharmacokinetic uncertainty for the RfCs based on the
three studies with renal endpoints [(Woolhiser et al., NCI (1976), and NTP (1988)], they
provided additional support for the RfC.
The Panel noted that the same final RfC, 0.001 ppm, was supported by the p-cRfCs based
on both the three principal studies (0.0003 ppm, 0.0004 ppm, and 0.003 ppm) and the supporting
(kidney) studies (0.0006 ppm, 0.001 ppm, and 0.002 ppm), and concluded that the use of p-
cRfCs for multiple critical effects to derive the final recommended RfC reduced uncertainty and
better characterizes variability. It was noted that, in general, this approach may create more
work for the risk assessors and the users of the risk assessment than use of the single most
sensitive endpoint. However, it was recognized that, even if the RfC were to be based on the
single most sensitive endpoint, it would be necessary to develop p-cRfCs for multiple endpoints
in order to rigorously determine which study and endpoint provides the most sensitive RfC. It
was also noted that a single RfC value was provided to users of the risk assessment.
Reference Dose
As discussed in the draft document, the proposed RfD, 0.0004 mg/kg/day, was within 25% of
the p-cRfDs for the four critical endpoints selected (toxic nephropathy (NTP, 1988), decreased
thymus weight [(Keil et al, 2009), developmental immunotoxicity (Peden-Adams et al., 2006),
and cardiac malformations (Johnson et al., 2003)]. All four p-cRfDs were based on oral
exposure, and three of them were based on drinking water exposure, a route relevant to
environmental exposures. EPA stated in the draft document (p. 5-83) that there was high
confidence in the p-cRfD based on a renal endpoint (toxic nephropathy, (NTP, 1988)) because of
the clearly adverse nature of the effects in a chronic study and the high confidence in the
estimate of the dose metric which was clearly related to toxicity, while there was somewhat less
confidence in the three p-cRfCs based on other endpoints [decreased thymus weight (Keil et al.,
2009), developmental immunotoxicity (Peden-Adams et al., 2006), and cardiac malformations
(Johnson et al., 2003)]. As stated in the response to (g), TCE could clearly cause significant renal
toxicity when administered in sufficient doses. Thus, as in the RfC discussion above, the Panel
agreed that kidney toxicity was indisputably a key effect of TCE from a hazard identification
perspective. However, as discussed above, the Panel concluded that the p-cRfD for the kidney
endpoint was based on an uncertain dose metric in regard to the relative rate of formation of the
toxic metabolite in humans and animals. Although there was somewhat less confidence in the
immune and cardiac malformation endpoints from a hazard identification perspective, for
reasons discussed extensively in other sections of this response, there was sufficient confidence
in them to consider them critical endpoints to support the RfC. While the confidence in these
three endpoints was less than for the kidney endpoints as far as hazard identification, the three p-
cRfCs for these endpoints were based on relatively certain dose metrics.
Although there was greater pharmacokinetic uncertainty for the p-cRfD based on the renal
endpoint (NTP, 1988), it provided additional support for the final RfD.
The Panel noted that the same final RfD, 0.0004 mg/kg/day was supported by the p-cRfCs
based on both the three principal studies (0.0004 mg/kg/day, 0.0005 mg/kg/day, and 0.0005
mg/kg/day) and the supporting (kidney) study (0.0003 mg/kg/day), and concluded that the use of
40
-------
p-cRfDs for multiple critical effects to derive the final recommended RfD reduced uncertainty
and better characterizes variability. As discussed above for the RfC, it was noted that, in general,
this approach may create more work for the risk assessors and the users of the risk assessment
than use of the single most sensitive endpoint. However, it was recognized that, even if the RfD
were to be based on the single most sensitive endpoint, it would be necessary to develop p-cRfCs
for multiple endpoints in order to rigorously determine which study and endpoint would give the
most sensitive RfD. It was also noted that a single RfD value was provided to users of the risk
assessment.
Recommendations:
• The two endpoints for immune effects from Keil et al. (2009) and the cardiac malformations
from Johnson et al. (2003) should be considered the principal studies supporting the RfC.
• The endpoints for immune effects from Keil et al. (2009) and Peden-Adams et al. (2009) and
the cardiac malformations from Johnson et al. (2003) should be considered as the principal
studies supporting the RfD.
41
-------
9. Cancer Dose-Response Assessment
In accordance with the approach outlined in the U.S. EPA Cancer Guidelines and
Supplemental Guidance (U.S. EPA, 2005a; U.S. EPA, 2005b), EPA's dose-response
assessment includes the development of an inhalation unit risk and oral unit risk for the
carcinogenic potency of TCE. Please address the following methods, results, and
conclusions from EPA's cancer dose-response assessment in terms of the extent to
which they are clearly and transparently described and technically/scientifically
adequate to support EPA's draft inhalation and oral unit risks:
a. the estimation of unit risks for renal cell carcinoma from the Charbotel et
al. (2006) case-control study;
b. the adjustments of renal cell carcinoma unit risks to account for the
added risk of other cancers using the meta-analysis results and
Raaschou-Nielsen et al. (2003);
c. the estimation of human unit risks from rodent bioassays;
d. in accordance with the approach in the U.S. EPA Cancer Guidelines (U.S.
EPA, 2005a) and the conclusions as to MOA (above), the use of linear
extrapolation from the point of departure (POD) for the cancer dose-
response assessment of TCE;
e. the applications of PBPK modeling, including the selection of dose
metrics and the use of PBPK model predictions for inter-species, intra-
species, and route-to-route extrapolation based on internal dose, and
their preference over default approaches based on applied dose;
f. the qualitative and quantitative characterization of uncertainty and
variability;
g. the conclusion that the unit risk estimates for TCE based on human
epidemiologic data and those based on rodent bioassay data are
consistent overall; and,
h. the preference for the unit risk estimates for TCE based on human
epidemiologic data over those based on rodent bioassay data
9a Estimation of Unit Risks for Renal Cell Carcinoma
The Panel agreed that the analysis of the Charbotel et al. (2006) data was well
described and scientifically appropriate and that the study should be used to estimate unit
risks. The Panel did, however, agree that some more discussion was needed on cutting oils
and whether or not it was necessary to adjust for exposure to cutting oils when computing an
odds ratio or relative risk relating TCE exposure to kidney cancer. As noted in the document
(p. 5-136), Charbotel et al. (2006) found a marginally significant relationship between cutting
and petroleum oils and RCC (p-value < 0.1) though the relationship disappeared after
adjustment for other variables. Given that there was some suggestion of a relationship, the
Panel recommended that the EPA take a closer look at the literature to determine if there
were other studies which suggested that exposure to cutting oils was a risk factor for kidney
cancer.
42
-------
Recommendations:
• The Panel believed that the EPA should provide a more detailed discussion of the
limitations of their analysis. In particular, the model described on p. 5-131 made some
very restrictive assumptions: linear dose-response and exposure was measured without
error. In addition, the life table analysis applied the same estimated RR to each age
interval; another restrictive assumption. While the Panel understood that these
assumptions were necessary due to limited data, there was inadequate discussion of how
violations of these assumptions may affect the results.
• Finally, in constructing the life table, the EPA used background kidney cancer rates in
the US though the Charbotel et al. (2006) data were based on a French cohort. Hence, a
comparison of background cancer rates in France and the U.S. would be helpful in
supporting their conclusions.
9b Adjustment of Renal Cell Carcinoma Unit Risks
The Panel agreed that the analysis and presentation should be accepted in its current
form.
9c Estimation of Human Unit Risks from Rodent Bioassays
EPA also calculated cancer unit risk estimates based on chronic bioassays on rats and
mice. Five inhalation bioassays and 7 oral bioassays were selected for dose-response
analyses. Dose-response modeling using the linearized multistage model was performed
using applied doses as well as PBPK model-based internal doses. Bioassays for which time-
to-tumor data were available were analyzed using a Multistage Weibull model. A cancer
potency estimate for different tumor types combined were derived from bioassays in which
there was more than one type of tumor response in the same sex and species. Unit risk
estimates based on PBPK model-estimated internal doses were then extrapolated to human
population unit risk estimates using the human PBPK model. Based on these results, the
most sensitive bioassay (i.e. the one with the greatest unit risk estimate) was considered as a
candidate unit risk estimates for TCE.
Recommendations:
• The Panel agreed that the analysis and results were appropriate but recommended that
the EPA providemore details about their implementation and potential biases. For
instance, in bioassays in which mortality occurred before time to first tumor, the authors
simply adjusted their denominators to equal the number alive at time to first tumor. This
approach assumed that drop-out prior to time to first tumor was unrelated to future risk
of a tumor which could result in biased estimates.
• In addition, more information was needed on the priors used in their Bayesian analysis
of combined risk across tumor types.
9d Use of Linear Extrapolation for Cancer Dose-Response Assessment
The Panel agreed that the analysis was consistent with current cancer guidelines.
There was sufficient evidence to conclude that a mutagenic MOA was operative for TCE-
43
-------
induced kidney tumors, so linear extrapolation was used to derive unit risk estimates for this
site. For all other tumor types, linear extrapolation was used as the default approach, in
accordance with EPA's cancer guidelines. Hence, the Panel recommended accepting the
analysis and presentation of the results in its present form.
9e Application of PBPK Modeling
The Panel agreed that the PBPK models provided valuable information to the risk
assessment and agreed that the internal dose should be preferred over applied dose as it was
the only way one could, at the mechanistic level, combine information about
pharmacokinetics and pharmacodynamics.
9f Qualitative and Quantitative Characterization of Uncertainty and Variability
The Panel agreed that consideration of uncertainty and variability was adequate. The
Panel believed that the characterization of uncertainty and variability in the PBPK models
was exceptionally strong. Use of AIC to select the best fit model was an adequate way to
address model uncertainty. However, the authors' use of a 0.05 significance level for
goodness of fit tests was inappropriate; typically, larger type-I error rates are used in such
tests (e.g., values between 0.1 and 0.2) since one usually does not want to reject the null
hypothesis that the model fits the data.
9g Conclusion on the Consistency of Unit Risk Estimates Based on Human
Epidemiologic Data and Rodent Bioassay Data
The Panel agreed with this conclusion. For inhalation, the most sensitive rodent
bioassay responses based on the preferred dose metrics ranged from 2.6 x 10"3 per ppm to 8.3
x 10"2 per ppm across the sex/species combinations. For oral exposure, the most sensitive
bioassay responses based on the preferred dose metrics ranged from 2.3 x 10"3 per mg/kg/d to
2.5 x 10"1 per mg/kg/d across the sex/species combination. For both routes of exposure, the
most sensitive sex/species response was male rat kidney cancer based on the preferred dose
metric. When the human epidemiologic data were considered, a cancer inhalation unit risk
estimate of 2.2 x 10"2 per ppm and oral unit risk estimate of 5 x 10"2 per mg/kg/d were
obtained, which are both within the ranges reported in the aforementioned animal studies.
9h Preference for the Unit Risk Estimates based on Human Epidemiologic Data
The Panel agreed that human data, when available, should be preferred over rodent
data when estimating unit risk, since within-species uncertainty was easier to address than
between-species uncertainty.
44
-------
10. Age-Dependent Adjustment Factors
Based on the conclusions that the weight of evidence supports a mutagenic MOA for
TCE-induced kidney cancer and that the MOAs for TCE-induced liver cancer and
lymphomas are not known, the Age-Dependent Adjustment Factors (ADAFs) are only
applied to the kidney cancer component of the unit risk estimates. Please address the
extent to which the recommended approach to applying the ADAFs in this situation is
clearly, transparently, and accurately described.
Response
The Panel concluded that EPA has done an excellent job of describing and presenting
the ADAF computations for both oral and inhalation situations. Application of ADAFs in
the TCE analysis consistently followed recommendations in U.S. EPA Cancer Guidelines
(U.S. EPA, 2005a) and Supplemental Guidance (U.S. EPA, 2005b). All of the steps were
clearly presented for inhalation exposure. However, the discussion for the oral exposure
route was shortened and referred back to the inhalation section, making understanding of the
example less easy to follow.
EPA supplemental guidance recommends adjustment for children based on the
presumption that children <16 years of age are intrinsically more susceptible than adults to
mutagenic carcinogens because of biochemical and physiological factors related to the
development of many organs and tissues during this time period; the rationale for the
application of an ADAF is not based on the assumption that children have greater exposure
on a per body weight basis than adults.
The Panel recognized that EPA wished to maximize utility in its IRIS database for TCE and
other chemicals for which ADAFs were applied by providing slope factors and unit risk
factors that allow users to compute risks for situation-specific drinking water intake values
and for exposures to different age groups. Drinking water concentrations for specified
lifetime cancer risk levels (10~4, 10~5, 10~6) are routinely included in IRIS assessments in
which ADAFs are not applied; this information is very helpful to public health professionals
who use the IRIS database to evaluate situations of water contamination. For IRIS
assessments in which ADAFs are applied, as in TCE, it would be useful to users to include
this information, using representative drinking water intakes for various age groups. Other
drinking water estimates may be used if determined to be more applicable.
The Panel was somewhat concerned that the use of ADAFs was in conflict with the
assumptions that underlie the life-table analysis described in Section 5.2.2.1.2 and Appendix
H. As indicated on p. 5-131, lines 25-28, the life-table method used to calculate lifetime
extra risks from the Charbotel et al. (2006) study assumed that relative risk (RR) was
independent of age; as seen in Table H-l, the same estimate of RR was used in each age
interval of the life-table to compute the exposed RCC hazard rate (column L). However,
ADAFs were applied under the assumption that children were more susceptible to the
mutagenic effects which implied that RRs were age-dependent. The Panel recommended
that EPA clarify whether this conflict in assumptions truly exists and if so, what impact it
45
-------
might have on risk estimation and how it may be resolved in the future. For example, it
might make more sense to apply ADAFs during the life-table analysis instead of at the end of
the analysis, following estimation of the unit risk.
Recommendations:
• The Panel recommended that the statement on page 5-151, lines 14-18, be expanded to
better explain why age-dependent adjustment factors were used for <16 years of age, but
not for the elderly, and why EPA did not directly produce age dependent unit risks per
mg/kg/d.
• Include all details presented for the inhalation sample calculations as was done for the
oral exposure sample calculations.
• IRIS assessments in which ADAFs are applied, such as TCE, should include estimated
drinking water concentrations for specified lifetime cancer risk levels (10~4, 10"5, 10"6),
using representative drinking water intakes for various age groups, while noting that
other drinking water estimates may be used if preferred.
• Include in the documentation a discussion of the perceived conflict between the use of
ADAFs and the assumptions underlying the life table analysis of the Charbotel et al.
(2006) data.
46
-------
11. Additional key studies
Please identify any additional studies that would make a significant impact on the
conclusions of the Toxicological Review and should therefore be considered in the
assessment of the noncancer and cancer health effects of TCE.
Response
The Panel has identified additional studies to be considered in the assessment:
lla Fetal Cardiac Effects
Some recent publications confirm and reinforce the results obtained in the Johnson et al.
(2003) study, so maybe they could be cited to make a stronger argument. They are listed as
follows:
Caldwell, PT; Thorne, PA; Johnson, PD et al. (2008) Trichloroethylene disrupts cardiac gene
expression and calcium homeostasis in rat myocytes. Toxicol Sci 104: 135-143.
Gyorke S, Terentyev D. (2008) Modulation of ryanodine receptor by luminal calcium and
accessory proteins in health and cardiac disease. Cardiovasc Res. 77(2):245-55. Epub 2007
Oct 15. Review. PubMed PMID: 18006456.
Lehnart SE, Mongillo M, Bellinger A,et al.(2008) Leaky Ca2+ release channel/ryanodine
receptor 2 causes seizures and sudden cardiac death in mice. J Clin Invest. 118(6):2230-45.
PubMed PMID: 18483626; PubMed Central PMCID: PMC2381750.
Lebeche D, Davidoff AJ, Hajjar RJ. (2008) Interplay between impaired calcium
regulation and insulin signaling abnormalities in diabetic cardiomyopathy. Nat
Clin Pract Cardiovasc Med. 5(11):715-24. Epub 2008 Sep 23. Review.
PubMed PMID: 18813212.
Pace, BM; Lawrence, DA; Behr, MJ; etal. (2005} Neonatal lead exposure changes quality
of sperm and number of macrophages in testes of BALB/c mice. Toxicology 210: 247-
256.
Yano M, Yamamoto T, Kobayashi S. et al. (2008) Defective Ca2+ cycling
as a key pathogenic mechanism of heart failure. Circ J. 72 Suppl A:A22-30.
Epub Sep 4. Review. PubMed PMID: 18772523.
lib Kidney Effects
Jacob, S; Hery, M ; Protois, JC ; et al. (2007) New insight into solvent-related end-stage renal
disease : occupations, products and types of solvents at risk. Occup Environ Med 64: 843-
848.
47
-------
12. Research Needs
Please discuss research likely to substantially increase confidence in the database for future
assessments of TCE.
Response
12a PBPK Model
The Panel concluded the analysis presented in the TCE Review Document defined how
EPA expects to use PBPK models to integrate what is known about animal and human biology
with TCE mode of action information and available animal and human study data to improve the
transparency and accuracy of chemical risk assessments. This is a substantial piece of research
and the EPA is to be applauded for this effort. The Panel discussed additional research, which
could further improve the TCE risk assessment as well as influence the broader use of PBPD
models in risk assessment.
The current model does not account for the temporal variability of the inputs and outputs
in humans. Future development of the trichloroethylene PBPK model requires accommodation
in the model for inter-individual temporal variability in the population. This is particularly
important for modeling both sub-chronic and chronic exposures. If anything, the model should
be most accurate in modeling the effects of human exposure over an extended period. Support
for adding an inter-individual temporal component to the model can be found in a number of
places in the report. For example on page 3-108 (lines 14-16) the text reads: "However, data
from Chiu et al. (2007) indicated substantial interoccasion variability, as the same individual
exposed to the same concentration on different occasions sometimes had substantial differences
in urinary excretion." In this paper Chiu et al. (2007), found that there was variability in urinary
excretion from the same individual exposed to the same concentration on different occasions.
Also, Fisher et al. (1998) (see Table 3-45, page 3-111) documents an occasion in which a female
was exposed to both 50 and 100 ppm. Assuming the same subject-specific estimates across the
two occasions at different doses resulted in over-prediction at the higher exposure.
To substantially improve the PBPK model for trichloroethylene, EPA should perform a
global sensitivity analysis. A formal Bayesian sensitivity analysis is one approach available, but
even a more traditional approach to model sensitivity would provide useful information. In
addition, the impact of changing priors and/or incorporating correlations among parameters
should be examined. Because key dose metrics include upper tails from the predicted posterior
distribution, future work should evaluate the sensitivity of the predictions to distributional
assumptions for the random effects, for example by replacing uniform priors with normal or
lognormal priors or by modifying the bounds on the priors. In future studies, the EPA should
perform at least a limited analysis of sensitivity of results to model form (especially sensitivity to
different assumed GSH pathways).
However, the hierarchical approach formulated in this report also made important
assumptions about the relationship between the PBPK model parameters across the different
48
-------
species. These assumptions should be used consistently throughout the model development and
not just in the case where there is limited prior information about a particular species.
Recommendations:
• Continue to look for data to support further refinement of priors, especially improving non-
informative priors to informative priors and wide priors to narrower priors.
• Develop more efficient sophisticated model algorithms/environments to improve the
simulation and reduce run time.
• Incorporate inter-individual temporal variability in future enhancements of the PBPK model
for TCE.
• Perform a sensitivity analysis that ranges from the traditional assessment of the impact of
parameter changes on final model predictions to an examination of the effect of changing
prior distributions.
12b Derivation of RfD and RfC
Recommendations:
• The uncertainty about the rate of human glutathione conjugation found in Lash et al. (1999a)
versus Green et al. (1997a) should be highlighted in the current assessment and addressed by
sensitivity analysis in future refinements of this assessment.
• The variability/uncertainty for the toxicokinetically-sensitive individual could be quantified
in future work by considering distributions in addition to the distribution of the 99th
percentile, such as the 95th percentile. A quantile regression looking simultaneously at
several quantiles could be developed in the future and presented in future refinements of this
assessment.
• In future work, EPA could develop an approach using distribution to characterize uncertainty
in a Bayesian framework.
12c Susceptibility Factors
Recommendations:
• There is a need for data examining potential modulation of health effects of TCE by t factors
such as genetics, lifestage, background, co-exposures, and pre-existing conditions).
• Modulation of TCE exposure-related hypersensitivity dermatitis by genetic variation may be
relevant for future study, given results of the study of hypersensitivity dermatitis in Asian
workers reported in Li et al. (2007) and increasing industrial chemical exposures in China.
49
-------
12d Male Reproductive System
Recommendations:
• For in utero exposure studies in rodents using lower doses of TCE and metabolites, where
effects (carcinogenic and non-carcinogenic) can be observed trans-generationally, attention
should be directed to epigenetic changes as possible MOA for TCE-mediated effects on the
reproductive systems.
50
-------
REFERENCES
Barrett, L; Torch, S; Leray, CL; et al. (1992) Morphometric and Biochemical studies in
trigeminal nerves of rat after trichloroethylene or dichloroacetylene oral administration.
Neurotoxicology 13:601-614.
Blechzcz, BRA; Gores, GJ (2008) Cholangiocarcinoma. Clin Liver Disease 12 :131-150.
Boyer, As; Finch, WT; Runyan, RB (2000) Trichloroethylene inhibits development of embryonic
heart valve precursors in vitro. Toxicol Sci 53: 109-117.
Carney, EW; Thorsrud, BA; Dugard, PH; et al. (2006) Developmental toxicity studies in Crl:Cd
(SD) rats following inhalation exposure to trichloroethylene and perchloroethylene. Birth
Defects Research (Part 8)77:405-412.
Charbotel, B; Fevotte, J; Hours, M; et al. (2006) Case-control study on renal cell cancer and
occupational exposure to trichloroethylene. Part II. Epidemiological aspects. Ann Occup Hyg
50(8): 777-787.
Chiu, WA; Bois, FY (2007) An approximate method for population toxicokinetic analysis with
aggregated data. J. Agric Biol Environ Stat 12: 346-363.
Chiu, WA; Okino, MS; Evans, MV (2009) Characterizing uncertainty and population variability
in the toxicokinetics of trichloroethylene and metabolites in mice, rats, and humans using an
updated database, physiologically based pharmacokinetic (PBPK) model, and Bayesian
approach. Toxicol Appl Pharmacol. Doi:10.1016/n.taap.2009.07.032.
Crofton, KM; Zhao, X (1997) The ototoxicity of trichloroethylene: extrapolation and relevance
of high-concentration, short-duration animal exposure data. Fundam Appl Toxicol 38(1): 101-
106.
El-Serag, HB (2007) Epidemiology of hepatocellular carcinoma in USA. Hepatology Res
37(Suppl2):S88-S94.
Evans, MV; Chiu, WA; Okino, MS; et al. (2009) Development of an updated PBPK model for
trichloroethylene and metabolites in mice, and its application to discern the role of oxidative
metabolism in TCE-induced hepatomegaly. Toxicol Appl Pharmacol 236: 329-340.
Eveillard, A; Lasserre, F; de Tayrac, M; et al. (2009) Identification of potential mechanisms of
toxicity after di-(2-ethylhexyl)-phthalate (DEHP) adult exposure in the liver using a systems
biology approach. Toxicol Appl Pharmacol 236(3): 282-292.
Fisher, JW; Mahle, D; Abbas, R (1998) A human physiologically based pharmacokinetic model
for trichloroethylene and its metabolites, trichloroacetic acid and free trichloroethanol. Toxicol
Appl Pharmacol 152(2): 339-359.
51
-------
Green, T; Dow, J; Ellis, MK; et al. (1997a) The role of glutathione conjugaton in the
development of kidney tumours in rats exposed to trichloroethylene. Chem Biol Interact
105(2):99-117.
Greenland, S (2004) Model-based estimation of relative risks and other epidemiologic measures
in studies of common outcomes and in case-control studies. Am J Epidemiology 160: 301-305.
Guyton, KZ; Chiu, WA; Bateson, TF; et al. (2009) A reexamination of the PPAR-a activation
mode of action as a basis for assessing human cancer risks of environmental contaminants.
Environ Health Perspect 117(11): 1664-1672.
Hack, CE; Chiu, WA; Zhao, QJ; et al. (2006) Bayesian population analysis of a harmonized
physiologically based pharmacokinetic model of trichloroethylene and its metabolites. Regul.
Toxicol Pharmacol 46: 63-83.
Isaacson, LG; Spohler, SA; Taylor, DH (1990) Trichloroethylene affects learning and decreases
myelin in the rat hippocampus. Neurotoxicol Teratol 12(4):375-381.
Ito, Y; Yamanoshita, O; Asaeda, N; et al. (2007) Di(2-ethylhexyl)phthalate induced hepatic
tumorigenesis through a peroxisome proliferator-activated receptor a-independent pathway. J
Occup Health 49:172-182.
Jacob, S; Hery, M ; Protois, JC ; et al. (2007) New insight into solvent-related end-stage renal
disease : occupations, products and types of solvents at risk. Occup Environ Med 64: 843-848.
Johnson, PD; Goldberg, SJ; Mays, MZ; et al. (2003) Threshold of trichloroethylene
contamination in maternal drinking waters affecting fetal heart development in the rat. Environ
Health Perspect 111(3): 289-292.
Keil, DE; Peden-Adams, MM; Wallace, S; et al. (2009) Assessment of trichloroethylene (TCE)
exposure in murine strains genetically-prone and non-prone to develop autoimmune disease. J.
Environ Sci Health A Environ Sci Eng Toxic 44:443-453.
Kjellstrand, P; Holmquist, B; Mandahl, N; et al. (1983b) Effects of continuous trichloroethylene
inhalation on different strains of mice. Acta Pharmacol Toxicol (Copenh) 53(5):369-374.
Kim, S; Kim, D; Pollack, GM; et al. (2009) Pharmacokinetic analysis of trichloroethylene
metabolism in male B6C3F1 mice: Formation and disposition of trichloroacetic acid, S-(l,2-
dichlorovinyl)glutathione and S-(l,2-dichlorovinyl)-L-cysteine. Toxicol Appl Pharmacol
238(l):90-99.
Kulig, BM. (1987) The effects of chronic trichloroethylene exposure on neurobehavioral
functioning in the rat. Neurotoxicol Teratol 9(2): 171-178.
Lash, LH; Lipscomb, JC; Putt, DA; et al. (1999a) Gluthione conjugation of trichloroethylene in
human liver and kidney: kinetics and individual variation. Drug Metab Dispos 27(3):351-359.
52
-------
Li, H; Dai, Y; Huang, H; et al. (2007) HLA-B*1301 as a biomarker for genetic susceptibility to
hypersensitivity dermatitis induced by trichloroethylene among workers in China. Environ
Health Perspect 115(11): 1553-1556.
Mishima, N; Hoffman, S; Hill, EG; et al. (2006) Chick embryos exposed to trichloroethylene in
an ex vivo culture model show selective defects in early endocardial cushion tissue formation.
Birth Defects Res A Clin Mol Terratol 76:517-527.
NCI (National Cancer Institute). (1976) Carcinogenesis bioassay of trichloroethylene. Division
of Cancer Cause and Prevention, National Cancer Institute, U.S. Department of Health,
Education, and Welfare, DHEW Publication No. (Nffl) 76-802, Technical Report Series No.
2,218 pages, NCI-CG-TR-2; NTIS PB254122. Available online atNeghab, M; Qu, S; Bai, CL;
et al. (1997) Raised concentration of serum bile acids following occupational exposure to
halogenated solvents, l,l,2-trichloro-l,2,3-trifluoroethane and trichloroethylene. Int Arch Occup
Environ Health 70: 187-194.
Niklasson, M: Tham, R; Larsby, B; et al. (1993) Effects of toluene, styrene, trichloroethylene
and trichloroethane on the vestibule and opto-oculo motor system in rats. Neurotoxicol Teratol
15(5): 327-334.
Morimura, K; Cheung, C; Ward, JM; et al. (2006) Differential susceptibility of mice humanized
for peroxisome proliferator-activated receptor {alpha} to Wy-14,643 induced liver
tumorigenesis. Carcinogenesis 27:1074-1080.
NRC (National Research Council) (2006) Key scientific issues in the health risk assessment of
trichloroethylene. Washington: National Academy Press.
NTP (National Toxicology Program). (1988) Toxicology and Carcinogenesis studies of
trichloroethylene (CAS no. 79-01-6) in four strains of rats (ACI, August, Marshall, Osborne-
Mendel) (gavage studies). Public Health Service, U.S. Department of Health and Human
Services; NTP TR-273; Nffl Publication No. 88-2529. Available from the Information Service,
Springfield, VA; PB88-218896. Available online at
http://ntp/niehs.nih.gov/htdocs/LT rpts7tr273.pdf
Pace, BM; Lawrence, DA; Behr, MJ; etal. (2005} Neonatal lead exposure changes quality of
sperm and number of macrophages in testes of BALB/c mice. Toxicology 210: 247-256.
Peden-Adams, MM; Eudaly, JG; Lee, AM et al. (2008) Developmental effects of
trichloroethylene (TCE) on immunity: studies in MRL (+/+) mice. J. Environ Sci Health A Tox
Hazard Subst Environ Eng A43 (12): 1402-1409.
Raaschou-Nielsen, O; Hansen, J; McLaughlin, JK; et al. (2003) Cancer risk among workers at
Danish companies using trichloroethylene: a cohort study. Am J Epidemiol 158(12): 1182-1192.
53
-------
Ramdhan, DH; Kamijima, M; Yamada, N; et al. (2008) Molecular mechanisms of
trichloroethylene-induced hepatoxicity mediated by CYP2E1. Toxicol Appl Pharmacol 231:300-
307.
Schwetz, BA; Leong, BKJ; Gehring, PJ (1975) The effect of maternally inhaled
trichloroethylene, perchloroethylene, methyl chloroform, and methylene chloride on embryonal
and fetal development in mice and rats. Toxicol Appl Pharmacol 32: 84-96.
Sweeney, LM; Kirman, CR; Gargas, ML; et al. (2009) Contribution of trichloroacetic acid to
liver tumors observed in perchloroethylene (perc)-exposed mice. Toxicology 260(1-3):77-83.
Takashima, K; Ito, Y; Gonzalez, FJ; et al. (2008) Different mechanisms of DEHP-induced
hepatocellular adenoma tumorigenesis in wild-type and PPAR alpha-null mice. J Occup Health
50(2): 169-180.
Tham, R; Bunnfors, I; Eriksson, B; et al. (1984) Vestibule ocular disturbances in rats exposed to
organic solvents. ActaPharmacol Toxicol 54(l):58-63.
U.S. EPA (Environmental Protection Agency). (2005a) Guidelines for carcinogen risk
assessment. Risk Assessment Forum, Washington, DC; EPA/630/P-03/001B. Available online
at http://www.epa.gov/iris/backgr-d.htm.ES
U.S. EPA (Environmental Protection Agency). (2005b) Supplemental guidance for assessing
susceptibility from early-life exposure to carcinogens. Risk Assessment Forum, Washington,
DC; EPA/630/R-03/003F. Available online at http:www.epa.gov/iris/backgr-d.htm
Woolhiser, MR; Krieger, SM; Thomas, J; et al. (2006) Trichloroethylene (TCE): Immunotoxicity
potential in CD rats following a 4-week vapor inhalation exposure. Dow Chemical Company,
Toxicology & Environmental Research and Consulting, Midland, MI, Study ID 031020, July 5,
2006, unpublished.
Yang, Q; Ito, S; Gonzalez, FJ (2007) Hepatocyte-restricted constitutive activation of PPAR alpha
induces hepatoproliferation but not hepatocarcinogenesis. Carcinogenesis 28(6): 1171-1177.
Zandi-Nekad, K; Eddy, AA; Glassock, RJ; et al (2004) Why is proteinuria an ominous biomarker
of progressive kidney disease? Lodmeu Omt 66: S76-89.
Zhang, J and Yu, KF (1998) What's the relative risk? A method of correcting the odds ratio in
cohort studies of common outcomes. JAMA 280: 1690-1691.
54
-------
Appendix A: Editorial Comments
Chapter 4
Typographical corrections - In the section on vestibular function - (headaches, dizziness, nausea)
there is a typo on p 4-101 that should be corrected. LOAEL 1000 ppm human study (Kylin et
al., 1967); 2700 ppm in rats (Tham et al 1984, Niklasson et al., 1993) and rabbits (Tham et al,
1983).
In the kidney section, there needs to be added mention of the 18% increase in kidney weight (in
male mice only) seen in the largely immunotoxicity study conducted by Peden-Adams (2008).
Editorial Footnote #1 on page 146: "Elevation of NAG in urine is a sign of proteinuria, and
proteinuria is both a sign and a cause of kidney malfunction (Zandi-Nejad et al., 2004). " Beta -
N-acetylglucosaminidase (NAG) is an enzyme released by the proximal tubules. Usually total
NAG is measured, however, this is comprised of NAG B, which reflects necrosis, and NAG A,
which reflects milder forms of proximal tubule perturbation.
The last sentence on p4-173 line 32, 33 needs to be reworded as it is unclear. Additionally, there
is a double period on line 23, p4-199.
Chapter 5
p. 5-33, line 25. Does "delivered dose" mean "administered dose"? If so, the term
"administered dose" would be clearer.
p. 5-37, line 17. Should "kidney tumors" be changed to "kidney toxicity", since this section
discusses non-cancer effects?
p. 5-10, line 9, Barrett et al., 1992, was referred to as an "acute study". On p.4-91, Table 4-21, it
was shown that Barrett et al., 1991, was acute and Barrett et al., 1992, was subchronic (10
weeks). This should be corrected.
p. 5-2, point (7), the use of the 99th percentile HEC and HED estimates was discussed. The
reason for choosing 99th percentile instead of 95th percentile was explained later in the chapter
(p. 5-45). A reference to this discussion (p. 5-48) here would be helpful for clarification, since
the 95th percentile was more commonly used in other risk assessments.
Table 5-23, NCI (1976), last bullet. 0.9 ug/m3 should be corrected to 9 ug/m3.
p. 5-24, lines 31-32. Change to "within 2-fold of each other" (1.1-1.9 mg/kg/day).
55
------- |