EPA/600/R-05/022 '
February 2005
U.S EPA Headquarters Library
Mail code 3404T
1200 Pennsylvania Avenue NW
Washington, DC 20460
202-566-0556
Tiichloroethylene Issue Paper 1:
Issues in Trichloroethylene Pharmacokinetics
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC 20460
-------
DISCLAIMER
This issue paper does not represent and should not be construed to represent any agency
determination or policy. This issue paper has not been externally reviewed. The information is
being provided to assist the National Academy of Sciences in their review of the scientific issues
surrounding trichloroethylene health risks.
n
-------
CONTENTS
LIST OF FIGURES Jv
LIST OF ABBREVIATIONS AND ACRONYMS v
PREFACE vii
AUTHORS AND CONTRIBUTORS viii
THE TCE ISSUE PAPERS ix
1. INTRODUCTION AND PURPOSE 1
2. TCE METABOLISM 2
2.1. DCA FORMATION AND THE ROLE OF TCE OXIDE (EPOXIDE) 3
2.2. PATHWAYS OF GSH CONJUGATION AND SUBSEQUENT
METABOLISM 4
2.3. EXTRAHEPATIC METABOLISM 5
2.3.1. Oxidative Metabolism in the Lung 5
2.3.2. Metabolism in the Male Reproductive System .6
3. EXISTING TCE PBPKMODELS 6
3.1. PUBLISHED TCE PBPK MODELS USED IN EPA'S 2001 DRAFT RISK
ASSESSMENT 6
3.1.1. Models Had a Common Basis 7
3.1.2. "Second Generation" Fisher Models 7
3.1.2.1. Updated Mouse PBPK Model 7
3.1.2.2. Updated Human PBPK Model 8
3.1.3. Clewell Model 8
3.1.4, Bois Reparameterizations of Fisher and Clewell Models 9
3.2. U.S. EPA/U.S. AIR FORCE-SPONSORED TCE PBPK MODEL
DEVELOPMENT 10
4. CONTINUING SCIENTIFIC UNCERTAINTIES IN TCE PBPK MODELING 11
4.1. UNCERTAINTY AND VARIABILITY IN PBPK MODELING 11
4.2. SCIENTIFIC UNCERTAINTIES RELATED TO TCE
PHARMACOKINETIC MODELING 14
4.2.1. Remaining Parameter Uncertainties in Oxidative Metabolism 14
4.2.2. Enterohepatic Recirculation 15
4.2.3. Wash-In/Wash-Out for Inhalation 15
4.2.4. Diffusion Limited Fat and Liver Compartments 16
4.2.5. Plasma Binding 16
4.2.6. Metabolites With Low Circulating Concentrations and/or Extrahepatic
Metabolism 17
5. PLANS FOR CONTINUED TCE PBPK MODEL DEVELOPMENT 18
5.1. MODEL PURPOSE AND SCOPE 18
5.2. SOURCES OF PHYSIOLOGICAL AND KINETIC DATA 20
•
iii
-------
CONTENTS (continued)
5.3. MODEL EVALUATION 22
REFERENCES 27
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION .... '. 33
APPENDIXB: COMPUTER IMPLEMENTATION 37
LIST OF FIGURES
Figure 2-1. Metabolism of trichloroethylene (TCE) 25
Figure 3-1. Typical simplified metabolism scheme for modeling 26
Figure 4-1. Possible liver metabolism scheme to model potential dose metrics 26
IV
-------
LIST OF ABBREVIATIONS AND ACRONYMS
ACSL Advanced Continuous Simulation Language
AUC Area under the concentration curve
BL Cysteine conjugate p-lyase
CDH Chloral dehydrogenase (aldehyde oxidase)
CGDP Cysteinyl-glycine dipeptidase
CH Chloral hydrate
CHL Chloral
CYP Cytochrome P450
DCA Dichloroacetic acid
DCVC Dichlorovinylcysteine
DCVCS DCVC sulfoxide
DCVG Dichlorovinyl glutathione
DCVSH Dichlorovinyl mercaptan
EHR Enterohepatic recirculation
ERDEM Exposure Related Dose Estimating Model
EPA U.S. Environmental Protection Agency
FA Formic acid
FMO Flavin-containing monooxygenase
GA Glyoxylic acid
GC/MS Gas chromatography/mass spectroscopy
GGTP y-glutamyl transpeptidase
GSH Glutathione
GST Glutathione-S transferase
iv Intravenous
KM Michaelis-Menten affinity parameter
MCA Monochloroacetic acid
MFO Mixed-function oxidase (P450)
MOA Mode of action
NADCVC N-aceryl dichlorovinylcysteine
NAS National Academy of Sciences
NAT N-acetyl transferase
NCEA National Center for Environmental Assessment
NERL National Exposure Research Laboratory
-------
LIST OF ABBREVIATIONS AND ACRONYMS (continued)
NRC National Research Council
OA Oxalic acid
ORD Office of Research and Development
PBPK Physiologically based pharmacokinetic
PEN Phenyl-tert-butyl nitroxide
PD Pharmacodynamic
RfD Reference dose
TCA Trichloroacetic acid
TCE Trichloroethylene
TCE-O-P450 Oxygenated TCE-cytochrome P450 transition state complex
TCOG TCOH glucuronide
TCOH Trichloroethanol
UDP Uridine diphosphate
UGT UDP glucuronosyl transferase
V^ Michaelis-Menten maximum velocity of reaction parameter
VI
-------
PREFACE
Publication of these issue papers is a part of EPA's effort to develop a trichloroethylene
(TCE) human health risk assessment. These issue papers were developed to provide scientific
and technical information to the National Academy of Sciences (NAS) for use in developing their
advice on how to best address the important scientific issues surrounding TCE health risks. As
such, these papers discuss a wide range of perspectives and scientific information (current
through Fall 2004) on some of these important issues, highlighting areas of continuing
uncertainty and data that may be relevant. They are intended to be useful characterizations of the
issues, not a presentation of EPA conclusions on these issues. The papers have undergone
internal review within EPA, but they have not been externally reviewed. The concepts presented
in these papers will eventually be addressed in EPA's revised risk assessment of TCE, after the
advice from the NAS, along with comments from the EPA Science Advisory Board and the
public, as well as recently published scientific literature, have been incorporated.
vii
-------
AUTHORS AND CONTRIBUTORS
Many individuals contributed to the completion of this set of tichloroethylene (TCE)
issue papers. The TCE Risk Assessment Team identified the topics covered by the papers and
prepared them for submission to the National Academy of Sciences. The authors wish to thank
Dr. Peter Preuss, Dr. John Vandenberg, Mr. David Bussard, Mr. Paul White, Dr. Bob Sonawane,
Dr. Hugh Barton, Dr. Apama Koppikar, Mr. David Bayliss, Dr. William Wood, and Dr. Ila Cote
for their input and comments.
TCE Risk Assessment Team
Jerry Blancato, ORD/NERL
Jane Caldwell*, ORD/NCEA
Chao Chen, ORD/NCEA
Weihsueh Chiu*. ORD/NCEA (TCE Chemical Manager)
Marina Evans, ORD/NHEERL
Jennifer Jinot, ORD/NCEA
Nagu Keshava*, ORD/NCEA
John Lipscomb*, ORD/NCEA
Miles Okino*. ORD/NERL
Fred Power, ORD/NERL
John Schaum, ORD/NCEA
Cheryl Siegel Scott*, ORD/NCEA
* Primary authors of issue papers. Entire team contributed via review and comment.
Vlll
-------
THE TCE ISSUE PAPERS
BACKGROUND
In August 2001, a draft, Trichloroethylene (TCE) Health Risk Assessment: Synthesis and
Characterization, was released for external review. This draft assessment drew on 16 "state-of-
the-science" papers published as a supplemental issue of Environmental Health Perspectives
(Volume 108, Supplement 2, May 2000). Subsequent to its release, EPA's 2001 draft assessment
underwent a peer review by a panel of independent scientists through EPA's Science Advisory
Board (SAB), which provided a peer review report in December 2002. In addition, the public
submitted more than 800 pages of comments to EPA during a 120-day public comment period.
There are a number of important issues that EPA will need to examine as it moves
forward in revising the draft TCE assessment. These include issues raised not only in the SAB
peer review and public comments, but also by new scientific literature published since the release
of the state-of-the-science papers and EPA's 2001 draft assessment. Some of this research is
specific to the study of TCE or its metabolites while some of it describes advances in scientific
fields more generally but which have potential relevance to characterizing the human health risks
from TCE.
In February 2004, EPA held a symposium so that authors of some of the TCE-specific
research that had been published since the release of the draft assessment could present their
findings in more detail. This symposium represented only a limited cross section of recently
published research, but was reflective of the breadth of new relevant science that EPA will
consider in revising the assessment (the presentation slides and a transcript of the meeting are
available separately on EPA's website and have already been sent to the NAS).
In 2004, EPA, in cooperation with a number of other federal agencies, initiated a
consultation with the National Academy of Sciences (NAS) to provide advice on scientific issues
related to the health risk assessment of TCE. It was recognized that a review by an NAS panel of
the important scientific issues would be beneficial and informative to clarify the state-of-the-
science as EPA moves forward in completing its health risk assessment. A charge was
developed for the NAS through an Interagency Workgroup led by the White House Office of
Science and Technology Policy.
PURPOSE OF THE TCE ISSUE PAPERS
Although EPA will need to address all of the issues identified in the charge to the NAS
panel in updating its assessment, EPA would like to focus the NAS panel's attention on a subset
of issues that EPA believes to be most critical in developing a revised risk assessment, as
summarized in four issue papers developed by EPA staff:
ix
-------
1. Issues in trichloroethylene pharmacokinetics;
2. Interactions of trichloroethylene, its metabolites, and other chemical exposures;
3. Role of peroxisome proliferator-activated receptor agonism and cell signaling in
trichloroethylene toxicity; and
4. Issues in trichloroethylene cancer epidemiology.
Each paper provides an overview of the science issues, a discussion of perspectives on
those issues (including the SAB and public comments), and an outline of some of the recently
published scientific literature. The pharmacokinetics issue paper also summarizes results from a
recent collaboration with the U.S. Air Force on TCE pharmacokinetics, as well as EPA's planned
approach for further refinement of the pharmacokinetic modeling of TCE and its metabolites.
These scientific areas were selected because they are (a) critical to the hazard and/or dose-
response characterization of TCE; (b) scientifically complex and/or controversial; and (c) areas in
which substantial important scientific literature has been recently published. The input from the
NAS on the topics described in the issue papers, as well as other topics put forth in the charge to
the NAS, should help to strengthen EPA's revised TCE assessment.
NEXT STEPS
The advice from the NAS, along with comments already received from the EPA SAB and
the public, as well as recently published scientific literature, will be incorporated into a revised
EPA risk assessment of TCE, strengthening its scientific basis. Because of the substantial
amount of new information and analysis that is expected, the revised draft of the assessment will
undergo further peer review and public comment prior to completion.
-------
1. INTRODUCTION AND PURPOSE
Understanding trichloroethylene (TCE) pharmacokinetics—absorption, distribution,
metabolism, and elimination—is critical to both the qualitative and quantitative assessment of
human health risks from environmental exposures. On a qualitative level, pharmacokinetic
information can help to identify the chemical species that may be causally associated with
observed toxic responses. In addition, the delineation of inter- and intraspecies differences can
provide insights into how laboratory animal and epidemiological data may inform overall human
health risks and into how individuals may differ in their susceptibility. Quantitatively, this
information may allow the development of physiologically based pharmacokinetic (PBPK)
models to describe the relationship between external measures of exposure and internal measures
of toxicologically relevant dose. However, it should be noted that the selection of appropriate
dose metrics depends not only on the reliability of PBPK data and models but also on the
understanding of the mode of action (MO A) for a particular toxic effect (issues regarding the
interactions of metabolites are discussed in a separate issue paper). With an adequate database,
testing, and evaluation, quantitative results from PBPK models may then be used along with
MOA information to develop appropriate alternatives to default procedures for addressing the
pharmacokinetic component of a number of risk assessment issues, including extrapolation to
different exposure conditions (e.g., exposure routes, time-concentration patterns, co-exposures);
extrapolation across species (e.g., developing human equivalent doses or concentrations);
dose-response relationships (e.g., low-dose extrapolation, use in pharmacodynamic modeling);
uncertainty (e.g., due to pharmacokinetic complexity and/or data limitations); and variability
(e.g., differences in metabolism or clearance).
This document is intended to provide an overview of TCE metabolism and PBPK
modeling to focus the National Research Council (NRC) committee's advice on specific
scientific issues and approaches to addressing TCE pharmacokinetics for the purposes of risk
assessment. This document, in discussing issues in TCE pharmacokinetics, refers to both past
work on TCE pharmacokinetics and to some of the recent studies that may be relevant for risk
assessment, but it is not intended to provide a complete survey and synthesis of the scientific
literature. Sections 2 through 4 summarize the issues and uncertainties surrounding TCE
metabolism and PBPK models with particular attention to information from recent studies and
modeling efforts. Section 5 describes the U.S. Environmental Protection Agency's (EPA's)
plans for continued development of a TCE PBPK model for use in a revised TCE risk
assessment. Input from the National Academy of Sciences (NAS) regarding interpretation of the
information presented in the following sections, the availability of additional data sources, and
-------
possible improvements to EPA's planned approach to PBPK modeling would help strengthen the
basis of EPA's revised TCE assessment
2. TCE METABOLISM
Lash et al. (2000a) presented a comprehensive review of the absorption, distribution,
metabolism, and elimination of TCE and its metabolites as part of the State of the Science series
on TCE published in 2000, and a brief summary was presented in EPA's 2001 draft risk
assessment (U.S. EPA, 2001). TCE is rapidly and extensively absorbed by all routes of
environmental exposure—ingestion, inhalation, and dermal contact. Once absorbed, TCE
distributes via the circulatory system throughout the body. Because it is lipophilic, it can
accumulate in fat and other tissues, although the half-life of TCE in fatty tissues (on the order of
hours or days) is still much shorter than more persistent substances such as dioxins (on the order
of years). The majority of TCE taken into the body is metabolized; direct exhalation is the other
major route of elimination of the parent (Lash et al., 2000a). There are a number of complexities
regarding the full spectrum of metabolic pathways. An understanding of the formation of TCE
metabolites and the pharmacokinetics of each of them is motivated by the potential toxicological
significance of many of these metabolites. In particular, for many endpoints, the toxicity of TCE
is hypothesized to be attributable to one or more of these metabolites. (The toxicology of TCE
metabolites and their interactions is discussed in more detail in a separate issue paper.) Figure
2-1 presents a postulated scheme for the pathways of TCE metabolism, adapted from the work of
Lash et al. (2000a) and Clewell et al. (2000). As shown in the figure, TCE metabolism occurs
through two main pathways—oxidation via the microsomal mixed-function oxidase (MFO)
system (i.e., P450s) and conjugation with glutathione (GSH) by glutathione-S transferases
(GSTs). Several important issues related to these metabolic pathways are discussed below.
Particular attention is given to recently published literature that may be informative. Input from
NAS regarding the interpretation of this new information and its potential utility for quantitative
analysis would be helpful to EPA as it revises its draft assessment. Of particular importance
across all the issues described below is whether sufficient information exists both within and
across species to quantify rates of TCE metabolism as well as the factors that may influence
differential flux through the various metabolic pathways. These issues are critical to PBPK
model development and use because they inform the formation and relative concentrations of
metabolites in potential target organs. Issues more directly related to pharmacokinetic modeling
are discussed in Sections 3 and 4.
-------
2.1. DCA FORMATION AND THE ROLE OF TCE OXIDE (EPOXIDE)
As noted in the Lash et al. (2000a) review, considerable uncertainty exists as to the extent
of DCA formation from TCE exposure. Two potential sources of dichloroacetic acid (DCA)
formation, from trichloroethanol (TCOH) and from dechlorination of trichloroacetic acid (TCA),
were discussed in Lash et al. (2000a). As reviewed in Bull (2000), DCA is one of the TCE
metabolites that has been hypothesized to be involved in liver tumor induction. Although some
researchers have suggested that DCA levels are too low in mice and humans after TCE exposure
to play a significant role in TCE toxicity (e.g., Barton et al., 1999), it has also been noted that
DCA-induced toxicity has been observed at exposure levels for which DCA cannot be detected
in vivo (Bull et al., 2002). Pharmacokinetic issues related to DCA formation are discussed
below; recent toxicological information regarding the potential role of DCA in TCE toxicity is
discussed in a separate issue paper.
Detection of DCA production in vivo after TCE administration has been complicated by
reported problems with analytical methodologies that have led to artifactual formation of DCA
ex vivo when samples contain significant amounts of TCA (Ketcha et al., 1996). Following the
discovery of these analytical issues, Merdink et al. (1998) re-evaluated the formation of DCA
from TCE, TCOH, and TCA in mice, with particular focus on the hypothesis that DCA is formed
from dechlorination of TCA. They were unable to detect blood DCA in naive mice following
administration of TCE, TCOH, or TCA, and, even with pretreatment with DCA to reduce
clearance rates, they were unable to detect DCA following TCA administration. They concluded
that "[although there is significant uncertainty in the amount of DCA that could be generated
from TRI [TCE] or its metabolites, our experimental data and pharmacokinetic model
simulations suggest that DCA is likely formed as a short-lived intermediate metabolite."
However, it has been noted that when directly administered, DCA can produce significantly
elevated liver rumor incidence in mice at doses for which DCA blood levels remain below
analytical detection limits owing to DCA's rapid metabolism (Bull et al, 2002; Kato-Weinstein
et al., 1998; Merdink et al., 1998). Fisher et al. (1998) reported the results of a controlled human
exposure study in which DCA was detected in some, but not all, human blood samples. To
minimize ex vivo formation of DCA resulting in chemical artifacts, the investigators analyzed
plasma rather than whole blood. However, it is still difficult to determine whether the observed
inter-individual differences are due to intrinsic variability (e.g., differences in DCA degradation
via GST-zeta), measurement errors, or a combination of each.
Much of the focus on DCA formation following TCE administration has been on
dechlorination of TCA. For instance, Merdink et al. (2000) report trapping of a DCA radical
with the spin-trapping agent phenyl-tert-butyl nitroxide (PBN), identified by gas
chromatography/mass spectroscopy (GC/MS), in both a chemical Fenton system and rodent
-------
microsomal incubations with TCA as substrate. On the other hand, the work by Guengerich and
colleagues has suggested that the source of DCA may be through a TCE oxide (epoxide)
intermediary. Although oxidation of TCE by P450s results predominantly in chloral (CHL)
(Lash et al., 2000a), previous work of Miller and Guengerich (1983) had reported evidence of
formation of the epoxide as an independent oxidative pathway (i.e., not leading to formation of
CHL). The epoxide itself may be of toxicological importance, either by itself through formation
of protein or DNA adducts or through subsequently formed reactive species (Guengerich, 2004).
In addition, Cai and Guengerich (1999) recently reported that a significant amount of DCA
(about 35%) is formed from aqueous decomposition of TCE oxide via hydrolysis in an almost
pH-independent manner. Because this reaction forming DCA is a chemical process rather than
mediated by enzymes, and because evidence suggests that some epoxide was formed from TCE
oxidation, Guengerich (2004) noted that DCA would be an expected product of TCE oxidation.
2.2. PATHWAYS OF GSH CONJUGATION AND SUBSEQUENT METABOLISM
As discussed in Lash et al. (2000b), TCE metabolism through the GSH pathway is
hypothesized to be involved in renal toxicity, but processing of GSH conjugates is complex and
poorly understood relative to the processing of oxidative metabolites. The first stable product of
the conjugation of TCE is S-(l,2 dichlorovinyl)glutathione (DCVG). A postulated scheme for
subsequent processing to dichlorovinylcysteine (DCVC), corresponding mercapturates
(N-acetyl-DCVC), and other compounds, is shown in Figure 2-1. Evidence for the in vivo
activity of the GSH pathway in humans comes from Lash et al. (1999a), who reported detection
of DCVG in the blood of human volunteers exposed to TCE. However, DCVC was not detected
in blood, and the mercapturates were detected only sporadically in urine. Bloemen et al. (2001)
measured GSH pathway metabolites in the urine of human volunteers and occupationally
exposed workers and reported that levels were below detection limits in all cases.
DCVC is thought to be a critical intermediate in the fate of GSH conjugates of TCE.
Although one potential fate of DCVC is detoxification via N-acetylation to yield mercapturates,
bioactivation to a toxic form is a potential parallel pathway. Thus, data on detoxification (e.g.,
urinary mercapturates) do not capture the total flux through the GSH pathway, and, in particular,
the data are not informative regarding the amount bioactivated (Lash et al., 2000a). It has been
hypothesized that bioactivation is through the renal beta-lyase metabolism of DCVC, producing
reactive metabolites that may contribute to renal toxicity. Recent in vitro data (Krause et al.,
2003; Lash et al., 2003) indicate that flavin-containing monooxygenases (FMO) also may be
toxicologically important for the bioactivation of DCVC, particularly in the human kidney.
Moreover, there are multiple ways in which DCVC may become available in the kidney for
bioactivation. GSH conjugates produced in the liver maybe exported directly to the blood into
-------
systemic circulation, or to the bile, where they can be reabsorbed through the gut. Although the
liver is the primary site of GSH conjugation, most tissues, including the kidney, contain GSTs
(Lash et al., 2000a). In vitro studies (Cummings et al., 2000a, b; Cunimings and Lash, 2000)
have reported GSH conjugation of TCE in rat and human kidney cells, suggesting a potential role
for local metabolism. This work has also identified several GST isoforms in kidney cells that
may be involved in TCE metabolism.
Although the work cited above may help lead to a better understanding of complex
pathways and the metabolism that results from TCE exposure, it appears to be limited for
developing a firm quantitative understanding of the relative rates of in vivo processing and the
bioactivation of conjugative metabolites.
2.3. EXTRAHEPATIC METABOLISM
Although it is generally thought that the liver is the major site of TCE metabolism, P450s,
GSTs, and other metabolizing enzymes are distributed at varying levels of activity throughout
other tissues (Lash et al., 2000a). Although extrahepatic metabolism may not contribute
significantly to overall mass balance (Lash et al., 2000a), it may be important locally in terms of
the toxicological effects from in situ production of metabolites. Two sites of potential
importance are the lung and the male reproductive system (metabolism in the kidney was
discussed above).
2.3.1. Oxidative Metabolism in the Lung
As discussed in Green (2000), the oxidative pathway of TCE metabolism in mouse lung
Clara cells is-hypothesized to be responsible for the accumulation of CHL in mouse lungs,
leading to cytotoxicity. Forkert and colleagues had previously reported cytotoxicity in mouse
lung Clara cells from TCE exposure (Forkert and Forkejt, 1994; Forkert and Birch, 1989; Forkert
et al., 1985). Green (2000) suggested that although the activity of enzymes is lower in the lung
as a whole than in the liver, the activity of P450 in the lung appears to be relatively higher than
the activity of enzymes involved in clearing CHL and TCOH (believed to be alcohol
dehydrogenase and uridine diphosphate (UDP)-glucuronosyl transferase [UGT]). Hence, these
two metabolites may accumulate in the mouse lung and lead to toxicity. Green (2000) suggests
that such a mechanism in mice may not be relevant to humans because there is little CYP2E1
activity in the human lungs as a whole. In the draft TCE assessment, it was noted that metabolic
activity from whole lungs may give misleading results because of the variety of cell types in
which high activity in a few may be diluted by others with low activity. Boers et al. (1999)
reported the number of Clara cells in the human lung and indicated that Clara cells both
contribute substantially to cell renewal and are important in the development of lung
-------
adenocarcinoma in humans. In addition, the differential activities of the relevant enzymes in
human lung tissues and cell types have not been examined to date.
2.3.2. Metabolism in the Male Reproductive System
Reports of TCE exposure affecting the male reproductive system (reviewed in the 2001
draft TCE assessment), including the observation of Leydig cell tumors in rats exposed to TCE
(Maltoni et al., 1988, 1986), have led to the investigation of metabolism and toxicity of TCE and
its metabolites in the male reproductive system. Forkert et al. (2003,2002) report several studies
that indicate TCE oxidative metabolism occurs in the male reproductive tract. They detected
CYP2E1 activity in the epididymal epithelium and testicular Leydig cells in mice, monkeys, and
humans. Analysis of seminal fluid from eight human subjects diagnosed with clinical infertility
and exposed to TCE occupationally was also performed and showed the presence of TCE, CHL,
and TCOH in all eight subjects; DCA in two subjects; and TCA in one subject. TCA and/or
TCOH were identified in urine samples from only two subjects. Although the lack of detailed
exposure information limits the use of these data for development of a quantitative
pharmacokinetic understanding, this evidence is qualitatively informative regarding the potential
for local metabolism of TCE in the male reproductive tract.
3. EXISTING TCE PBPK MODELS
Multiple PBPK models published for TCE and its metabolites show varying levels of
detail and data consistency. The focus of most of these models has been on the oxidative
pathway and the major oxidative metabolites TCA, TCOH, and TCOH glucuronide (TCOG),
reflecting the limited quantitative understanding described above for the other metabolic
pathways. Section 3.1 briefly describes the models previously used in developing EPA's 2001
draft risk assessment. The draft assessment noted a substantial amount of model uncertainty
because the models sometimes provided disparate internal dose predictions, differing in some
cases by an order of magnitude. Section 3.2 describes recent efforts sponsored by EPA and the
U.S. Air Force to develop a revised interim TCE PBPK model.
3.1. PUBLISHED TCE PBPK MODELS USED IN EPA'S 2001 DRAFT RISK
ASSESSMENT
The PBPK models used in EPA's 2001 draft risk assessment were described in detail
elsewhere (Fisher, 2000; Clewell et al., 2000; Bois, 2000a, b, and references therein) and are
briefly summarized below.
-------
3.1.1. Models Had a Common Basis
All the models described below are extensions of the earlier published models of Fisher et
al. (1991) for rats and mice and Allen and Fisher (1993) for humans. The structures of these
"first generation" models were similar to that for styrene, reported by Ramsey and Andersen
(1984), which has been used as the basis for a number of PBPK models for volatile organic
solvents. Common characteristics of these models include the following:
• Physiological compartments included the liver and fat, lumped rapidly, and slowly
perfused tissues, where transport in and out is perfusion limited with rapid
equilibrium partitioning between the tissues and the venous blood leaving the tissues.
• Gas exchange in the lung occurring via rapid equilibrium partitioning between
alveolar air and arterial blood.
• Oxidative metabolism, modeled as a saturated (Michaelis-Menten) process, occurring
in the liver, with the metabolite TCA lumped into an equivalent volume of
distribution.
The models described below built on this early work to include additional physiological
compartments and a more detailed description of metabolism.
3.1.2. "Second Generation" Fisher Models
3.1.2.1. Updated Mouse PBPK Model
Abbas and Fisher (1997) and Greenberg et al. (1999) developed updated PBPK models
for TCE and its metabolites in mice for both oral (corn oil gavage) and inhalation exposure,
respectively. The number of parent compartments was expanded with the addition of a lung,
kidney, and gut compartment. Additional metabolites included in the model were CHL/chloral
hydrate (CH), TCA, TCOH, TCOG, and dichloroacetic acid (DCA). Physiological submodels
for each metabolite, with liver, kidney, lung, and a lumped body compartment, were linked to the
parent model and to each other through liver metabolism. Physiological parameters were taken
or derived from the literature; chemical-specific parameters were measured experimentally or
inferred from the experimental data. New time-course data, which included extensive metabolite
measurements in multiple tissues, were used for calibration and/or validation. For the inhalation
experiments, previously collected time-courses (Fisher et al., 1991) were also used for validation
purposes. Fisher (2000), in reviewing these efforts, noted several unresolved discrepancies in
model parameters, including the following:
-------
• In the inhalation experiments, a fractional uptake of 53% in the lung was
hypothesized that significantly improved the model fit to the data.
• Several metabolic rate constants derived from model calibration differed significantly
between oral and inhalation exposures.
3.1.2.2. Updated Human PBPK Model
Fisher et al. (1998) developed an updated human PBPK model for TCE and its
metabolites. The model structure was similar to the mouse model, with an expanded
physiological model for TCE and physiological submodels for metabolites. However, the
metabolic scheme included only TCA, TCOH, and TCOG (Figure 3-1) because those were the
only metabolites consistently and reliably detected in experimental data. New time-course data
were collected on human subjects exposed to 50 ppm and 100 ppm of TCE in air. Data included
individual measurements of parent and metabolite concentrations in exhaled air, blood, and urine
as well as covariates, such as age, body weight, and percentage of body fat. Physiological
parameters other than those measured were taken or derived from the literature; chemical-
specific parameters were measured experimentally or inferred through calibration to the
experimental data. Urinary excretion parameters were optimized on a subject-specific basis
owing to high observed variability. The observed variability and the sparseness of the data set
(i.e., limited measurements relative to the model complexity) required that all the data be used
for calibration, and traditional validation was not performed. Additional comparisons were
performed against previously published data from Monster et al. (1976) and Muller et al. (1974),
but these required adjustment of metabolic parameters to obtain adequate fits. Overall, there was
some overprediction of the exhaled vapor concentration after cessation of exposure and some
underprediction of TCE blood concentrations at 50 ppm.
3.1.3. Clewell Model
Clewell et al. (2000) reported on the development of an updated PBPK model for TCE in
mice, rats, and humans. The same model structure was used for all three species. The primary
extension of the model from the original Fisher et al. (1991) and Allen and Fisher (1993) models
was the inclusion of additional metabolites. Like the updated Fisher models, TCOH, TCOG, and
DC A were described. However, at the time the model was developed (circa 1996), the more
recent data from Fisher's laboratory (Greenberg et al., 1999; Fisher et al., 1998; Abbas and
Fisher, 1997) were not yet available. Clewell et al. (2000) also included CHL in the
tracheobronchial region and DCVC production in the liver with bioactivation in the kidney.
Unlike the updated Fisher models, all circulating metabolites were modeled using
one-compartment (volume of distribution) descriptions. Biliary excretion of TCOG and
8
-------
enterohepatic recirculation of TCOH were also included as part of the metabolite model
structure. As was the case for the Fisher models, physiological parameters were taken or derived
from the literature, and chemical-specific parameters were derived from previously published
measurements or inferred through calibration to previously published data sets. Specifically,
evaluations were made against a subset of measurements for mouse oral and inhalation studies
(Fisher et al., 1991; Templin et al., 1993; Prout et al., 1985), rat oral studies (Templin et al.,
1995; Larson and Bull, 1992b), and human inhalation studies (Monster et al., 1979; Muller et al.,
1975, 1974; Stewart et al., 1970). In addition, in vitro data were used to estimate some
metabolism parameters related to GHL in the lung and DCVC in the kidney. Note that these data
sets did not include any of the data collected by Fisher and colleagues reported above as having
been used in the second-generation Fisher models. Finally, allometric scaling was used
extensively to convert parameters across species from those for which calibration data were
available to those for which no calibration data existed.
Validation was not performed in the strict sense because not enough experimental data
were available for all the metabolites across species. Clewell et al. (2000) visually inspected the
model simulations and the data and concluded that simulation results were generally reasonable,
although it was clear that complete agreement between the model and each study investigated
could not be obtained with a single set of parameters for each species. The results of the
sensitivity and uncertainty analyses that were performed indicated that dose metrics for TCE and
for the major metabolites TCA and TCOH could be expected to be reasonably precise (Clewell et
al., 2000). Dose metrics related to CHL in the lung and DCVC bioactivation in the kidney, on
the other hand, were reported to be highly uncertain owing to a lack of adequate pharmacokinetic
data across species (Clewell et al., 2000). The DCA metrics also were considered to be uncertain
because known analytical errors (Ketcha et al., 1996) existed in some measurements.
3.1.4. Bois Reparameterizations of Fisher and Clewell Models
Bois (2000a, b) performed reparameterizations of the Fisher and Clewell models using a
Bayesian statistical framework. The basic approach was to develop a hierarchical statistical
model for the population distribution of each model's parameters rather than to use the single
values determined by Fisher and Clewell. Therefore, population variability and the uncertainty
surrounding that variability were incorporated into the analysis. A Markov Chain Monte Carlo
technique was used to perform the high-dimensional numerical sampling and integration
necessary to derive the joint probability distribution of the parameters given the available data.
These analyses also provided estimates of the uncertainty and variability surrounding individual
dose estimates, although, in many cases, uncertainty and variability could not be disentangled
because the data were aggregated. For the analysis of rodent data, the population model
-------
described "inter-lot" or "inter-lab" variability, which included measurement errors, rather than
actual inter-individual variability.
The Bois analysis of the Clewell model was performed first (Bois, 2000b). The statistical
analysis required some minor modifications of the Clewell model. The data used included most
of those reported in Clewell et al. (2000) and some additional published data sets. Again, the
data used in Fisher's second-generation models were not included here. Overall, this formal
statistical method led to predictions that were systematically better fitting than those (already
reasonable) fits obtained by Clewell et al. (2000). However, as found in Clewell et al. (2000), a
number of data sets still showed a relatively poor fit, even with allowance for variability supplied
by the hierarchical statistical framework.
The marginal posterior distributions for a number of parameters generated from this
analysis were used as input priors to the Bois analysis of the Fisher model (Bois, 2000a). Some
minor modifications of the Fisher model were required to accommodate the population analysis,
and the data sets used included most of the data used by Fisher. However, the analysis did not
include the Greenberg et al. (1999) mouse inhalation data that show evidence of fractional
absorption in the lung (see Section 4.2.3). The mouse model showed good fits to TCE and TCA,
while residuals for other metabolites often reached one or two orders of magnitude. Some of the
differences may have been due to variability, as there seemed to be substantial "noise" in the data
for some metabolites, but systematic differences were still apparent. The human model showed a
substantially better fit, particularly for TCA, with residuals generally of a factor of 2 or less.
Because the Clewell model posterior distributions were used as input priors here, the Bois
analysis of the Fisher model actually incorporates to some degree all the data from both analyses.
The incorporation is imperfect not only because the two model structures are different but also
because the covariance structure of the "Bois-Clewell" posterior distributions was lost in the use
of marginal posterior distributions for the new priors. Therefore, somewhat different results may
have been obtained if each model were calibrated to the entire data set.
3.2. U.S. EPA/U.S. AIR FORCE-SPONSORED TCE PBPK MODEL DEVELOPMENT
Throughout 2004, EPA and the Air Force have jointly sponsored an initial attempt at
combining elements from the TCE PBPK modeling efforts described above. The interim results
of this effort are briefly summarized here. A more detailed description will be available in a
separate interim report. The development of this model included a peer consultation conducted
in June 2004, a summary report for which is also available separately (TERA, 2004).
The goal of this effort was to address several important issues for the first time,
particularly the following:
10
-------
• A single interim model structure combining features from both the Fisher and Clewell
models was developed and used for all three species of interest (mice, rats, and
humans). An effort was made to combine structures in as simple a manner as
possible; the evaluation of alternative structures was left for future work.
• The Fisher and Clewell models, as well as the Bois analyses of those models,
reflected substantially different databases of information. This effort evaluated the
revised model against a combination of the data sets previously used, to the extent
applicable. However, a comprehensive review of all published data was left for future
work.
• Similar to the Bois (2000a, b) analyses, a hierarchical Bayesian population analysis
using Markov Chain Monte Carlo techniques was performed on the combined model
with the combined database of kinetic data to provide estimates of parameter
uncertainty and variability.
• Species- and dose-dependent TCA plasma binding was implemented to evaluate its
effects on the associated dose metrics (see discussion in Section 4,2.5). Equilibrium
binding, as reported in the in vitro study of Lumpkin et al. (2003), was incorporated
into the TCA metabolite submodel. It was assumed that the on/off rates were fast
compared to transport, as existing TCA studies have not reported the time-scale of
TCA binding kinetics.
Interim results from this effort seem to suggest that a single model structure can fit a
variety of data evaluated for TCE and its major oxidative metabolites TCA, TCOH, and TCOG,
although in some cases different parameter values, particularly for metabolism, are required for
different studies. This interim model represents a major step in the development of TCE PBPK
models, particularly because the model was evaluated against a larger database of kinetic data
than was any previous model. However, a number of the issues described below remain to be
investigated; therefore, EPA plans additional model development, evaluation, and
characterization. The genera! approach for these continued efforts is described in Section 5.
4. CONTINUING SCIENTIFIC UNCERTAINTIES IN TCE PBPK MODELING
4.1. UNCERTAINTY AND VARIABILITY IN PBPK MODELING
Uncertainty and variability in PBPK modeling are discussed in this section. (See
Bernillon and Bois, 2000, for a more detailed discussion of statistical issues in PBPK modeling.)
Specific issues related to TCE are discussed in Section 4.2. The terms "uncertainty" and
"variability," as used here, refer to distinct concepts: uncertainty refers to a lack of knowledge
that may be reducible with additional data or study, and variability refers to inherent
11
-------
heterogeneity that is irreducible. Uncertainty in the characterization of variability may exist
because the data are limited or because understanding of the interrelationships within complex
biological systems is lacking. Although one would ideally like to separate uncertainty and
variability, this is not always possible. For instance, when data from individuals are aggregated,
then measurement error and inherent variability may not be separable, even assuming the
underlying model is valid.
Given the complexity of TCE pharmacokinetics and the concomitant complexity of TCE
PBPK models, it is important to characterize the uncertainty in the PBPK modeling results so as
to inform the uncertainty in dose-response assessment. This characterization can have both
qualitative and quantitative components. Model uncertainty—which is the uncertainty due to the
structural features and assumptions that underlie a particular model—is usually addressed
primarily qualitatively through a critical evaluation of model features. Parameter
uncertainty—which is the uncertainty in the values of input parameters that are required by the
model—lends itself to more quantitative analysis. Although the use of PBPK models in risk
assessment is intended to provide more accurate (i.e., less uncertain) estimates of dose relative to
default procedures, it should be recognized that PBPK modeling cannot be expected to
completely eliminate pharmacokinetic uncertainty. In some cases, rigorous analysis of PBPK
models may actually reveal pharmacokinetic uncertainties that were not previously understood or
characterized, or that are greater in magnitude than assumed through default procedures.
Therefore, it is possible that analysis using PBPK models may appear to increase uncertainty.
Even in this case, however, PBPK models can help to identify data that may have the greatest
impact on reducing these uncertainties.
Owing to their biological basis, PBPK model parameters may also exhibit significant
inter-individual variability. Parameters for PBPK models fall into several categories, with
varying levels of independent information as their variability. Physiological parameters (such as
organ weights and blood flows) are generally measurable, with some having a priori information
as to their variability. Chemical-specific parameters, such as partition coefficients, are also
measurable using in vitro methods. Generally less information is available about variability for
these parameters. However, in these cases, experiment-specific data (other than body weights)
are generally not available—that is, the individuals for which physiological and chemical-
specific parameters are measured differ from the individuals for which pharmacokinetic data are
collected. Parameters such as metabolic and clearance rates are usually inferred through
comparison of model predictions with pharmacokinetic data—i.e., the models are calibrated by
changing these parameters to fit experimental data, such as time-courses of chemical
concentrations. In addition, they may be inferred from in vitro measurements (e.g., microsomal
preparations) or data from other chemicals, but with an additional uncertainty that is not easily
12
-------
quantified. Pharmacokinetic data for both humans and laboratory animals show considerable
variability, even within single experiments that use inbred rodent strains under identical
conditions (e.g., Prout et al., 1985). Variability in parameters can thus be inferred from
variability in the kinetic data and, in some cases, independent measures obtained from in vitro
data on enzyme content and/or activities.
PBPK models may be useful in characterizing this variability through a population
approach. The basic idea is to fit the variability in individual data by assigning different
parameter values to different individuals. Although this is a relatively new approach in human
health risk assessment, it has been applied routinely in the development of pharmaceutical drugs;
however, the structural models used in pharmaceutical research are typically empirical (e.g., one
or two compartments) rather than physiologically based (Ette and Williams, 2004a, b, c;
Davidian and Giltinan, 2003; Yuh et al., 1994; Sheiner et al., 1972). One commonly used
approach is the "two-stage" method, which involves fitting a model to each separate individual's
data and then obtaining population parameter estimates in a second step based on the individual
results from the first step. Because of the difficulty of parameter estimation for a complex
model, it is common practice either to fix all but a few parameters to estimated values or to set
up a very simple model with only a few parameters (as is common for pharmaceuticals). Fisher
et al. (1998) used this two-stage approach in their analysis of human data on TCE, calibrating the
urinary excretion separately for different individuals. However, this procedure may attribute too
much variability to the parameter being adjusted rather than to other parameters that also may be
uncertain and/or variable (Bernillon and Bois, 2000; Woodruff and Bois, 1993). In addition, this
procedure is ill-suited for the situation where multiple data sets provide information on
overlapping sets of parameters.
A second approach involves developing a statistical model for simultaneously
characterizing population variability and overall parameter uncertainty. This approach has been
variously described as nonlinear mixed-effect modeling (e.g., Davidian and Giltinan, 2003;
Sheiner et al., 1977) and Bayesian population modeling (e.g., Gelman et al., 1996), depending on
whether the analysis uses a frequentist (typically maximum likelihood) or Bayesian statistical
framework. The use of population models (in both pharmaceutics and toxicology) has focused
on single (mostly human) data sets where individual measurements are available (Jonsson and
Johanson, 2003, 2002, 2001a, b; Jonsson et al., 2001a, b; Smith et al., 2001; Bois et al., 1999;
Gelman et al., 1996; Bois et al., 1996a, b). The use of this methodology for simultaneously
calibrating parameters using multiple data sets, including laboratory animal data that are typically
aggregated, is a relatively new area with few published results other than those of Bois (2000a, b)
for TCE, described above. However, because of the ability to incorporate prior information on
the analysis of any particular data set, the Bayesian approach is conceptually well suited for the
13
-------
situation with TCE in which multiple data sets are being analyzed together. A commonly used
method to implement this Bayesian population approach has been the use of Markov Chain
Monte Carlo techniques.
4.2. SCIENTIFIC UNCERTAINTIES RELATED TO TCE PHARMACOKINETIC
MODELING
Although the EPA/Air Force-sponsored effort to combine elements of the Fisher,
Clewell, and Bois models into an interim PBPK model has examined some of the differences
between these previous approaches (see the separate interim report for additional details), a
number of uncertainties remain as to model assumptions, structure, and parameters. These issues
are briefly summarized below. It is unclear at this time which of these uncertainties are
important with respect to fitting available data or predicting internal dose; our proposed approach
to assessing these sensitivities is summarized in Section 5.
4.2.1. Remaining Parameter Uncertainties in Oxidative Metabolism
The analyses described above reported a number of parameter uncertainties.
Uncertainties related to modeling of low-concentration metabolites, such as DCA, DCVC, and
CHL, are discussed below, as they are intertwined with structural issues. Even for the "well
calibrated" oxidative pathway, two significant issues surrounding the metabolic parameter
remain:
Fisher (2000) reported substantial differences in parameter values for TCE oxidative
metabolism in mice between inhalation and gavage dosing. Of particular note was
that the Michaelis-Menten (KM) parameter used to fit the corn oil gavage data was
greater by over an order of magnitude than that used to fit the inhalation data.
Moreover, the estimate for KM based on the gavage data was substantially greater than
that based on previously published studies, a finding also reported in the Bois (2000a)
analysis of the Fisher models. Fisher (2000) suggested that this adjustment reflected
the model's oversimplification of the oral uptake of TCE in com oil rather than a true
change in the KM for oxidation of TCE. Smaller changes of up to a factor of two in
other parameters, such those for as glucuronidation of TCOH, were also noted by
Fisher (2000) between gavage and inhalation studies.
Fisher (2000) reported a substantially lower value for the BW3/4-scaled Michaelis-
Menten Vjnax, or maximum velocity of reaction, parameter (denoted "VmaxC") in
humans based on the Fisher et a!. (1998) data than was reported in previous studies
using previously published data. This finding was also reported in the Bois analysis
of the Fisher models, with Bois (2000a) noting that the difference could be due to
differences in the data, rather than in the modeling.
14
-------
Currently, it is not clear whether these differences are due to inherent variability or
structural misspecification of the model.
4.2.2. Enterohepatic Recirculation
In the liver, chemicals can be secreted into the bile and circulated into the gut, where they
are reabsorbed into the portal blood. Recirculation of metabolites increases their effective
half-life in the body, and this is reflected in urinary output of the chemicals. TCOG and TCA
have been measured in the bile of rats (Stenner et al., 1997), and bile-cannulated rats showed
different blood profiles of the chemicals. A PBPK model developed based on this work (Stenner
et al., 1998) included enterohepatic recirculation and showed a reasonable match to rat
concentration profiles after oral doses of TCE (in 2% Tween 80) and intravenous doses of TCA
and TCOH. Difficulties exist in extrapolating the rat data to other species because
pharmaceutical studies have shown that biliary excretion does not scale uniformly (Mahmood
and Sahajwalla, 2002).
The significance of recirculation on important dose metrics is uncertain, as existing
PBPK models have generally shown reasonable fits to blood and urine data without recirculation.
For instance, even though Clewell et al. (2000) implemented recirculation structurally,
reabsorption in the gut was set to zero for comparison to most data. Bois (2000a) noted,
however, that urinary excretion data for TCOG in mice was not well fit by the Fisher model,
which did not include recirculation. Overall, model fit and the sensitivity of dose metrics with
and without enterohepatic recirculation have not been evaluated. It is likely that TCA and TCOH
metrics are sensitive to enterohepatic recirculation, but a quantitative characterization has not yet
been reported.
4.2.3. Wash-In/Wash-Out for Inhalation
For different exposure routes in a particular species, the distribution, metabolism, and
elimination of chemicals are expected to be the same, with the only difference being absorption
and first-pass clearance for the particular route. Inhalation is commonly modeled as being
complete, whereas for volatile chemicals, the blood:air partition coefficient determines both the
uptake and elimination of a chemical.
As mentioned above, when the best-fit model parameters based on oral mouse data
(Abbas and Fisher, 1997) were used in an inhalation exposure simulation (Greenberg et al.,
1999), the model overpredicted the absorption of TCE as reflected in TCE blood and exhalation
concentrations. The model could be made consistent with the data if fractional uptake were
implemented, where only a fraction of the chemical is available for transfer to the plasma during
inhalation exposure. Physiologically, this could occur if the lung epithelium were to act as a
15
-------
reservoir for TCE during inhalation exposure, so only a fraction of the TCE inhaled reaches the
alveolar region. Fractional uptake has been reported for other water-soluble solvents (Pastino et
al., 1997; Perkins et al., 1995; Johanson, 1991), and because the lung tissue is exposed to high
local concentrations, it has a potential impact on risk.
4.2.4. Diffusion Limited Fat and Liver Compartments
The PBPK models for TCE described above all assume perfusion-limited distribution of
chemicals to the physiologic compartments. The representation assumes that the compartments
are well mixed over the time-scales of blood flow, so that the compartment concentration can be
described by the blood concentration and a partition coefficient. Some discrepancies have been
noted, however. For instance, Bois (2000a) reported that the measured adiposity of the
individual subjects from Fisher et al. (1998) did not correlate well with the posterior estimates for
the model parameter for percentage body weight as fat. Bois suggested that one possible
explanation was that the pharmacokinetic compartment for fat was not well estimated by external
adiposity measurements. However, model error has also been proposed as an explanation.
The liver and fat are known to be heterogenous tissues (Albanese et al., 2002; Andersen
et al., 1997), and based on their importance to PBPK models of volatile organics, investigations
have been conducted on the impact of diffusion limitations in these tissues. For instance, Keys et
al. (2003) recently developed a PBPK model for TCE parent kinetics in rats and mice that
includes more complex descriptions of the fat and liver compartments. For the fat, transport
between the blood and the compartment was changed from flow- to diffusion-limited. For the
liver, a second "deep" compartment was added with transport via diffusion to and from the
"shallow" liver, which is also the site of metabolism. Keys et al. (2003) concluded that TCE
parent concentrations are better simulated by this more complex model, and that although other
dose metrics were not evaluated, metabolite concentrations would not be expected to be
significantly changed.
4.2.5. Plasma Binding
The binding of chemicals to proteins in plasma affects their availability to other tissues
and ultimately their effective half-life in the body. Typical descriptions of relatively weak serum
protein binding assume fast rates compared to the other relevant time-scales.
The TCE metabolites TCA and DCA are known to bind to plasma proteins. Templin et
al. (1995) measured the extent of TCA binding in humans, mice, and rats over limited
concentration ranges. More recently, Lumpkin et al. (2003) measured TCA binding in humans,
mice, and rats over a wide concentration range that spans reported TCA plasma concentrations
from experimental studies. These data showed significant species differences, with humans
16
-------
exhibiting the most binding and mice exhibiting the least. As mentioned above, existing TCA
studies have not reported the time-scale of TCA-binding kinetics. For modeling, one typical
assumption is that the ratio of bound-to-free is in equilibrium in arterial blood, but only the free
fraction is available for exchange with tissues.
Schultz et al. (1999) measured the extent of DCA binding in rats at a single concentration
of about 100 uM and found a binding fraction of less than 5%. However, these data are not
greatly informative for TCE exposure in which DCA levels are significantly lower, and limitation
to a single concentration precludes fitting to standard binding equations from which the binding
at low concentrations could be extrapolated. Furthermore, there is insufficient information on
cross-species differences for the extrapolation of rat DCA data to other species.
4.2.6. Metabolites With Low Circulating Concentrations and/or Extrahepatic Metabolism
As mentioned in Section 2, significant uncertainty surrounds the metabolic pathways for
metabolites with low circulating concentrations and for which extrahepatic metabolism may be
important toxicologically. These include DCA, CHL, and the metabolites from GSH
conjugation. Previous modeling of TCE metabolism indicates that the relative formation rates of
these chemicals are small, so that they are not appreciably constrained by the total TCE mass
balance (Clewell et al., 2000). However, it would be desirable to model the metabolism to these
chemicals (e.g., Figure 4-1) because of their potential toxicological importance.
Pharmacokinetic studies of DCA exposure (Barton et al., 1999; Abbas et al., 1997;
Larson and Bull, 1992a) provide a picture of the behavior of DCA once formed, but, as discussed
in Section 2, the magnitude of the amount formed from TCE is uncertain. Circulating DCA in
mice has been measured in a variety of TCE exposure studies (Greenberg et al., 1999; Merdink et
al., 1998; Abbas and Fisher, 1997), The only DCA data for humans following TCE exposure are
from Fisher et al. (1998). For all of these studies, the extent to which analytical artifacts of DCA
remain (Ketcha et al., 1996) is unclear, so these data may be useful only as a maximum
constraint.
As discussed in Section 2, a complication for CHL and GST metabolites is that local
metabolism to these compounds can occur in the lung and kidney, respectively, and the
contributions of local and liver metabolism to the concentration at the site of action are unknown.
CHL (in equilibrium with CH) is rapidly metabolized into TCOH and TCA, so circulating CHL
levels are expected to be low. Circulating CH/CHL was measured after high-dose TCE
exposures in mice (Greenberg et al., 1999; Abbas and Fisher, 1997; Prout et al., 1985) and rats
(Prout et al., 1985). In Abbas and Fisher (1997) and Greenberg et al. (1999), the concentration of
CH was also measured in lung homogenate, although the concentration in the lung Clara cells,
which are believed to be a site of local production and toxicity (Odum et al., 1992), was not
17
-------
accessible. CH was administered to mice (Abbas et al., 1996) and humans (Muller et al., 1974)
in controlled exposures, which may be useful for characterizing its behavior after formation from
TCE. The relative contributions of hepatic and extrahepatic metabolism remain uncertain.
As mentioned in Section 2, Lash et al. (1999a) measured DCVG concentrations in blood
following TCE exposure. In addition, the results of a rat study measuring both oxidative and
GST metabolites are in preparation (Lash, 2004). However, the kidney is believed to be the site
of action for metabolites of DCVG, and the contribution of circulating DCVG produced in the
liver relative to local production of DCVG is uncertain. Bernauer et al. (1996) and Birner et al.
(1993) have measured the urinary metabolites of DCVG, providing an indicator of the amount of
metabolism that occurred, at least through the N-acetyl transferase (NAT) detoxification
pathway, hi vitro studies exist that measure metabolism by GST in liver and kidney cells (Lash
et al., 1999b) and for activity of beta-lyase (Lash et al., 1990), which has been associated with the
nephrotoxicity (Anders et al., 1988). The relative in vitro kinetics can be used to inform
reasonable parameter values in the liver and kidney, although, as discussed in Section 2, the
potential role of FMO complicates the quantification of GST metabolite bioactivation across
species.
5. PLANS FOR CONTINUED TCE PBPK MODEL DEVELOPMENT
The issues raised in the previous sections suggest a number of potential approaches for
PBPK model development. The models of Fisher, Clewell, and Bois provide a relatively
consistent basis for development of a single base model structure for mice, rats, and humans.
Additional structural features can then be evaluated against the database of existing data. The
model parameters can then be refined and further characterized through a Bayesian statistical
framework. Given the complexity of the model and the large database of studies with which to
compare model predictions, this process will need to proceed in an iterative manner. An interim
model developed through an EPA/Air Force collaboration, described above, will be used as a
starting point for additional model development and evaluation. The following sections are
intended to inform the NRC committee of EPA's plans for continued TCE PBPK model
development. NRC feedback on this approach, and suggestions for improving it, would help to
strengthen the basis of EPA's revised TCE risk assessment
5.1. MODEL PURPOSE AND SCOPE
The main objective of EPA's TCE PBPK modeling effort is to estimate biologically
relevant internal doses for use in risk assessment. These dose estimates may be used for a variety
of extrapolations (e.g., high to low dose, inter- and intra-species, exposure route or regime) and
18
-------
are subject to both uncertainty and variability. Selection of the appropriate dose metric for use in
risk assessment is dependent not only on the reliability of the data and models but also on the
understanding of the MOA for a particular toxic effect—i.e., what may be the "biologically
relevant" internal doses. Because a revised assessment of TCE MOA hypotheses has not yet
been completed, selection of the most appropriate dose metric(s) is not possible at this time
(although metabolite interactions are discussed in a separate issue paper). Therefore, the
approach EPA is planning concentrates first on characterization of the PBPK model structure and
parameters. Previous PBPK modeling of TCE is used as a guide to develop a model that is both
flexible and reliable to the extent scientifically supported by available data.
How PBPK modeling was used in the 2001 draft assessment provides an important first
step in consideration of the scope of continued PBPK model development. Uses of PBPK
modeling considered in the 2001 draft assessment included the following:
• Cross-species extrapolation of cancer risk estimates based on mouse and rat
bioassays. Use of PBPK modeling was considered for analyzing mouse liver tumors
(based on TCA and DCA area under the concentration curve [AUC] dose metrics), rat
kidney tumors (based on bioactivation of DCVC), and mouse lung tumors (based on
CHL). Because of the significant model and parameter uncertainty of these three
endpoints, PBPK model-based cancer risk estimates were developed only for mouse
liver tumors.
• Route-to-route extrapolation of human cancer risk estimates, based on either TCA or
DCA AUCs.
• Development of noncancer risk estimates based on mouse and rat noncancer studies.
Dose metrics considered included TCA and DCA AUC, TCOH peak concentration,
and bioactivation of DCVC, depending on the study and endpoint. Because the draft
reference dose (RfD) was based on liver weight changes, TCA and DCA AUC dose
metrics were used. These dose metrics were used both for characterizing the
pharmacokinetic component of cross-species extrapolation and human variability,
based on the Bois analysis of the Clewell model.
Therefore, the basic lexicological database, as reviewed in the 2001 draft, suggests that
the PBPK models of TCE be developed in mice, rats, and humans. These are also the species in
which numerous pharmacokinetic studies have been conducted. EPA's modeling effort will
consider the dose metrics that have been previously considered for use in risk assessment,
although additional dose metrics may need to be evaluated if the MOA evaluation suggests
alternatives not listed here. Regarding routes of exposure, oral and inhalation kinetic studies are
most common in mice and rats, and nearly all controlled human kinetic studies are based on
inhalation exposure. A few rodent studies were reported for injection exposures (Stenner et al.,
19
-------
1997; Abbas et al. 1997, 1996; Lee et al.3 1996; D'Souza et al., 1985), and only one group has
reported results for controlled dermal exposure in rats and humans (Poet et al., 2000). However,
the general human population may be exposed orally and dermally as well. Therefore, for risk
assessment purposes, the model should ideally simulate inhalation and oral exposures for all
species, as well as injection for rodents and dermal exposure for humans.
Although the 2001 draft assessment noted that co-exposures with other chemicals may
modulate TCE toxicity, the PBPK modeling effort is not planning to include explicit modeling of
chemical mixtures beyond the extent to which TCE metabolism leads to a mixture of circulating
metabolites. Modeling additional exposure from some of the metabolites themselves would
presumably be straightforward, but there appear to be insufficient data on more complex
chemical mixtures with TCE to model with a PBPK model. Some exploratory analyses have
been reported (Dobrev et al., 2002, 2001), but the data sets are sparse, where only a few
chemicals and compartments were measured. Also, the previously mentioned uncertainties with
TCE alone seem significant enough so that evaluation of a model of mixtures seems impractical
at this time. However, the mechanistic framework of the PBPK model illustrates the parameters
that have an important impact on the dose metrics and observable quantities, thereby facilitating
experimental design for future investigations of mixtures.
The 2001 draft assessment also noted the potential for susceptible subpopulations with
differential risks from TCE exposure. However, the data appear to remain largely qualitative, so
potential subpopulations will not be explicitly modeled within the PBPK models, although they
may be addressed through other means in the revised risk assessment Although physiologic
differences are modestly characterized for segments of the population, substantial uncertainties
are associated with the TCE-specific biochemical differences associated with age, disease state,
and genetic polymorphisms. These questions can to a degree be addressed in the individual
model evaluation, where the sensitivity analysis will reveal the parameters that are sensitive with
respect to toxicologically important dose metrics. These parameters can be compared to the
knowledge of the represented biological process to identify potential subpopulations that may
respond differently and to develop hypotheses for the identification and eventual quantification
of risk metrics for those populations.
5.2. SOURCES OF PHYSIOLOGICAL AND KINETIC DATA
Parameters for PBPK models include three distinct types of data: physiological data,
chemical-specific parameters, and parameters for determining the stochastic behaviors of models.
The physiological data are independent of the chemical being modeled and refer to such areas as
organ volumes and blood flows. Chemical-specific parameters include the partition coefficients,
20
-------
metabolic rate constants, and coefficients for protein binding. Important stochastic behaviors to
be modeled are derived from inter-individual and experimental variances.
The specific physiological compartments considered in EPA's modeling effort are
selected based on information available for exposure, toxicology, and metabolic profile related to
a particular chemical and potential active metabolites. Important information includes
measurements of parent or metabolite concentrations, known metabolism or elimination, or a
known toxic effect within specific tissues. Distribution within, between, and among organs,
tissues, and fluid is modeled according to compartmental volumes, blood flow rates, and
blood:tissue partitioning.
As in previous PBPK modeling efforts (Clewell et al., 2000; Fisher et al., 1998),
physiological parameters and chemical-specific partition coefficients are estimated from
independent studies, such as standard physiology references or in vitro experiments. The kinetic
parameters associated with absorption, metabolism, and elimination are fit to kinetic data,
although in some cases they may be informed by in vitro data as well.
Studies on the formal analysis of variability using PBPK models are limited (Bois, 2000a,
b). There is no established method to characterize the variability in a heterogeneous population.
Even physiologic references, such as ILSI-RSI (1994), Brown et al. (1997), and ICRP (2003),
focus on determining the range and/or central tendency of reasonable values, not on developing
the distributions that are necessary for probabilistic analyses, hi addition, the covariance between
parameters (e.g., between tissue volumes and blood flows) are not characterized but may need to
be accounted for to avoid unrealistic combinations of parameter values.
For the chemical-specific parameters, the variability of parameters can be estimated from
in vitro experiments (partition coefficients, metabolism) and individual pharmacokinetic data.
As discussed previously, formal evaluation of variability against data can be performed using a
hierarchical Bayesian population model employing a Markov Chain Monte Carlo algorithm.
However, for much of the laboratory data on TCE, individual data were pooled, making it
difficult to characterize inter-individual variation and to separate variability from measurement
error. An additional challenge is that human data sets tend to be relatively homogeneous and
performed under controlled conditions, and thus may not represent the range of inter-individual
variability present in the full population. Therefore, in the evaluation of variability for the target
populations of the risk assessment, posterior distributions from the Bayesian analysis could tend
to underestimate full population variability and thus may not always be appropriate.
21
-------
53. MODEL EVALUATION
Model evaluation involves the consideration of alternative model structures, the
estimation of model parameters, and characterization of uncertainty and variability. The EPA
effort will be conducted in three broad phases: exploratory model development, sensitivity
analysis, and Bayesian population analysis of uncertainty and variability. The resulting
individual models with "mean" parameter values estimated from the Bayesian population
analysis will be made available to the public through the Exposure Related Dose Estimating
Model (ERDEM) system, developed by the U.S. EPA (Appendix B). The Bayesian population
analysis using the Markov Chain Monte Carlo algorithm will also be done using publicly
available software, MCSim (Appendix B).
The first phase will involve exploratory analysis in which a single model structure (or set
of structures) will be built with separate parameterization for each species. During this phase, the
uncertainties described above surrounding alternative model structures will be investigated.
These include enterohepatic recirculation, wash-in/wash-out in the lung, diffusion-limited fat and
liver compartments, and plasma binding. In addition, as described above, considerable
uncertainty exists regarding the metabolic pathways and parameters for low-concentration
metabolites and those locally produced. The goal of this phase is to settle on one or a few model
structure(s) that best fit(s) the available data while maintaining biological plausibility. These
exploratory models will be calibrated with available kinetic data. The evaluation of model fit,
performed either statistically and/or by visual inspection depending on the nature of the available
data, will guide the refinement of model structure and parameters. Appendix A provides a table
of the candidate pharmacokinetic studies against which the PBPK models are to be evaluated.
For simple models and a limited number of data sets, a typical "validation" after
calibration would be to compare the fit model with data not used for calibration. However, the
data sets for TCE include different exposure scenarios and measurements on different scales that
do not present a clear metric to evaluate the performance of an individual model. Also, despite
the number of studies, the data are sparse relative to the qualitatively important compartments
and metabolites. Thus, traditional "validation" will not be feasible. An arguably more relevant
question for risk assessment is a characterization of the confidence with which a model can
predict dose metrics of interest. This is related to the characterization of uncertainty and
variability, described below.
The second (intermediate) phase is sensitivity analysis of the individual models to
determine the important parameters. Two types of analyses will be performed: sensitivity of the
fit to data and sensitivity of the dose metrics. Sensitive parameters are critical to characterize the
analysis of uncertainty and variability. The priors assigned to those parameters will reflect the
certainty to which the value is known. For example, the tissue volumes for which we have
22
-------
independent measurements will have greater precision than the kinetic rate constants that were fit
to data. The parameters that impact the dose metrics but not the fit to data require other
constraints to characterize their uncertainty. The model can also be analyzed around these
parameters to determine the measurable quantities that would constrain the value; this provides
the basis for future targeted experimental studies to reduce the uncertainty. The sensitivity
analysis also highlights the biological processes that affect risk by identifying the corresponding
model parameters. These parameters can be evaluated against the knowledge of the biological
pathway to identify other chemicals that would impact the dose metrics (i.e., sharing the same
saturable metabolic pathway) or to identify subpopulations that would exhibit different risks due
to differences in a pathway.
The third phase of evaluation is a Bayesian population analysis (using a Markov Chain
Monte Carlo algorithm) that incorporates the variability in the available data to characterize the
uncertainty and variability in the model parameters and their impact on model predictions.
Examples of the use of the Bayesian population approach with PBPK models were discussed
earlier. The previous section described the sources of data for developing prior estimates of the
population means and variances and their attendant uncertainties. Other published uncertainty
analyses, including both the Bayesian analyses cited above and traditional Monte Carlo analyses
(e.g., Clewell et at, 2000; Cronin IV et al., 1995) will also be used as a guide for reasonable prior
distributions. As noted above, however, the populations for which kinetic data are available (i.e.,
Appendix A) may not be wholly representative of the target populations of interest for the risk
assessment.
Before using the results of this analysis to perform inferences as to the uncertainty and
variability in dosimetry predictions, it is necessary to confirm that the model is consistent with
the data and existing biological knowledge. As was the case for the exploratory analysis, a
formal model "validation" will not be performed owing to the need to use all the available data in
calibrating the model. However, "model fit" does need to be assessed, particularly because
posterior parameter estimates may be overconstrained if the model does not fit the data (e.g.,
discussion in Bois, 2000a). Gelman et al. (2004) provide an extensive discussion of techniques
for "posterior predictive checking," or the assessment of whether the model is consistent with the
data. The general basic idea is to generate simulated data from the posterior parameter
distribution for comparison with the actual observed data. Bois (2000a, b) provided some
examples of this type of analysis. In addition, posterior estimates of parameters and predictions
should also be checked for biological plausibility because there is biological knowledge that is
not formally included in either the prior estimates or in the likelihood. For instance, it can be
difficult to translate biological knowledge quantitatively into formal prior distributions,
particularly with respect to parameter covariances. Thus, parameter combinations that are
23
-------
unrealistic biologically may be easier to check a posteriori. The results of these analyses may
lead to additional refinement of the model and/or parameters and may thus necessitate iteration
of the entire model evaluation process. It should be noted, however, that it may not be possible
to completely disentangle the combination of data errors, model errors, and parameter variability.
Finally, the results of the Bayesian population analysis will be used along with available
information on the target populations of interest (see Section 5.2) to characterize the uncertainty
and variability in the pharmacokinetic modeling results. This characterization will be an
important input, along with the appropriate mode-of-action and hazard characterization, into the
selection of dose metrics as well as species- and route-extrapolation methods for use in
quantitative dose-response analysis.
24
-------
GGTP
CGDP
TCE
GST / VMFO
^S* FA GA
DCVG TCE-O-P450, ' * OA
/
* EHR
\
DCA
MCA
DCVC
BL/ 1 ^Vv NAT
if JFMOX^
DCVSH * NADCVC
i DCVCS 1
\ * \
reactive species
CHL
urine
urine A. TCA
UGT
urine
\
\
OA 4— DCA TCOG
\ 1
MCA
urine urine
Figure 2-1. Metabolism of trichloroethylene (TCE).
CDH = chloral dehydrogenase (aldehyde oxidase); EHR = enterohepatic
recirculation; FA = formic acid; FMO = flavin-containing monooxygenase;
GA = glyoxylic acid; OA = oxalic acid; TCE-0-P450 = oxygenated TCE-
cytochrome P450 transition state complex; TCOG = TCOH glucuronide;
UGT = UDP glucuronosyl transferase; BL = cysteine conjugate p-lyase;
CGDP = cysteinyl-glycine dipeptidase; DCVCS = DCVC sulfoxide; DCVG =
dichlorovinyl glutathione; DCVSH = dichlorovinyl mercaptan; GGTP = y-
glutamyl transpeptidase; NADCVC = N-acetyl dichlorovinylcysteine;
NAT = N-acetyl transferase.
Source: Adapted from Clewell et al. (2000) and Lash et al. (2000a).
25
-------
TCE
TCA
TCOH
TCOG
Figure 3-1. Typical simplified metabolism scheme for modeling.
TCE = trichloroethylene; TCA = trichloroacetic acid; TCOH = trichloroethanol;
TCOG = TCOH glucuronide.
TCE
DCA
DCVG
CH
TCA
TCOH
TCOG
Figure 4-1. Possible liver metabolism scheme to model potential dose
metrics.
TCE = trichloroethylene; DCA = dichloroacetic acid; DCVG = dichlorovinyl
glutathione; CH = chloral hydrate; TCA = trichloroacetic acid; TCOH =
trichloroethanol; TCOG = TCOH glucuronide.
26
-------
REFERENCES
Abbas, R; Fisher, JW. (1997) A physiologically based pharmacokinetic model for trichloroethylene and its
metabolites, chloral hydrate, trichloroacetate, dichloroacetate, trichloroethanol, and trichloroethanol glucuronide in
B6C3F1 mice. Toxicol Appl Pharmacol 147:15-30.
Abbas, RR; Seckel, CS; Kidney, JK; et al. (1996) Pharmacokinetic analysis of chloral hydrate and its metabolism in
B6C3F1 mice. Drug Metab Dispos 24(12):1340-1346.
Abbas, RR; Seckel, CS; MacMahon, KL; et al. (1997) Determination of kinetic rate constants for chloral hydrate,
trichloroethanol, trichloroacetic acid, and dichloroacetic acid: a physiologically based modeling approach.
Toxicologist 36:32-33.
Albanese, RA; Banks, HT; Evans, MV; et al. (2002) Physiologically based pharmacokinetic models for the transport
of trichloroethylene in adipose tissue. Bull Math Biol 64:97-13 1.
Allen, BC; Fisher, JW. (1993) Pharmacokinetic modeling of trichloroethylene and trichloroacetic acid in humans.
Risk Anal 13:7 1-86.
Anders, MW; Lash, LH; Dekant, W; et al. (1988) Biosynthesis and metabolism of gtutathione conjugates to toxic
forms. Crit Rev Toxicol 18:311-341.
Andersen, ME; Eklund, CR; Mills, JJ; et al. (1 997) A multicompartment geometric model of the liver in relation to
regional induction of cytochrome P450s. Toxicol Appl Pharmacol 144:135-144.
Barton, HA; Bull, RJ; Schultz, I; et al. (1999) Dichloroacetate (DCA) dosimetry: interpreting DCA-induced liver
cancer dose response and the potential for DCA to contribute to trichloroethylene-induced liver cancer. Toxicol Lett
Bernauer, U; Bimer, G; Dekant, W; et al. (1996) Biotransformation of trichloroethylene: dose-dependent excretion
of 2,2,2-trichlor-metabolites and mercapturic acids in rats and humans after inhalation. Arch Toxicol 70(6):338-346.
Bernillon P; Bois, FY. (2000) Statistical issues in toxicokinetic modeling: a Bayesian perspective. Environ Health
Perspect 108 (Suppl 5):883-893.
Birner, G; Vamvakas, S; Dekant, W; et al. (1993) Nephrotoxic and genotoxic N-acetyl-S-dichlorovinyl-L-cysteine is
a urinary metabolite after occupational 1,1,2-trichloroethylene exposure in humans: implications for the risk of
trichloroethylene exposure. Environ Health Perspect 99:281-284.
Bloemen, LJ; Monster, AC; Kezic, S; et al. (2001) Study on the cytochrome P-450- and glutathione-dependent
biotransformation of trichloroethylene in humans. Int Arch Occup Environ Health 74(2): 102-108.
Boers, JE; Ambergen, AW; Thunnissen, FB. (1999) Number and proliferation of Clara cells in normal human airway
epithelium. Am J Respir Crit Care Med 1 59(5 Pt. 1 ): 1 585-1 59 1 .
Bois, FY. (2000a) Statistical analysis of Fisher et al. PBPK model of trichloroethylene kinetics. Environ Health
Perspect 108:275-282.
Bois, FY. (2000b) Statistical analysis of Clewell et al. PBPK model of trichloroethylene kinetics. Environ Health
Perspect 108(Suppl 2):307-3 16.
Bois, FY; Jackson, ET; Pekari, K; et al. (1996b) Population toxicokinetics of benzene. Environ Health Perspect
104(Suppl6):1405-1411.
27
-------
Bois, FY; Smith, TJ; Gelman, A; et al. (1999) Optimal design for a study of butadiene toxicokinetics in humans.
Toxicol Sci 49(2):213-224.
Bois, FY; Gelman, A; Jiang, J; et al. (1996a) Population toxicokinetics of tetrachloroethylene. Arch Toxicol
70(6):347-355.
Brown, RP; Delp, MD; Lindstedt, SL; et al. (1997) Physiological parameter values for physiologically based
pharmacokinetic models. Toxicol Ind Health 13(4):407-484.
Bull, RJ. (2000) Mode of action of liver tumor induction by trichloroethylene and its metabolites, trichloroacetate
and dichloroacetate. Environ Health Perspect 108(Suppl 2):241-259.
Bull, RJ; Orner, GA; Cheng, RS; et al. (2002) Contribution of dichloroacetate and trichloroacetate to liver tumor
induction in mice by trichloroethylene. Toxicol Appl Pharmacol 182(l):55-65.
Cai, H; Guengerich, FP. (1999) Mechanisms of aqueous decomposition of trichloroethylene oxide. J Am Chem Soc
(121): 11656-11663.
Clewell, HJ; Gentry, PR; Allen, BC; et al. (2000) Development of a physiologically based pharmacokinetic model of
trichloroethylene and its metabolites for use in risk assessment. Environ Health Perspect 108(suppl 2):283-305.
Cronin, IV, WJ; Oswald, EJ; Shelley, ML; et al. (1995) A trichloroethylene risk assessment using a Monte Carlo
analysis of parameter uncertainty in conjunction with physiologically-based pharmacokinetic modeling. Risk
Analysis 15(5):555-565.
Cummings, BS; Lash, LH. (2000) Metabolism and toxicity of trichloroethylene and S-(l,2-dichlorovinyl)-L-cysteine
in freshly isolated human proximal tubular cells. Toxicol Sci 53(2):458-466.
Cummings, BS; Lasker, JM; Lash, LH. (2000a) Expression of glutathione-dependent enzymes and cytochrome
P450s in freshly isolated and primary cultures of proximal tubular cells from human kidney. J Pharmacol Exp Ther
293(2):677-685i
Cummings, BS; Parker, JC; Lash, LH. (2000b) Role of cytochrome P450 and glutathione S-transferase alpha in the
metabolism and cytotoxicity of trichloroethylene in rat kidney. Biochem Pharmacol 59(5):531-543.
Davidian, M; Giltinan, DM. (2003) Nonlinear models for repeated measurements: an overview and update. Editor's
invited article. J Agricult Biolog Environ Stat 8(4):387-419.
Dobrev, ID; Andersen, ME; Yang, RSH. (2001) Assessing interaction thresholds for trichloroethylene and 1,1,1-
trichloroethane using gas uptake studies and PBPK modeling. Arch Toxicol 75:134-144.
Dobrev, ID; Andersen, ME; Yang, RSH. (2002) In silico toxicology: simulating interaction thresholds for human
exposure to mixtures of trichloroethylene, tetrachloroethylene, and 1,1,1-trichloroethane. Environ Health Perspect
110(10)1031-1039.
Ette, El; Williams, PJ. (2004a) Population pharmacokinetics I: background, concepts, and models. Ann
Pharmacother 38(10):1702-1706.
Ette, El; Williams, PJ. (2004b) Population pharmacokinetics II: estimation methods. Ann Pharmacother
38(11):1907-1915.
Ette, El; Williams, PJ. (2004c) Population pharmacokinetics III: design, analysis, and application of population
pharmacokinetic studies. Ann Pharmacother 38(12):2136-2144.
28
-------
Fisher, JW. (2000) Physiologically based pharmacokinetic models for trichloroethylene and its oxidative metabolites
Environ Health Perspect 108(suppl 2):265-273.
Fisher, JW; Mahle, DA; Abbas, R. (1998) A human physiologically based pharmacokinetic model for
trichloroethylene and its metabolites, trichloroacetic acid and free trichloroethanol. Toxicol Appl Pharmacol
152:339-359.
Fisher, JW; Gargas, ML; Allen, BC; et al. (1991) Physiologically based pharmacokinetic-modeling with
trichloroethylene and its metabolite, trichloroacetic acid, in the rat and mouse. Toxicol Appl Pharmacol
109:183-195.
Forkert, PG; Birch, DW. (1989) Pulmonary toxicity of trichloroethylene in mice. Covalent binding and
morphological manifestations. Drug Metab Dispos 17(1):106-113.
Forkert, PG; Forkert, L. (1994) Trichloroethylene induces pulmonary fibrosis in mice. Can J Physiol Pharmacol
72(3):205-210.
Forkert, PG; Sylvestre, PL; Poland, JS. (1985) Lung injury induced by trichloroethylene. Toxicology
35(2):143-160.
Forkert, PG; Lash, LH; Nadeau, V; et al. (2002) Metabolism and toxicity of trichloroethylene in epididymis and
testis. Toxicol Appl Pharmacol 182(3):244-254.
Forkert, PG; Lash, L; Tardif, R; et al. (2003) Identification of trichloroethylene and its metabolites in human seminal
fluid of workers exposed to trichloroethylene. Drug Metab Dispos 31 (3):306-311.
L.
Gelman, A; Bois, F; Jiang, J. (1996) Physiological pharmacokinetic analysis using population modeling and
informative prior distributions. J Amer Stat Assoc 91:1400-1412.
Gelman, A; Carlin, JB; Stern, HS; et al. (2004) Bayesian data analysis. 2nd. ed. Boca Raton, FL: Chapman &
Hall/CRC.
Green, T. (2000) Pulmonary toxicity and carcinogenicity of trichloroethylene: species differences and modes of
action. Environ Health Perspect 108(Suppl 2):261-264.
Greenberg, MS; Burton, GA; Fisher, JW. (1999) Physiologically based pharmacokinetic modeling of inhaled
trichloroethylene and its oxidative metabolites in B6C3F1 mice. Toxicol Appl Pharmacol 154:264-278.
Guengerich, FP. (2004) Metabolism of trichloroethylene and covalent binding of reaction products. Presented at the
Symposium on New Scientific Research Related to the Health Effects of Trichloroethylene. (February 26-27,2004 at
the Renaissance Mayflower Hotel, Washington, DC). Available from:
http://cfbub2.epa.gov/ncea/cfm/recordisplay.cfm?deid=75934.
ICRP (International Commission on Radiological Protection). (2003) Annals of the ICRP basic anatomical and
physiological data for use in radiological protection: reference values. 32(3-4). Publication 89.
ILSI-RSI (International Life Science Institute, Risk Science Institute). (1994) Physiological parameters for PBPK
models. Report prepared by the ILSI Risk Science Institute Working Group on Physiological Parameters.
December.
Johanson, G. (1991) Modelling of respiratory exchange of polar solvents. Ann Occup Hyg 35:323-329.
Jonsson, F; Johanson, G. (200la) Bayesian estimation of variability in adipose tissue blood flow in man by
physiologically based pharmacokinetic modeling of inhalation exposure to toluene. Toxicology 157(3): 177-193.
29
-------
Jonsson, F; Johanson, G. (2001b) A Bayesian analysis of the influence of GSTT1 polymorphism on the cancer risk
estimate for dichloromethane. Toxicol Appl Pharmacol 174(2):99-112.
Jonsson, F; Johanson, G. (2002) Physiologically based modeling of the inhalation kinetics of styrene in humans
using a Bayesian population approach. Toxicol Appl Pharmacol 179(l):35-49.
Jonsson, F; Johanson, G. (2003) The Bayesian population approach to physiological toxicokinetic-toxicodynamic
models—an example using the MCSim software. Toxicol Lett 38(1-2): 143-150.
Jonsson, F; Bois, FY; Johanson, G. (2001b) Assessing the reliability of PBPK models using data from methyl
chloride-exposed, non-conjugating human subjects. Arch Toxicol 75(4}:189-199.
Jonsson, F; Bois, F; Johanson, G. (2001 a) Physiologically based pharmacokinetic modeling of inhalation exposure of
humans to dichloromethane during moderate to heavy exercise. Toxicol Sci 59(2):209-218.
Kato-Weinstein, J; Lingohr, MK; Orner, GA; et al. (1998) Effects of dichloroacetate on glycogen metabolism in
B6C3F1 mice. Toxicology 130(2-3): 141-154.
Ketcha, MM; Stevens, DK; Warren, DA; et al. (1996) Conversion of trichloroacetic acid to dichloroacetic acid in
biological samples. J Anal Toxicol 20:236-241.
Keys, DA; Bruckner, JV; Muralidhara, S; et al. (2003) Tissue dosimetry expansion and cross-validation of rat and
mouse physiologically based pharmacokinetic models for trichloroethylene. Toxicol Sci 76(1):35-50.
Krause, RJ; Lash, LH; Elfarra, AA. (2003) Human kidney flavin-containing monooxygenases and their potential
roles in cysteine s-conjugate metabolism and nephrotoxicity. J Pharmacol Exp Ther 304(1): 185-191.
Larson, JL; Bull, RJ. (1992a) Metabolism and lipoperoxidative activity of trichloroacetate and dichloroacetate in
rats and mice. Toxicol Appl Pharmacol 115:268-277.
Larson, JL; Bull, RJ. (1992b) Species differences in the metabolism of trichloroethylene to the carcinogenic
metabolites trichloroacetate and dichloroacetate. Toxicol Appl Pharmacol 115:278-285.
Lash, LH. (2004) Metabolism and mechanisms of renal cellular injury induced by TCE. Presented at the
Symposium on New Scientific Research Related to the Health Effects of Trichloroethylene. (February 26-27,2004 at
the Renaissance Mayflower Hotel, Washington, DC). Available from: .
http://cfpub2.epa.gov/ncea/cfm/recordisplay.cfm?deid=75934.
Lash, LH; Parker, JC; Scott, CS. (2000b) Modes of action of trichloroethylene for kidney tumorogenesis. Environ
Health Perspect l08(Suppl 2):225-240.
Lash, LH; Fisher, JW; Lipscomb, JC; et al. (2000a) Metabolism of trichloroethylene. Environ Health Perspect
108(Suppl 2): 177-200.
Lash, LH; Lipscomb, JC; Putt, DA; et al. (1999b) Glutathione conjugation of trichloroethylene in human liver and
kidney: kinetics and individual variation. Drug Metab Dispos 27(3):351-359.
Lash, LH; Nelson, RM; Van Dyke, R; et al. (1990) Purification and characterization of human kidney cytosolic
cysteine conjugate beta-lyase activity. Drug Metab Dispos 18:50-54.
Lash, LH; Putt, DA; Hueni, SE; et al. (2003) Roles of necrosis, apoptosis, and mitochondrial dysfunction in
S-( l,2-dichlorovinyl)-L-cysteine sulfoxide-induced cytotoxicity in primary cultures of human renal proximal tubular
cells. J Pharmacol Exp Ther 305(3): 1163-1172.
30
-------
Lash, LH; Putt, DA; Brashear, WT; et al. (1999a) Identification of S-(l,2-dichlorovinyl)glutathione in the blood of
human volunteers exposed to trichloroethylene. J Toxicol Environ Health 56(1):1-21.
Lee, KM; Bruckner, JV; Muralidhara, S; et al. (1996) Characterization of presysteraic elimination of
trichloroethylene and its nonlinear kinetics in rats. Toxicol Appl Pharmacol 139:262-271.
Lumpkin, MH; Bruckner, JV; Cambell, JL; et al. (2003) Plasma binding of trichloroacetic acid in mice, rats, and
humans under cancer bioassay and environmental exposure conditions. Drug Metab Disp 31(10): 1203-1207.
Mahmood, I; Sahajwalla, C. (2002) Interspecies scaling of biliary excreted drugs. JPharmSci 91(8):1908-1914.
Maltoni, C; Lefemine, G; Cotti, G. (1986) Experimental research on trichloroethylene carcinogenesis. In: Maltoni,
C; Mehlman, MA, eds. Archives of research on industrial carcinogenesis. Princeton, NJ: Princeton Scientific
Publishing; pp. 1-393.
Maltoni, C; Lefemine, G; Cotti, G; et al. (1988) Long-term carcinogenic bioassays on trichloroethylene
administered by inhalation to Sprague-Dawley rats and Swiss and B6C3F1 mice. Ann N Y Acad Sci 534:316-351.
Merdink, JL; Gonzalez-Leon, A; Bull, RJ; et al. (1998) The extent of dichloroacetate formation from
trichloroethylene, chloral hydrate, trichloroacetate, and trichloroethanol in B6C3F1 mice. Toxicol Sci 45:33-41.
Miller, RE; Guengerich, FP. (1983) Metabolism of trichloroethylene in isolated hepatocytes, microsomes, and
reconstituted enzyme systems containing cytochrome P-450. Cancer Res 43(3):1145-1152.
Monster, AC; Boersma, G; Duba, W. (1976) Pharmacokinetics of trichloroethylene in volunteers, influence of
workload and exposure concentration. Intl Arch Occup Environ Health 38:87-102.
Monster, AC; Boersma, G; Duba, W. (1979) Kinetics of trichloroethylene in repeated exposure of volunteers. Int
Arch Occup Environ Health 42:283-292.
Muller, G; Spassovski, M; Henschler, D. (1974) Metabolism of trichloroethylene in man. II. Pharmacokinetics of
metabolites. ArchToxikol 32:283-295.
Muller, G; Spassovski, M; Henschler, D. (1975) Metabolism of trichloroethylene in man. III. Interaction of
trichloroethylene and ethanol. ArchToxikol 33:173-189.
Odum, J; Foster, JR; Green, T. (1992) A mechanism for the development of Clara cell lesions in the mouse lung
after exposure to trichloroethylene. Chem Biol Interact 83:135-153.
Pastino, GM; Asgharian, B; Roberts, K; et al. (1997) A comparison of physiologically based pharmacokinetic model
predictions and experimental data for inhale'd ethanol in male and female B6C3F1 mice, F344 rats, and humans.
Toxicol Appl Pharmacol 145:147-157.
Perkins, RA; Ward, KW; Pollack; GM. (1995) A pharmacokinetic model of inhaled methanol in humans and
comparison the methanol disposition in mice and rats. Environ Health Perspect 103(7-8):726-733.
Poet, TS; Corley, RA; Thrall, KD; et al. (2000) Assessment of the percutaneous absorption of trichloroethylene in
rats and humans using MS/MS real-time breath analysis and physiologically based pharmacokinetic modeling.
Toxicol Sci 56(l):61-72.
Prout, MS; Provan, WM; Green, T. (1985) Species differences in response to trichloroethylene. Toxicol Appl
Pharmacol 79:389-400.
Ramsey, JC; Andersen, ME. (1984) A hysiologically based description of the inhalation pharmacokinetics of styrene
in rats and humans. Toxicol Appl Pharmacol 72:15 9-175.
31
-------
Schultz, 1R; Merdink, JL; Gonzalez-Leon, A; et al. (1999) Comparative toxicokinetics of chlorinated and brominated
haloacetates in F344 rats. Toxicol Appl Pharmacol 158:103-114.
Sheiner, LB; Rosenberg, B; Melmon, KL. (1972) Modeling of individual pharmacokinetics for computer-aided drug
dosage. Comput Bioraed Res 5(5):411-459.
Sheiner, LB; Rosenberg, B; Marathe, VV. (1977) Estimation of population characteristics of pharmacokinetic
parameters from routine clinical data. J Pharmacokinet Biopharm 5(5):445-479.
Smith, TJ; Lin, YS; Mezzetti, M; et al. (2001) Genetic and dietary factors affecting human metabolism of
1,3-butadiene. Chem Biol Interact 135-136:407-28.
Stenner, RD; Merdink, JL; Fisher, JW; et al. (1998) Physiologically-based pharmacokinetic model for
trichloroethylene considering enterohepatic recirculation of major metabolites. Risk Anal 18(3):26i-269.
Stenner, RD; Merdink, JL; Templin, MV; et al. (1997) Enterohepatic recirculation of trichloroethanol glucuronide as
a significant source of trichloroacetic acid in the metabolism of trichloroethylene. Drug Metab Disp 25(5):529-535.
Stewart, R; Dodd, H; Gay H; et al. (1970) Experimental human exposure to trichloroethylene. Arch Environ Health
20:64-71.
Templin, MV; Parker, JC; Bull, RJ. (1993) Relative formation of dichloroacetate and trichloroacetate from
trichloroethylene in male B6C3F1 mice. Toxicol Appl Pharmacol 123:1-8.
Templin, MV; Stevens, DK; Stenner, RD; et al, (1995) Factors affecting species differences in the kinetics of
metabolites of trichloroethylene. J Toxicol Environ Health 44:435-447.
TERA (Toxicology Excellence for Risk Assessment). (2004) Report of the peer consultation of harmonized PBPK
model for trichloroethylene. Available at
http://www.tera.org/vera/TCE/TCE%20PBPK%20Peer%20Consultation%20Meeting%20Report%20final.pdf.
U.S. EPA (Environmental Protection Agency). (2001) Trichloroethylene health risk assessment: synthesis and
characterization (external review draft). National Center for Environmental Assessment, Washington Office,
Washington, DC, EPA/600/P-01/002A.
Woodruff, TJ; Bois, FY. (1993) Optimization issues in physiological toxicokinetic modeling: a case study with
benzene. Toxicol Lett 69(2): 181-196.
Yuh, L; Beal, S; Davidian, M; et al. (1994) Population pharmacokinetic/pharmacodynamic methodology and
applications: a bibliography. Biometrics 50(2):566-575.
32
-------
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION
Lead author
R.R. Abbas
R.R. Abbas
R.R. Abbas
H.A. Barton
H.A. Barton
U. Bernauer
U. Bernauer
G. Bimer
L.J. Bloemen
C.E. Dallas
R.W. D'Souza
J.G. Fernandez
J.W. Fisher
J.W. Fisher
J.W. Fisher
J.W. Fisher
Reference
Toxicol Appl
Pharmacol 147: 15-30,
1997.
Toxicologist 36:32-33,
1997.
Drug Metab Dispos
24(12): 1340-1346,
1996.
Toxicol Appl
Pharmacol
130:237-247, 1995.
Toxicol Lett
106(1):9-21, 1999.
Arch Toxicol
70(6):338-346, 1996.
Arch Toxicol
70(6):338-346, 1996.
Environ Health
Perspect 99:28 1-284,
1993.
Int Arch Occup
Environ Health
74:102-108,2001.
Toxicol Appl
Pharmacol
110:303-314,1991.
J Toxicol Environ
Health 15:587-601,
1985.
BrJIndMed
34(l):43-55, 1977.
Risk Anal
13(l):87-95, 1993.
Toxicol Appl
Pharmacol
109(2): 183-1 95, 1991.
Toxicol Appl
Pharmacol
109(2): 183-1 95, 1991.
Toxicol Appl
Pharmacol
109(2): 183-1 95, 1991.
Species
Mice-Male B6C3F1
Mice-B6C3Fl
Mice-B6C3Fl
Rats-Male SD
Mice-Male B6C3F1
Humans-Male
Rats-Wistar
Rats-Wistar, F344
Mice-NMRI
Humans-Male
Rats-Male SD
Rats-Male SD
Humans-Male
Mice-Male and
female B6C3F1
Mice-Female
B6C3F1
Mice-Male B6C3F1
Rats-Female F344
Exposure
seen aria
TCE oral
TCA, TCOH,
DCA, CH iv
CHiv
TCE inhalation
DCA iv and oral
(aqueous)
TCE inhalation
TCE inhalation
TCE oral gavage
TCE inhalation
TCE inhalation
TCEiv
TCE inhalation
TCE oral gavage
in com oil
TCE inhalation
TCE inhalation
TCE inhalation
Measurements
TCE and metabolites in
blood, liver, lung,
kidney, fat, and urine
Compounds in blood
and urine
Compounds in blood
and urine
TCE closed chamber
concentrations
DCA in blood
Oxidation and GST
metabolites in urine
Oxidation and GST
metabolites in urine
NADC, TCA in urine
Oxidation and GST
metabolites in urine
TCE in blood, breath
TCE in blood
Alveolar, excreted
TCE
TCE in blood; TCA in
plasma
TCE and metabolites in
blood and chamber
TCE and metabolites in
blood and chamber
TCE and TCA in blood
33
-------
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION
(continued)
Lead author
J.W. Fisher
J.W. Fisher
J.W. Fisher
M.S. Greenberg
I. Jakobson
T. Kaneko
D.A. Keys
S. Lapare
J.L. Larson
J.L. Larson
J.L. Larson
J.L. Larson
L.H. Lash
L.H. Lash
Reference
Toxicol Appl
Pharmacol
109(2): 183-195, 1991.
Toxicol Appl
Pharmacol
152(2):339-359, 1998.
Toxicol Appl
Pharmacol
152(2):339-359, 1998.
Toxicol Appl
Pharmacol
154(3):264-278, 1999.
Acta Pharmacol
Toxicol (Copenh)
59(2): 135-143, 1986.
Toxicology
143(2):203-208, 2000.
Toxicol Sci
76(1):35-50, 2003.
Int Arch Occup
Environ Health
67(6):375-394, 1995.
Toxicol Appl
Pharmacol
115(2):278-285, 1992.
Toxicol Appl
Pharmacol
1 15(2):278-285, 1992.
Toxicol Appl
Pharmacol
115(2):268-277, 1992.
Toxicol Appl
Pharmacol
H5(2):268-277, 1992.
J Toxicol Environ
Health A. 56(1): 1-21,
1999.
J Toxicol Environ
Health A. 56(1): 1-21,
1999.
Specie*
Rats-Male F344
Humans-Female
Humans-Male
Mice-Male B6C3F1
Rats-Female SD
Rats-Male Wistar
Rats-Male SD
Humans
Mice-Male B6C3F1
Rats-Male Sprague-
Dawley
Mice-Male B6C3F1
Rats-Male F344
Humans-Female- 1 -
BW 66.5
Human-Male- 1-BW
71.4
Exposure
scenario
TCE inhalation
TCE inhalation
TCE inhalation
TCE inhalation
TCE inhalation
TCE inhalation
TCE inhalation
TCE inhalation
TCE oral
(aqueous)
TCE oral
(aqueous) .
DCA, TCA oral
DCA, TCA oral
TCE inhalation
TCE inhalation
Measurements
TCE and TCA in blood
TCE and metabolites in
blood and urine;
exhaled TCE
TCE and metabolites in
blood and urine;
exhaled TCE
TCE and metabolites in
blood, liver, lung, fat,
and kidney; TCE
chamber
concentrations
TCE in blood
TCE in blood; TCA,
TCOH in urine
TCE in blood and
tissues
TCE in blood; TCA in
urine
TCE and metabolites in
blood
TCE and metabolites in
blood
DCA, TCA in plasma
DCA, TCA in plasma
DCVG in blood
DCVG in blood
34
-------
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION
(continued)
Lead; author
K.M. Lee
K.M.Lee
J.L. Merdink
J.L. Merdink
A.C. Monster
A.C. Monster
G. Muller
G. Muller
Z.V. Paykoc
T.S. Poet
T.S. Poet
!
M.S. Prout
M.S. Prout
Reference
Toxicol Appl
Pharmacol
164(l):55-64, 2000.
Toxicol Appl
Pharmacol
139(2):262-271, 1996.
Toxicol Sci
45(1):33-41, 1998.
J Toxicol Environ
Health A.
57(5):357-368, 1999.
Int Arch Occup
Environ Health
38(2}:87-102, 1976.
Lit Arch Occup
Environ Health
42(3^4):283-292,
1979.
Arch Toxicol
32(4):283-295, 1974.
Arch Toxicol
33(3):173-189, 1975.
J Pharmacol Exp Ther
85:289, 1945.
Toxicol Sci
56(1):6 1-72, 2000.
Toxicol Sci
56(l):61-72, 2000.
Toxicol Appl
Pharmacol
79(3):389-400, 1985.
Toxicol Appl
Pharmacol
79(3):389^MX>, 1985.
Species
Rats-Male SD
Rats-Male SD
Mice-Male B6C3F1
Rat-Male F344
Humans
Humans
Humans-Male
Humans-Male
Humans
Humans
Rats-Male F344
Mice-Male B6C3F1
and Swiss Webster
Rats-Male Osbome-
Mendel and
Alderley Park
Wistar
Exposure .
scenario
TCE stomach
injection
TCE arterial,
venous, portal,
stomach
injections
CH,TCEiv
CH, TCOH iv
TCE inhalation
TCE inhalation
TCE inhalation;
CH.TCA,
TCOH oral
TCE inhalation
TCAiv
TCE dermal
TCE dermal
TCE gavage
TCE gavage
Measurements
TCE in arterial blood
TCE in arterial blood
TCA, CH, TCOH in .
blood
CH, TCOH in blood;
TCOG, CH, TCA in
bile
TCE in breath; TCA
and TCOH in blood
and urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCA in blood and
urine
TCE in breath
TCE in chamber
TCE and metabolites in
blood; I4C elimination
TCE and metabolites in
blood; 14C elimination
35
-------
should require no additional recompilation of code to run the-model as described in the
document.
Model 2: PBPK Development Using MCSim Language
A second model is being developed in the MCSim language. MCSim is an open-source
statistical modeling package initially developed by Frederic Bois and others for the application of
modern Monte Carlo statistical methods in complex nonlinear models. Since MCSim includes a
sublanguage for describing dynamic models in terms of their component differential equations
and typical time-varying inputs, it has been particularly valuable in the application of Markov
Chain Monte Carlo methods to estimating Bayesian posterior distributions for parameters of
PBPK models.
Dynamic models in MCSim are written in an algebraic language. Model specification
includes predefining all the parameters for the model, declaring all the variables whose dynamics
are governed by differential equations, declaring all the variables whose values need to be output,
specifying input variables whose values will be determined by special functions that provide for
periodic or episodic inputs, and specifying the differential equations for the model. This model
specification file is translated by the MCSim software into the C programming language. Then it
is compiled and linked to libraries that provide routines for integrating the differential equation
system, carrying out the required Monte Carlo simulations, and doing the input and output
functions. The resulting executable file is then run with specially formatted input files that can
change parameter values and specify the nature of the desired simulation, whether it is a
numerical integration of the differential equation system, a Monte Carlo simulation of parameter
variability or uncertainty, or a Markov Chain Monte Carlo estimate of Bayesian posterior
distributions for model parameters.
MCSim models are portable at several levels. At the lowest level, since MCSim itself is
open source, and since open-source C-language compilers are available for almost all computing
platforms (e.g., UNIX, Microsoft Windows, and Apple OS-X), models can be distributed as
model source and recompiled and run with little additional cost to reviewers. Compiled models
are also executable files and can be run without any additional software (although the executables
are specific to particular operating systems and computing hardware). Thus, the compiled
models can be distributed and their behavior evaluated without the installation of additional
software.
38
-------
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION
(continued)
Lead author
K.M. Lee
K.M. Lee
J.L. Merdink
J.L. Merdink
A.C. Monster
A.C. Monster
G. Muller
G. Muller
Z.V. Paykoc
T.S. Poet
T.S. Poet
M.S. Prout
M.S. Prout
Reference
Toxicol Appl
Pharmacol
164(l):55-64, 2000.
Toxicol Appl
Pharmacol
139(2):262-271, 1996.
Toxicol Sci
45(1}:33^1, 1998.
J Toxicol Environ
Health A.
57(5):357-368, 1999.
Int Arch Occup
Environ Health
38(2):87-102, 1976.
Int Arch Occup
Environ Health
42(3-4):283-292,
1979.
Arch Toxicol
32(4):283-295, 1974.
Arch Toxicol
33(3):173-189, 1975.
J Pharmacol Exp Ther
85:289, 1945.
Toxicol Sci
56(1):6 1-72, 2000.
Toxicol Sci
56(l):61-72, 2000.
Toxicol Appl
Pharmacol
79(3):389^tOO, 1985.
Toxicol Appl
Pharmacol
79(3):389^»00, 1985.
Species
Rats-Male SD
Rats-Male SD
Mice-Male B6C3F1
Rat-Male F344
Humans
Humans
Humans-Male
Humans-Male
Humans
Humans
Rats-Male F344
Mice-Male B6C3F1
and Swiss Webster
Rats-Male Osborne-
Mendel and
AlderleyPark
Wistar
Exposure
scenario
TCE stomach
injection
TCE arterial,
venous, portal,
stomach
injections
CH,TCEiv
CH,TCOHiv
TCE inhalation
TCE inhalation
TCE inhalation;
CH.TCA,
TCOH oral
TCE inhalation
TCAiv
TCE dermal
TCE dermal
TCE gavage
TCE gavage
Measurements
TCE in arterial blood
TCE in arterial blood
TCA, CH, TCOH in
blood
CH, TCOH in blood;
TCOG,CH,TCAin
bile
TCE in breath; TCA
and TCOH in blood
and urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCE in blood and
breath; TCA and
TCOH in blood and
urine
TCA in blood and
urine
TCE in breath
TCE in chamber
TCE and metabolites in
blood; 14C elimination
TCE and metabolites in
blood; 14C elimination
35
-------
APPENDIX A: CANDIDATE STUDIES FOR MODEL EVALUATION
(continued)
Lead author
S.A. Saghir
A. Sato
I.E. Simmons
RX>. Stenner
R.D. Stewart
M.V. Templin
M.V. Templin
K.D, Thrall
Reference
Environ Health
Perspect 1 10:757-763,
2002.
BrJIndMed
34(l):56-63, 1977.
Toxicol Sci
69(1):3-15, 2002.
Drug Metab Dispos
25(5):529-535, 1997.
Arch Environ Health
20(1):64-71, 1970.
Toxicol Appl
Pharmacol 23(l):l-8,
1993.
J Toxicol Environ
Health 44(4):435^147,
1995.
J Toxicol Environ
Health A.
59{8):653-670,2000.
Species
Rats-Male F344
Humans-Male
Rats-Male long-
evans
Rats-Male F344
Humans
Mice
Rats-Male F344
Rats-Male F344
Exposure
scenario
DCA iv, oral
TCE inhalation
TCE inhalation
TCOH, TCA iv;
TCE
intraduodenal
TCE inhalation
TCE oral
(aqueous)
TCE oral
(Tween 80
solution)
TCEiv
Measurements
DCA in plasma
TCE 'in blood, expired
air; TCA in urine
TCE in chamber, liver,
blood, brain, fat
TCA, TCOH in blood
Exhaled TCE; TCA,
TCOH in urine
TCE and metabolites in
blood
TCE in blood; TCA
and TCOH in blood
and bile
TCE in breath,
chamber
36
-------
APPENDIX B: COMPUTER IMPLEMENTATION
/
The general modeling strategy described above will be implemented in different
languages in two separate modeling efforts. This activity provides a quality control check on the
modeling software and the coding of the model so that outputs of the two models given the same
input should be similar. Divergence of the two model outputs would indicate improper coding in
at least one of the models. Since the two languages that are being used differ in syntax and in
how the model code is structured, it is unlikely that a coding error would be made in both
programs that would be similar enough to go undetected (i.e., both program outputs would be the
same). Moreover, the capabilities of the two modeling languages differ, and features unique to
each program add to the overall ability to develop and test the model. The two models are
described below.
Model 1: Exposure Related Dose Estimating Model
The U.S. Environmental Protection Agency's (EPA's) National Exposure Research
Laboratory (NERL) has developed the Exposure Related Dose Estimating Model (ERDEM) as a
platform for the application of physiologically based pharmacokinetic (PBPK) and
PBPK/pharmacodynamic (PD) models. The heart of ERDEM
(http://www.epa.gov/heasdweb/erdem/erdem.htm) is a PBPK model that simulates the
absorption, distribution, metabolism, and elimination of chemicals in mammalian systems.
Simulated chemicals are introduced into the physiological system by any of several
routes, including injection, ingestion, inhalation, and/or dermal absorption. The ERDEM system
contains a large set of potential compartments and processes, with over 30 physiological
compartments, such as arterial and venous blood, brain, skin (surface and dermis), fat, kidney,
liver, rapidly and slowly perfused tissues, lung, stomach, and intestine. Any given model is
derived by selecting the compartments and processes that are most applicable to the kinetics of
the chemical(s) and endpoint of interest. ERDEM is programmed using the Advanced
Continuous Simulation Language (ACSL). Model-specific parameter values are entered into
ERDEM based on the physiological, biological, and biochemical modeling data specific to the
chemical and/or scenario of interest. Any PBPK model, including ERDEM, is made up of a
series of the differential equations that describe the rates of inflow, distribution, metabolism, or
outflow of a chemical and various metabolites in each separate biological compartment.
ERDEM consists of an ACSL-based model engine and a power builder front end. Both
of these components will be made available to the public as executables from EPA's Office of
Research and Development (ORD)-NERL. No special software is required. An ACSL software
license is needed to recompile the code and cannot be provided by EPA. However ERDEM
37
-------
should require no additional recompilation of code to run the model as described in the
document
Model 2: PBPK Development Using MCSim Language
A second model is being developed in the MCSim language. MCSim is an open-source
statistical modeling package initially developed by Frederic Bois and others for the application of
modern Monte Carlo statistical methods in complex nonlinear models. Since MCSim includes a
sublanguage for describing dynamic models in terms of their component differential equations
and typical time-varying inputs, it has been particularly valuable in the application of Markov
Chain Monte Carlo methods to estimating Bayesian posterior distributions for parameters of
PBPK models.
Dynamic models in MCSim are written in an algebraic language. Model specification
includes predefining all the parameters for the model, declaring all the variables whose dynamics
are governed by differential equations, declaring all the variables whose values need to be output,
specifying input variables whose values will be determined by special functions that provide for
periodic or episodic inputs, and specifying the differential equations for the model. This model
specification file is translated by the MCSim software into the C programming language. Then it
is compiled and linked to libraries that provide routines for integrating the differential equation
system, carrying out the required Monte Carlo simulations, and doing the input and output
functions. The resulting executable file is then run with specially formatted input files that can
change parameter values and specify the nature of the desired simulation, whether it is a
numerical integration of the differential equation system, a Monte Carlo simulation of parameter
variability or uncertainty, or a Markov Chain Monte Carlo estimate of Bayesian posterior
distributions for model parameters.
MCSim models are portable at several levels. At the lowest level, since MCSim itself is
open source, and since open-source C-language compilers are available for almost all computing
platforms (e.g., UNIX, Microsoft Windows, and Apple OS-X), models can be distributed as
model source and recompiled and run with little additional cost to reviewers. Compiled models
are also executable files and can be run without any additional software (although the executables
are specific to particular operating systems and computing hardware). Thus, the compiled
models can be distributed and their behavior evaluated without the installation of additional
software.
38
------- |