vvEPA
United States
Environmental Protection
Agency
Office of Health and
Environmental Assessment
Washington, DC 20460
EPA/600/3-90/019
August 1989
Research and Development
Biological Data for
Pharmacokinetic
Modeling and Risk
Assessment
Report of a Workshop
Convened by the U.S.
Environmental Protection
Agency and I LSI Risk
Science Institute
-------
-------
EPA/600/3-90/019
August 1989
BIOLOGICAL DATA FOR PHARMACOKINETIC MODELING AND
RISK ASSESSMENT
Report of a Workshop Convened by the U.S. Environmental Protection Agency
and ILSI Risk Science Institute
Asheville, North Carolina
May 23-25, 1988
Sponsored by:
ILSI Risk Science Institute
U.S. Environmental Protection Agency
The Dow Chemical Company
American Industrial Health Council
American Petroleum Institute
Mobil Oil Company
The Procter and Gamble Company
Shell Oil Company
Report Prepared by:
Eastern Research Group, Inc.
6 Whittemore Street
Arlington, MA 02174
Contract No. CX-814628-02-2
-------
DISCLAIMER
The views expressed in this report are those of the authors and do not necessarily reflect the views
or policies of the U.S. Environmental Protection Agency or other sponsors. Mention of trade names or
commercial products does not constitute endorsement by the Agency or recommendation for use.
ii
-------
TABLE OF CONTENTS
PREFACE ,vi
ACKNOWLEDGEMENTS vii
ABSTRACT ~-i viii
INTRODUCTION 1
SESSION SUMMARIES 5
Session I: Overview of Risk Assessment 6
Dr. William Farland, U.S. Environmental Protection Agency, Session Chair
Dr. Hugh Spitzer, American Petroleum Institute, Session Summary
Session II: Modeling Biological Phenomena 9
Dr. Kenneth B. Bischoff, University of Delaware, Session Chair and Session Summary
Session III: Overview of Pharmacokinetic Models 11
Dr. Robert L. Dedrick, National Institute of Health, Session Chair
Dr. Kenneth B. Bischoff, University of Delaware, Session Summary
Session IV: The Use of Experimental Data in Pharmacokinetic Modeling 14'
Dr. Richard W. D'Souza, Miami Valley Labs, The Procter and Gamble Company,
Session Chair and Session Summary
Session V: The Use of Pharmacokinetic Modeling in Risk Assessment: Case Studies 17
Dr. Linda Birnbaum, National Institute of Environmental Health Sciences,
Session Chair and Session Summary
CASE STUDY COMMENTS 23
Benzene, Dr. Richard Irons, University of Colorado 24
Butadiene. Dr. Richard Irons, University of Colorado 29
Methvlene Chloride. Dr. Rory B. Conolly, Northrop Services, Inc 32
RAPPORTEUR REPORTS 36
Dr. Angelo Turturro, National Center for Toxicological Research 37
Dr. Hugh L. Spitzer, American Petroleum Institute 43
CONCLUSION 47
Dr. Kenneth B. Bischoff, University of Delaware, and
Dr. Irene B. Glowinski, ILSI Risk Science Institute
APPENDIX A: AGENDA A-l
APPENDIX B: ABSTRACTS OF PRESENTATIONS B-l
in
-------
SESSION I: OVERVIEW OF RISK ASSESSMENT B-2
Concepts: Academic or Theoretical Discussion of Risk Assessment B-2
Dr. Nicolas A. Ashford, Massachusetts Institute of Technology
Use of Data in Risk Assessment
Dr. Joseph V. Rodricks, ENVIRON Corporation
The Implications of Using More Complicated Models or Risk Estimation
in Carcinogenesis
B-3
B-5
Dr. Christopher J. Portier, National Institute of Environmental Health Sciences
SESSION II: MODELING BIOLOGICAL PHENOMENA B-7
Introduction to Modeling . . B -
Dr. Mary Davis, West Virginia University Medical Center
SESSION III: OVERVIEW OF PHARMACOKINETICS MODELS B-10
Development and History of Pharmacokinetic Modeling B-10
Dr. Kenneth Bischoff, University of Delaware
What Models Can Do and Their Limitations B-12
Dr. Robert Dedrick, National Institute of Health
Data Needs for Pharmacokinetic Modeling in Risk Assessment B-15
Dr. Harvey Clewell III, Wright Patterson Air Force Base
Major Uncertainties in Pharmacokinetic Modeling and Sensitivity Analysis . . . B-18
Dr. Murray Conn, Consumer Product Safety Commission
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC
MODELING B-20
Hepatic Metabolism
Dr. Marilyn Morris, State University of New York
Cross-species Scaling
Dr. Joyce Mordenti, Genentech
Biotransformation of Environmental Chemicals: Development of In Vitro Svstei
to Predict In Vivo Events
Dr. I. Glenn Sipes, University of Arizona
B-20
B-24
B-32
Target Tissue/Cell Dose of Chemical Carcinogens B-35
Dr. George Lucier, National Institute of Environmental Health Sciences
IV
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK
ASSESSMENT: CASE STUDIES B-38
Benzene Pharmacokinetic Model: Case Study B-38
Dr. Michelle Medinsky, Lovelace ITRI
Implications of Pharmacokinetic Data and Models in a Risk Assessment of
Benzene B-44
Dr. A. John Bailer, National Institute of Environmental Health Sciences
The Use of Pharmacokinetic Analysis in Risk Assessment - The Case of
Butadiene B'46
Dr. Dale Hattis, Massachusetts Institute of Technology
Cancer Risk Assessment of 1.3-Butadiene B-53
Dr. Steven Bayard, U.S. Environmental Protection Agency
Methvlene Chloride Model: Inhalation Data B-56
Dr. Richard Reitz, The Dow Chemical Company
Methvlene Chloride Model: Ingestion Data B-58
Dr. Michael Angelo, Merck and Company
The Impact of Pharmacokinetics on the Risk Assessment of
Dichloromethane B-60
Dr. Jerry Blancato, U.S. Environmental Protection Agency
APPENDIX C: LIST OF PARTICIPANTS C-l
-------
PREFACE
On May 23-25, 1988, the U.S. Environmental Protection Agency (EPA) and the ILSI Risk Science
Institute held a workshop in Asheville, North Carolina, titled "Biological Data for Pharmacokinetic
Modeling and Risk Assessment." The goal of this workshop was to encourage a better understanding
between the communities in toxicology, risk assessment, biological modeling, and pharmacokinetics. The
workshop was divided into five sessions, each covering a specific topic related to pharmacokinetics.
Within each session, speakers from academia, industry, and government presented their expertise. At the
conclusion of the workshop, participants identified research and data needs for pharmacokinetic models
in risk assessment. This report summarizes the proceedings of the workshop.
VI
-------
ACKNOWLEDGEMENTS
This document was prepared by Eastern Research Group, Inc., Arlington, Massachusetts, based on
submissions by speakers, commentors, and rapporteurs of the workshop. It was reviewed by Lawrence
Fishbein and Carol J. Henry of the Risk Science Institute and all workshop chairpeople and rapporteurs.
Their time and contributions are gratefully acknowledged.
Vll
-------
ABSTRACT
This report summarizes the information presented at the "Workshop on Biological Data for
Pharmacokinetic Modeling and Risk Assessment" held by EPA and the Risk Science Institute on May
23-25, 1988, in Asheville, North Carolina. This report provides a general background of risk assessment
and reviews how biological data are used in risk assessment and the ways pharmacokinetic modeling can
reduce the uncertainties in risk assessment. Different biological models and their value are described,
and the differences between predictive and descriptive models are explained. The report reviews the
development and history of pharmacokinetic models, the strengths and limitations of the models, the
types of biological data necessary to use pharmacokinetic models, and the role of sensitivity analysis in
incorporating pharmacokinetic data into risk assessment. Hepatic metabolism data is used as an example
of incorporating in vitro and in vivo data into pharmacokinetic models. Two approaches used for cross-
species scaling are included, as well as biotransformation and the correlation of in vitro and in vivo data.
Actual case studies showing pharmacokinetic modeling used in risk assessment are also described, and
research needs are identified.
viii
-------
INTRODUCTION
-------
The Workshop on Biological Data for Pharmacokinetic Modeling and Risk Assessment was held
May 23-25, 1988, in Asheville, North Carolina. The overall goal of the workshop was to encourage a
better understanding among the research communities in toxicology, risk assessment, biological modeling,
and pharmacokinetics and to encourage a greater understanding of the needs of each group.
Most research communities recognize that there are a number of critical uncertainties inherent in
the major assumptions underlying current risk assessments that utilize modeling techniques based on
animal bioassays. The principal uncertainties relate to the necessity of extrapolating experimental results
across species from rats or mice to humans, from high dose regions (laboratory animal exposure) to low
dose regions (human exposure in the environment), and across routes of administration. Also, current
risk assessment methodology provides no insight for determining which dosage or active material should
be modeled. In addition, risk assessors are not assured that modeling administered dosage data has any
direct relationship to delivered or target dose either for the test compound or an active metabolite.
Hence, there is an overwhelming need for an evaluation of the scientific base for the assumptions (and
underlying uncertainties as described above) used in the risk assessment process. Pharmacokinetics
provides such a tool.
The goals of the Workshop on Biological Data for Pharmacokinetic Modeling and Risk Assessment
were:
To communicate to biologists the value of pharmacokinetic models and the kinds of data
needed.
To provide insight and guidance to risk assessors with respect to the rigor of analysis.
To focus on making the uncertainties in modeling more explicit.
To identify research/data needs for pharmacokinetic models in risk assessment.
The workshop focused on five themes. The first session, "Overview of Risk Assessment," provided a
general background of risk assessment and reviewed how biological data are used in risk assessment and
the ways pharmacokinetic modeling can reduce the uncertainties in risk assessment. The second session,
"Modeling Biological Phenomena," was designed to acquaint biologists, biochemists, pharmacologists, and
lexicologists with different biological models and their value. The use of models in biology and the
differences between the predictive and descriptive models were explained. Session Three, "Overview of
Pharmacokinetic Modeling," described the development and history of pharmacokinetic models and
included discussions on the strengths and limitations of pharmacokinetic models, the types of biological
-------
data necessary for these models, and the role of sensitivity analysis in incorporating pharmacokinetic data
into the risk assessment process. The fourth session, "The Use of Experimental Data in Pharmacokinetic
Modeling," used hepatic metabolism data as an example of incorporating in vitro and in vivo data into
pharmacokinetic models. Two approaches used for cross-species scaling were reviewed, as well as
biotransformation and the correlation of in vitro and in vivo data. The final presentation in this session
was a review of molecular dosimetry.
The last session presented actual case studies showing how pharmacokinetic modeling has been used
in risk assessment. These case studies described modeling information on three chemicals shown to be
carcinogenic in rodents: benzene (also a known human leukemogen), butadiene, and methylene chloride.
Both an inhalation model and an ingestion model were described for methylene chloride. These
presentations were followed by explanations of how such data have been used in risk assessments.
Benzene risk assessments used data from oral carcinogenicity studies conducted by the National
Toxicology Program. The butadiene presentation provided a modification of the 1985 U.S.
Environmental Protection Agency (EPA) risk assessment for butadiene, which incorporates new
experimental and epidemiological data. The presentation on methylene chloride described the impact of
pharmacokinetic models on risk assessment.
This document contains summaries of each workshop session that provide comprehensive overviews
of the capabilities and problems of using physiologically based pharmacokinetic (PBPK) modeling at the
present time. The details of how to actually do this are given in the abstracts and the listed references.
Reports from two rapporteurs focus on the chemical-specific needs for PBPK models and the application
of pharmacokinetic data in resolving or reducing the uncertainty in risk assessment.
The concluding section of the document outlines research needs for pharmacokinetic modeling
identified by the participants that will help decrease the scientific uncertainties in risk assessment.
The workshop was organized and supported by a cooperative agreement between the ILSI Risk
Science Institute and the EPA The Workshop Planning Committee consisted of Dr. Kenneth B.
Bischoff, University of Delaware (Workshop Chairperson); Dr. Linda Birnbaum, National Institute for
Environmental Health Sciences; Dr. Richard D'Souza, The Procter and Gamble Company; Dr. Irene B.
Glowinski, ILSI Risk Science Institute; Dr. Dale Hattis, Massachusetts Institute of Technology; Dr. Stan
Lindstedt, University of Wyoming; Dr. Hugh Spitzer, American Petroleum Institute; and Dr. Catherine
St. Hilaire, ENVIRON Corporation.
-------
We wish to gratefully acknowledge additional financial support from the following organizations
which made possible the attendance of selected presenters and attendees: The Dow Chemical Company,
American Industrial Health Council, American Petroleum Institute, Mobil Oil Company, The Procter
and Gamble Company, and Shell Oil Company.
-------
SESSION SUMMARIES
-------
SESSION I: OVERVIEW OF RISK ASSESSMENT
William Farland, Session Chair
Hugh Spitzer, Session Summary
Dr. William Farland opened this session by reviewing the current risk assessment methodology used
by the U.S. Environmental Protection Agency (EPA) and by explaining how the physiologically based
pharmacokinetic (PBPK) modeling can fit into this methodology. He pointed out that there is an
increasing interest in developing better biologically based models for assessing risk in an effort to address
many of the assumptions inherent in the mathematical models. PBPK models are critical in this effort
since they provide the methodology to assess metabolite distribution, differences in metabolism rates
between organs and across species, and better estimates of dose-response. In addition, the use of PBPK
in risk assessment can have a significant impact on exposure assessment and dose-to-target issues. Dr.
Farland noted that EPA has already used PBPK data in assessing risk for some halogenated compounds,
thus demonstrating a willingness to use such information in the risk assessment process.
Dr. Nicholas Ashford reviewed the risk assessment process within a legal/political context. He
reflected on the risk assessor's obligation to the risk manager, stressing that hidden values that can
possibly govern the selection of the type of models and data sets used in assessing a particular chemical.
Because of these hidden values, the bottom line in risk assessment should not be a number but a
distribution of possibilities. Dr. Ashford noted that, for high exposures to a substance, the decisions and
methods for control are relatively simple. However, for low levels of exposure, the risk assessment must
take into account a great many uncertainties and thus becomes very complex. The risk assessor should
give the risk manager a sense of the uncertainties and explain the impact of the uncertainties on the
final estimates.
Dr. Ashford elaborated further on the risk assessment process, stating the need for a uniform
intellectual approach with its demand for rigor, not a uniform set of formulae. He also emphasized the
need for balance, objectivity, and a professional review process, i.e., a review conducted by various types
of experts. In closing, Dr. Ashford touched on the social issues involved in the regulatory process and
discussed how society must deal with uncertainty. He pointed out that the key question for the risk
manager is: "How much can I afford to be wrong?"
Dr. Joseph Rodricks discussed the practice of risk assessment. He pointed out that, in the absence
of perfect science, efforts must be made to improve the methodologies. Without perfect data or a "best"
-------
answer, science policy guidance must use all available data. Dr. Rodricks stressed that the risk assessor
has an obligation to use all information and to clearly communicate the uncertainties.
Dr. Rodricks then reviewed current risk assessment practices, pointing out their advantages and
disadvantages. The current practices suggest the use of: (1) a generic method for all chemicals; (2) a
compound-specific approach; or (3) a combination of the generic and compound-specific methodologies.
Because of the large number of chemicals to assess, the generic approach has the advantages of avoiding
complexity and promoting consistency. However, this simplification discourages the use of chemical-
specific data as well as the development of novel or alternative assessment methodologies. Perhaps more
important, however, the generic approach tends to obscure the uncertainties in the analysis.
Dr. Rodricks then turned his attention to new approaches and the burden these new or alternative
approaches place on the risk assessor. He pointed out the need to keep pressing the research
community for new and better data for risk assessment. He also stressed that the risk assessor had to
more clearly explain to the research community the assumption/hypothesis to be addressed by the
research and the type of data needed. In using alternative approaches, the risk assessor must understand
that the selection of assumptions and models, i.e., particular data sets, or the decision to use or not use
information on mechanism of action, is not based on science but on science policy. Thus, as the risk
assessor uses particular pieces of information, he or she must describe the reasons for the selection and
the uncertainties that accompany it. The analysis, therefore, can be very complex for certain chemicals.
In using this approach, the risk assessor puts an enormous burden on the risk manager. Because of the
uncertainties involved, the risk manager must decide how much uncertainty is tolerable in any given risk
assessment.
Dr. Rodricks ended his talk by pointing out the three critical uncertainties/limitations in the use of
PBPK modeling that the risk assessor must address. First, the risk assessor must determine if the
correct metabolite or surrogate was selected for the chemical in the risk assessment. Second, the risk
assessor must recognize the limitations on the use of the data.caused by the differences in dosing pattern
(e.g., acute exposure for PBPK modeling vs. chronic dosing in the bioassay). Finally, the risk assessor
must determine if the appropriate interspecies scaling factors (translation from one species to another)
were used in the extrapolations.
Dr. Portier ended the session by reviewing the mathematical models used in carcinogenic risk
assessment and the uncertainties of each model. He pointed out that currently there are three levels of
modeling: (1) data interpretation; (2) tumor incidence/dose-response; and (3) species conversion. He
-------
discussed some of the questions the biostatistician addresses in the analysis of data, describing the
various uncertainties and how each can be approached mathematically. He then turned his attention to
the use of PBPK modeling to improve the accuracy and, therefore, reduce the uncertainty of the risk
questions regarding variability. He explained that first, the investigator must learn how to deal with
individual vs. population variability; and second, the investigator must be aware that since dose is not
known without error, variability must be estimated as a parameter. Dr. Portier ended his talk by
pointing out the number of assumptions and, therefore, sources of variability in the EPA's use of PBPK
models using the Andersen approach to assessing methylene chloride.
In the discussion that followed the formal talks, Dr. Richard D'Souza asked the panel to discuss the
issue of uncertainty vs. variability. The panel members agreed that it was an issue of concern and that
as uncertainty decreased, variability would increase.
-------
SESSION II: MODELING BIOLOGICAL PHENOMENA
Kenneth B. Bischoff, Session Chair and Session Summary
Dr. Mary Davis presented an introduction to the uses of modeling in biology to: (1) summarize
data; and (2) attempt to relate that data to the1 theoretical processes proposed in the model (i.e.,
elimination by renal clearance vs. hepatic metabolism). Models can be quantitative or qualitative, and
can include physical, chemical, and biological processes.
Pharmacokinetics uses two main classes of models: predictive and descriptive. The latter are
classical compartmental models and are primarily associated with blood or plasma concentrations and
elimination. In these models, several compartments (one of which includes plasma) are postulated, and
the compartment volumes (of distribution) and intercompartment fluxes are determined to fit the
observed (plasma) data. Parameter estimation techniques,are important in defining and using these
models. A descriptive model is able to show the effects of different dosage regimens, steady-state levels,
etc.; to relate elimination information; and to detect differences in the handling of compounds.
However, the model is limited because the compartments have no physiologic meaning, and the kinetic
constants are often hybrids of several processes. Therefore; it is difficult to extrapolate classical
compartmental models across species, sex, or age.
Predictive models are physiologically based pharmacokinetic (PBPK) models jthat utilize organ
volumes to establish compartments, blood flows and clearance rates, protein binding, diffusion and
permeabilities, and enzyme kinetics. Certain organs are also utilized because they are sites of toxicity,
metabolism, elimination, or storage. From this information, mass balance equations are derived and
simultaneously solved.
Extending the PBPK philosophy to pharmacodynamics (drug effects) can account for biological
responses such as DNA repair and macromolecular turnover. Some parameter estimation is involved,
but if a PBPK model does not describe the data well, it must be refined by adding parameters that were
presumably left out. Predictive models are able to scale between species; to extrapolate to low or high
doses and different routes of exposure; and to accommodate differences in organ function and size, and
metabolic and clearance processes from .disease or growth. The main disadvantages of this model type
are that a large variety of data and knowledge of the underlying processes are needed, and the models
usually must be solved numerically using computers.
-------
It is useful to realize that models can be used in an "investigative mode," as well as a "simulation
mode." The latter mode uses model solutions to generate predicted tissue drug levels as a function of
exposure and then uses these levels to manipulate dosage regimens to achieve some desired result. This
mode is the most common interpretation of "using a model." The investigative mode uses the model to
quantitatively analyze the data in terms of the proposed processes and focuses on better understanding
those processes. In this method, a whole-body model may not even be necessary. Examples of both
modes can be found in the literature (see Appendix B).
10
-------
SESSION III: OVERVIEW OF PHARMACOKINETIG MODELS
Robert L. Dedrick, Session Chair
., . Kenneth B. Bischoff, Session Summary
Dr. Kenneth Bischoff began this session by reviewing the development and history of
pharmacokinetic models. The purpose of pharmacokinetic models is to quantitate by mathematical
models absorption, disposition, metabolism, excretion, and biological responses to drugs and chemical
agents for clinical applications and toxicology and risk assessment. Dr. Bischoff presented examples of
classical compartmental models as well as the more recent, physiologically based pharmacokinetic
(PBPK) models.
Early work on pharmacokinetic models focused on the use of anesthetic agents to predict brain
levels. Ethanol also was studied by using Michaelis-Menten kinetics for hepatic metabolism. The
concepts of clearance and volume of distribution in pharmacokinetics were defined and applied. The
term "pharmacokinetics" is derived from "pharmacokinetik," a term (in German) apparently coined by F.
Dost (1953).
The PBPK approach attempts to base the models on real biological information concerning the
physicochemical, physiological, and pharmacological properties of the drug, rather than using, for
example, abstract definitions of central and peripheral compartments. Organ blood flow rates,
solubilities, protein binding, and local membrane permeability are all specifically incorporated into the
models. This, information permits the model to translate results between animal species and predict
effects of different routes of administration and different doses on the tissue levels of the chemical.
<
Dr. Bischoff described several specific examples of the application of this approach to thiopental,
methotrexate, and other drugs and provided references to both the history and review articles on PBPK.
Dr. Robert Dedrick reviewed the strengths and limitations of pharmacokinetic models. Many forms
of pharmacokinetic models have been extensively used in pharmacology for the design and analysis of
preclinical and clinical trials, as well as for guidance of therapy. Although classical compartmental,
physiological, and specific model-independent approaches all have important applications, Dr. Dedrick
focused on the mechanistic insight that can be obtained from the physiologic models.
11
-------
Concerning in vivo-an vitro correlations, the benefit of using systems ranging from isolated enzymes
to perfused organs is the uncoupling of the local effects from the complexities of chemical and
metabolite distribution and interactions in the intact animal. Although published examples are
supportive of the concept and predictability of the events of interest, no comprehensive methodology
exists.
The extensive literature on animal anatomy and physiology can be used as a basis for physiologic
pharmacokinetic models. This information provides many model parameter values (e.g., organ blood
flow) in different species, often through empirical allometric correlations with body weight. However,
other aspects, such as xenobiotic metabolism, can have large and unpredictable variations between
species, and experimental data will be required for a particular chemical and animal.
The use of pharmacokinetic models for determining bioavailability and dosing schedules is well
established. Application of the design to regional drug administration has occurred to some extent,
focusing on obtaining some pharmacokinetic advantage over systemic administration. Diffusion,
convection, and mixing often need to be considered in more detail than usual to truly optimize the
advantages.
Pharmacodynamics (drug effects) as well as pharmacokinetics must be considered in order to
associate drug concentrations with a biological effect. The paucity of biologically based
pharmacodynamic models limits their use in quantitative risk assessments since pharmacokinetics alone
tells nothing about the probability Of a chemical causing, for example, cancer in humans, or even about
the appropriate dose metric necessary for predicting human cancer risk based on animal experiments.
Dr. Harvey Clewell's presentation focused on data needs for modeling. The purpose of using
models in risk assessment is to increase the scientific basis and to reduce or better quantify the levels of
uncertainty. Models are helpful for designing bioassays and should be used in initial experiments.
Qualitative information (e.g., mechanism of action) must be gathered to define the model structure (i.e.,
the physiological and biochemical basis for defining the quantitative relations), and quantitative data are
needed to define the model parameters.
When using biologically based mathematical models in risk assessment, pharmacokinetics needs to
be coupled with pharmacodynamics (toxic endpoints, e.g., carcinogenesis). For the latter, information
such as cell turnover rates, mutation, and DNA repair rates must be known in humans and in animals.
There are seven aspects required for developing the risk assessment: (1) classification of the chemical as
12
-------
a carcinogen; (2) mechanism of carcinogenesis; (3) nature of the proximate carcinogen; (4) model
structure; (5) model parameters; (6) species differences; and (7) extrapolation to low doses. Only the
model parameters, and to some extent, the model structure have been extensively treated. The first four
aspects of risk assessment often are decided on rather ill-defined bases; the extrapolation to low doses is,
by U.S. Environmental Protection Agency policy, handled by the linearized multistage model, and species
differences are a refuge for empirical "safety factors."
There are two extremes in risk assessment: (1) the "safe" risk, which avoids underestimating by
using conservative methods that do not rapidly change over a period of years; and (2) the "accurate" risk
which avoids erroneous risk management decisions by using new techniques, avoiding conservative safety
factors, etc. These two approaches need to be brought together.
Dr. Clewell recommended that: (1) models should be used in the development of bioassays; (2)
legislative agencies should have the support of modeling groups; (3) basic data should be collected and
organized; and (4) formal methodologies should be used to incorporate uncertainty into the quantitative
outcome.
Dr. Murray Conn focused on one role that sensitivity analysis can play when incorporating
pharmacokinetic data into the risk assessment process. The well-studied example of methylene chloride
can exemplify this role. Major uncertainties concern high-to-low dose extrapolation and species-to-
species extrapolation. Some specifics include assumptions about mechanism of action, e.g., the roles of
the parent compound and the possible reactive metabolic intermediates. Other problems can be with
possible nonunique parameter estimates for the models and the effect on predictions from animal data.
Another issue in the risk assessment process can be the methods for scaling to humans, especially for
the metabolic parameters.
Some uncertainty is reduced or better defined through incorporating pharmacokinetic considerations
into dose-response analysis. Also, the sensitivity analysis can provide a basis for corrections to the above
results. It is not clear how the uncertainties can be reduced compared to the traditional methods
utilizing extrapolation factors of "mg/kg/day" or "mg/m2/day." Finally, the ill-defined question "What is
the "target dose?" also is an issue in the risk assessment process.
13
-------
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC MODELING
Richard W. D'Souza, Session Chair and Session Summary
Dr. Marilyn Morris presented an overview of hepatic metabolism. Enzyme activity (VmM, K^) in the
metabolizing organ is an important determinant of the rate of metabolism of a compound. In addition
to enzyme activity, a number of factors influence metabolism rates including blood flow, cofactor
availability (e.g., thiol or sulfate), enzyme distribution patterns, and diffusion rate limitations. The
relative importance of these factors changes with the dose level of the compound, the extraction ratio,
and other variables.
Most physiological pharmacokinetic models assume a spatially uniform model for hepatic
concentration and metabolism. Other models used are the parallel-tube model (sinusoidal model),
dispersion model, and distributed model. Although the spatially uniform model is an oversimplification
and is conceptually incompatible with our knowledge of liver anatomy and physiology, it has worked well
in describing the kinetics of several drugs, for example, lidocaine, phenytoin, and propranolol. It is
important to consider these different metabolism models when developing physiological pharmacokinetic
mode'.s, to select the most appropriate.
Enzyme activity parameters obtained in vitro may be misleading, as they may not be sensitive to the
same factors as those obtained in vivo. Some in vivo validation, therefore, is necessary.
One in vitro system, the intact perfused-liver preparation, may offer definite advantages over other
in vitro systems (e.g., microsomal preparations) as it maintains liver architecture and can control
perfusion rate to simulate in vivo blood flow rate. Also, the effects of various factors mentioned above
can be studied by appropriately perturbing the system.
Dr. Joyce Mordenti reviewed two approaches to cross-species scaling (translation from one species
to another): the allometric approach and the physiological pharmacokinetic approach. Both approaches
have some advantages and some limitations. The allometric approach is a body-weight power
relationship that typically uses data from the literature, making this approach quick and inexpensive.
Because of the ease and simplicity of use, this model is valuable and often recommended for a first
approximation. The allometric approach is well established for scaling anatomic/physiological parameters
(e.g., volumes, flow rates) and also for scaling physiological processes (e.g., renal clearance). This
approach has had some success in scaling first-order metabolism rates, except for oxidation (e.g.,
14
-------
antipyrene), where humans may metabolize more slowly than was predicted from a single-term (body-
weight) power function. In this case, a two-variable power function, for example, body-weight and brain-
weight, can be used. The allometric approach is empirical and not well understood, so there is little
recourse when it fails. It is recommended that pharmacokinetic data from different species first be
plotted on a log-log grid before proceeding to other methods.
Physiological scaling is accomplished by adjusting for important differences in various model
parameters between species. It is, therefore, more realistic biologically than allometry and, thus, is the
recommended method. However, it can be time consuming and expensive when obtaining or validating
parameters experimentally. Hybrid allometric-physiological scaling can be performed and has had both
successes (e.g., with phenobarbital, phenytoin) and failures (diazepam).
Dr. Glenn Sipes discussed biotransformation and the development of in vitro systems to predict in
vivo events. In vitro systems provide the opportunity to study metabolism of xenobiotics in humans.
Several systems of varying scope and complexity have been used including subcellular fractions, cell
suspensions, and precision-cut liver slices.
In vitro precision-cut liver slices may have the potential to mimic in vivo metabolism. This system
has been tested for viability with protein synthesis and glucose release and can perform integrated Phase
I (e.g., oxidation, de-ethylation) and Phase II (e.g., glucuronidation, sulfation) metabolism. Data from
this system have also been consistent with in vivo literature data (e.g., sulfate conjugation in the rat,
slow and fast human acetylators, etc.) and have demonstrated potential for studying enzyme inhibition,
cellular damage, and other processes. Due to a lack of commercially available instrumentation, only one
laboratory currently has the capability to perform these studies. The system is still in the developmental
stage and other factors such as glutathione conjugation will be studied in the near future.
Data from the various in vitro systems are not expected to give exactly the same results (e.g., for
kinetic constants), nor are they expected to have a one-to-one quantitative relationship with in vivo data.
One in vitro-to-in vivo correlation compares the results of different in vitro studies with in vivo data in
animals. Once a data base is broadly established, it may be possible to correctly predict in vivo
metabolism rate constants in humans from in vitro data in humans. Furthermore, metabolism and
toxicity parameters could be directly scaled from animal to human, based solely on in vitro data. Such a
data base would be very useful for developing physiological models.
15
-------
The final presentation in this session was a review of molecular dosimetry by Dr. George Lucier.
Dr. Lucier explained that molecular dosimetry (using DNA and other adducts) can be used effectively
with physiological pharmacokinetic models to assemble a more comprehensive picture of the events
taking place from a carcinogen at the site of absorption to the toxic/carcinogenic response.
DNA adducts must not be studied in isolation; the complex processes that include metabolic
activation/deactivation, DNA repair and persistence, and the concentration of chemicals at the receptors
must be understood to correctly interpret DNA adduct data. Besides specific organs and tissues, specific
cell types should be studied for DNA adduct formation, repair, etc. In humans, adducts can be studied
in accessible fluids to gain information on precarcinogenic events at nonaccessible sites. For example,
hemoglobin and lymphocyte adducts can be quantitated. Several methods can be used for measuring
adducts, including immunoassays on monoclonal and polyclonal antibodies, 32P post-labeling, and in vitro
lymphocyte uptake.
16
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
Linda Birnbaum, Session Chair and Session Summary
Development of pharmacokinetic models can lead to improved dose-setting and study design. Some
model predictions can also lead to new questions. An important use of pharmacokinetic models is in
the area of risk assessment, where such models can be used to extrapolate between doses, sexes, and
even species of animals. The focus of this session was to show how pharmacokinetic models (in this
case all physiological models) could be used in the assessment of risk. Three chemicals of major
industrial importance, all of which have been shown to be carcinogenic in rodents, were examined:
benzene, butadiene, and methylene chloride. The physiological model for benzene and one of the
models for methylene chloride were based upon the model for styrene developed by M. Anderson and
coworkers. The models for butadiene and one /or methylene chloride were developed independently of
the "Anderson" paradigm.
Benzene
Benzene has been known for many years to cause cancer in humans. Only recently, however, has it
been shown to be carcinogenic in rodents. Dr. Michelle Medinsky presented a pharmacokinetic model
describing the metabolism of benzene in rats and mice. Her model, based on extensive oral and
inhalation metabolism data, indicated that differences between the species were observed in the pathways
for the metabolism of benzene. She concluded that after both oral and inhalation exposures, mice form
relatively more of the hydroquinone and muconic acid metabolites - markers for the putative toxic
pathways of metabolism - than do rats. In contrast, rats form relatively more of the metabolites that
serve as markers of the detoxification pathways, phenyl and mercapturic acid conjugates, than do mice.
At higher exposure concentrations, due to the lower Kn for the toxic metabolite pathways, metabolism
shifts more toward detoxification. The relatively greater production of toxic metabolites at low doses
suggests that linear extrapolation seen at high exposure concentrations might obscure the potential
toxicity occurring at low exposure concentrations. This is a good example of the utility of a
pharmacokinetic model that incorporates physiological, chemical, and metabolite parameters. Another
good example is the demonstration by simulation that the species differences in metabolite production
are due to differences in the metabolism rates, not differences in physiological parameters such as
alveolar ventilation rate.
17
-------
Dr. John Bailer presented a risk assessment of benzene using different dosing metrics with the data
generated from the oral carcinogenicity studies conducted by the National Toxicology Program (NTP).
He computed the internal dose (ID) received by the rodents as a function of the administered dose
(AD), using the information available in the literature about levels of benzene metabolites. The
relationship of internal to AD was clearly nonlinear because there was a nonlinear increase in toxic
metabolites as a function of AD. The model used for the relationship between the ID and AD was a
Michaelis-Menten function, in which:
ID =
/3+AD
Using the multistage model of Crump and Allen, Dr. Bailer's risk assessment, derived from the NTP
gavage carcinogenicity studies, was based upon both the ID predicted from the Michaelis-Menten
function and the AD; the ID led to a safe dose less than one order of magnitude smaller than a safe
dose based upon the AD. However, the species differences were highlighted using ID as a measure of
exposure. In rats, the ID was roughly linear relative to the AD; hence, a safe dose derived from a risk
assessment based upon either metric should be roughly the same. In contrast, in mice the nonlinear
relationship between ID and AD resulted in a lower safe dose being estimated when ID was used. Of
course, measures of ID other than total metabolites could be used, one of these being the concentration
of toxic metabolites associated with some AD. If the relationship between ID and AD proved to be the
same in humans as in mice, or conversely in rats, a risk assessment for humans could be done using ID
as the metric. The estimate of risk in this case would be similar to that made using several human
epidemiological studies.
Dr. Richard Irons commented on the use of models in risk assessment of benzene. He focused on
the assumptions generally made in pharmacokinetic modeling and how they are used to extrapolate to a
given endpoint, i.e., toxicity or cancer. In his opinion, experimental data suggest that, for benzene, the
assumption that exposure is cumulative with risk is not valid, in part because the metabolic and~
interactive pathways used by that chemical are complete. Another point that should be incorporated
into modeling for benzene is the cell-cycle dependence of the toxic response: benzene affects only
dividing cells. Thus, continuous administration of a carcinogen is less effective at evoking a response
than is intermittent exposure. The recent discoveries that, at a given time, only one (or at most a few)
stem cell is populating the entire hematopoietic system mean that modeling leukemogenesis requires
attention to a very small target. The initial event in leukemogenesis may not involve mutation of a
single stem cell but may involve expansion of the number of stem cells at risk via physiological
18
-------
mechanisms such as altered growth factor responses or differentiation. Such a model would require the
incorporation of more than just metabolite levels in the target tissues or single-hit models leading to
carcinogenesis.
Butadiene
Butadiene is a potent carcinogen in mice, inducing tumors at multiple sites, especially lymphomas,
hemangiosarcomas of the heart, and lung adenomas/carcinomas, all within 1 year of exposure. Although
oncogenic in rats as well, the response is much weaker than in mice, and the neoplastic tissues are
mainly endocrine. Using a readily available computer software package, Dr. Dale Hattis developed a
physiological model to describe the pharmacokinetics of butadiene. His objective was to insert more
causal mechanistic information into the mathematical models used for risk assessment. He proposed
that the development of a physiological model would uncover anomalies between the experimental data
and the theory, leading to the development of additional data, followed by modification of the theory.
In fact, said Dr. Hattis, modeling often raises as many questions as it answers. In his pharmacokinetic
model, he attempted to reinterpret the dosages of active metabolites actually delivered to the rats and
mice during a 2-year bioassay. Given data on metabolism in humans, as well as in rats and mice, the
model allowed better predictions to be made about the dosages across species. For butadiene, the
difference in effect is nearly an order of magnitude. There was also an appreciable difference in the risk
projections derived from the experiment in rats and mice, which may have related to the role of an
endogenous retrovirus in the mice. Dr. Hattis stressed the need to test the sensitivity of the model
conclusions using reasonable alternative estimates for the parameters. He also called on experimentalists
to produce the "data that the modelers need, for example, multiple time points in disposition
experiments.
Dr. Steven Bayard presented a modification of the 1985 U.S. Environmental Protection Agency risk
assessment of butadiene; his modification incorporated new experimental as well as new epidemiology
data. He focused on the mouse-to-human extrapolation model and the comparison of cancers predicted
from this model with those actually observed in humans. Recent studies, indicating increases in
hematopoietic system cancers in humans, are consistent with the extreme sensitivity of mice to the
development of lymphomas following exposure to relatively low concentrations of butadiene. The levels
of two active butadiene metabolites, epoxide and diepoxide, which have been measured in animal models,
were incorporated into the risk assessment models. However, the lack of human metabolism data makes
this exercise quite speculative.
19
-------
Dr. Irons commented on the use of pharmacokinetic data in the risk assessment of butadiene
carcinogenicity. He felt that the differences in rat/mouse carcinogenicity by butadiene are difficult to
explain by metabolite differences alone since the spectrum of tumors varies, as does the incidence. The
qualitative metabolite profiles are also the same in both species, although there are quantitative
differences. The potent leukemogenic activity of butadiene in mice may, in fact, be related to the
presence of an endogenous retrovims, since mice that do not have this virus are much less susceptible to
butadiene-induced lymphomas (all are T-cell in origin while spontaneous lymphomas in the mice are of
B-cell origin). In fact, butadiene treatment has been shown to specifically activate a single endogenous
retrovirus. The parallels to human T-cell leukemia viruses are obvious. Those mice that do not have
the retrovirus do exhibit the typical butadiene-induced bone-marrow toxicity of the virus-carrying strain.
Therefore, while bone-marrow toxicity may be necessary for the carcinogenic response, it may not be
sufficient. Thus, modeling carcinogenesis based on toxicity may not be appropriate for this compound.
Methvlene Chloride
Methylene chloride has recently been shown to be a carcinogen in rodents following inhalation
exposure. However, exposure to the chemical in drinking water did not result in tumors. Several such
issues, including effect-of-exposure route and dose-to-target tissues, have received a great deal of
attention. Extrapolation of the rodent data to humans has been quite controversial. Dr. Richard Reitz
described a physiological model which calculates the "delivered dose" to the target organs, liver and
lungs. The model incorporated two metabolic pathways: a high-affinity, low-capacity pathway involving
mixed function oxidation and a low-affinity, high-capacity pathway involving glutathione conjugation.
Enzyme levels measured in various species were incorporated into the model, which also accounted for
different routes of exposure, both oral and inhalation. A parallelogram approach involving in vivo and
in vitro rodent (rat, mouse, hamster) results was used to extrapolate from in vitro human data for both
enzymatic pathways to the in vivo situation. The predictions of the model explained the negative
carcinogenicity results of the drinking water study as compared to the positive tumor response observed
after inhalation of methylene chloride. The model predicts that dose effects will be nonlinear and, in
this case, that extrapolation from high dose to low dose may result in overestimating risk. Development
of this physiological model, which was validated by extrapolating from one route of exposure to another
and by determining that the experimental data fit the predictions, provides a perspective for risk analysis
since it allows calculation of the dose delivered to the target tissues. It does not describe the mechanism
of toxicity for methylene chloride. Dr. Reitz stressed that a quantitative risk assessment cannot provide
precise estimates of risk, but rather can put a plausible upper bound on risk. The real risk remains
unknown and, in fact, may even be zero. The usefulness of quantitating risk is to allow rank ordering of
20
-------
hazards. Incorporation of a physiological model for dose extrapolation in risk assessment may obviate
the need for interspecies scaling.
A different physiological model for the pharmacokinetic behavior of methylene chloride was
presented by Dr. Michael Angelo. The objectives of this model were to compare ADs between bolus
gavage treatment and drinking water exposure. Pharmacokinetic information was needed to understand
the different internal disposition patterns that resulted from two different oral dosing regimens. Data
were obtained by varying the dose, route, dosing vehicle, and frequency of administration. Dose-
dependent metabolism from methylene chloride was observed, in agreement with the enzymatic data
presented by Dr. Reitz. Metabolism did not appear to be affected by repeated dosing. However, there
was an apparent species difference between rats and mice in eliminating methylene chloride from blood,
which may reflect the difference in the levels of the two metabolic pathways. The liver concentration
profile of methylene chloride was also not reflected in the blood levels, emphasizing that
pharmacokinetic data for blood may not be truly representative of distribution in a specific tissue. This
may be explained by the occurrence of membrane-limited transport between the liver and blood
compartments. The model developed was flow limited and predicted that a vehicle such as corn oil
would slow the absorption and intestinal transport of methylene chloride and thus affect absorption.
Model simulations demonstrated that a corn oil carrier had a very large effect on the distribution and
metabolism patterns of methylene chloride following high dose administration as compared to the
pharmacokinetic profile resulting from dosing in water. Thus, use of a pharmacokinetic model can help
to predict how a controllable factor, such as choice of dosing vehicle, can influence the disposition
characteristics of a compound. Using the model, oral and inhalation exposures were simulated using
equivalent "internal" doses based on the AUC (area under the blood level curve, related to the total
fraction of the dose that was absorbed) for a given tissue or total metabolites formed by either the
monoxygenase or glutathione pathways. Which of the two measures of ID is used, however, cannot be
decided by the model but must be determined by the fit of the data. Dr. Angelo stressed that
physiological models can help to explain physiological behaviors that are influenced by such factors as
dosing vehicle and dose level. Tissue data are essential to ensure that the model represents true
pharmacokinetic behavior.
The impact of pharmacokinetic models on the risk assessment of methylene chloride was discussed
by Drs. Jerry Blancato and Lorenz Rhomberg. They stressed that pharmacokinetic models cannot
replace dose-response models. Instead, pharmacokinetic information can be viewed as a dose assessment
tool. In the case of methylene chloride, the pharmacokinetic models can predict the delivered dose,
which can then be related to risk. However, allometric scaling continues to be incorporated into risk
21
-------
assessments and surface-area scaling is used to account for the different lifespans of rodents and man.
The question of equivalent target tissue dose for animals and humans, given their different lifespans, still
requires attention.
Dr. Rory Conolly concluded by commenting on the two different physiological models and risk
assessments presented. He stressed that a major issue in risk assessment is the measure of dose, and
, one of the major uses of physiological models is to estimate the ID received by the target tissue. He
stressed that optimum models will contain as few parameters (experimentally measurable variables) as
possible. A question he asked was, if additional structures are added to the model, do they describe real
compartments? While it is easy to extrapolate from high to low doses and from route to route using a
properly constructed model, it is more difficult to extrapolate between species because it is frequently
impossible to measure some of the parameters in humans. In such cases, allometric scaling can be used,
with a certain degree of caution. The innate variability of living creatures introduces a degree of
variability into any of the models which would be improved by incorporating a distribution of parameter
values, rather than single point measurements. Methods to incorporate distributions are available.
Overall, it is clear that validated models will improve our ability to make quantitative comparisons.
22
-------
CASE STUDY COMMENTS
23
-------
BENZENE CASE STUDY
Dr. Richard Irons: Commentor
I think it is important that we consider what assumptions must be made in the process of
introducing extrapolation of risk. There are two issues on which my comments will focus: (1) what
assumptions are made about pharmacokinetic modeling and its application to benzene toxicity; and (2)
what assumptions are made about the mechanisms of carcinogenesis. Unlike most of the compounds
discussed at this conference, benzene is an accepted and well-known human leukemogen. It is associated
with an increased risk of Acute Myelogenous Leukemia. In that context, modeling represents an
opportunity to test the relevance 6f animal data to the situation in man. So far, we have assumed that,
in terms of dosimetry, we are focusing on a stationary target. I suggest that in the case of benzene, we
are not dealing with a stationary target. I also suggest that we may not, as previously assumed,
completely or, perhaps, even remotely understand what the metabolites are doing. And I also suggest
that the assumption that a single phenomenon (i.e., a direct structural relationship such as an adduct) is,
with respect to either dosimetry or mechanism, somehow related to cancer may not be true for benzene.
There are at least three different levels of metabolism that are involved in benzene-induced toxicity
to the bone marrow. Primary metabolism of benzene occurs predominantly in the liver and is a mixed
function oxidase-dependent activity. Although it has been argued that a ring-opened product may occur,
it is clear that the principal product is phenol. Secondary phenolic metabolism also occurs within the
liver to produce polyphenolic metabolites; this also appears to be a cytochrome-dependent process.
Hydroquinone and catechol are relatively stable and are transported to the bone marrow, where they
accumulate independent of metabolism for reasons which are not understood. The accumulation of
these metabolites does not correlate with the tissue partition coefficients. It appears that tertiary
metabolism at the level of hydroquinone in the bone marrow correlates very well with toxicity produced
by dibenzoquinone. This involves monoperoxidase activity which is concentrated in the bone marrow.
Although we and others have provided some evidence of a capability for direct metabolism of benzene
by bone marrow, it is extremely limited and dwarfed by the ability of the liver with respect to primary
metabolism.
One of the basic assumptions for risk assessment and modeling is that some product of dose x
duration is equivalent or at least proportional to risk. A corollary assumption, predicted, I think,
primarily on its mathematical simplicity, is that exposure is cumulative with respect to risk. For benzene
we have data that, at an acute and chronic level, may not be consistent with this particular assumption.
24
-------
The first issue that can be addressed is the potential complexity of benzene metabolism. We know
that if we administer hydroquinone and measure a crude endpoint of acute toxicity or subacute toxicity
(bone marrow cellularity), we can produce a transient suppression of bone marrow. With continued
exposure, a refractory response is observed. We have learned that if phenol is administered alone, no
toxicity is seen; but if phenol and hydroquinone are administered concomitantly, a marked increase in
toxicity associated with the administration of hydroquinone is seen. This appears to correlate with an
enhancement of monoperoxidase-mediated metabolism of hydroquinone in the presence of phenol. This
is a possibility that should be considered when metabolites influence the subsequent bioactivation or
metabolism of other metabolites. A complex situation arises that can't be related simply to primary
metabolism of the parent compound.
Even with concomitant administration of phenol and hydroquinone, and although the magnitude of
toxicity is greater and of longer duration, toxicity is only transient, and with continued treatment on a
daily basis, eventually a refractory response in bone marrow is observed. One explanation for this is that
benzene is known to be a cycle-specific agent. When we talk about bone marrow, we are talking about a
highly proliferative tissue, producing about a kilogram of tissue or more per day in an average adult.
Thus, issues such as regimen-dependence become very important in understanding benzene toxicity. If
we look at the tissue kinetics in rapidly and asynchronously dividing cells, such as bone marrow, there
will be a proportion of cells in each phase of the cycle at any given time. If, in fact, toxicity is phase-
specific, then at any given time there will only be a fixed proportion of the cells that are susceptible to
toxicity.
Bone marrow diffusion cultures performed by John Marsh at Yale, which enable a direct
comparison of a variety of different tissues in mouse, dog, and human bone marrow, illustrate this point.
Adriamycin yields a typical log dose-response curve in an acute experiment where an increasing dose of
adriamycin increases cytotoxicity. In contrast, increasing the dose of methotrexate, a cycle-specific agent
exhibiting schedule-dependent effects, will not increase cytotoxicity. After a given point, toxicity will not
increase because the cell population will become synchronized and will be protected from the effects of
the agent. With respect to the toxicity of benzene metabolism, when the same doses of hydroquinone
and phenol are administered repeatedly 3 days a week with a 4-day interval in between, no refractory
response is observed. Instead, increased toxicity and the development of a very severe bone marrow
aplasia is seen. Thus, cell-cycle dependence^appears to be a fairly important issue with respect to
modeling acute benzene toxicity, at least as far as its metabolites are concerned.
25
-------
How does this relate to long-term or chronic exposure and the issue of leukemogenesis? For years
the lack of an experimental model for leukemogenesis associated with benzene has been a limiting factor.
This may be as much a consequence of experimental design as it is of species differences. A number of
inhalation studies with benzene have been conducted in rats and mice. In CD-I, B6C3F!, and C57BL/6
mice, where animals were exposed to benzene concentrations between 300 and 3,000 ppm for 2 years or
a lifetime, the data with respect to leukemia were either negative or inconclusive. Occasionally a
leukemia will show up in a" study with a large number of animals, but not enough to convince anybody.
Following initial regimen-dependence studies, Cronkite (1987) decided to look at a limited duration
exposure of 100 to 300 ppm in C57BL/6 and CBA/Ca mice. When mice were exposed to benzene for 8
weeks or 16 weeks and held for a lifetime, a very efficient leukemia response was observed. However, if
animals were exposed to 3,000 ppm benzene for 8 days as opposed to 80 days (the same product in
terms of dose duration), no cancer of any sort was observed over a lifetime. In the C57BL/6 mouse, the
12-week exposure to the same concentration of benzene, 3 or 6 days per week, is very efficient in
inducing lymphoid leukemia or myelodysplastic changes in the spleen. These lesions do not appear
immediately after cessation of exposure but appear if the animals are then held for a lifetime. Thus, the
lifetime maximum tolerated dose does not appear to produce an efficient or measurably convincing
leukemogenic response whereas a high exposure of more limited duration does.
Now I'd like to briefly address the developing area of experimental leukemogenesis. In general,
leukemogenesis is often thought to be a more complicated process than carcinogenesis and is widely held
to be a multifactorial process. Whether leukemogenesis and carcinogenesis in solid tissues are similar, I
think we know more about leukemogenesis at this stage than about other types of cancer. In addition,
some of the simplifying assumptions we apply to solid tissue cancers cannot be applied to
leukemogenesis.
Leukemogenesis is a multifactorial process involving critical events in more than one compartment.
Virtually all leukemias, with very few exceptions, originate in stem cells, independent of the particular
cell line in which the leukemia is expressed. Multipotent and certainly committed stem cells are a very
rapidly proliferating cell population. Although it has been known for a long time that the number of
pluripotent stem cells that are active at any one time is minute, it is not known how small that number
is. In the last year, the ability to insert a foreign gene into individual stem cells has allowed the tracking
of the activation and regulation of individual stem cells (Lemischka et al., 1986; Snodgrass and Keller,
1987). In mouse studies conducted at the Whitehead Institute and in Europe, it has become evident
that the number of pluripotential stem cells giving rise to all the cells of the entire hematopoietic system
26
-------
can be as few as one. Conceptually, this can only mean that nature places a very high priority on
protecting these cells at all costs, keeping them at rest while virtually all proliferative activity takes place
at a very fast rate. Why is this important? It is important because the vast majority of leukemias are
known to originate in cells at different levels within the stem cell compartment. These cells may all be
aptotic. In contrast, cells in the rapidly proliferating precursor cell population appear to be aptotic (i.e.,
irreversibly committed to terminal differentiation). As such, they are sensitive targets for cytotoxic
agents but are incapable of neoplastic transformation. When we measure the effects of toxicity on these
cells and attempt to draw relevance with respect to leukemogenesis, at best we're looking at an indirect
population, and at worst, the results may have no relevance at all.
The implication for benzene toxicity is two-fold. First, if one assumes the same type of
phenomenon as a single hit to define increased risk or the initiation of cancer, there is a problem:
What is the target cell? Second, all secondary and tertiary benzene metabolites recapitulate alpha- and
beta- unsaturated diketone groups that react almost selectively with sulfhydryl groups. They only
reluctantly react with nitrogens, carbons, or other potential target molecules, and there has been no
definitive demonstration of a DNA adduct of benzene in vivo. Whether or not this occurs to any
appreciable extent remains to be seen. It now is clear that if a DNA adduct can be identified in vivo, it
will be at doses that are several-fold in excess of the minimum carcinogenic dose. In addition, there is
considerable evidence suggesting that benzene alters the differentiation of stem cell populations in
response to the damage occurring in the cycling population. I would like to suggest a hypothesis that is
consistent with paradigms currently being explored in experimental leukemogenesis. The initial event
that may predispose a subject to increased risk of leukemia may simply involve a shift in the relative
numbers of stem cells from the resting pluripotential stem cell compartment to the multipotent or
committed stem cell compartment. Cytotoxic agents and benzene in particular are known to increase
proliferation in the stem cell compartment and to deplete the number of pluripotent stem cells. The
initiation event may, in fact, occur second or third in the process of the progression of events that leads
to leukemogenesis.
I have explored this hypothesis because we are all familiar with either a single- or two-hit model
involving initiation already. It is, however, worthwhile to consider that there are potentially different
mechanisms other than a one-hit model to explain various types of carcinogenesis.
I'd like to conclude by suggesting that the biological complexity of various animal model systems, as
well as the differences in metabolism, pharmacokinetics, distribution, and the interspecies differences in
tumorigenesis may require us to consider creating composite models in which we carefully construct
27
-------
V
what, in fact, attorneys refer to as "the reasonable man" from bits and pieces of various models from
other species. No one simple paradigm is likely to adequately reflect the real situation in humans.
References
Cronkite, E.P., 1987. "Chemical leukemogenesis: Benzene as a model." Seminars in Hematology 24:2-11.
Lemischka, I.R., Rauler, D.H., and Mulligan, R.C., 1986. "Developmental potential and dynamic
behavior of hematopoietic stem cells." Cell 45:917-927.
Snodgrass, R. and Keller, G., 1987. "Clonal fluctuation within the hematopoietic system of mice
reconstituted with retrovirus-infected stem cells." EMBO J. 6:3955-3960.
28
-------
BUTADIENE CASE STUDY
Dr. Richard Irons: Commentor
One of the principal issues with respect to butadiene carcinogenicity and toxicity is the dramatic
species differences. Multiple mechanisms may be involved in explaining carcinogenesis, and this may be
of some importance in understanding the relevance of butadiene as a potential hazard for man.
Although we have just begun to study butadiene carcinogenesis in experimental animals, it is an area
that illustrates the potential for chemical-biological interactions that are relatively unexplored in modern
toxicology. It also illustrates some of the limitations of pharmacokinetic modeling as it exists today.
As has been discussed, there are differences in metabolism, differences in disposition, and
differences in carcinogenicity with respect to butadiene across species. I would differ somewhat from Dr.
Bayard in that I am not convinced that the metabolic changes between mice and rats can account for the
marked species differences, certainly in leukemogenesis. If one looks at the high dose required for
carcinogenesis in rats versus the low dose effective in mice, metabolite concentrations are almost the
same. And you still have marked differences in the pattern.of carcinogenicity.
The extensive disposition data that has been collected in butadiene metabolism studies do not
appear to correlate with the established major target organs. In fact, using different genetic models,
acute target organ toxicity does not correlate with carcinogenesis, in this case, leukemia.
Many of the recent butadiene studies have shown marked species differences as shown by Owen
(1981) and Huff et al. (1985). Butadiene is a very potent leukemogen; it is probably the most potent
mouse leukemogen that has been studied in the context of an environmental or occupational hazard. In
expanding Dr. Bayard's comments, all of the tumors associated with butadiene exposure are T-cell
lymphomas. They are either leukemias or lymphomas in the mouse. They are all derived from the
thymus, but, as is the case for this lesion in the mouse, the bone marrow is the principal target organ.
Therefore, if it fits the paradigm for all leukemias, bone marrow toxicity would be expected. I don't
think it is biologically correct to compare the background incidence for lymphomas in the B6C3Ft mouse
with butadiene-induced lymphomas that are T-cell in origin. Background lymphomas in the B6C3Ft mice
are of B-cell origin. They can be considered two separate diseases. Thus, butadiene-induced lymphoma
in the B6C3Fi mouse is a very significant lesion with an incidence of 60 percent.
29
-------
One of the problems with the B6C3Fl mouse, and virtually all laboratory strains of mice with one
or two exceptions, is that they carry retroviruses that are associated with leukemia, under a variety of
conditions. In order to explain the potential species differences, we decided to examine whether one
could model butadiene on the basis of what is known about viral-induced leukemias in the mouse. It
was found that leukemia in these B6C3Ft mice is preceded over a 20-week period by a marked activation
of the specific retrovirus that has been implicated as playing a role in leukemogenesis in mouse models.
This is not an all-or-none situation, and it is very complex.
This type of activation is very intriguing because this has not been described previously for agents
such as butadiene. There are many endogenous retroviruses in the mouse, and butadiene appears to be
turning on only one. However, it is one that has been implicated in the past in leukemia. Retroviral
production is increased approximately 10,000-fold as a function of exposure to butadiene.
In order to obtain an indication of whether or not the virus was playing a role, we decided to do
comparative studies of leukemogenesis in B6C3Ft and NIH Swiss mice. The NIH Swiss mouse is a
mouse in which we know the endogenous ecotropic retrovirus associated with playing a role in
leukemogenesis in murine strains is truncated and not expressed, A large, but not absolute, difference in
the incidence of leukemogenesis in these two strains was observed. The B6C3Fj mouse exhibits the
typical 60 percent incidence of lymphoma, while the NIH Swiss mouse has an incidence of 13 to 14
percent. Thus, the rate of leukemogenesis in these two strains differs by four- to five-fold. The
incidence of leukemia in the NIH Swiss mouse cannot, presumably, be related to the retrovirus. Since
we could not rely on pharmacokinetics or disposition to help us at that point, we looked at the bone
marrow which is the target organ. When micronuclei production, cellularity, and chromosomal
aberrations were examined, the data in the B6C3Fi and the NIH Swiss mice were superimposable.
Quantitatively and qualitatively, the target organ toxicity was the same. This, along with the viral
activation, suggests that, in fact, there is some, albeit presumptive, role for a retrovirus in this model
system. It also suggests that typical biomarkers associated with toxicity that might also relate to
leukemogenesis do not correlate with leukemogenesis in this mouse model.
This leukemia type is preceded by a marked selective activation of an ecotropic retrovirus. We
characterized the most likely potential gene interactions in the mouse that could be predicted to lead to
specific activation of retrovirus. These would not be extrapolatable to any other species because there
are unique specific transactivating genes that regulate virus expression in the mouse. We found that they
are not altered by butadiene. The mechanism of activation thus appears to be the de novo activation of
ecotropic retroviral sequences and not an indirect mechanism involving alteration in mouse host
30
-------
resistance or virus restriction. This type of direct activation is the only potential mechanism that might
possibly relate to activation of retrovirus in other species such as man. Thus, to date, we have not been
able to exclude the mouse as a model for retroviral activation independent of its relevance as a model
for leukemogenesis.
It is simply too early to say whether the mouse model is relevant. However, we are pursuing the
issue of whether it is possible for butadiene or its metabolites to alter the latency of biology of human
retroviruses as well.
A number of additional studies need to be performed before we can provide any definitive
conclusion as to whether or not the lymphoma seen in the mouse is directly related to a retrovirus. At
this point, our paradigm seems to be reduced to two likely scenarios, and there are in vitro models for
both. Given the possibility that retroviruses and butadiene are independently leukemogenic, we are
looking at additive mechanisms. On the other hand, either butadiene alters or influences
leukemogenicity associated with a mutagen, the retrovirus, or the retrovirus is acting in a cocarcinogenic
manner associated with butadiene exposure.
From the standpoint of biology, what is the relevance, or the potential relevance, of understanding
these models? For leukemogenesis associated with butadiene, it is possible that the B6C3Fi mouse may
not be the most appropriate model for extrapolation to the general population. If retroviral genes are
implicitly related to mechanisms underlying the high incidence of lymphoma, then they are probably not
the most reliable indicators of the potential for butadiene to cause leukemia in man. In addition, one
must evaluate the potential relevance of the mouse model for risk associated with a defined population
in man that may be carrying specific retroviruses. In this population, the compound might affect the
latency of behavior of those viruses. Thus, this model raises several issues with respect to potential
health risks in man and the process of evaluating hazard in general.
References
Huff, I.E., et al., 1985. "Multiple organ carcinogenicity of 13-butadiene in B6C3Ft mice after sixty weeks
of inhalation." Science 227:548-549.
Owen, P.E., 1981. "The Toxicity and Carcinogenicity of Butadiene Gas Administered to Rats by
Inhalation for Approximately Twenty-four Months." Final Report, Hazelton Laboratories,
Harrogate, England.
31
-------
METHYLENE CHLORIDE CASE STUDY
Dr. Rory B. Conolly: Commentator
Introduction
A major issue in carcinogen risk assessment is the measure of dose. For example, oral gavage dose
and exposure concentration have been commonly used (though the latter is not a measure of dose per
se). There may be significant nonlinearities, however, between these external doses and internal
surrogates, e.g., the concentration of parent compound in target tissue. Furthermore, within the target
tissue, saturable and/or competing pathways for metabolism of the proximate and ultimate carcinogens
may introduce additional nonlinearities. Internal dose surrogates that reflect these nonlinearities can be
a meaningful basis for risk assessments of chemical carcinogens. This discussion will address the use of
physiologically based pharmacokinetic (PBPK) models for computer simulation of dose surrogates. After
some introductory comments on the PBPK approach, the PBPK models described by Drs. Angelo and
Reitz for simulation of methylene chloride (MC) dose surrogates will be discussed.
PBPK Models
PBPK models contain mathematical descriptions of the target species physiology and carcinogen
chemistry which determine pharmacokinetic behavior. The optimum models are parsimonious, i.e., they
contain no more structural detail than is needed. Digital computers are used to exercise these models to
obtain simulations of the internal dose surrogates. These simulations are accurate when: (1) model
structure is representative of the essential elements of the actual, real-world structure; and (2) the
parameter values used are accurate. (It should be emphasized that in PBPK models, parameters
correspond to measurable entities in the target species.) When these conditions are met, the model is
"valid" and can be used predictively.
Interspecies Extrapolation
When PBPK models are validated with data obtained from laboratory animals, simulation of dose
surrogates for humans requires scaling of the model across species. The physiological and biochemical
structure encoded in a PBPK model can be adjusted for changes in body size using allometric
relationships. It is always preferable, however, to actually measure parameter values in the different
species. This, of course, is not always possible. Allometric scaling is valid to the degree that: (1) the
32
-------
experimental species are physiologically similar to humans; and (2) the allometric relationships are
properly defined.
Dose Extrapolation
It is often necessary to simulate pharmacokinetic behavior for doses well below the range over
which experimental data are available. If model structure is correct and has been properly validated,
there is no dose constraint on the accuracy of such low dose simulations. By its nature, a physiologically
based model contains the information needed for simulation of carcinogen pharmacokinetics at low doses
as well as high. It is possible, therefore, to use a valid PBPK model to simulate internal dose surrogates
for any exposure scenario.
Variability
Concern was expressed during the workshop over the variability of model predictions. This
variability arises because the models are physiologically, or more generally speaking, biologically
structured. Parameters in such models do not have single correct values simply because the real world
correlates of the parameters are also variable. A group of 10 people or mice, for example, will typically
have 10 different cardiac outputs, liver volumes, and so on. In the mathematical terms of the PBPK
model, it is most useful to think of a range of possible values for any given parameter with a distribution
of values around the mean. The shape(s) of the distribution for parameters commonly used in PBPK
models is unknown and this problem needs investigation. Still, model output is a function of the
particular set of parameter values used for a given simulation and naturally varies as different values are
used.
Concern that this aspect of PBPK modeling is a failing is unfounded. Variability arises because
PBPK models are analogues of real-world systems such as mice, rats, humans, etc. These analogues
enable us to examine the variability inherent in responses of these systems to toxic chemicals, which is
an advantage. Methods for incorporating distributions of parameter values into PBPK models are
available. Drs. Portier and Reitz both presented Monte Carlo approaches to obtaining distributions of
model behaviors. This type of work is innovative and more is needed. Of particular concern are
assumptions about the shape of the distribution of parameter values and estimates of parameter variance.
33
-------
PBPK Models and MC
The PBPK models for MC described by Drs. Reitz and Angelo have been used to simulate internal
dose surrogates. The questions raised by their presentations are addressed below using the overall
paradigm for PBPK modeling outlined above.
In the model described by Dr. Angelo, body organs have two subcompartments, a blood-flow-limited
vascular region and a diffusion-limited extravascular region. This structure allows accurate simulations of
MC blood and tissue concentration data and was developed after preliminary versions of the model,
lacking the subcompartment structure, produced inaccurate simulations.
The use of subcompartments is a significant complication of model structure and must be evaluated
against the need for parsimony noted above. Could some other less drastic change in the model suffice?
In our laboratory (NSI Inc., Dayton, Ohio), a similar blood-tissue concentration discrepancy was found
with chloropentafluorbenzene (CPFB). In this case, simulation accuracy was improved by "unlumping"
the rapidly perfused compartment. Explicit descriptions for kidney, lung, testes, and brain were used and
the respective partition coefficients obtained. These were quite different from each other and the
"unlumped" model then accurately simulated blood and tissue levels of CPFB. Related to the question
of parsimony in the Angelo model is the issue of model structure. Is the two-subcompartment structure
real in the sense that it is isomorphic with target species for MC, or is it a nonphysiological
modification? If it is the latter, then the model cannot be used with confidence for the interspecies and
high dose to low dose extrapolations which are the great strength of PBPK models.
It is a misconception to speak of PBPK models for particular routes of exposure. A properly
structured PBPK model should work for any and all routes of exposure. Of course, an appropriate
description must be included in the model for each route. A model that accurately simulates MC
pharmacokinetics after intravenous dosing will work for oral dosing, provided that the rate of oral
absorption and any vehicle effects are properly described. A PBPK model that works for one route of
exposure but not another suffers from either a structural defect or from a validation failure.
Dr. Reitz described a PBPK model in which MC metabolism by the glutathione S-transferase (GST)
pathway in the lung was used as the dose surrogate. This work represents the state of the art in the use
of PBPK models and computer simulation for estimation of dose surrogates. As such, it also serves to
identify the current limits of the approach. It was necessary to estimate in vivo GST activity toward MC
for the human lung. This was accomplished by measuring human GST activity towards MC in vitro and
34
-------
extrapolating to an in vivo value based on an in vivo/in vitro ratio obtained from animal experiments.
This approach is reasonable, but much more data are needed on these types of correlations. A
mechanistic explanation of in vivo/in vitro differences in enzyme activity would also be useful. A better
data base and an understanding of mechanism would together help to identify when the type of
correlation used by Reitz is acceptable and when it is not. This particular problem is just part of the
larger issue of how physiological parameters scale across species. The use of PBPK models in risk
assessment is creating a growing need for experimental work specifically directed to this question.
The simulations developed by Anderson et al. (1987) show a good correlation of the GST pathway
dose surrogate with MC carcinogenesis. These data are only correlational, however. Mechanistic studies
illustrating the role of the GST pathway in MC tumorigenesis are needed.
Finally, the issue of interspecies dose scaling for MC must be addressed. As noted above, there is
little doubt that MC carcinogenicity is due to bioactivation, and there is good evidence suggesting that
glutathione conjugation is the activation step. Scaling of the physiological parameters of the model was
according to well-established allometric relationships which are adequately supported by experimental
data. Also, Andersen et al. (1987) estimated the rates of the relevant metabolic processes in several
different species. The addition of interspecies dose correction factors over and above the scaling rules
built into the PBPK model ignores the logic of the PBPK approach. Moreover, it decreases the accuracy
of the dose surrogate simulation.
There are, of course, uncertainties in the experimental estimation of parameter values, in the
specification of allometric relationships, and also with the accuracy of model structure. These
uncertainties should be addressed by inclusion of a safety factor applied to the simulation of the dose
surrogate for the human case.
References
Andersen, M.E., Clewell, H.J., III, Gargas, M.L., Smith, F.A, and Reitz, R.H., 1987. "Physiologically
based pharmacokinetics and the risk assessment process for methylene chloride." Toxicol. Appl.
Pharmacol. 87:185-205.
35
-------
RAPPORTEUR REPORTS
36
-------
APPLIED RESEARCH NEEDS
Angelo Turturro
Introduction
The major focus of my comments will be on the application of pharmacokinetic data to the problem
of limiting uncertainty in a risk assessment. Murray Conn brought up a number of cautions in using this
information, to which I will add a few of my own. However, use of information such as this is part of
the evolving science of risk assessment, and the question is not whether to use this information, but how
to use it.
As an aside, this workshop included many participants trained in the discipline of engineering, and
the overall discussion has been well served by their presence. People trained in engineering especially
understand why conservatism is built into risk assessment because the engineer is responsible if a
chemical reactor vessel collapses or a process does not work. In a sense, risk managers in a regulatory
agency are in the same position. If people are hurt, or even if the public thinks that people will be hurt,
risk managers are held accountable. Thus, there is an inherent conservatism in risk assessment, which I
think is necessary and practical.' When one is protecting the public health, one has to be especially
careful about the consequences of one's actions.
General Rules
One effort that would especially help in risk estimation is the definition of some general rules for
addressing questions about toxicity. Efforts in cross-species scaling and physiologically based models
reported in this workshop are attempting to define these rules. I know this is anathema to most of us
who spend our research time looking at a particular compound, but it would be useful to define some
general way a regulatory agency can estimate toxicity in order to avoid being overwhelmed by the
number of chemicals that must be addressed. And this number is growing. Because of the great
diversity of compounds, I assume that there will not be a general rule for all chemicals, however some
chemical class-specific rules or guidelines related to different mechanisms of toxic action may be possible.
For instance, one "class" that could be addressed is chemicals that undergo minimal metabolism and do
not bind well to macromolecules. The pharmacokinetics of chemicals such as these may be relatively
straightforward, e.g., may be estimated by simple clearance models. These models may allow fairly
37 .
-------
simple extrapolation across species of target dose to a cell. Obviously, all chemicals would not fit into
this class. But if one can define the target dose and limits of the class, regulation would become easier.
An important part of the data base for this task will be derived from the tremendous amount of
study done on the metabolism of proprietary drugs. Pharmaceutical companies have provided
information on pharmacokinetics, both for defining drug efficacy and toxicity, as well as in the
development of drugs by increasing delivered dose. This information, especially pharmacokinetic data
involving toxicity estimates among different species, can be crucial for making general risk estimation
rules.
Risk Assessment
One question that can be asked is "Why do risk assessment at all?" Risk assessment is a critical
part of the risk management process. The key factor is to practice appropriate risk management. In
3500 B.C., risk managers, such as high priests, estimated risk of public actions by looking at animal
entrails. Today, with the scientific method, we still look at entrails, but with more detailed analysis.
Unfortunately, there is a good deal of uncertainty in the process of assessing risk to humans. If we
decrease this uncertainty, we can be less conservative in our estimates, reducing the "insurance" we
presently use. Series of conservatisms are built into models basically to protect us from our own
ignorance; we want to eliminate that ignorance without endangering public health.
Much of the discussion at this workshop centered around cancer risks because that is the area that
risk assessments have focused on in the past. It is useful to emphasize that there can be risk
assessments for a number of toxic endpoints besides tumors, such as reproductive and neurotoxic
endpoints, that are important to public health and well-being. These risk assessments share many of the
same uncertainties in pharmacokinetics and pharmacodynamics as cancer estimations and can also benefit
from pharmacokinetic analysis." For instance, as was discussed in the workshop presentation describing
the relationship between benzene exposure and aplastic anemia, a good pharmacokinetic model for the
anemia may be much more helpful for evaluating the effect of benzene on public health and its risk to
the public than a cancer model. An example of an area that needs exploration is the effects of
methotrexate (used in chemotherapy) on tumor cells, a subject that even experts who use the compound
are unclear about. We have to understand that pharmacokinetics are not simply for cancer, and the
pharmacodynamic patterns we should be concerned about are not strictly those for tumorigenesis.
38
-------
Risk assessment is usually separated into four parts:
(1) Hazard identification, which identifies the potential human adverse effect
(2) Exposure assessment, which estimates human exposure
(3) Dose-response assessment, which develops a dose-response curve for the adverse effect
(4) Risk characterization, which applies all of the above information to make an estimate of
human risk.
I will use this framework of risk assessment to structure this discussion.
Hazard Identification. The key to using pharmacokinetics in hazard identification is to clearly identify
the toxic agent or agents. Developing pharmacokinetic data for this step is not very useful unless one
can truly demonstrate that the chemical species being investigated is key to the toxic effect. This
demonstration must be convincing to everyone. If this is demonstrated, and if it is shown that the active
species is not produced in humans, then the results of an animal test need not be extrapolated to
humans. Thus, pharmacokinetics can tremendously reduce risk. Alternatively, it could indicate that
humans are more sensitive to the effects of an agent.
One important consideration, especially for chronic endpoints, is that agents may have multiple
effects. For instance, for a genetic toxin that is also cytotoxic, there is no reason to believe, a priori,
that only one metabolite causes both the genetic damage and the cytotoxicity. A single agent may thus
impact both initiation and promotion with different pharmacokinetics. If one is studying cancer, one has
to consider possibilities such as metabolites which increase proliferation or impair the immune function
system.
Another significant aspect to emphasize is that the sum of activation, deactivation or detoxification,
and reactivation is important, especially in carcinogenesis. An example of this is the toxicity of N-
hydroxyarylamine conjugates (Young and Kadlubar, 1982). In this case, the agent was metabolized and
the conjugate stored in the bladder prior to voiding. This conjugate was reactivated by urine pH and
time of retention in the bladder, parameters that can be correlated with the species susceptibility to
carcinogenicity. Pharmacokinetics could have been determined with every possible blood-borne agent
metabolite, yet would have been of very little significance. Situations such as these are reminders that
information on mechanisms is important in order to understand the significance of data on disposition.
39
-------
And finally, the physiology or the metabolism of the animal can be changed by a toxic compound.
An example of this is the effect on an animal's consumption of calories. In some studies, because of
factors such as agent palatability, animals will consume fewer calories. As little as a 20 percent
reduction in calories compared to the control will eliminate the effects of a high fat diet in promoting
carcSnogenesis (Boissoneault et al., 1986). With caloric restriction, the animal's whole body physiology is
changed (Turturro and Hart, 1989). Thus, using standard protocols (for instance, measuring bolus
injections in a normally fed animal) may not be relevant to the physiological state of the animal in
which tumors develop. At the higher doses, the alteration of animal physiology, chemistry, or
metabolism should be prevented. This is the type of issue that will cause a regulatory agency to be
concerned and that our research should address.
Exposure Assessment. Exposure assessment was not discussed much in this workshop. However, this is
the part of risk assessment in which most resources are used and major uncertainties exist. It is very
difficult to accurately estimate, for instance, what the time course of concentration is in a single place if
there are fifteen point sources producing Chemical A. Compounding this by the variability in human
activities leads to tremendous uncertainty in the estimate of exposure. For instance, try to determine
your exposure to 1-nitropyrene (in air and food) this morning or over the last year and then try to apply
this analysis to a population. One way to address this problem is to use biomarkers of exposure. These
markers, such as certain DNA damages, may not be directly toxic. If a target agent and/or a target dose
cannot be adequately defined, other information, like total metabolites, agent incorporated, etc., can be
used to help define the relationship of administered to applied dose. This will be a valuable area of
research if simple useful exposure markers can be developed to deal with the uncertainty in estimating
dose.
Dose-Response Assessment. Estimating the dose-response relationship describes the heart of risk
assessment, i.e., assessing risk under practical conditions of use and exposure. The variability of a
population response is critical, both as a function of metabolism and susceptibility to the effects of the
agent. Seven of ten humans may be similar to the average inbred rat we use in estimating metabolism.
However, and this was brought home by George Lucier, there may be a substantial number of people
who can either metabolize more efficiently or not at all. The ratio of the mean metabolism of these
subpopulations, therefore, may be orders of magnitude and there may be a subset of people who are at
risk. Given the bias toward public health, in order to show that a chemical will not be a danger, the
regulatory agency must be convinced that a sensitive subpopulation either does not exist, or is so small
40
-------
that any risk to them can be managed. Even the perception that such a population could exist would be
a problem. Research in this area is based on the basic issue of extrapolation of animal data to humans.
Another issue is enzyme induction. Although a number of toxic agents are used to produce specific
cellular fractions with induced enzymes, it is often not appreciated that many compounds will induce the
enzymes that metabolize them. These will change the pharmacokinetics, suggesting that chronic effects
may vary significantly from acute ones.
The central problem in many situations is extrapolation from high dose to low dose. When a high
dose is given, sometimes effects are induced that are different than at the low dose. Since,
unfortunately, it is necessary to extrapolate to a region on the dose-response curve that is not
observable, one really has to make a case that the high dose and low dose mechanisms are similar to use
information from one range to the other. Saturation, cytotoxicity, hormone imbalance, and accumulation
of toxic products are all possibilities that may distort the dose-response relationship, especially after
chronic administration.
Finally, there was a comment previously that once one has a few interacting parameters, the
statistics become formidable. Fortunately, there is a new area of mathematics that is addressing
problems such as these. It is called sensitive dependence on an initial condition and known popularly as
"chaos" (Tsonis and Eisner, 1989). "Chaos" provides methods to calculate some of the recurrent factors
that arise in these seemingly chaotic systems. It could be interesting to apply chaos mathematics that
have been devised for problems such as weather prediction and other very complex systems to the
biology of multiple parameters interacting in a statistical sense.
Risk Characterization. It should never be forgotten that risk assessment will be applied to people where
the interaction between factors such as common dietary agents, environmental agents, and physiological
condition must be considered. For example, if we know that certain ingested foods induce the aryl
hydrocarbon hydroxylase and a certain compound uses that enzyme for its metabolism, we should take
this into account when applying the results of the risk assessment to the real world. Interaction with
diet as well as genetic variability can produce the real world responses to chemical compounds. The
same is true about common environmental agents. For instance, it would be of limited value to discuss
the risk associated with a phototoxic agent applied to the skin without also discussing exposure to
sunlight. Physiological state is also a factor. Pregnancy is sometimes addressed, but old age and
diseased states are almost never considered, despite the known effects that disease, such as renal
pathology, can have on pharmacokinetics.
41
-------
References
Boissoneault, G., Elson, C, and Pariza, M., 1986. "Net energy effects of dietary fat on chemically
induced mammary carcinogenesis in F344 rats." J. Natl. Cancer Inst. 76:335-338.
Tsonis, A., and Eisner, J.B., 1989. "Chaos, strange attractors, and the weather." Bull. Am. Meteor. Soc.
70:14.
Turturro, A. and Hart, R.W. "Overview of the caloric restriction project." Mech. Aging Dev. In press.
Young, J.F. and Kadlubar, F.F., 1982. "A pharmacokinetic model to predict exposure of the bladder
epithelium to urinary N-hydroxyarylamine carcinogens as a function of urine, pH, voiding interval,
and resorption." Drug Metab. Dispos. 10:641-644.
42
-------
CHEMICAL-SPECIFIC NEEDS FOR PHYSIOLOGICALLY BASED
PHARMACOKINETIC MODELS
Hugh L. Spitzer
Introduction
I would like to start my comments by acknowledging the key role Richard Reitz from Dow
Chemical Company has played in legitimatizing the use of physiologically based pharmacokinetics in risk
assessment. I think, more than anything else, it was his willingness to have frank and open discussions
with scientists from the regulatory agencies, be available to discuss and test the model, and go back into
the laboratory and do additional experiments that resulted in the use of physiologically based
pharmacokinetic (PBPK) models in the methylene chloride risk assessment. Dr. Reitz might not be the
father of PBPK, but he certainly is the doctor of record that brought it into the regulatory process.
I think you are going to be disappointed if you think the effort behind the use of pharmacokinetics
will result in greater exposure without increased risk. If the discussion over the last few days is a
reasonable indication, PBPK can clearly result in either an increase or decrease in the risk associated
with a given exposure. This is a healthy situation because it demonstrates that the approach is neutral.
Overview
Traditionally, for the purpose of risk estimation, the U.S. Environmental Protection Agency makes a
direct extrapolation of tumor incidence from rodents exposed to high concentrations of a chemical to
humans exposed to low concentrations. These extrapolations are based on a series of conservative
assumptions that incorporate only limited consideration of some physiological parameters in risk
estimation.
The PBPK model offers an exciting new opportunity to reduce the uncertainties in high to low dose
and cross-species extrapolation. The PBPK model, unlike classic compartmental models, gives a more
realistic representation of the behavior of chemicals in biological systems. The PBPK model differs from
the conventional compartmental analysis in that the actual physiology of the animal serves as the basis of
a compartmental description. Thus, the physiological model represents the mammalian system in terms
of specific organs or groups of organs/tissues based on common characteristics. This allows each organ
or group of organisms to be considered with respect to intrinsic volumes, blood flow rates, partition
43
-------
coefficients, and biochemical constants specific to the species and chemical under investigation. The
value of this approach is three-fold. First, the model can utilize a wide array of data not previously
considered, but which could significantly impact extrapolation from species to species. Second, the
model has the ability to provide a better estimate of dose to target tissue. Finally, the model's predictive
power is not restricted to the range of conditions or animal species for which the experimental
observation exists. Thus, this process, if used correctly, could reduce the uncertainty for estimating dose
to target tissue.
The model is usually pictured as ia flow chart of blood through the various organ systems. These
systems may be identified specifically, as is usually done with liver and lung, or as general groups such as
fat, muscle, or gut. Absorption and desorption in the tissues can be modeled by assuming attainment of
equilibrium between blood and organ,' as zero, or first order kinetic processes (ed: or any other rate
form). If equilibrium is assumed, the important parameters used in estimating transfer of the chemical
are: organ volume, blood/organ partition coefficient, and blood flow through the organ. Metabolism is
measured in various ways (e.g., vapor uptake, metabolite production), or estimated by choosing values
which provide the best agreement with experimental data on disposition of the parent compound.
Metabolic kinetics may be first-order, but Michaelis-Menten kinetics are usually required to fit the data.
The simulated processes assigned to each organ system are incorporated into a computer program
that provides estimates of chemical or metabolite concentrations in all of the organ systems of the model
for any input parameter (inhalation concentration, oral, intravenous, or intraperitoneal dose). Known or
postulated species differences in chemical metabolism or elimination may be incorporated in the model.
Because one PBPK model is assumed to be valid for all mammalian systems, extrapolation across species
is easily accomplished by substitution of those physiological constants characteristic of the species of
interest. Extrapolation from high to low doses is obtained directly from the model.
The goal of PBPK is to predict the concentrations of the active metabolite of a toxic or
carcinogenic chemical in the target organ as exposure doses, exposure regimens, and routes of exposure
vary. Using the list of parameters summarized in Table 1, it is possible to assess which critical
elements/assumptions necessary for PBPK modeling have been addressed. This type of analysis will
become even more important as investigators begin to develop organ-specific metabolic data and
genotoxic data and begin to understand the transfer of metabolites between organs. Thus, ultimately, the
goal of PBPK modeling must be to significantly reduce the uncertainties involving the potential risk
associated with human exposure to a chemical. If successful, we will then be able to better explain the
44
-------
public risk associated with a given exposure and, with some confidence, hopefully reduce the need to
express risk estimates using 95 percent upper-bound confidence limits.
Chemical-Specific Needs
Part of this workshop has been devoted to a discussion of the use of PBPK modeling on selected
chemicals. This discussion is useful in that we are now beginning to better understand how PBPK fits
into risk assessment and the data required for its acceptance. I think it is important to understand that
the criteria for acceptance will change as we gain confidence in the process. This, to a large degree, is
the nature of consensus building in the scientific community.
The use of PBPK can impact two elements of a risk assessment, exposure and dose-response. It is
important to understand that for each element, the data required will differ. Over the past few days,
workshop presentations describing the use of PBPK for methylene chloride demonstrated how data on
metabolism impact dose-response. Also, PBPK for 1,3-butadiene provided for adjustments in exposure
parameters. Lastly, presentations were given on the risk associated with exposure to benzene. These
presentations were particularly interesting because they demonstrate how one needs to consider both
route of exposure and differences in species.
The presentations in total'show how the use of increased/decreased assumptions can impact PBPK
in risk assessment. Indeed, the concern voiced by Chris Portier should be kept in mind as we go from
very simple to complex systems: are we creating greater uncertainties? An example of the problem we
face can be demonstrated from the presentations on benzene. While the use of metabolic data clearly
advance the assessment, the need to assume that liver metabolism reflects the risk of processes going on
in bone marrow may not be appropriate.
I had hoped that this workshop would provide some guidance on minimal data needs for
pharmacokinetics. After listening to the presentations and discussions over the last three days, I am not
sure that the request for guidance is a wise one at this time. Right now, we appear to be unconstrained
in our approach, which is healthy. In time, I think scientific discussion will begin to point us in a
direction which will allow us to define the limitations on the use of PBPK in risk assessment.
45
-------
TABLE 1
PARAMETERS AFFECTING THE PHYSIOLOGICALLY BASED
PHARMACOKINETIC ASSESSMENT OF CHEMICALS
PARAMETER
CONSIDERATION
Uptake
First pass effect
Tissue equilibrium
Biological half-life of parent
compound and "active" metabolites
Transport of "active" metabolites
between compartments and to
target tissue(s)
Specificity of enzymes and
enzyme induction
Carcinogenic potential
How does concentration (exposure) affect uptake?
Is metabolism during exposure of primary
concern?
Does the partition into lipid-rich compartments
effectively sequester the chemical?
How long is post exposure in the animal or
person at risk?
Does the amount metabolized affect transport?
How does excretion affect amount transported
via the blood?
What consideration must be made for species
differences? Do endogenous substrates affect
rates of metabolism?
Does an unmetabolized chemical play a role in
the process? Is there an unidentified metabolite
of concern?
46
-------
CONCLUSION
47
-------
BIOLOGICAL DATA FOR PHARMACOKINETIC MODELING AND RISK ASSESSMENT:
RESEARCH NEEDS
Kenneth B. Bischoff and Irene B. Glowinski
One of the goals of the workshop was to identify research or data needs for pharmacokinetic
models in risk assessment. The following list of research and data needs, compiled from the workshop
transcript, summarizes comments made by workshop participants during the conference summation and
the final overview and wrap-up sessions.
Suggestions given by workshop participants are grouped into several broad categories, beginning
with pharmacokinetic models, which was the focus of the workshop. Research that defines
pharmacodynamic models, especially with a stronger basis in biology, is the topic of the next two
categories and possibly constitutes a greater need than pharmacokinetic models. The fourth group of
research needs, model parameters, is divided into physiological and physiochemical parameters, and
metabolic parameters which are required for both pharmacokinetic and pharmacodynamic models.
Research Needs
1. Pharmacokinetic Models
The basis for and the necessary detail required in any pharmacokinetic model, as well as the
uncertainties, must be better defined.
More care is needed in delineating assumptions used in a particular model.
There is a need to define the most critical assumptions and whether these can be validated.
Sophisticated statistical techniques should be applied to physiologically based pharmacokinetic
models in addition to classical pharmacokinetic models.
Physiologically based pharmacokinetic models must be developed for toxic compounds with very
long lives, such as heavy metals.
Pharmacokinetic models must be developed at the intracellular level.
48
-------
2. Basis for Pharmacodynamic Effects
The basis for appropriate definition of metabolic or other biological markers as surrogates for a
toxic agent must be better understood.
The relationship between the delivered dose to a tissue and the biologically effective dose for
the endpoint in question must be better defined. For example, researchers should determine
whether their conclusions are species-dependent.
A more ^fundamental basis is needed for choosing the appropriate dose-metric as the proper
measure of toxicity. Examples of the dose-metric could be dose rate (mg/kg/day), a peak
concentration (mg/L), a lifetime dose (mg/kg/Iifetime), or, for carcinogenesis, related to the cell
cycle time. Also, it must be determined if the dose-metric is species-dependent.
There is a need to analyze the assumptions involved when using a marker of toxicity that may
not correctly identify other organ systems. For example, DNA adducts may not adequately
predict effects on the immune system.
Known human carcinogens must be studied in animals and the results of these studies should
be correlated.
3. Pharmacodynamic Models
Biologically motivated pharmacodynamic models must be developed using the same philosophy for
physiologically based pharmacokinetic models, i.e., the use of biochemical and toxicological concepts in
formulating pharmacodynamic models must increase.
Biologically based pharmacodynamic models must be developed that account for possible
changes in the model and associated parameters during different biological states. Also,
carcinogenesis models other than the two-stage model should be considered.
It must be realized that there are endpoints other than cancer for which models must be
developed.
4. Model Parameters
Both pharmacokinetic and pharmacodynamic models require various types of parameters that need
to be numerically identified from experimental measurements. Some of these parameters can often be
obtained from literature, but others are more substance- and/or species-specific, and must be directly
measured.
49
-------
More studies should examine the effect of modifying factors on pharmacokinetic and
pharmacodynamic parameters such as sex, strain, age, dietary factors, environmental factors, and
other changes in physiological conditions.
Modern sophisticated, noninvasive techniques such as NMR and PET scanning should be used
for measurement of pharmacokinetic parameters.
Methods of model reduction and any subsequent loss of information should be studied.
Model predictive uncertainty (range of outputs) should be examined in more depth using
statistical simulation methods ;such as the Monte Carlo method.
A. Physiological and Phvsiochemical Parameters
More accurate values of anatomical-physiological parameters, such as blood flow rates and
tissue volumes in several species, must be developed.
A consensus must be reached for the most representative anatomical and physiological
parameters for different species.
In order to have a model adequately predict desired tissue levels, researchers need to
understand if there is a need to. account for minute-to-minute changes in organ blood flow and
volumes.
Existing partition coefficients and binding data should be evaluated for accuracy.
Standardized measurement techniques must be developed for physiological and physicochemical
parameters.
B. Metabolic Parameters :
With respect to Michaelis-Menten kinetics, the relationship between Vmax and K,n values in
various systems must be understood, ranging from isolated enzymes up to whole organs and
animals.
There is a need to understand how to assemble the kinetics of isolated enzymes into networks
of enzyme reactions.
The translation of Vmax and K^ values between species should be better understood.
Rate equations other than those of the Michaelis-Menten form must be considered.
More information on the distribution of enzymes in different species should be obtained.
Enzyme induction must be further defined, especially concerning acute vs. chronic toxicity and
high vs. low dose.
50
-------
5. Other Sources of Information
Valuable data exist in other fields that could be utilized in toxicology and risk assessment. Detailed
techniques that take advantage of existing data in these fields (e.g., radiation biology, cell kinetics, and
toxic effects of drugs) should be explored.
6. Glossary
In order to promote cross-disciplinary research in pharmacokinetics and pharmacodynamics, a
glossary of modeling terms with appropriate examples should be developed.
51
-------
-------
APPENDIX A
AGENDA
A-l
-------
WORKSHOP ON BIOLOGICAL DATA FOR PHARMACOKINETIC
MODELING AND RISK ASSESSMENT
May 23-25, 1988
Sponsored by:
ILSI Risk Science Institute
U.S. Environmental Protection Agency
The Dow Chemical Company
American Industrial Health Council
American Petroleum Institute
Mobil Oil Corporation
The Procter & Gamble Company
Shell Oil Company
WORKSHOP AGENDA
Monday. May 23. 1988
8:30am Welcome
Kenneth B. Bischoff
8:45am I. Overview of Risk Assessment
Chairperson - William Farland
Overview/Introduction
Panel Presentations
8:55am (1) Concepts - academic or theoretical discussion of
risk assessment.
Nicholas A. Ashford
Discussion
9:30am (2) Practices - how data are used in a risk
assessment.
Joseph V. Rodricks
Discussion
10:05am (3) Assumptions - highlight and identify where
pharmacokinetic modeling can reduce the uncertainty.
Christopher Portier
Discussion
10:40am COFFEE BREAK
10:55am General Discussion
ll:25am II. Modeling Biological Phenomena
Chairperson - Kenneth B. Bischoff
Overview/Introduction
A-2
-------
ll:40am Introduction to Modeling
Mary E. Davis
12:10pm General Discussion
12:30pm LUNCH BREAK
l:45pm III. Overview of Pharmacokinetics Models
Chairperson - Robert L. Dedrick
Overview/Introduction
Panel Presentations
l:55pm (1) Development and history of pharmacokinetic modeling.
Kenneth B. Bischoff
Discussion
2:30pm (2) Purposes: What models can do and their limitations.
Robert L. Dedrick
Discussion
3:05pm COFFEE BREAK
3:20pm (3) Data needs for modeling.
Harvey Clewell III
Discussion
3:55pm (4) Major uncertainties in pharmacokinetic modeling and
sensitivity analysis.
Murray S. Cohn
Discussion
4:30pm General Discussion
5:00pm ADJOURN
Tuesday. May 24. 1988
8:45am IV. The Use of Experimental Data in Pharmacokinetic
Modeling
Chairperson - Richard W. D'Souza
Introduction
Panel Presentations
9:00am (1) Hepatic metabolism (rates/pathways).
Marilyn E. Morris
Discussion
A-3
-------
9:35am (2) Cross-species scaling.
Joyce Mordenti
Discussion
10:10am COFFEE BREAK
10:25ara
(3) Correlation of in vitro and in vivo data.
I. Glenn Sipes
Discussion
ll:00am (4) Target tissue/cell dose/chemical carcinogens.
George Lucier, Steve Belinsky, and
Claudia Thompson
Discussion
ll:35am General Discussion
12:00pm LUNCH BREAK
l:30pm V. The Use of Pharmacokinetic Modeling in Risk Assessment/Case Studies
Chairperson - Linda S. Birnbaum
Introduction
Analysis of Different Models and Compounds
l:40pm (1) Benzene
Model - Michele A Medinsky
Risk Assessment - A. John Bailer
Comment - Richard Irons
2:50pm COFFEE BREAK
3:05pm (2) Butadiene
Model - Dale Hattis
Risk Assessment - Steven Bayard
Comment - Richard Irons
4:30pm ADJOURN
Wednesday, May 25. 1988
8:45am VI. The Use of Pharmacokinetic Modeling in Risk Assessment/Case Studies
(continued)
Introduction
A-4
-------
Analysis of Different Models and Compounds (cont'd.)
8:55am (3) Methylene Chloride
Model - Inhalation Data - Richard H. Reitz
Model - Ingestion Data - Michael J. Angelo
Risk Assessment - Jerry Blancato and
Lorenz Rhomberg
Comment - Rory B. Conolly
10:35am COFFEE BREAK
10:50am General Discussion
ll:20am VIII. Conference Summation
Chairperson - Kenneth B. Bischoff
Introduction
(Reports of Rapporteurs)
ll:30am (1) Basic biological research needs.
Alan Wilson
ll:50am (2) Chemical-specific needs.
Hugh Spitzer
12:10pm
12:30pm
12:45pm ADJOURN
(3) Applied research needs.
Angelo Turturro
OVERVIEW AND WRAP-UP
Kenneth B. Bischoff
A-5
-------
-------
APPENDIX B
ABSTRACTS OF PRESENTATIONS
B-l
-------
SESSION I: OVERVIEW OF RISK ASSESSMENT
CONCEPTS: ACADEMIC OR THEORETICAL DISCUSSION OF RISK ASSESSMENT
Nicholas A. Ashford
This presentation provides a framework for considering values in the use of science in the
regulatory process. The science in question includes both the assessment of technological risk and the
assessment of technological options to reduce those risks. The focus of the inquiry is on the role of the
scientist and engineer as analysts or assessors. The difficulties in separating facts and values are
addressed by focusing on the central question: What level of evidence is sufficient to trigger a
requirement for regulatory action? For the purposes of this discussion, the regulatory process will
include notification of risks to interested parties, control of technological hazards, and compensation for
harm caused by technology. The discussion addresses the problems in achieving both a fair outcome and
a fair process in the regulatory use of science.
References
Ashford, N.A, 1988. "Science and values in the regulatory process," Statistical Science. Institute of
Mathematical Statistics (to be published).
Ashford, N.A., 1984. "Advisory committees in OSHA and EPA: Their use in regulatory decision-
making." Science. Technology, and Human Values Winter:72-81.
Hattis, D. and Smith, J.A. Jr., 1986.: "What's wrong with quantitative risk assessment?" In: Biomedical
Ethics Review 1986; Almeder, R.I. and J.M. Huber (eds.). The Humana Press.
!
Nader, R., 1974. "Obligation of scientists to respond to society's needs." In: Scientist in the Legal
stem; Thomas, W.A. (ed.). Ann Arbor, MI: Ann Arbor Science Publishers, Inc.
Rushefsky, M.E., 1986. Making Cancer Policy. State University of New York Press.
Whittemore, A.S., 1983. "Facts and values in risk analysis for environmental toxicants." Risk Analysis
(March): 22-23.
B-2
-------
SESSION I: OVERVIEW OF RISK ASSESSMENT
USE OF DATA IN RISK ASSESSMENT
Joseph V. Rodricks
At our present stage of understanding, the completion of risk assessment requires not only the use
of traditional forms of scientific data, but also the imposition of a number of assumptions and models
that have not been subjected to thorough empirical verification. The risk assessor is forced to adopt
such assumptions and models because no direct methods are available to measure many small, or even
moderately large, risks that have proven to be of sufficient social concern to have prompted the
enactment of regulatory laws.
To achieve consistency in the absence of scientific certainty, regulatory agencies have typically
adopted assumptions and models that are applied generically to all agents. It is becoming clear,
however, that increased understanding of the chemical and biological behavior underlying the production
of toxicity can lead to the substitution of agent-specific information for generic models and assumptions.
Nowhere is the potential for using such information to improve risk assessment greater than in the area
of pharmacokinetics.
Because no data concerning underlying mechanisms of toxicity (in which I include pharmacokinetic
and metabolism data) can be expected to answer all questions, the use of such data in risk assessments
will necessarily introduce uncertainties. Regulators will be reluctant to use such information if the risks
of specific agents are revealed to be less than those predicted under the use of generic models (i.e.,
because there is uncertainty associated with the lower risks, the fear arises that risks have been
underestimated). Of course, as long as regulatory agencies reject the use of data on mechanisms of
toxicity because they do not answer all questions, research scientists will have little incentive to develop
such data. This unhealthy situation can be remedied if risk assessors strive to use all available data on
mechanisms in their assessments and then exhaustively describe for decision-makers the relative degrees
of scientific support merited by results based on the use of generic models and those based on the use of
mechanistic information. Decision-makers are then free to choose from among several possible risk
estimates, based on the scientists' best judgment regarding their relative merits.
If research scientists ask that risk assessors adopt such an approach, the research scientists should
be sensitive to the types of dilemmas that the use of mechanistic information creates for the risk
B-3
-------
assessor. In the area of pharmacokinetics, three major questions are typically encountered in the use of
data in risk assessment:
1. How can we be certain that the metabolites whose pharmacokinetic behavior we are measuring
or modeling are responsible for the observed toxicity?
2. How can we be certain that pharmacokinetic behavior observed over a relatively short period of
dosing can be expected to hold over periods of long-term exposure?
3. How can we be certain that pharmacokinetic behavior observed in experimental animals closely
resembles behavior in humans?
Perhaps we shall find answers to these questions at this workshop.
References
Anderson, M.E., Clewell, H.J. II, Gargas, M.L., Smith, F.A., and Reitz, R.H., 1987. "Physiologically
based pharmacokinetics and the risk assessment process for methylene chloride." Toxicol. Appl.
Pharmacol. 87:185-205.
Anderson, M.W., Hoel, D.G., and Kaplan, N.L., 1980. "A general scheme for the incorporation of
pharmacokinetics in low-dose risk estimation for chemical carcinogenesis: Example - vinyl
ihloride." Toxicol. Appl. Pharmacol. 55:154-161.
Gehring, P.J., Watanabe, P.O., and Young, J.D., 1977. "The relevance of dose-dependent
pharmacokinetics in the assessment of carcinogenic hazard of chemicals." In: Origins of Human
Cancer. Book A; Hiat H.H., Watson, J.D., and Winsten, J.A. (eds.). Cold Spring Harbor
Laboratory, Cold Spring Harbor, pp. 187-203.
Hoel, D.G., Kaplan, N.L., and Anderson, M.W., 1983. "Implications of nonlinear kinetics on risk
estimation in, carcinogenesis." Science 219:1032-1037.
Whittemore, A.S., Grosser, S.C., and Silvers, A., 1986. "Pharmacokinetics in low dose extrapolation
using animal cancer data." Fundam. Appl. Toxicol. 7:183-190.
B-4
-------
SESSION I: OVERVIEW OF RISK ASSESSMENT
THE IMPLICATIONS OF USING MORE COMPLICATED MODELS FOR
RISK ESTIMATION IN CARCINOGENESIS
C.J. Portier and N.L. Kaplan
Quantitative risk assessment of mathematical models that have biologically interpretable parameters
is becoming more common. The current practice is to obtain estimates of these parameters and ignore
their intrapopulation variability when estimating safe-dose levels. If some of the model parameters have
large intrapopulation variability (e.g., metabolic parameters in humans), then this procedure is likely to
substantially underestimate the population variability of the safe dose.
The purpose of this investigation is to describe a method based upon Monte Carlo resampling
(Efron, 1982; Portier and Bailer,> 1987) for including the intrapopulation variability of biologically
interpretable parameters in the risk estimation process. As an example, the physiologically based
pharmacokinetic model used by Anderson et al. (1987) for estimating a safe-exposure dose for methylene
chloride was reexamined. Their model included 23 biologically interpretable parameters consisting of
tissue weights, blood flow rates, partition coefficients, and metabolic constants. The results indicate that
the intrapopulation variability of the model parameters can have a substantial effect on the distribution
of the safe doses, by broadening the range of possible values and thus increasing the standard deviation
and lowering the fifth percentile. Where the intrapopulation variability was assumed to be small in both
animals and humans, the standard deviation of the safe dose more than doubled, and the fifth percentile
decreased by a factor of 1.5. Where the metabolic parameters in humans were assumed to have a high
degree of variability, the standard deviation of the safe dose increased by a factor of more than 10, and
the fifth percentile decreased by a factor of more than 7, actually falling below the corresponding EPA
lower bound (U.S. EPA, 1985).
Variability analysis is an important tool in the estimation of risk from exposure to chemical
carcinogens. The methods described here can be used to study the variability of safe-dose estimates for
mathematical models with biologically interpretable parameters estimated from different studies. As our
understanding of the carcinogenic process improves, the models used for quantitative risk assessment will
become more complex and there will be an even greater need for the numerical methods described here.
B-5
-------
The results of our study suggest that intrapopulation variability of the model parameters can
increase the variability of safe dose estimates an appreciable amount. This increased variability
represents a better estimate of the overall uncertainty in the risk assessment process.
References
Anderson, M., Clewell, H., Gargas, R, Smith, R, and Reitz, R., 1987. "Physiologically based
pharmacokinetics and the risk assessment process for methylene chloride." Toxicol. Appl.
Pharmacol. 87:185-205.
Efron, B., 1982. "The jackknife, bootstrap, and other resampling plans." Philadelphia: SIAM.
Mennear, J.H., McConnell, E.E., Huff, J.E., Renne, R.A., and Giddens, K, 1988. "Inhalation toxicology
and carcinogenesis studies of methylene chloride (dichloromethane) in F344/N rats and B6C3Fi
mice. Ann. NY Acad. Sci. 534:343-351.
National Toxicology Program, 1985. Technical report on the toxicology and carcinogenesis studies of
dichloromethane in F344/N rats and B6C3Ft mice (inhalation) studies. NTP Technical Report
Series No. 306, Research Triangle Park, NC.
Portier, C. and Bailer, A.J., 1987. "Simulating failure times when the event of interest is unobservable
with emphasis on animal carcinogenicity experiments." Comput. Biomed. Res. 20:458-466.
Portier, C. and Kaplan, N., 1989. "The variability of safe dose estimates when using complicated models
of the carcinogenic process. A case study: methylene chloride." Unpublished study.
U.S. Environmental Protection Agency, 1985. "Addendum to the health assessment document for
dichloromethane (methylene chloride). Updated carcinogenicity assessment for dichloromethane."
Office of Health and Environmental Assessment, Washington, DC. EPA/600/8-82/004F.
U.S. Environmental Protection Agency, 1987. "Technical analysis of new methods and data regarding
dichloromethane hazard assessments." Office of Health and Environmental Assessment,
Washington, DC. EPA/600/8-87/029A.
U.S. Environmental Protection Agency, 1988. "A cancer risk-specific dose estimate for 2,3,7,8-TCDD.
Office of Health and Environmental Assessment, Washington, DC. EPA/600/6-86/007a and b.
B-6
-------
SESSION II: MODELING BIOLOGICAL PHENOMENA
INTRODUCTION TO MODELING
Mary E. Davis
Mathematical or quantitative models are used in our attempts to understand nature, to summarize
data, and to relate that data to the processes or systems included in the models. Mathematical or
quantitative models are rules describing how the component parts of a system interact with each other,
often in response to external forces or perturbations; it is the interaction between these component parts
that is of interest. Models are formulated from data that characterize a system. If the model predicts
the actual behavior of that system then there is some confidence that the relationship among the
components of the model is similar to their relationship in reality. This belief is strengthened if the
model can predict behavior beyond the extremes of data used to formulate the relationship the model
describes.
Classical, descriptive models are appealing because they require relatively modest amounts of data
and yet reveal much understanding of how a compound is handled. To develop a compartmental model
for absorption, distribution, and elimination of a compound, the data required are derived from the
concentration of that compound in plasma or its excretion. A model of one or more compartments is
then fit to the data, allowing the number and size of the compartments and the transfer between
compartments to vary in order to fit the data more closely. This type of model describes the time-
course-of-exposure and steady-state conditions well and can detect differences in the handling of a
compound between experimental groups. However, because the compartments have no physiologic
meaning, and the kinetic constants are hybrids reflecting several processes, one cannot determine what
changes have occurred to cause experimental groups to differ. And because the constants are not related
to distinct biological processes, it is not possible to extrapolate beyond the conditions of measurement
(that is, across species or sexes, or to where animal growth is occurring).
In contrast to classical pharmacokinetic models are physiologically based pharmacokinetic models.
The two types differ in several important respects: (1) the input data that drive the model; (2) the
amount of prior knowledge needed to develop a good model; and (3) the type of information the model
produces and how the model, with its information, can be used. In essence, physiologically based
pharmacokinetic models describe the disposition of the compound of interest as it actually occurs in the
organism. Instead of using amorphous compartments of variable size, the compartments used have an
B-7
-------
anatomical basis, representing individual organs, with their respective volumes. Rather than considering
single rates of elimination, movement of the compound into and out of the organ is taken into account,
e.g., blood flow to the organ, as well as the compound's intrinsic ability to cross the membranes
involved. Several organs can be grouped together if the organs handle the compound similarly and do
not need to be modeled separately for other reasons (e.g., a target organ of a site of metabolism or
elimination). The detail of the model depends upon the compound being modeled and the model's uses.
For some chemicals, it is important to include the turnover of macromolecules involved in defending
against toxicity (e.g., glutathione, binding proteins), and for carcinogenesis, the rates of DNA repair and
cell replication are needed to accurately describe the response. Physiologically based pharmacokinetic
models require more and different types of information, including physiological and biochemical data
(e.g., organ volumes, blood flow, rates of metabolism and excretion), physical data (e.g., diffusion,
permeability, protein binding), and an understanding of the biology of the risk being modeled.
After the data are collected, the model is formulated and compared to actual observations. If the
predictions differ from the observed values, then the model is refined, adding the processes that were
omitted but that are important to the compound's disposition. Thus, physiologically based
pharmacokinetic models help guide experimental design. Because the constants used relate to
physiological entities that can be measured in other animal species, physiologically based pharmacokinetic
models can predict behavior of a compound in other species as well, and do particularly well if there is
information on the metabolism of those species. In vitro experiments can provide much of the needed
metabolic information, so physiologically based pharmacokinetic models are attractive for extrapolating
from experimental animals to humans. Similarly, the metabolic and clearance parameters can deal with
saturation of those processes, and thus physiologically based pharmacokinetic models are amenable to
extrapolating beyond the exposures actually used, or into populations with impaired metabolic functions
such as the young and old, and other sensitive populations. Exposure to chemical mixtures is another
area in which physiologically based pharmacokinetic models can help guide research efforts because the
interactions between many chemicals are due to competition for metabolism or elimination, or other
changes that alter distribution.
For physiologically based pharmacokinetics models to achieve full potential in risk assessment there
needs to be a better understanding of the biological processes underlying risks. Often only a small
proportion of an exposed population develops tumors; the mechanisms for individual differences in
susceptibility are not understood, yet are important for minimizing societal risk. Physiologically based
pharmacokinetics has the potential to make great contributions in applying our understanding of species
B-8
-------
differences in reproduction and development, for example, to protecting developing organisms while
administering beneficial therapies to the parent.
References
Committee of Methods for the in vitro Testing of Complex Mixtures, 1988. "Interpretation and
modeling of toxicity-test results." In: Complex Mixtures: Methods for In Vitro Toxicity Testing.
National Academy Press, pp. 99-124.
Lutz, R.J. and Dedrick, R.L., 1985. "Physiological pharmacokinetics: relevance to human risk
assessment." In: New Approaches in Toxicity Testing and Their Application in Human Risk
Assessment. Li, A.P. (ed.). New York: Raven Press, pp. 129-149.
Subcommittee on Pharmacokinetics in Risk Assessment, 1987. In: Pharmacokinetics in Risk
Assessment. Vol. 8 of Drinking Water and Health. Washington, DC: National Academy Press.
B-9
-------
SESSION III: OVERVIEW OF PHARMACOKINETIC MODELS
DEVELOPMENT AND HISTORY OF PHARMACOKINETIC MODELS
Kenneth B. Bischoff
This presentation will describe the origins of pharmacokinetics and the reasons why
pharmacokinetics has been found useful in research and applications of pharmacology and toxicology.
The main purpose is to quantitate the results and to formulate predictive mathematical models for
absorption, disposition, metabolism* excretion, and biological response for clinical applications and risk
assessment.
A brief review of the history of pharmacokinetics will be given, focusing on the introduction of key
concepts. (A comprehensive history is given by J. Wagner (1981).) Then the more recent development
of physiologically based pharmacokinetic (PBPK) models will be described, again focusing on introducing
key ideas.
Early work, beginning around the turn of the century, was concerned with anesthetic agents, since
knowledge of specific brain concentrations seemed important. Ethanol and salicylate were also studied,
and Michaelis-Menten kinetics and renal clearance were defined. The one-compartment open model was
developed by E. Widmark and J. Tandberg (1924), and R. Dominquez (1934) studied creatinine and
defined volume of distribution. F. Dost (1953) apparently coined the term "pharmacokinetik" (in
German). During the 50s and 60s, 'many researchers were involved: B. Brodie, S. Riegelman, E. Nelson,
J. Wagner, E. Garrett, P. Wiegand, E. Kruger-Thiemer, G. Levy, J. Gillette, M. Gibaldi, M. Rowland, L.
Benet, and R. Jelliffe. (See the excellent text by Gibaldi and Perrier (1982).)
Physiological pharmacokinetics attempts to base models on real biological data concerning the
physicochemical, physiological, and pharmacological properties of a drug, rather than using abstract
definitions of, for example, central and peripheral compartments. Thus, organ blood flow rates, liquid
solubility, protein binding, and local membrane permeability are all specifically incorporated into the
models, and this level of detail permits translation of results between animal species and prediction of
effects from different routes of administration and different doses.
This concept was actually proposed 50 years ago by T. Teorell (1937), whose model consisted of the
circulatory system, drug depot, fluid volume, kidney elimination, and tissue inactivation. However, he
B-10
-------
could mathematically solve only a greatly simplified form of model. J. Jacquez, R. Bellman, and R.
Kalaba (1960) considered a two-region model with six compartments, but as computers became readily
available, K. Bischoff and R. Brown (1966) developed a multiregion computer model based on chemical-
engineering mass-balance concepts. Adding Michaelis-Menten kinetics and nonlinear Langmuir protein
binding, K. Bischoff and R. Dedrick (1968) formulated a model for thiopental, and then in 1971, they
studied methotrexate pharmacokinetics in several species. R. Dedrick (1973) specifically discussed i
interspecies scaling and first used the term "physiological pharmacokinetics." Since then, many
investigators have applied these concepts to pharmacology and toxicology. Several reviews are provided
below.
References
Bischoff, K.B., 1987. "Physiologically based pharmacokinetic modeling." In: Proc. Workshop
Pharmacokinetics in Risk Assessment. Vol. 8 of Drinking Water and Health. National Academy
Press, pp. 36-61.
Chen, H.-S.G. and Gross, J.F., 1979. "Physiologically based pharmacokinetic models for anticancer
drugs." Cancer Chemother. Pharmacol. 2:88-94.
i
Gerlowski, L.E. and Jain, R.K., 1983. "Physiologically based pharmacokinetic modeling: Principles and
applications." J. Pharm. Sci. 72:1103-1127.
Gibaldi, M. and Perrier, D., 1982. Pharmacokinetics. 2nd ed. New York, NY: Marcel Dekker.
Himmelstein, K.J. and Lutz, R.J., 1979. "A review of the applications of physiologically based
pharmacokinetic modeling." J. Pharmacokin. Biopharm. 7:127-145.
Wagner, J.G., 1981. "History of Pharmacokinetics." Pharmacol. Ther. 12:537-562.
B-ll
-------
SESSION III: OVERVIEW OF PHARMACOKINETIC MODELS
WHAT MODELS CAN DO AND THEIR LIMITATIONS
Robert L. Dedrick
Pharmacokinetics has been used extensively in pharmacology for the design and analysis of
preclinical and clinical studies and the guidance of therapy. Classical compartmental, physiologic, and
model-independent approaches have been used. All have important applications depending on the
purpose of the investigation. This presentation emphasizes physiologic models because these generally
provide greater mechanistic insight than other commonly used approaches. Several topics are included.
A. In Vitro-In Vivo Correlations
Xenobiotic metabolism is frequently studied in model systems such as extracted enzymes, tissue
homogenates, cell suspensions, tissue slices, and isolated perfused organs. These systems permit
investigator control of the reaction conditions and uncouple the metabolism from the complexities of
chemical and metabolite distribution in the whole animal. Data that predict rates of metabolism in the
body must be used correctly to account properly for the actual conditions that exist at the site(s) of
metabolism with respect to such effects as blood flow, membrane transport, and binding. Published
literature has been generally supportive of the concept that reasonable predictability may be obtained,
but the literature is still limited.
B. Species Similarities and Differences
An extensive literature exists on the anatomy and physiology of animals. Some of this can be
adapted to physiologic pharmacokinetic models directly; alternatively, established correlations can be used
to derive plausible estimates of relevant parameters. Capillary permeability, for example, appears quite
similar among mammals for a particular capillary or tissue type. Empirical allometric correlations have
successfully displayed the orderly interspecies variation with body weight of a variety of organ sizes and
physiologic functions; however, ,intraspecies correlations of this type may differ or be unsuccessful.
Xenobiotic metabolism often shows large and unpredictable variations among species; experimental data
generally will be required for each particular chemical and animal.
B-12
-------
C. Dose Schedule and Unavailability
Pharmacokinetic theory fpr determining bioavailability is well established. Also, methodology exists
for predicting plasma concentrations as a function of dose schedule. A single bolus dose, for example,
can be used to characterize drug pharmacokinetics in a particular subject and the parameters derived
used to calculate an infusion schedule with a particular concentration objective. Errors can be made if
nonlinearities are not correctly described or if pharmacokinetic parameters change with time as a result
of drug effects, growth of the animal, or other causes.
D. Regional Drug Administration
Drugs have been applied to the skin, infused into arteries and tissues, instilled in a variety of body
cavities, and inhaled into the lungs. Some interesting and important considerations relate to the
distributed character of regional drug administration because the compartments are not well mixed.
Diffusion, convection, and mixing are involved, and these processes are less well understood than the
basic idea of pharmacokinetic advantage between compartments with uniform concentrations.
E. Biological Response Models
Pharmacokinetics is limited to describing and predicting the physical and chemical aspects of the
distribution of drugs and other chemicals in the body. Pharmacodynamic information is required to
associate drug concentrations (or concentration history) with a biological effect. The paucity of adequate
biological models is a major impediment to the rational use of pharmacokinetic theory in risk
assessment. Current pharmacokinetic theory can provide much information and often substantial
predictive power concerning questions of dose-to-dose, route-to-route, and species-to-species
pharmacokinetics. Pharmacokinetics by itself tells us nothing about the probability of a chemical causing
cancer in humans or even the appropriate dose metric to be used for predicting human cancer risk from
studies on experimental animals.
References
Dedrick, R.L., 1986. "Interspecies scaling of regional drug delivery." J. Pharm. Sci. 75:1047-1052.
Dedrick, R.L. and Bischoff, K.B., 1980. "Species similarities in pharmacokinetics." Fed. Proc. 39:54-59.
B-13
-------
Dedrick, R.L., Forrester, D.D., and Ho D.H.W., 1972. "In vitro-in vivo correlation of drug metabolism
deamination of l-B-D-arabinpfuranosylcytosine." Biochem. Pharmacol. 21:1-16.
King, F.G., Dedrick, R.L., and Farris, F.F., 1986. "Physiological pharmacokinetic modeling of cis-
dichlorodiammineplatinum(II) (DDP) in several species." J. Pharmacokinet. Biopharm. 14:131-155.
B-14
-------
SESSION III: OVERVIEW OF PHARMACOKINETICS MODELS
DATA NEEDS FOR PHARMACOKINETIC MODELING IN RISK ASSESSMENT
Harvey Clewell, III
The purpose of applying pharmacokinetic models to chemical risk assessment is to increase the
scientific basis of the process and to reduce the level of uncertainty in the result. Pharmacokinetic
models can aid in the experimental design of bioassays. They also permit the incorporation of biological
information into quantitative risk estimates, serving not only to document the approach taken by the risk
assessor but also to highlight those areas where the approach is most sensitive to error. The best time
to develop a model is in parallel with initial data collection, so that the data collection can be directed
by the model and discrepancies between the data and the model can be verified and used to refine the
model. A considerable body of data is needed to develop a physiologically based pharmacokinetic model.
Qualitative information on the toxic effects of the chemical, its metabolism, and the mechanism of action
is needed to define the model structure. Quantitative information must also be collected both to
determine the parameters of the model and to check the model's validity.
The next logical step in the application of biologically motivated mathematic models for risk
assessment is the linking of pharmacokinetic models to models of carcinogenesis. Crucial to this next
step is the development of a body of data on typical values of the parameters used in these models.
These needed data include cell turnover rates, mutation rates, and DNA repair rates in both humans and
rodents, as well as measures of the impact of specific chemicals on these rates. The result will be a
quantitative risk assessment that incorporates not only the uptake, distribution, and metabolism of a
chemical but also the specific mechanism by which it leads to tumor formation.
There are at least seven levels of uncertainty in quantitative chemical risk assessment. They are:
1. Classification of the chemical as a carcinogen
2. The mechanism of carcinogenesis
3. The nature of the proximate carcinogen
4. The structure of the PK model
5. The parameters in the PK model
B-15
-------
6. Species differences in sensitivity
7. Extrapolation to ultra-low doses
Of these, only the fifth has been treated in any quantitative fashion to date. The seventh is, by policy,
handled by the linearized multistage model. The first four are generally treated simply as decisions to be
made, and there are not as yet any accepted methodologies for reflecting the uncertainty in these
decisions in the final outcome. The \ sixth has become the refuge for a safety factor that was originally
derived inappropriately from the toxicity across species of a particular class of chemicals. As PK models
are extended into the process of carcinogenesis, all of these uncertainties will become quantifiable, at
least to some extent.
There are several related issues that tend to divide the risk assessment community. On the one
extreme are the risk-averse, those who are concerned primarily with estimating "safe" absolute risk
estimates on chemicals. Given the greater perceived cost of underestimating risk as opposed to
overestimating it, these individuals tend to compound conservative assumptions and resist any changes
that would lower the risk estimate because of their fear that the residual uncertainty in the process could
led to an overall underestimate of risk. On the other extreme are the risk-tolerant, those who are
concerned primarily with accurate relative risk estimates. Motivated by a fear that inaccurate
information could lead to erroneous risk-management decisions, these individuals tend to favor the rapid
incorporation of new techniques in the risk assessment process and are opposed to redundant safety
factors and risk assessment calculations that ignore chemical differences. Since the true risk for most
regulated carcinogens may never be known, it is imperative that these two communities work together to
reduce or deal with uncertainties in a mutually satisfactory manner.
There are four areas in particular that could greatly improve the process of chemical risk
assessment:
The usefulness of carcinogenic bioassays would be greatly improved if PK models were used in
the design of the studies and if animals were included in the study specifically for the
assessment of changes in PK parameters throughout the entire duration of the study. In
addition, time-to-tumor and time-to-death data should always be recorded.
The legislative agencies charged with performing chemical risk assessments should have the
support of PK research groups possessing both modeling expertise and PK data acquisition
capability. !
Basic data are critically needed on parameters required for biologically motivated cancer
models, particularly stem cell population data, cell turnover rates, and mutation/repair rates in
rodents and man.
B-16
-------
Formal methods are needed for incorporating the uncertainty in qualitative risk assessment
decisions into the quantitative outcome.
References
Greenfield, R.E., Ellwein, L.B., and S.M. Cohen, 1984. "A general probabilistic model of carcinogenesis:
Analysis of experimental urinary bladder cancer.11 Carcinogenesis 5:437-455.
Moolgavkar, S.H., 1986. "Carcinogenesis modeling from molecular biology to epidemiology." Ann Rev
Public Health 7:151-169. ?
Moolgavkar, S.H. and AG. Knudson, Jr., 1981. "Mutation and cancer: A model for human
carcinogenesis." J. Natl. Cancer Inst. 66:1037-1052.
Munro, I.C. and D.R. Krewski, 1981. "Risk assessment and regulatory decision making.11 Food Cosmet
Toxicol. 19:549-560.
National Academy of Science, 1987. Pharmacokinetics in Risk Assessment. Vol. 8 of Drinking Water
and Health. Washington, DC: National Academy Press.
Office of Science and Technology Policy, 1985. "Chemical carcinogens; a review of the science and its
associated principles, February 1985." Fed. Reg. Part 11:10371-10442 (March 14, 1985).
B-17
-------
SESSION III: OVERVIEW OF PHARMACOKINETIC MODELS
MAJOR UNCERTAINTIES IN PHARMACOKINETIC MODELING AND SENSITIVITY ANALYSIS
Murray S. Cohn
Pharmacokinetics is useful in the risk assessment process, provided that data exist for extrapolations
between routes -- from one exposure scenario to another, from high to low dose, and from species to
species. It is important to realize the uncertainties inherent in the use of pharmacokinetic information;
some of these uncertainties are examined in the referenced works. The most important research,
however, will concern the use of pharmacokinetic information for species-to-species extrapolation in the
carcinogenic risk assessment process. At this time, it is my opinion that pharmacokinetics can be used
to describe the "target site" dose in various species, but that the relative effects of such doses, i.e., tissue
sensitivity in various species (pharmacodynamics), are unknown. This gap precludes using such
information for risk assessment purposes.
Additional research is needed in two areas. The first is examination of pharmacodynamic processes,
either through empirical examination or by modeling, to try to ascertain the relationship between target-
site dose and effect. This approach can be very complex; research areas include DNA repair, cell cloning
rates (of transformed cells), cell death rates, normal cell turnover rates, immunological responses to
transformed cells, number of cells at risk etc., all of which obviously depend upon a thorough
understanding of the mechanism of carcinogenic transformation in all species being considered.
Although this approach should be pursued, any meaningful output is years away.
The second approach may have more immediate use. There are a number of known human
carcinogens, all of which are carcinogenic in animals when appropriately tested. If proper
pharmacokinetic studies were to be done for humans and all relevant animal species tested with these
various carcinogens, one might begin to understand the magnitude of pharmacodynamic effects, since
both total effects (responses to applied dose in humans and animals) and pharmacokinetic effects will
have been determined or estimated. We may find that pharmacodynamic effects are fairly constant for
various species pairs of a given general mechanistic type (e.g., initiation, cytotoxicity). This approach
requires pharmacokinetic research and modeling of the type described at this workshop, and a careful
analysis of the correlations that may become apparent. Once we know the approximate magnitudes of
such effects, it may be easier to plan the studies needed to conduct the necessary experimentation via the
first approach described above. :
B-18
-------
References
Cohn, M.S., 1987. "Sensitivity analysis in pharmacokinetic modeling." In: Proc. Workshop
Pharmacokinetics in Risk Assessment. Vol. 8 of Drinking Water and Health. Washington DC-
National Academy Press, pp. 265-272. '
Cohn, M.S., 1987. "Updated risk assessment for methylene chloride (dichloromethane)." Included as
part of the June 1987, Briefing Package to the Commissioners of the Consumer Product Safety
Commission on methylene chloride.
U.S. Environmental Protection Agency, 1987. "Technical analysis of new methods and data regarding
dichloromethane hazard assessments." Office of Health and Environmental Assessment
Washington, DC. EPA/600/8-87/029A.
B-19
-------
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC MODELING
HEPATIC METABOLISM
Marilyn E. Morris
In risk assessment there is a need to evaluate the hepatic elimination of toxic compounds or the
formation of toxic metabolites at varying doses. It is important to realize, though, that the Michaelis-
Mcnten parameters, which are determined in most in vitro studies, can be used to predict hepatic
metabolism in vivo only after considering the physiological determinants of hepatic metabolism and what
factor is rate limiting for the metabolism of a compound. In risk assessment one also needs to
extrapolate that information on hepatic metabolism from animals to humans. These two areas represent
current research needs with respect to the assessment of hepatic metabolism in risk assessment.
A. Determinants of Hepatic Metabolism
Insight into hepatic elimination requires an understanding of the physiological determinants of the
process and how these relate to each other in a quantitative manner. Determinants of hepatic
metabolism include the following:
1. Organ bloodflow (transit time)
2. Drug binding
3. Enzymatic activities: K^ and VmM for formation of metabolites
4. Cosubstrate availability
5. Diffusional barrier
6. Localization of enzyme systems
t
7. Competing metabolic pathways
An important consideration is that the rate-limiting step for the metabolism of a compound may change
with changing dose.
B-20
-------
B. Use of the Perfused-liver Preparation in the Study of Hepatic Metabolism
Numerous in vitro techniques including perfusion systems, liver homogenate, liver slices, isolated
hepatocytes, and subcellular fractions, as well as purified enzyme preparations, have been widely used in
the examination of hepatic metabolism. We have used predominantly the perfused-liver preparation
since, in contrast to other in vitro preparations, organ structure, spatial heterogeneity, and architecture
are preserved with this technique. Biliary excretion of drugs and metabolites also may be quantitated.
Liver perfusion techniques have been applied to many animal species and have been used extensively in
studies of drug metabolism and drug toxicity.
C. Assessment of Metabolite Formation Rates and Pathways with the Model Substrate, Gentisamide
We have used the perfused-rat-liver preparation in the assessment of the pathways and rates of
metabolism of the model substrate, gentisamide (2,5-dihydroxybenzamide, GAM). Gentisamide-5-sulfate
(GAM-5S) and gentisamide-5-glucuronide (GAM-5G) are major metabolites, and gentisamide-2-sulfate
(GAM-2S) is a minor metabolite. Single-pass rat-liver perfusions were used to examine the effect of
stepwise increases or decreases of input GAM concentration (Qn) on the extraction ratio of GAM and
steady-state formation rates of metabolites. Steady-state metabolite-formation rates were determined by
the sum of the steady-state efflux in perfusate and bile. Fitting the steady-state metabolite formation
rates and the logarithmic mean drug concentration at the various C!n's to the Michaelis-Menten equation
furnished parameter estimates for the three metabolic pathways.
D. Prediction of Hepatic Metabolism
Hepatic clearance models have been derived to allow a method for the prediction of, first, hepatic
clearance of drugs, and second, the effect of perturbations in the physiological determinants of
elimination on clearance and blood-concentration time profiles. A number of models of hepatic
clearance have been proposed that have been shown to be compatible with experimental data. These
models generally use the parameters hepatic blood flow, free fraction of drug in blood, and hepatic
intrinsic clearance (the intrinsic ability of the liver to eliminate the drug). The models include:
1. Venous Equilibrium (Well-Stirred) Model
The liver is assumed to be a single well-stirred compartment with the concentration of unbound
drug in effluent blood in equilibrium with that in the liver, expressed as follows:
B-21
-------
Q fB Clu
C1H=
Q + fa Clu,int
where C1H is hepatic clearance, Q is hepatic blood flow, fB is the free fraction in blood, and Clu>int is the
intrinsic clearance of free drug.
2. Sinusoidal Perfusion (Parallel-Tube) Model
The parallel-tube model assumes that the liver is composed of a large number of identical tubes or
sinusoids, arranged in parallel, with enzymes uniformly distributed in parenchymal cells surrounding the
cylinder. Tube concentrations decline in an exponential manner as drug is eliminated so that the mean
hepatocyte concentration is taken as the logarithmic mean of the unbound input and output
concentrations. Modifications of this model include a statistical distribution of sinusoid lengths
(distributed model) and heterogeneous enzyme distribution. The basic model is configured as follows:
-fBClu,int/Q
C1H = Q (1-e )
3. Dispersion Model
The dispersion model was so named because of its analogy to nonideal flow in a packed-bed
chemical reactor. The model is characterized by two main parameters: the efficiency number (RN),
which describes the efficiency of drug removal by the liver, and an axial dispersion number (DN). The
DN is a measure of the dispersion or spread in residence times of drug molecules moving through the
liver. As DN approaches infinity, C1H is identical to the result predicted by the well-stirred model; as DN
approaches O, C1H is the same as that predicted by the undistributed sinusoidal perfusion model.
Since the relationships between Q, fB, and Clu,int vary, the prediction of C1H will vary with the
different models of hepatic clearance. However, a change in any determinant has the same general
effect, regardless of the model used. Significant differences between models occur only when drugs with
high total intrinsic clearance are examined.
B-22
-------
References
Morris, M.E., Yuen, V., Tang, B.K., and Pang, K.S., 1988. "Competing pathways in drug metabolism. I.
Effect of input concentration on the conjugation of gentisamide in the once-through in situ perfused
rat liver preparation." J. Pharmacol. Exp. Ther. 245:614-624.
Pang, K.S., 1984. "Liver perfusion studies in drug metabolism and toxicity." In: Drug Metabolism and
Drug Toxicity; Mitchell, J.R. and M.G. Horning (eds.). New York: Raven Press, pp. 331-352.
Roberts, M.S. and Rowland, M., 1986. "A dispersion model of hepatic elimination: 1. Formulation of
the model and bolus considerations." J. Pharmacokinet. Biopharm. 14:227-260.
Wilkinson, G.R., 1987. "Clearance concepts in pharmacology." Pharmacol. Rev. 39:1-47.
Wilkinson, G.R. and Shand, D.G., 1975. "A physiological approach to hepatic drug clearance." Clin.
Pharmacol. Ther. 18:377-390.
B-23
-------
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC MODELING
. CROSS-SPECIES SCALING
Joyce Mordent!
I. Allometric Approach
A. Equations
1. One independent variable: Y = aWb, where W = body weight
2. Two independent variables: Y = cWd(BW)e, where BW = brain weight
B. Data Requirements '
1. Enough data are available to obtain statistical significance.
2. If protein binding is greater, than 80%, it is essential that the percentage of protein
binding is similar in air species and linear over the concentration range of interest. When
in doubt, scale unbound drug concentrations.
3. Physical methods of elimination (renal, biliary, pulmonary) are preferred for one
independent variable; oxidative metabolism scales with two independent variables.
C. Uses
1. Estimation of pharmacokinetic parameters (clearance, volume of distribution, half-life, etc.)
(Figure 1)
2. Estimation of entire pharmacokinetic profiles (Figure 2)
3. Estimation of toxicity endpoints (Figure 3)
D. Advantages: Easy and fast
E. Limitations
1. Empirical
2. No attempt is made to give physiological meaning to the pharmacokinetic parameters.
3. May not work well for metabolized compounds, although there are data to support
allometric approach for some metabolized compounds.
II. Physiologic Approach
A. Equations: Series of differential equations that are solved simultaneously
B. Data Requirements for Model (Figure 4)
1. Need anatomical, physiological, biochemical, and binding data for drug disposition, such as:
a. Blood flow to eliminating organs
b. Tissue and fluid volumes
c. Blood-to-plasma and tissue-to-plasma drug concentration ratios
d. Drug protein binding
e. Enzyme kinetics
2. Linear or nonlinear pharmacokinetic data
3. Linear or nonlinear protein binding
B-24
-------
C. Uses
1. Estimates of drug disposition in other animal species are obtained once the model is
described in detail for one species (Figure 5)
2. Estimation of toxicity endpoints
3. Low-dose to high-dose extrapolation (and vice versa)
D. Advantages
1. Pharmacokinetic parameters have physiological meaning
2. Method of choice for metabolized compounds
3. The parameters in the model can be modified to make a priori predictions of
pharmacokinetic changes associated with disease states, age, pregnancy, or drug-drug
interactions (Figure 6).
E. Limitations
1. Requires sophisticated computer programs, mathematical prowess, lots of experimental
time, and money
2. Data in humans may not be available; therefore, may need to rely on in vitro experiments,
allometric estimates, or may need to assume same value as animals.
III. Recommendations: Develop a set of criteria for deciding when the physiologically based
pharmacokinetic model is needed and when simpler models will suffice.
References
Boxenbaum, H., 1986. "Time concepts in physics, biology, and pharmacokinetics.11 J. Pharm. Sci.
75:1053-1062.
Calabrese, E., 1986. "Animal extrapolation and the challenge of human heterogeneity." J. Pharm. Sci.
75:1041-1046.
Dedrick, R., 1986. "Interspecies scaling of regional drug delivery." J. Pharm. Sci. 75:1047-1052.
Mordenti, J., 1986. "Man versus beast: pharmacokinetic scaling in mammals." J. Pharm. Sci.
75:1028-1040.
Subcommittee on Pharmacokinetics in Risk Assessment, 1987. Pharmacokinetics in Risk Assessment.
Vol. 8 of Drinking Water and Health. Washington, DC: National Academy Press.
Yates, F. and Kugler, P., 1986. "Similarity principles and intrinsic geometries: Contrasting approaches
to interspecies scaling." J. Pharm. Sci. 75:1019-1027.
B-25
-------
1000 -
c
£ 100
5
CO
LU
X
O
M
LU
U
10
Obsorved
(trong*)-,
(2)
(3)
X
1
E
.01
1
BODY WEIGHT (kg)
10
100
Figure 1. Log-log plot of ceftizoxime half-life versus body weight. The solid circles represent the
half-life of ceftizoxime in various animal species. The solid line is the least squares linear
regression line. The prediction of ceftizoxime half-life in humans is read off the regression
line at 70 kg. The triangle represents the reported ceftizoxime half-life in humans (mode),
and the bars represent the range of values from the literature.. Numbers in parentheses
indicate numbers of data points. Source: Mordenti, J., 1985. "Forecasting cephalosporin
and monobactam antibiotic half-life in humans from data collected in laboratory animals."
Antimicrob. Agents Chemother. 27:887-891.
B-26
-------
1000
O)
z
O 100
u
8 I0
Of
x
O
M
-End
Infusion
5 I
_L
678
TIME (hours)
10
12
Figure 2. Comparison of the predicted and experimental ceftizorime concentrations in human serum
after a 4-gram, 30-minute iv infusion. KEY: solid circles represent ceftizoxime serum
concentrations; the solid and dashed lines represent three prediction methods. Source:
Mordenti, J., 1985. "Pharmacokinetic scale-up: accurate prediction of human
pharmacokinetic profiles from animal data." J. Pharm. Sci. 74:1097-1099.
B-27
-------
1000
D>
.^
0)
or
o
100 ..
10
BDF. mouse
Swiss
mouse rat
LD10
MAXIMUM TOLERATED DOSE
monkey
human
0.01
0.1 1 10
Weight (kg)
100
Figure 3. Log-log plot of toxic doses versus body weight data for 5-FUDR (floxuridine). The solid
line is from an equation fitted using the method of least squares on unweighted
logarithmically transformed data. Source: Mordenti, J., 1986. "Dosage regimen design for
pharmaceutical studies conducted in animals." J. Phartn. Sci. 75:852-857.
B-28
-------
Figure 4. Compartmental model for IgG pharmacokinetics. Antibody is distributed to each organ
according to arterial plasma flow to that organ (Q;). Source: Covell, D. et al., 1986.
"Pharmacokinetics of monoclonal immunogiobulin O,, F(ab')2, and Fab' in mice." Cancer
Res. 46:3969-3978.
B-29
-------
r
TOTAL PT
1
z"
g
cc
t-
Ul
u
o
o
TOTAL FILTERABLE PT
3 4
TIME HOURS
Figure 5. Physiologic model predictions of platinum (Pt) containing species in human plasma
following an iv dose of 100 mg/m2. Each point represents the mean +. SD for 5-6 patients
for total Pt, total filterable Pt, and ciy-dichlorodiammineplatinum (DDP). Source: King,
F., Dedrick, R., and Farris, F. 1986. "Physiologic pharmacokmetic modeling of cis-
dichlorodiammineplatinum in several species." J. Pharmacokinet. Biopharm. 14:131-156.
B-30
-------
4.00
RENAL FUNCTION
O normal
Q impaired
anuric
234
DAYS
6
Figure 6. Predicted and observed plasma digoxin concentrations in patients with normal renal
function, in patients with moderate renal impairment, and in patients with severe renal
failure. Predictions based on scale-up of a dog physiologic model. Source: Harrison, L.
and Gibaldi, M., 1977. "Physiologically based pharmacokinetic model for digoxin
disposition in dogs and its preliminary application to humans." J. Pharm. Sci. 66:1679-
1683.
B-31
-------
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC MODELING
BIOTRANSFORMATION OF ENVIRONMENTAL CHEMICALS: DEVELOPMENT OF IN VITRO
SYSTEMS TO PREDICT IN VIVO EVENTS
Glenn Sipes
Differences in rates and/or routes of biotransformation of chemicals often explain species differences
in the pharmacologic/toxicologic responses to the chemicals. Therefore, knowledge of the metabolism of
a chemical is critical in the early phases of its' toxicity testing. Obtaining such data is relatively
straightforward for laboratory animals:; However, with respect to humans, such data are difficult to
obtain, particularly for chemicals that are not to be used therapeutically and that have toxic potential
(i.e., pesticides, solvents, chemical intermediates, etc.). In estimating the human risk as a result of
exposure to environmental chemicals, knowledge of human metabolism is also needed. Therefore, in
vitro systems are needed that can determine how humans metabolize particular chemicals. These in vitro
systems can then be used to compare directly how animals and humans biotransform chemicals. From in
vitro findings it will be possible to make pharmacokinetic and limited toxicological predictions. These
predictions can then be tested in laboratory animals. If the in vitro findings accurately predict in vivo
findings in animals, then the limited in vitro data obtained with human metabolism preparations can be
given more weight in estimating human risk from an exposure to a particular chemical.
A number of in vitro preparations can be used to obtain metabolic data. These include purified or
partially purified enzymes, subcellular preparations (i.e., mitochondria, microsomes, cytosol, etc.) or tissue
homogenates, slices, or wedges. The type of preparation used depends on the availability of tissue (a
problem with humans) or the question being asked.
Over the past few years, my colleagues and I have used subcellular preparation of human liver,
particularly microsomal preparations, to determine if humans can metabolize particular polychlorinated
biphenyls. Data from human hepatic microsomal preparations were then compared to data obtained
from animal preparations. These simple metabolic systems showed that humans cannot readily
metabolize 2,2',4,4',5,5'-hexachlorobiphenyl (2,4,5-HCB). This inability to metabolize 2,4,5-HCB explains
why this compound accumulates in human adipose tissue. Human hepatic microsomes were able to
metabolize 2,2',3,3',6,6'-hexachlorobiphenyl and 4,4'-dichlorobiphenyl, two biphenyls that are known to be
eliminated readily by rats, dogs, and monkeys and probably by humans. As a sidelight, the in vitro
studies explained why dogs can metabolize and more readily eliminate 2,4,5-HCB. Dog microsomes
B-32
-------
contain a particular cytochrome P-450 with high catalytic activity (as compared to other species) towards
this environmental contaminant (Duignan et al., 1987).
Although subcellular preparations are useful to answer certain questions (the intrinsic ability to
metabolize a particular chemical), they do not provide data on the nature of metabolism that occurs in
intact cellular systems. Under these conditions coupled and competing pathways are operative and will
influence the rates and routes of metabolism. We refer to this as integrated biotransformation.
We have recently developed a dynamic liver culture system to study the integrated biotransformation
and potential toxicity of chemicals. The system requires the use of a mechanical slicer to produce slices
from sections of liver bathed in oxygenated buffer. The precision-cut slices that are produced are thin
(250 /tin) and of uniform size and weight (25 mg/slice). For incubation the slices are placed on a wire
mesh screen, which is inserted into an incubation vial. The vials are then incubated on a temperature-
controlled roller system, so that the slices rotate in and out of the culture media at a preset rate and are
maintained at 37° C. This dynamic culture system enhances oxygenation of cells throughout the slice.
These innovations have resulted in slices that are viable for over 20 hours, when assessed for
maintenance of protein synthesis, potassium, and ATP content and leakage of lactic acid dehydrogenase
into the culture medium. Chemical-induced changes in these indices of viability also are useful markers
of chemical-induced toxicity (Sipes et al., 1987).
Human and rat slices prepared as described and incubated in various buffers can biotransform a
number of xenobiotics. Both the cytochrome P-450 system and various conjugating enzymes are stable
for 12-20 hours of incubation (depending on the particular liver). Substrates tested to date include 7-
ethoxycoumarin, biphenyl, chlorobenzene, 1,2-, 1,3- and 1,4-dichlorobenzenes, and phenacetin for P-450
mediated reactions, and the hydroxylated metabolites of several of these for assessment of
glucuronidation and sulfation capacity. These slices can also acetylate a number of aromatic amines and
be used to phenotype humans as fast or slow acetylators (Gunawardhana et al., 1988). When allyl
alcohol, bromobenzene, or the isomers of dichlorobenzene were incubated with rat liver slices, dose-
dependent toxicity was observed as assessed by changes in protein synthesis, slice K+ and ATP content,
and LDH leakage (Sipes et al., 1987). Factors that modify the toxicity of these compounds in vivo
resulted in similar alterations in the hepatotoxicity produced by these chemicals in this in vitro system.
Human liver slices also revealed dose-dependent toxicity for those chemicals tested.
Clearly, further refinement and testing of this dynamic liver slice system are necessary. However, it
should become a useful tool to help elucidate mechanisms of toxicant-induced liver injury. In addition,
B-33
-------
it will aid in the extrapolation of in vitro data to the in vivo situation, as well as in the extrapolation of
animal data to humans. Such systems are particularly needed for the study of environmental chemicals,
since ethical considerations preclude their administration to humans.
Particular research needs are four-fold: to assess the in vitro metabolism of more chemicals by in
vitro human preparations; to develop appropriate kinetic parameters in vitro that can be used to predict
in vivo clearance; to test these predictions in vivo (in laboratory animals); and to develop methods to
cryopreserve human tissues so that they can be used as needed to assess human chemical metabolism.
(Supported by NIEHS-N01-ES-55112).
References
Duignan, D.B., Sipes, I.G., Leonard, T.B., and Hapert, J.R., 1987. "Purification and characterization of
the dog hepatic cytochromes P-450 isozyme responsible for the metabolism of 2,2',4,4',5,5'-
hexachlorobiphenyl." Arch. Biochem. Biophvs. 225:290-303.
Gunawardhana, L., Barr, J., Weir, A.J., Brendel, K., and Sipes, I.G., 1988. "Liver slices: An in vitro
system for determination of N-acetylation in human liver." Proc. West. Pharmacol. 31:137-141.
Sipes, I.G. and Schnellmann, R.G., 1987. "Biotransformation of PCB's." In: Metabolic Pathways and
Mechanisms Environmental Toxin Series 1." S. Safe and O. Hutzinger (eds.). Berlin, Heidelburg:
Springer-Verlag.
Sipes, I.G., Fisher, R.L., Smith, P.F., Stine, E.R., Gandolfi, A.J., and Brendel, K., 1987. "A dynamic
liverculture system: A tool for studying chemical biotransformation and toxicity." Arch. Toxicol.
(Supp. II):20-33.
B-34
-------
SESSION IV: THE USE OF EXPERIMENTAL DATA IN PHARMACOKINETIC MODELING
TARGET TISSUE/CELL DOSE OF CHEMICAL CARCINOGENS
George Lucier, Steven Belinsky, and Claudia Thompson
There is a great deal of uncertainty regarding estimation of human risks from low dose exposure to
chemical carcinogens when using high dose animal data as the basis for extrapolation. It is becoming
increasingly evident that molecular approaches can contribute a great deal to reducing the uncertainty
that is inherent in the risk assessment process when gross biological endpoints such as tumors are used.
For example, macromolecular interactions such as DNA adducts can be detected at doses much lower
than those needed to produce significant increases in tumors.
There are several issues that affect the use of DNA adducts as a molecular dosimeter. These
include:
1. Adduct heterogeneity - Different adducts have markedly different capacities to produce genetic
damage. Therefore, measuring total DNA adducts can produce misleading data.
2. Cell specificity - Different cell types have different capacities to form and repair DNA adducts.
Therefore, measuring DNA adducts in whole-organ preparations can produce misleading data.
3. Surrogate markers - In molecular epidemiologic studies, it is customary to quantify DNA
adducts in lymphocytes. In some cases, lymphocytes will be good surrogate markers'for adduct
concentrations in target cells and in other cases they will not.
4. Indirect adduction - Some chemicals, such as estrogens, form no DNA adducts or only small
amounts, but they do dramatically influence the DNA adduction of dietary constituents or
other endogenous compounds. In other words, a carcinogen that does not bind to DNA
directly is not always a nongenotoxic carcinogen.
5. Adduct detection - The two most sensitive methods for detecting DNA adducts are antibody
techniques and 32P-postlabeling methods. In general, 32P methods are best for bulky aromatic
adducts, whereas antibody methods are more suitable for alkylated adducts.
DNA adducts have been used as a molecular dosimeter in some cases. One of the best examples is
NNK (a carcinogenic metabolite of nicotine). The promutagenic adduct of NNK, O6-methylguanine, is
formed more efficiently at low doses than high doses in lungs; this finding is consistent with dose-
response relationships for carcinogenicity. Moreover, the Clara cell has considerably more
O6-methylguanine than other lung cell types, and this cell is considered to be the progenitor cell in
NNK-induced lung cancer. In other studies, DNA adducts have been detected in human lymphocytes by
B-35
-------
^P-postlabeling, but adduct profiles were not different between smokers and nonsmokers. A great deal
of interindividual variation exists in adduct concentrations, and this variation may reflect differences in
metabolic activation/deactivation reactions, including the presence of a glutathione transferase isozyme
that is polymorphic in human populations.
A recent NTP study revealed significant discordance between short-term tests for genetic toxicity
and in vivo carcinogenicity (lifetime exposures). These data suggest that up to 40% of the chemicals
that are positive for carcinogenicity in the NTP bioassay are acting through nongenotoxic mechanisms.
The implication is that effects on signal transduction pathways, receptor-mediated proliferative responses,
and cell cycle control are involved in the mechanism of action of many carcinogens, including dioxin.
Therefore, one research need is to evaluate the quantitative relationships between biochemical events
involved in tumor promotion and carcinogenic incidence. Another need is for a better characterization
of structure-activity relationships. A central underlying need in molecular dosimetry studies is increased
knowledge of the diverse mechanisms of chemical carcinogens.
References
Belinsky, S.A., White, CM., Devereux, T.R., and Anderson, M.W., 1987. "DNA adducts as a dosimeter
ibr risk estimation." Environ. Health Perspect. 76:3-8.
Guengerich, P.P., Peterson, L.A, Cmarik, J.L., Koga, N., and Inskeep, P.B., 1987. "Activation of
dihaloalkanes by glutathione conjugation and formation of DNA adducts." Environ. Health
Perspect. 76:15-18.
Hanawalt, P.C., 1987. "Preferential DNA repair in expressed genes." Environ. Health Perspect. 76:9-14.
Lucier, G.W. and Thompson, C.L., 1987. "Issues in biochemical applications to risk assessment: When
can lymphocytes be used as surrogate markers?" Environ. Health Perspect. 76:187-194.
Lucier G.W., Nelson, K.G., Everson, R.B., Wong, T.K., Philpot, R.M., Tiernan, T., Taylor, M., and
Sunahara, G.I., 1987. "Placental markers of human exposure to polychlorinated biphenyls and
polychlorinated dibenzofurans.": Environ. Health Perspect. 76:79-88.
Perera, R, 1987. "The potential usefulness of biological markers in risk assessment." Environ. Health
Perspect. 76:141-148.
Reddy, M.V. and Randerath, K., 1987. "32P-postlabeling assay for carcinogen-DNA adducts: nuclease P1-
mediated enhancement of its sensitivity and applications." Environ. Health Perspect. 76:41-48.
Swenberg, J.A, Richardson, F.C., Baucheron, J.A, Deal, F.H., Belinsky, S.A, Charbonneau, M., and
Short, B.G., 1987. "High- to low-dose extrapolation: Critical determinants involved in the dose
response of carcinogenic substances." Environ. Health Perspect. 76:57-64.
B-36
-------
Tennant, R.W., 1987. "Issues in biochemical applications to risk assessment: Are short-term tests
predictive of in vivo tumorigenicity?" Environ. Health Perspect. 76:163-168.
B-37
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
BENZENE PHARMACOKINETIC MODEL: CASE STUDY
Michele A. Medinsky
Studies on the chronic toxicity of benzene conducted by the National Toxicology Program indicated
that B6C3Fi mice were more sensitive to the carcinogenic effects of benzene than were F344 rats. A
physiological model was developed to describe the uptake and metabolism of benzene in rats and mice.
Our objective was to determine if differences in toxic effects could be explained by differences in
pathways for benzene metabolism or by differences in total uptake of benzene. Compartments
incorporated into the model included liver, fat, a poorly perfused tissue group, a richly perfused tissue
group, an alveolar or lung compartment, and blood (Figure 1). Metabolism of benzene was assumed to
take place only in the liver and to proceed by four major competing pathways (Figure 2). These include
formation of hydroquinone conjugates (HQC), formation of phenyl conjugates (PHC), ring-breakage and
formation of muconic acid (MUC), and conjugation with glutathione with subsequent mercapturic acid
(PMA) formation. Values for parameters such as alveolar ventilation, cardiac output, organ volumes,
blood flow, and partition coefficients were taken from the literature (Table 1). Metabolic rate constants
(Vma!c and Ka) were determined experimentally.
Model simulations confirmed that during and after 6-hour inhalation exposures, mice metabolized
more benzene, on a ^mole/kg body weight basis, than did rats. After oral exposure, rats metabolized
more benzene at doses above 50 mg/kg than did mice; this was due to more rapid absorption of benzene
by mice, resulting in more benzene being exhaled unmetabolized. Model simulations for PHC and PMA,
generally considered to be detoxification metabolites, were similar in shape and dose-response to those
for total metabolism. However, simulations for the metabolites representative of the putative toxication
pathways, HQC and MUC, indicated that after both oral and inhalation exposures at all concentrations,
mice would produce more of these metabolites than would rats. This trend was especially apparent for
HQC (Figure 3). Differences were due to the greater rates of metabolism in mice. Increased
metabolism of benzene via the HQC and MUC pathways in mice is consistent with the observed
susceptibility of this species to benzene toxicity.
The key gaps in our understanding of benzene metabolism include lack of information on the
concentrations of the unconjugated metabolites in liver, blood, and target tissues and insufficient
B-38
-------
understanding of the capacity of human tissue to metabolize benzene. In particular, the concentrations
of hydroquinone and phenol in bone marrow appear to be critical to the production of the reactive
metabolite, benzoquinone. Models that describe the flux of unconjugated hydroquinone and phenol
through the target organs and incorporate metabolism by human tissues will generally help us estimate
health risks of exposure to benzene.
References
Bechtold, W.E., Sabourin, P.J., and Henderson, R.F., 1988. "A reverse isotope dilution method for
determining benzene and metabolites in tissues." J. Anal. Toxicol. (in press)
Kalf, G.F, 1987. "Recent advances in the metabolism and toxicity of benzene." CRC Crit. Rev. Toxicol.
18:141-159.
Sabourin, P.J., Chen, B.T., Lucier, G., Birnbaum, L.S., Fisher, E., and Henderson, R.F., 1987. "Effect of
dose on the absorption and excretion of MC-benzene administered orally or by inhalation in rats and
mice." Toxicol. Appl. Pharmacol. 87:325-336.
i
Sabourin, P.J., Bechtold, W.E., Birnbaum, L.S., Lucier, G., and Henderson, R.F., 1988a. "Difference in
the metabolism of inhaled 3H-benzene by F344/N rats and B6C3Fj mice." Toxicol. Appl. Pharmacol.
94:128-140.
Sabourin, P.J., Bechtold, W.E., and Henderson, R.F., 1988b. "A high pressure liquid chromatographic
method for the separation and quantitation of water-soluble radiolabeled benzene metabolites."
Anal. Biochem. 170:316-327.
B-39
-------
TABLE 1
PARAMETERS USED IN THE PHYSIOLOGICAL SIMULATION MODEL FOR
BENZENE METABOLISM IN F344 RATS AND B6C3Fi MICE
PARAMETER
VALUE
RATS
MICE
Physiological
Alveolar ventilation (L/hr.kg)
Cardiac output (L/hr.kg)
Gastrointestinal transfer (fraction/hr)
Body weight (kg)
Blood flow (fraction of cardiac output)
To liver
To fat
To poorly perfused tissues
To richly perfused tissues
Organ volumes (fraction of body weight)
Blood
Liver
Fat
Poorly perfused tissues
Richly perfused tissues
Other (not perfused)
Chemical
32.4
19.4
0.25
0.288
0.25
0.09
0.15
0.51
0.042
0.03
0.11
0.66
0.076
0.082
55.3
34.8
3.0
0.030
0.25
0.09
0.15
0.51
0.042
0.03
0.11
0.66
0.076
0.082
Molecular weight (g/mol) *
Partition coefficients
Liver/blood
Fat/blood
Poorly perfused tissues/blood
Richly perfused/blood
Blood/air
Biochemical parameters
Vmax (fimoles/hr.kg)
Kn, (>moles/L)
VmM;phc Oonoles/hr.kg)
K^phe (/mioles/L)
Vmil)(,pmc (/tmole/hr.kg)
K^pmc Gimoles/L)
Vra^hqc Omioles/hr.kg)
K^hqc (/imoles/L)
Voa]t,muc Otmoles/hr.kg)
K^ue (/zmoles/L)
78.2
1.0
28.0
0.6
1.0
18.0
122
40
174
5
104
15
1.7
0.5
3.5
0.5
78.2
1.0
28.0
0.6
1.0
18.0
200
1.0
333
3.0
90
4.0
27
0.1
12
0.1
B-40
-------
INHALED
$ EXHALED
POORLY\
TISSUES 7
^ RICHLY
& PERFUSED
TISSUES/
OF
FAT
^
9 METABOLITES
Figure 1. Physiological model of benzene metabolism. Metabolism of benzene was presumed to
take place in the liver compartment. Pathways for benzene metabolism incorporated
into the model are shown in Figure 2.
B-41
-------
BENZENE
! BENZOQUINONE \
BENZENE '
i OXIDE
u
' HYDROQUINOL T
J
vmax.hqo
i 'Km, hqe
MUCONIC ACID
k
Xnai, IMC
*m, niHe
r 1
« MUCONALDEHYDE |
" PHENOL j
HYDROQUINONE
CONJUGATES
^ ^ «
^
»«,.. ^
vm«», phc
PHENYL
CONJUGATES
PHENYL
MERCAPTURIC
ACIDS
Figure 2. Scheme for metabolism of benzene. Biochemical rate constants are outlined in Table 1.
This scheme for metabolism of benzene takes place in the liver compartment described
in Figure 1. The solid boxes represent metabolites of benzene that were measured and
used in model simulations. The dotted boxes represent immediate metabolites that were
not incorporated into the model.
B-42
-------
1000
800
Q
UJ
a
40O
-.300
100 200
/xmol«« HQ GLUC and HQ SO4/kg
Figure 3. Hydroquinone conjugates excreted by rats and mice compared after single oral
administration (O) or a single 6-hour inhalation exposure (I) to benzene. Lines
represent results of model simulations for rats and mice for pinoles of hydroquinone
conjugates (glucuronide and sulfate) per kg of body weight.
B-43
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
IMPLICATIONS OF PHARMACOKINETIC DATA AND MODELS IN A RISK
: ASSESSMENT OF BENZENE
A. John Bailer
Risk assessments of benzene have been based upon both human and animal studies. In this report,
metabolite information is used to construct an internal dose (a surrogate of the biologically effective
dose) for a given administered dose. The relationship between the administered dose and the internal
dose is nonlinear and is well described by a Michaelis-Menten function. The administered doses from
the National Toxicology Program's rodent carcinogenicity study of benzene are transformed into internal
doses, and the internal doses are used in conjunction with a multistage model to perform a comparison
of previous estimated "virtually safe doses" (VSDs) associated with small added health risks. The ratio of
VSD for the administered-dose risk assessment to VSD for the internal-dose risk assessment was
approximately 1.0 for F344/N rats and ranged from 2.5 to 5.0 for B6C3F! mice in the National
Toxicology Program study. Given an occupational exposure of 1 ppm, a risk estimate of 0.7 excess
cancers/1,000 exposed with an upper bound of 3.5/1,000 was obtained. Risk estimates based upon
internal doses constructed from levels of toxic metabolites of benzene are in the same range as the total
metabolite-based risk estimates. A better characterization of the dose-response function for benzene may
be provided by obtaining data describing the molecular dosimeters of benzene. These molecular
dosimeters ideally will be a better surrogate for the target tissue dose.
References
Austin, H., Delzell, E., and Cole, P., 1988. "Benzene and leukemia: a review of the literature and a risk
assessment." Am. J. Epidemibl. 127:419-439.
Crump, K.S. and Allen, B.C., 1984. "Quantitative estimates of risk of leukemia from occupational
exposure to benzene." Occupational Safety and Health Administration, Docket H-059 B, Exhibit
152, May.
Henderson, R.F., Sabourin, P.J,., Bechtold, W.E., Griffith, W.C., Medinsky, M.A., Birnbaum, L.S., and
Lucier, G.W., 1988. "The effect of dose, dose rate, route of administration, and species of tissue
and blood levels of benzene metabolites." Environ. Health Perspect. 82.
B-44
-------
National Toxicology Program, 1986. "Technical report on the toxicology and carcinogenesis studies of
benzene." NTP Technical Report Series No. 289, Research Triangle Park, NC.
Oak Ridge National Laboratory, 1987. Toxicological Profile for Benzene. Oak Ridge, TN.
Sabourin, P.J., Chen, B.T., Lucier, G., Birnbaum, L.S., Fisher, E., and Henderson, R.F., 1987. "Effect
ofdose on the absorption and excretion of [14C]benzene administered orally or by inhalation in rats
and mice." Toxicol. Appl. Pharmacol. 87:325-336.
B-45
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
THE USE OF PHARMACOKINETIC ANALYSIS IN RISK ASSESSMENT -
THE CASE OF BUTADIENE
Dale Hattis
Introduction ,
;
A trend to put more causal mechanism information into the mathematical models used for risk
assessment has been gathering momentum for some time. This trend allows greater input from basic
experimental science into the risk assessment/policy formulation process, and it can provide a framework
for asking interesting and relevant questions within the framework for basic experimental science itself.
A philosophy of science issue results from this trend. Experimental scientists in Baconian tradition
are reluctant to build elaborate mathematical models; more mathematical/statistical workers who have
largely been in control of risk assessment procedures often do not have the detailed familiarity with
causal mechanisms to feel comfortable building realistic mechanism-based representations of complex
biological processes. In any event, doing so would complicate the use of their usual black-box curve-
fitting approaches to analysis. I want to advance today what may be a startling proposition to the
experimentalists among you - that by uncovering anomalies in the fit between data and theory, analysis
can be as fruitful in producing new "knowledge," in some cases, as additional data-gathering activities and
can properly be thought of as a complementary and synergistic enterprise in science.
The particular efforts I will discuss today take the form of using pharmacokinetic analysis to:
1. Reinterpret the dosages of active metabolites actually delivered in animals in the course of
chronic 2-year bioassays.
2. Improve the projection of dosage across species using data on metabolism in humans, rats, and
mice.
Models of this type will also be of importance for appropriately interpreting information on newer
biological markers for steps in the carcinogenic process, such as DNA aclduct formation (dynamic
processes of forhidtibn and repair of adducts).
B-46
-------
Overview of Lessons
Before getting immersed in the details of butadiene, I should foreshadow just what broad "lessons" I
think you should draw from them:
1.
2.
3.
Pharmacokinetic analysis is nobody's unambiguous quick-and-dirty cookie-cutter solution to the
problem of uncertainty in carcinogenic risk analysis. Each of the models I have developed to
date:
a.
b.
Needed to undergo serious structural modification in the light of the data available for the
specific case. The process of doing those modifications is a developing art, requiring
liberal doses of judgment rather than cookbook formulas.
Raised as many questions as it answered - often revealing unsuspected sources of
uncertainty, and nonobvious difficulties in fairly assessing the extent of the uncertainty.
As often as not, at the end of the day, the pharmacokinetic analysis doesn't make a major
difference in the final numerical projection of risks (particularly with ethylene oxide). The
exception is butadiene, where I got nearly an order-of-magnitude effect.
Nevertheless, in the long run, pharmacokinetic analysis can make it easier to ask "better'Vmore
relevant experimental scientific questions. It can also help make risk assessment models
somewhat better by incorporating realistic and experimentally testable information about the
causal processes underlying both carcinogenesis and other adverse health effects.
The Framework for Modeling
The work reported here has a number of features that distinguish it from other current efforts to
build pharmacokinetic analysis into quantitative risk assessments:
Our models are implemented in an easy-to-use Apple Macintosh microcomputer-based systems
dynamics modeling system (STELLA). Because the system makes extensive use of graphics to
represent models and quickly display results, it is straightforward for use by people with little
experience in programming, and it helps them understand the effects of changes in model
structure and parameters.
Our human models incorporate a realistic diurnal pattern of change in breathing rates and
blood flows to different tissues. Different assumptions can be built in for the level of activity
during waking hours and the timing of exposure relative to activity and sleep.
Our tissue partition coefficients were determined by regression analysis of data from other
compounds in relation to oil/air and water/air partition coefficients. Those oil/air and water air
partition coefficients were estimated from butadiene's structure using the methods and data of
Leo. Because the partition coefficients were estimated, it is particularly important to test the
sensitivity of the model conclusions to reasonable alternative estimates.
We produce "best estimate" as well as "plausible upper limit" projections of human risk.
B-47
-------
Butadiene
Past studies use net absorption of data of Bond et al. (1986) to determine "dose." PLEA TO
EXPERIMENTALISTS: DO MORE THAN ONE TIME PERIOD OF INHALATION EXPOSURE
FOR TOTAL DISPOSITION EXPERIMENTS.
Other Topics Discussed:
Implications of the work of Kreiling et al. (1986a,b) for model structure.
Comparison of model outputs with Bond and Kreiling experimental data, and methods for
setting the adjustable parameters.
Revisions of delivered dose estimates:
- Humans: indicated five-fold less absorption than estimated by the U.S.
Environmental Protection Agency (EPA) and ENVIRON.
-- Animals: indicated two- to five-fold greater absorption and metabolism than
estimated by EPA and ENVIRON.
Risk projections from different data sets, using different model variants (differing blood/air
partition coefficients).
!
Model parameters that do not make much difference (less than 10% in ultimate geometric
mean risk estimates).
Risk projections from experiments in rats and mice differ appreciably, even after correction for
the initial step in butadiene metabolism (possible viral difference, inability to predict which will
be the better model for humans).
Problems with low dose nonlinearity for the male rat data.
Comparison with the results of nonpharmacokinetic-based risk assessments (EPA and
ENVIRON, Table 1).
Tentative resolution of the low-dose nonlinearity/background interaction problem.
The use of human epidemiological data.
IISRP claims based on Matanoski data (humans 100- to 1,000-fold less sensitive than
mice).
"All Tumors" risk projection implicitly is assumed to be totally contained within the
subcategory of nonleukemia lymphopoietic cancers.
B-48
-------
Healthy worker effect.
Implicit assumption that in 18 years of follow-up, 18/50 of expected lifetime cancers
will appear.
References
Bolt, H M., Filser, J.G., and Stormer, R, 1984. "Inhalation pharmacokinetics based on gas uptake
ComParative Pharmacokinetics of ethylene and 1,3-butadiene in rats." Arch. Toxicol.
Bond ! J.A., Dahl, A.R., Henderson, R.F., Dutcher, J.S., Mauderly, J.L., and Birnbaum, L.S., 1986
Species differences m the disposition of inhaled butadiene." Toxicol. ADD!. Pharmaml 84:617-627.
Boxenbaum, H., 1982. "Interspecies scaling, allometry, physiological time, and the ground plan of
pharmacokinetics." J. Pharmacokinet. Biopharm. 10:201-227.
Brugnone, R, Perbellini, L., and Gaffuri, E., 1980. "N-N-Dimethylformamide concentration in
environmental and alveolar air in an artificial leather factory." Br. J. Ind. Med. 37:185-188.
Brugnone, R, Perbellini, L., Faccini, G.B., Pasini, R, Bartolucci, G.B. and DeRosa, E., 1986. "Ethylene
Citti, L., Gervasi, P.O., Turchi, G., Bellucci, G., and Bianchini, R., 1984. "The reaction of 3 4-
of 'the main
il' TSCA S" "f ^T- KW"/985- "Mortality amo»g ^kers at a butadiene production
ility. TSCA Section 8 submission, (unpublished)
ENVIRON Corporation, 1986. Assessment of the Potential Risks to Workers from Rvnn.,.^ to i 3
Butadiene. Washington, DC: ENVIRON Corporation, December, pp. 1-107. - - "~
Filser, J.G. and Bolt, H.M., 1983. "Inhalation pharmacokinetics based on gas uptake studies- IV The
endogenous production of volatile compounds." Arch. Toxicol. 52:123-133.
Filser, J.G. and Bolt, H.M., 1984. "Inhalation pharmacokinetics based on gas uptake studies- VI
SSSTS iVf rf^ °f 6thylene OXide 3nd bUtadiene monoxide as exhaled reactive metabolites of
ethylene and 1,3-butadiene in rats." Arch. Toxicol. 55:219-223.
Fiserova-Bergerova, V., Tichy, M., and di Carlo, F.J., 1984. "Effects of biosolubility on pulmonary
uptake and disposition of gases and vapors of lipophilic chemicals." Drug Metab. Rev. 15:1033-
Fiserova-Bergerova V. and Diaz, M.L., 1986. "Determination and prediction of tissue-gas partition
coefficients." Int. Arch. Occup. Environ. Health 58:75-87. P-ruuon
B-49
-------
Hattis, D., 1982. "From presence to health impacts: Models for relating presence to exposure to
damage." In: Analyzing the Benefits of Health. Safety, and Environmental Regulations. Report to
U.S. EPA under contract #68-01-5838. Cambridge: M.I.T. Center for Policy Alternatives Report
No. CPA-82-16.
Hattis, D., 1987. A Pharmacokinetic/Mechanism-Based Analysis of the Carcinogenic Risk of Ethvlene
Oxide. Cambridge: M.I.T. Center for Technology, Policy and Industrial Development, Report No.
CTPID 87-1, August.
Hattis, D. and Smith, J., 1986. "What's wrong with quantitative risk assessment?" In: Biomedical Ethics
Reviews 1986. Clifton, NJ: Humana Press.
Hattis, D., Tuler, S., Finkelstein, L.> and Luo, Z.Q., 1986a. A Pharmacokinetic/Mechanism-Based
Analysis of the Carcinogenic Risk of Perchloroethvlene. Cambridge: M.I.T. Center for Technology,
Policy and Industrial Development, No. CTPID 86-7, September.
Hattis, D., Erdreich, L., and DiMauro, T., 1986b. Human Variability in Parameters that are Potentially
Related to Susceptibility to Carcinogenesis - I. Preliminary Observations. Cambridge: M.I.T. Center
for Technology, Policy and Industrial Development, May.
Hazleton Laboratories Europe, Ltd:, 1981. The Toxicitv and Carcinogenicitv of Butadiene Gas
Administered to Rats by Inhalation for Approximately 24 Months. Prepared for the IISRP for
Hazleton Laboratories Europe, Ltd.
Howe, R.B. and Crump, K.S., 1982. "Global 82 - A computer program to extrapolate quantal animal
toxicity data to low doses." Report by K.S. Crump and Company to Office of Carcinogen Standards,
OSHA, May.
Irons, R.D., Stillman, W.S., Shah, R.S., Morris, M.S., and Higuchi, M., 1986. "Phenotypic
characterization of 1,3-butadiene (BD)-induced thymic lymphoma in male B6C3F1 mice." Society of
Toxicology, Abstracts for the 1986 Annual Meeting.
Irons, R.D., Smith, C.N., Stillman, W.S., Shah, R.S., Steinhagen, W.H., and Leiderman, L.J., 1987.
"Macrocytic-megaloblastic anemia in male B6C3F1 mice following chronic exposure to 1,3-
butadiene." Toxicol. Appl. Pharmacol. (in press)
Kirk-Othmer Encyclopedia of Chemical Technology. Vol. 4, 1982. New York: Wiley, pp. 313-337.
Kreiling, R., Laib, R.J., and Bolt, H.M., 1986a. "Alkylation of nuclear proteins and DNA after exposure
of rats and mice to [1.4-14C] 1,3-butadiene." Toxicol. Lett. 30:131-136.
Kreiling, R., Laib, R.J., Filser, J.G., and Bolt, H.M., 1986b. "Species differences in butadiene metabolism
between mice and rats evaluated by inhalation pharmacokinetics." Arch. Toxicol. 58:235-238.
Lee, P.N. and O'Neill, J.A, 1971. "The effect both of time and dose applied on tumor incidence rate in
benzopyrene skin painting experiments." Br. J. Cancer 25:759-770.
Lyman, W., Reeh, J., and Rosenblatt, D., 1982. Handbook of Chemical Properties Estimation. New
York: McGraw Hill, pp. 1-10 to 1-38.
Malvoisin, E. and Roberfroid, M., 1982. "Hepatic microsomal metabolism of 1,3-butadiene."
Xenobiotica 12:137-144.
B-50
-------
Malvoisin, E., Lhoest, G., Poncelet, R, Roberfroid, M., and Mercier, M., 1979. "Identification and
quantification of l,2-epoxybutene-3 as the primary metabolite of 1,3-butadiene." J Chromatoer
178:419-429. *-'
Matanoski, G.M. and Schwartz, L., 1987. "Mortality of workers in styrene-butadiene polymer
production." J. Occup. Med. 29:675-680.
Matanoski, G.M., Schwartz, L., Sperrazza, J., and Tonascia, J., 1982. "Mortality of workers in the
styrene-butadiene rubber polymer manufacturing industry." (Unpublished-cited by EPA (1985) and
ENVIRON (1986); also see Matanosky and Schwartz (1987) below).
Meinhardt, T.J., Lemen, R.A., Crandall, M.S., and Young, R.J., 1982. "Environmental epidemiologic
investigation of the styrene-butadiene rubber industry." Scand. J. Work Environ. Health 8:250-259.
National Toxicology Program (NTP), 1984. Toxicology and carcinoeenesis studies of 13-butadiene fCAS
106-99-0) in B6C3F1 mice ("inhalation studiesX National Toxicology Program, cited by EPA (1985).
Rosenthal, S.L., 1985. "The reproductive effects assessment group's report on the mutagenicity of 1,3-
butadiene and its reactive metabolites." Environ. Mutagen. 7:933-945.
Schmidt, U. and Loeser, E., 1985. "Species differences in the formation of butadiene monoxide from 13-
butadiene." Arch. Toxicol. 57:222-225. '
Sharief, Y., Brown, AM., Backer, L.C., Campbell, J.A, Westbrook-Collins, B., Stead, AG., and Allen,
J.W., 1986. "Sister chromatid exchange and chromosome aberration analyses in mice after in vivo
exposure to acrylonitrile, styrene or butadiene monoxide." Environ. Mutagen. 8:439-448.
Tasher, S.A and Kaufman, B.F., 1986a. "Comments of the International Institute of Synthetic Rubber
Producers on OSHA's advanced notice of proposed rulemaking to reduce worker exposure to 13-
butadiene." OSHA Docket No. H-041. December, pp. 1-26.
Tasher, S.A and Kaufman, B.F., 1986b. "Statement of position of the International Institute of Synthetic
Rubber Producers with respect to the proposed standard for 1,3-butadiene." OSHA Docket No H-
041, pp. 1-44.
Tessmer, W.E., 1987. "Comments of the International Institute of Synthetic Rubber Producers on the
ICF-Clement Report, 'Characterization of Risks Associated with Occupational Exposure to 1,3-
Butadiene.'" Submitted to the Office of Health Standards, Occupational Safety and Health
Administration, U.S. Department of Labor. OSHA Docket No. H-041. June, pp. 1-10.
U.S. Environmental Protection Agency, 1985. Mutagenicity and Carcinogenicitv Assessment of 1.3-
Butadiene. Office of Health and Environmental Assessment, Washington, DC. EPA/600/8-85/004F.
U.S. International Trade Commission, 1986. Synthetic Organics Chemicals-United States Production
and Sales, 1985. Washington, DC: U.S. Government Printing Office, International Trade
Commission. Publication 1892.
Van Duuren, B.L., Nelson, N., Orris, L., Palmes, E.D., and Schmitt, F.L., 1963. "Carcinogenicity of
epoxides, lactones, and peroxy compounds." J. Natl. Cancer Inst. 31:41-55.
B-51
-------
TABLE 1
COMPARISON OF 95% UPPER CONFIDENCE LIMIT PROJECTIONS FROM
OUR PHARMACOKINETIC-BASED ANALYSIS WITH THOSE DERIVED FROM
EPA (1985) AND ENVIRON (1986) (45-YEAR, 8 HOUR/DAY, EXPOSURE TO 1 PPM)
DATA SET
Male Rats
Female Rats
Male Mice
Female Mice
EPA (1985)"
6.43E-4
0.00853
0.131
0.0665
ENVIRON (1986)b
7.98E-4
9.29E-4
0.00644°
THIS STUDY (BEST
ESTIMATE MODEL)
1.55E-4
0.00119
0.0325
0.0243
'Derived from EPA's upper 95% qt coefficients after correcting for 45-year, 8 hours/day, 5 days/week
exposures.
bENVIRON'S projections have been multiplied by (45/40) and (52/50) to place them on a comparable
basis with our assumed 45-year, 52-week exposures.
These calculations excluded malignant lymphomas. Some of the rest of the difference in this line may
be accounted for by mg/kg interspecies projection, in contrast to rag/kg273 used by ourselves and EPA
(1985); and the use by ENVIRON (1986) of the Hartley-Sielken procedure for adjusting for less than
lifetime exposure and observation.
B-52
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
CANCER RISK ASSESSMENT OF 1,3-BUTADIENE
Steven Bayard
The U.S. Environmental Protection Agency's (EPA) risk assessment of 1,3-butadiene was published
in September, 1985 (U.S. EPA, 1985) and represents knowledge available on the mutagenicity and
carcinogenicity up to that time. The information on the carcinogenicity of 1,3-butadiene includes
positive cancer response studies in both rats and mice (Hazleton, 1981; National Toxicology Program,
1984). In the Hazleton study, groups of male and female Charles River (Sprague-Dawley) rats were
exposed for 2 years, 6 hours/day, 5 days/week to 1,000 and 8,000 ppm 1,3-butadiene via inhalation.
There was significantly decreased survival in both the male and female high dose groups. Increased
tumors were observed in males in leydig cell, pancreatic exocrine, and zymbal glands, with increases
being statistically significant only at the highest concentration and only for the two former sites. Female
rats showed mainly increased mammary and uterine tumors, with to a lesser degree increased Zymbal
gland and thyroid follicular cell tumors. In comparison to the response in rats, cancer response in both
male and female B6C3F! mice, at comparatively lower concentrations of 625 and 1,250 ppm, was both
massive and rapid. Survival in both sexes at both doses was affected such that the studies had to be
terminated after 60 and 61 weeks. Early tumors, especially malignant lymphomas and hemangiosarcomas
(mainly heart) were responsible for most of these deaths. Other statistically increased tumor sites
included lung and forestomach (both sexes) and liver, mammary and ovarian glands (females).
Human evidence for the carcinogenicity of 1,3-butadiene at the time of EPA's publication was
considered inadequate for classification, although excess cancers of the lymphatic and hematopoietic
tissues were seen in some studies (Meinhardt et al., 1982; Matanoski et al., 1982; and McMichael et al,
1976). More recent information presented at a 1988 conference on 1,3-butadiene (NIEHS, 1988; see
also Downs et al., 1987) provides additional confirmatory evidence that 1,3-butadiene exposure is
associated with these cancers in humans, but accurate contemporary exposure estimates are lacking.
Using the above evidence of positive cancer response in two animal species and inadequate
epidemiologic data, the EPA in 1985 classified 1,3-butadiene as a "probable" human carcinogen, Group
B2 according to EPA's Proposed Guidelines for Carcinogen Risk Assessment (U.S. EPA, 1984; see also
U.S. EPA, 1986).
B-53
-------
An estimate of the carcinogenic potency of 1,3-butadiene to humans was based on the most
sensitive animal sex-species. In this case, since the cancer responses in the male and female mice were
very similar, both sets of results were used by taking the geometric mean of the 95% upper-limit
estimates derived by the use of the linearized multistage model, using the computer program Global 82
(Howe and Crump, 1982).
To derive estimates of target tissue dose, the data from an unpublished study (NTP, 1985) on male
mouse total body burden were used in EPA's risk assessment (U.S. EPA, 1985). However, following
EPA's publication additional experiments and further analysis by NTP led to a publication of final
results (Bond et al., 1986) which differ from those in the unpublished report. These final results are
used for the body burden and absorption estimates in this risk assessment. Estimates of total body
burden were used instead of estimates of target organ dose, because of the high number of affected sites
and the different dose and time response characteristics of the different tumor sites., In estimating
internal dose from external exposure, estimates of the low exposure retention of 1,3-butadiene and/or
metabolites following 6 hour exposures were 20% over a two order-of-magnitude range of concentrations
(up to 7 ppm). At the higher exposures of 70 and 930 ppm, retention decreased to 8% and 4%
respectively. Since metabolic clearance of 1,3-butadiene in mice (and rats) follows linear
pharmacokinetics below exposure concentrations of about 1,000 ppm (Kreiling et al., 1986), the decreases
at high concentrations in micromoles body burden/ppm concentration after 6 hours exposure are assumed
to result from decreased lung absorption rates. When body burden doses are adjusted for this decreased
absorption at higher atmospheric concentrations, potency estimates in the female rat and female mouse
are within a factor of 3 of each other, but estimates based on the male mouse are still between one and
two orders of magnitude greater than those based on the male rat.
When the linearized multistage model is used to extrapolate from the cancer response in the
National Toxicology Program (NTP) mouse bioassay, a correction factor for early termination, and the
estimates of body burden based on the data of (Bond et al., 1986), the upper-limit estimate of
carcinogenic potency for humans (assuming a 70-year continuous exposure) is 0.25 per ppm. This
finding is very close to the potency estimate based on the analysis by Hattis, also presented at this
workshop (Hattis, 1988). Hattis' estimate of 0.20 per ppm reflects a decrease in risk due to his higher
body burden estimates than those of EPA's. This is partially offset, however, by his final selection of the
male mouse response and by his adding of risks for each tumor site separately compared with EPA's
risk, which is based on total number of animals with at least one of the increased tumor types. Other
factors used in the two risk assessment are similar.
B-54
-------
References
Bond, J.A., Dahl, A.R., Henderson, R.F., Butcher, J.S., Mauderly, J.L., and Birnbaum, L.S., 1986.
"Species differences in the disposition of inhaled butadiene." Toxicol. Appl. Pharmacol. 84:617-627.
Downs, T.D., Crane, M.M., and Kim, K.W., 1987. "Mortality among workers at a butadiene facility"
Am. J. Ind. Med. 12:311-329.
Hattis, D., 1988. "Use of pharmacokinetic analysis in risk assessment - case of 1,3-butadiene."
Presentedat the Workshop on Biological Data for Pharmacokinetic Modeling and Risk Assessment
Asheville, NC, May 1988.
Hazleton Laboratories Europe, Ltd., 1981. "The toxicity and carcinogenicity of butadiene gas
administered to rats by inhalation for approximately 24 months." Prepared for the International
Institute of Synthetic Rubber Producers, New York, NY.
Howe, R.B. and Crump, K.S., 1982. "Global 82--A computer program to extrapolate quantal animal
toxicity data to low doses." Report by K.S. Crump and Company to Office of Carcinogen Standards,
OSHA.
Kreiling, R., Laib, R.J., Filser, J.G., and Bolt, H.M., 1986. "Species differences in butadiene metabolism
between mice and rats evaluated by inhalation pharmacokinetics." Arch. Toxicol. 58:235-238.
Matanoski, G.M., Schwartz, L., Sperrazza, J., and Tonascia, J., 1982. "Mortality of workers in the
styrene-butadiene rubber polymer manufacturing industry." Johns Hopkins University School of
Hygiene and Public Health, Baltimore, MD. Unpublished.
McMichael, A.J., Spirtas, R., Gamble, J.F., and Tousey, P.M., 1976. "Mortality among rubber workers-
relationship to specific jobs." J. Occup. Med. 18:178-185.
Meinhardt, T.J., Lemen, R.A., Crandall, M.S., and Young R.J., 1982. "Environmental epidemiologic
investigation of the styrene-butadiene rubber industry." Scand. J. Work Environ. Health 8:250-259.
National Toxicology Program (NTP), 1984. "Toxicology and carcinogenesis studies of 1,3-butadiene
(CAS106-99-0) in B6C3FJ mice (inhalation studies)."
National Toxicology Program (NTP), 1985. "Quarterly report from Loveland Research Institute, January
1 through March 31, 1985." Interagency agreement 22-Y01-ES-0092. (L. Birnbaum, NTP Project
Officer.) 3
U.S. Environmental Protection Agency, 1984. "Proposed guidelines for carcinogen risk assessment"
Federal Register 49^227) :46?.Q4-46^ni
U.S. Environmental Protection Agency, 1985. Mutagenicitv and Carcinogenicitv Assessment of 1.3-
butadiene. Office of Health and Environmental Assessment, Washington, DC. EPA/600/8-85/004F.
U.S. Environmental Protection Agency, 1986. "Guidelines for carcinogen risk assessment" Federal
Register 51:33993-334005.
B-55
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
METHYLENE CHLORIDE MODEL - INHALATION DATA
Richard Reitz
Methylene chloride (dichloromethane, CH2C12) is a low-molecular-weight volatile solvent with slight
solubility in water. It is widely used in industrial processes, consumer products, and food processing. As
a consequence, the toxicity of CH2Cl2:has been widely studied. In general, this chemical has a low order
of subchronic toxicity, but recent reports have shown that chronic exposure to vapors of CH2C12 is
associated with increased incidences of lung and liver tumors in mice (NTP, 1985). The obvious
question raised by these new findings is whether humans exposed to CH2C12 are likely to develop the
same kinds of tumors seen in mice. While there are many biological factors to consider in answering
this question, one of the primary considerations involves differences in the delivery of the ultimate
toxicant (a metabolite of CH2C12) to the target tissue of different species under different exposure
paradigms. This subject has been the focus of a recent workshop conducted by the National Academy of
Sciences (NAS, 1987).
This paper reviews the development and validation of a physiologically based pharmacokinetic
(PBPK) model capable of providing a ^quantitative description of metabolite production in mice, rats,
hamsters, and humans. A detailed description of the model development and validation may be found in
the work of Andersen et al. (1987). Refinement of the model by inclusion of in vitro enzyme studies
with human and animal tissues has been described by Reitz et al. (1988a), and application of the model
to quantitative cancer risk assessments (which form the major basis of this case example) have been
described by Reitz et al. (1988b).
Incorporation of "internal dose" data from the PBPK model reduced quantitative estimations of
human risk by about two orders of magnitude. A portion of this reduction in the estimated risk arises
from the shift in metabolic pathways (from oxidation to conjugation) as doses of CH2C12 are increased,
and the remainder arises from the different levels of activating enzymes present in the different species.
Use of PBPK models in the future would be strengthened by collection of pharmacokinetic data in
chronically exposed animals, studies in animals of various ages, and development of in vitro systems for
estimating metabolic capacity in humans. Development of systems for estimating metabolic activity in
B-56
-------
humans may be the most important of these three areas, since it is likely that this information will be
lacking in most cases. Many uncertainties remain in the risk assessment process, but analyses that
properly consider the role of physiology and pharmacokinetics should be considerably more reliable than
those that do not.
References
Andersen, M.E., Clewell, H.J., Gargas, M.L., Smith, F.A., and Reitz, R.H., 1987. "Physiologically based
pharmacokinetics and the risk assessment process for methylene chloride." Toxicol. Appl.
Pharmacol. 87:185-205.
National Academy of Sciences (NAS), 1987. "Proceedings of the Pharmacokinetics in Risk Assessment
Workshop." Washington, DC, Oct. 7-9, 1986. Washington, DC: National Academy Press.
National Toxicology Program (NTP), 1985. "NTP technical report on the toxicology and carcinogenesis
studies of dichloromethane in F-344/N rats and B6C3F! mice (inhalation studies)." NTP Technical
Report Series No. 306.
Reitz, R.H., Mendrala, AM., and Guengerich, P.P., 1988a. "In vitro metabolism of methylene chloride
in human and animal tissues: Use in physiologically-based pharmacokinetic models." Toxicol. Appl.
Pharmacol. (in press).
Reitz, R.H., Mendrala, A.M., Park, C.N., Andersen, M.E., and Guengerich, P.P., 1988b. "Incorporation
of in vitro enzyme data into the PBPK model for methylene chloride: Implications for risk
assessment." Toxicol. Lett, (in press).
B-57
-------
SESSION V: THE USE OF PHARMACOK3NETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
METHYLENE CHLORIDE MODEL - INGESTION DATA
Michael J. Angelo
The presentation will be introduced with an historical perspective reviewing the 2-year bioassays
performed by the National Toxicology Program (NTP), in which rats and mice received daily oral gavage
administrations of methylene chloride in a corn oil carrier, and the 2-year bioassay sponsored by the
National Coffee Association (NCA), in which rats and mice were exposed to methylene chloride ad
libitum in drinking water.
NCA believed the drinking-water administration to be a more relevant exposure scenario, but
questions arose as to the manner in which administered dose levels were to be compared between a
bolus "nag/kg" dose and a "mg/kg/day" drinking-water dose, which was consumed gradually over a 24-hour
period.
Pharmacokinetic information on methylene chloride was believed to be very important in
understanding the different internal disposition patterns that resulted from these two dosing protocols.
Limited pharmacokinetic data did exist early in these studies, but not enough were available to quantify
the "delivered" doses to suspected sites or metabolic mechanisms of toxicity.
General Foods decided to contribute to the methylene chloride safety assessment effort by
contracting with the Huntingdon Research Centre to collect pharmacokinetic data that would be used to
understand tissue distribution and metabolism of methylene chloride as a function of several dosing
factors such as dose level, route of exposure, dosing vehicle, and frequency of administration. With this
information, a physiologically based pharmacokinetic model was to be constructed that would aid in
interpreting the different disposition phenomena, quantify "delivered" doses vs. administered doses, and
provide a mechanism for simulating the pharmacokinetics resulting from different exposure patterns in
studies which would otherwise be too complicated or costly to perform.
B-58
-------
References
Angelo, M.J., Bischoff, K.B., Pritchard, AB., and Presser, M.A, 1984. "A physiological model for the
pharmacokinetics of methylene chloride in B6C3F1 mice following i.v. administrations." J.
Pharmacokinet. Biopharm. 12(4):413-436.
Angelo, M.J. and Pritchard, AB., 1986. "Route-to-route extrapolation of dichloromethane exposure
using a physiological pharmacokinetic model." Pharmacokinetics in Risk Assessment Workshop.
National Research Council of the National Academy of Sciences. Washington, DC: National
Academy Press.
Angelo, M.J., Pritchard, AB., Hawkins, D.R., Waller, AR., and Roberts, A, 1986a. "The
pharmacokinetics of dichloromethane. I: Disposition in B6C3F1 mice following intravenous and oral
administrations." Food Chem. Toxicol. 24(9):965-974.
Angelo, M.J., Pritchard, AB., Hawkings, D.R., Waller, AR., and Roberts, A, 1986b. "The
pharmacokinetics of dichloromethane. II: Disposition in Fischer 344 rats following intravenous and
oral administrations." Food Chem. Toxicol. 24(9):975-980.
B-59
-------
SESSION V: THE USE OF PHARMACOKINETIC MODELING IN RISK ASSESSMENT:
CASE STUDIES
THE IMPACT OF PHARMACOKINETICS ON THE RISK
ASSESSMENT OF DICHLOROMETHANE
Jerry N. Blancato and Lorenz Rhomberg
Although it was not the U.S. Environmental Protection Agency's (EPA) first use of pharmacokinetic
data and physiologically based pharmacokinetic models (PBPK) for risk assessment purposes,
dichloromethane is certainly the most publicized and closely watched. Using administered dose as a
basis for risk extrapolation is not the method of choice when other means and data are available. For
example, administered dose ignores obvious pharmacokinetic differences between doses and species, and
in fact, it may not always result in the most conservative assessment, as is commonly thought.
The risk assessment process involves several extrapolation procedures, including going from
relatively high concentrations in the mouse bioassays to lower concentration for the expected potential
human exposure. In performing such extrapolations many assumptions are made and questions of
uncertainty arise. The understanding of pharmacokinetic differences between species and between doses
eliminates only some of these uncertainties.
We did not and do not hesitate to use PBPK models in the risk assessment process. Rather, we
ask questions regarding which model or models best represent the true process in the body, how such
models are validated, and how are the model's respective parameters most accurately determined. Even
after adequately addressing these pharmacokinetic issues, one is left with deciding how to apply the
results of such PBPK models to the risk,assessment process.
In formulating a PBPK model several things must be considered. First, what are the available data?
Are they adequate to support formulation and testing of a rational model that is congruent with known
physiology and anatomy? What is known about the mechanism of action? What delivered dose does
the risk assessor need to know? Is there evidence that there are significant nonlinearities between
administered dose and delivered dose? If not, then administered dose may serve as a very adequate
surrogate for the effective dose at the molecular level. Experience has shown, however, that most
internal biological processes are not linear with administered doses and, thus, some accounting for or
quantifying of the nonlinearities is crucial for proper assessment.
B-60
-------
Two physiologically based pharmacokinetic models have been formulated for dichloromethane, each
with its own peculiarities and each revealing interesting aspects of this compound's pharmacokinetic
behavior. One, the Reitz-Andersen model (Andersen et al., 1987), describes flow-limited conditions in
several organs. The other, developed by Angelo et al. (1984), is a hybrid model; for certain exposure
conditions it described some of the organs as having lipid-containing regions which would cause
dichloromethane to sequester for relatively long times.
In analyzing, validating, and applying pharmacokinetic models for the dichloromethane risk
assessment several important issues arise regarding model formulation and validation. In such
physiologically based pharmacokinetic models there are many crucial thermodynamic, biochemical, and
physiological parameters. The range of values for some, such as blood flows and organ sizes, are fairly
well documented in the literature. Others, such as partitioning ratios and metabolic rate constants, need
to be determined for each chemical of interest. Alternative methods for determining many of these may
exist. For dichloromethane some question arose regarding the reliability of in vitro determination of
both partition coefficients and metabolic rate constants. Before spending an inordinate effort to fine-
tune the exact values, sensitivity analyses may show the relative importance of particular parameters.
Those to which the model is most sensitive should be most carefully determined in the laboratory.
Particularly problematic are those parameters that must be extrapolated from in vitro to in vivo
conditions and those that must be extrapolated from one species to another. Generally, the literature
has shown that total body clearance scales along a body weight to the 0.7 to 0.75 power. It is not so
evident, however, how rate constants for individual metabolic pathways scale. The models for
compounds that are biotransformed will be most sensitive to such rate constants. A challenge exists
then, to develop guidance and methodologies to best determine these metabolic rate constants. It is
probable that species-to-species extrapolation will be accurate only with accompanying in vitro studies.
It is not clear, however, how to decide which in vitro experiment will provide the most suitable
information in a particular case and how to best utilize information gained from such studies in in vivo
pharmacokinetic models.
Model validation has only been lightly considered. One can easily adjust a number of parameters to
fit a particular endpoint, thus yielding a model that can be called, under specific circumstances, validated.
However, the choice of endpoint used for validation can have dramatic impact on the criteria of
validation. For example, as learned from this case, blood levels of dichloromethane do not accurately
reflect tissue levels. Thus, if the tissue level of dichloromethane is the desired delivered dose, validation
B-61
-------
of the model against blood levels might be considered to have limited usefulness. In this particular case,
the endpoint of concern was the amount of toxic metabolite produced. The difficulty here is that this
endpoint could not easily be measured in the laboratory; thus little or no data were available for direct
validation of the endpoint of concern. Because the metabolite was rapidly formed in the liver and lung,
it might be concluded that its formation is closely related to concentration of the substrate, in this case
parent dichloromethane, in the blood. | Under such conditions, one could use blood concentration of
dichloromethane as the product for validation. Since the original model was published, new data have
become available which may allow accurate estimation of metabolite formed by each pathway. If these
data are deemed reliable, they, rather than the blood level of dichloromethane, should be used as the
measure for evaluating the accuracy of the model.
Another area of potentially fruitful research is model simplification. Often in the process of model
formulation there is a tendency to formulate high resolution models, i.e., ones that include a multitude
of tissues and organs. In practical situations, however, sufficient data to support the estimatidn of the
necessary parameters are often lacking. This leads to either expensive and time-consuming laboratory
experiments or to models that are over-parameterized and thus no longer truly physiologic. Guidance is
needed that will counsel modelers on the best way to simplify models in a systematic way that is
consistent with sound mathematical principles and is congruent with the known physiology of the system.
The first step in the process of applying pharmacokinetic models to risk assessment is to decide
what should be used as the delivered dose. The appropriate definition of delivered dose may depend
upon the specific case. Delivered dose may simply be the administered dose corrected for less than
100% absorption efficiency. It may be a measure of the amount of metabolites resulting from
biotransformation of the compound entering the body. In fact, it may be a concentration of a chemical
or one or all of its metabolites in a particular organ, tissue, or cell type. PBPK models can be
formulated to describe and predict a variety of these delivered dose measures. The actual delivered dose
!
selected as the basis on which to conduct the risk assessment depends upon the understanding of the
mechanisms of toxicity, the test species, the exposure conditions, and the levels of data that are available.
For purposes of this discussion, effective dose will be defined as the dose causing the particular
mechanism at the molecular site of action. This effective dose should ideally be the desired endpoint of
any exposure assessment, but in reality it is almost never realized.
In the case of dichloromethane the evidence indicates that, while the exact mechanism is unknown,
the glutathione mediated biotransforming pathway leads to reactive metabolites that are carcinogenic.
Thus, for this particular case, the delivered dose used for risk assessment purposes was the total
B-62
-------
metabolite produced as a result of this biotransforming pathway. This pathway, although exhibiting
linear kinetics at the relevant exposure conditions, competes with another pathway, mediated by the P-
450 system, which exhibits saturation kinetics. The interaction between the two pathways, the difference
in predominance of the pathways at different doses and in different species affected the risk assessment
for this compound. A risk assessment based on administered dose, in contrast, would imply that the
kinetics (not the dose-response curve) were linear at all exposure conditions.
Once a satisfactory definition of delivered dose has been formulated, and once the ability to
estimate this dose using pharmacokinetic modeling is judged sufficient, one is faced with the task of
using this information in revising the extrapolation of risk from bioassay mice to humans at low doses.
The way to use pharmacokinetic information in risk extrapolation is not self evident. The experience
with dichloromethane has revealed several points of view and has helped to define and sharpen the
issues. The balance of the present paper outlines the steps used in EPA's revised cancer risk assessment
for methylene chloride. The rationale for the method is presented along with a discussion of its
implications and assumptions.
In EPA's pharmacokinetically based revision of the dichloromethane unit cancer risk (U.S. EPA,
1987) the following steps were taken: First, delivered dose was defined as the daily amount of reactive
intermediate formed during the course of biotransformation by the glutathione-S-transferase (GST)-
mediated pathway per liter of tissue in organs at risk in the NTP bioassay, namely liver and lung. The
chemical instability of this intermediate molecular species (which is presumably responsible for its
toxicity) indicates that it quickly and spontaneously reacts without leaving the site of formation. Thus,
the total amount formed per liter over the course of an exposure (a "virtual concentration" in Reitz's
terminology) is an index of target tissue exposure. (If the promoter toxin were a more stable metabolite,
its residence time in the tissue would also be at issue, and species differences in its clearance should
enter the definition of delivered dose.)
The second step was to use the Reitz-Andersen pharmacokinetic model to calculate delivered dose
estimates for liver and lung in mice exposed to dichloromethane vapor according to the protocol of the
NTP inhalation bioassay. Then, the multistage model procedure was used to define a low-dose upper
bound on the dose-response curve of tumor incidence as a function of this delivered dose. Separate
analyses were necessary for the two organs at risk, since they receive different delivered doses for a given
inhalation exposure. It should be noted that the need for low-dose extrapolation is not removed by
using delivered doses and that nonlinearities in the dose-response relationship do not arise solely from
pharmacokinetic nonlinearities.
B-63
-------
Finally, the human version of the pharmacokinetic model was used to determine the delivered doses
resulting from continuous exposure to 1 ppm of dichloromethane vapor. The risk engendered by such
human delivered doses (combined for both liver and lung) constitutes a unit risk per ppm for
dichloromethane. At such low exposures, delivered dose is linearly related to air concentration, and risk
is (by the low-dose extrapolation assumption) linearly related to delivered dose. The problematic and
controversial question is: What is the appropriate estimate of human risk for a given daily tissue
exposure vis-a-vis mice with the same delivered dose?
It is important to appreciate that pharmacokinetics does not obviate the question of scaling doses
across species. It addresses the relative delivered doses in mouse and man that result from a given
exposure to dichloromethane vapor, but it does not address the relative, risks arising from those tissue-
level exposures. It is not at all clear that equal "virtual concentrations" experienced for a lifetime (in
units of mg-eq metabolized/L of tissue/day) should engender equal risk in mouse and man; human organs
have about 2,000 times the volume, and presumably 2,000 times the number of cells at risk, only one of
which need be transformed to initiate a tumor. Moreover, the daily tissue exposure continues over a
lifetime that is 35 times longer in a human than in a rodent. These factors alone suggest that humans
should be extraordinarily sensitive to carcinogens ~ an assertion that is not supported by epidemiological
findings. Clearly, other factors are at work as well. Thus, fully assessing the impact of pharmacokinetics
on relative risk across species depends crucially on a remaining question, that of the comparative
pharmacodynamics of carcinogenic response across species. This constitutes a major research need in the
field of risk assessment.
Faced with this problem, we have examined the contribution of the pharmacokinetic component
alone. This entails defining a "usual" or "default" expectation about the contribution of pharmacokinetic
differences across species to the relative carcinogenic potencies of chemicals. Actual pharmacokinetic
results, then, should change our potency calculations to the degree that they show these prior
presumptions to be incorrect. We choose an assumption about the pharmacodynamic component that
will leave our old risk estimates (calculated on the basis of administered dose) unchanged when the
pharmacokinetics turn out to be in line with the prior expectations based on general principles.
First consider external or administered dose. This constitutes the amount of compound breathed in
(albeit not necessarily absorbed) during the course of exposure, and it forms the traditional basis of dose
calculation against which the pharmacokinetic approach is being compared. Administered dose is
calculated by multiplying the breathing rate (in L of air/day) times the vapor concentration (in mg/L),
B-64
-------
and dividing by body weight to give a dose in mg/kg/day. Since breathing rate varies in approximate
proportion to the surface area/volume ratio of the body, mice experience about a 13-fold higher
administered dose than humans for a given episode of breathing when contaminated with a given
concentration of vapor.
Starting from widely accepted general principles, one can show that the fraction of this administered
dose that is metabolized should be about equal across species, despite the allometric scaling of metabolic
rates. Briefly, during an inhalation exposure to a compound (such as dichloromethane) that is readily
absorbed and readily expired, the blood and tissue concentrations rise until they reach a steady state, at
which new net absorption is limited to the replacement of material lost to metabolism and
nonpulmonary excretion. In essence, steady state represents the equilibration of the amount of
compound dissolved in the external air and in the tissues. The concentration in the tissues is
determined by the relative solubilities of the compound in tissue, blood, and air. These solubilities are
physicochemical properties of the compound, and so should be approximately equal across species.
Hence, steady state tissue concentrations should be about equal across species, in spite of the mouse's
relatively high breathing rate (which only affects the speed with which equilibration is reached). Given
equal substrate concentrations, the rate of metabolism of the compound in metabolizing tissues (in
amount per liter of tissue per minute) conventionally scales in proportion to the animal's surface area to
volume ratio. Hence, mice have relatively fast metabolism compared to humans and are expected to
metabolize about 13 times more compound per liter of tissue per minute at steady state.
In sum, mice have a 13-fold higher administered dose, and they are expected, (on the basis of the
above argument from general principles) to metabolize 13-fold more compound than humans. Thus, the
fraction of their administered dose that is metabolized is expected to be about equal to that in humans.
Of course, the conventional allometric scaling of breathing rate and of metabolism of the compound, as
well as the steady state assumptions, need not be adhered to in a particular case. The above argument
provides a point of departure, based on the conventional and accepted allometric differences among
differently sized species, against which to judge the actual pharmacokinetic results. In a compound with
"typical" pharmacokinetics, the metabolized dose would be a constant fraction of the administered dose,
and the calculations based on delivered dose would give the same risk extrapolation as those based on
administered dose.
In the case of dichloromethane, the results of the Reitz-Anderson pharmacokinetic model show that
at similar air concentrations, humans and mice do metabolize about the same fraction of their
administered dose, as expected. But owing to changes in the relative activity of the GST and MFO
B-65
-------
metabolic pathways at high and low doses (arising from the saturation of MFO at high exposures) the
proportion of a dose metabolized by the GST pathway is lower at low doses than at high ones, in both
mice and humans. As a result, the low-dose human tissue exposures to the presumed proximate
carcinogen are a few fold (but not hundreds of times) lower than one would expect from the above
scaling argument. Accordingly, the EPA document advocates lowering the inhalation unit risk by 8.8-
fold because slightly less of the administered dose is metabolically activated in humans at low doses than
in mice at the bioassay levels of exposure. The difference is mostly due to the high dose to low dose
component and not to species differences in metabolism, although for other compounds this could be
different.
References
Andersen, M.E., Clewell III, H.J., Gargas ML., Smith F.A., and Reitz R.H., 1986. "Physiologically based
pharmacokinetics and the risk assessment process for methylene chloride." Toxicol. Appl.
Pharmacol. 87:185-205.
Angelo, M.J., Bischoff K.B., Pritchard AB., and Presser M.A., 1984. "A physiological model for the
pharmacokinetics of methylene chloride in B6C£i mice following I.V. administration." J.
Pharmacokinet. Biopharm. 12:413-436.
!
U.S. Environmental Protection Agency, 1987. "Update to the health assessment document and
addendum for dichloromethane (methylene chloride): Pharmacokinetics, mechanism of action, and
epidemiology." Office of Health and Environmental Assessment, Washington, DC. External Review
Draft. EPA/600/8-87/030A,
B-66
-------
APPENDIX C
LIST OF PARTICIPANTS
C-l
-------
WORKSHOP ON BIOLOGICAL DATA FOR PHARMACOKINETIC
MODELING AND RISK ASSESSMENT
May 23-25, 1988
LIST OF PARTICIPANTS
William T. Allaben
Scientific Coordination HFT-30
Department of Health and Human Services
Public Health Service
Food and Drug Administration
National Center for Toxicological Research
Jefferson, AR 72079
501-541-4000
Bruce C. Allen
Clement Associates
1201 Gaines St.
Ruston, LA 71270
318-255-4800
Harriet M. Ammann
U.S. Environmental Protection Agency
Environmental Criteria and Assessment Office
(MD-52)
Research Triangle Park, NC 27711
Michael J. Angelo
Director
Merck and Company
Technical Operations
P.O. Box 7
U.S. 340 South of Elkton
Elkton, VA 22827
703-298-1211
G.A. Shakeel Ansari
University of Texas Medical Branch
Division of Chemical Pathology
Galveston, TX 77550
409-761-3656
Angela D. Arms
Health and Safety Research Division
Oak Ridge National Laboratory
P.O. Box X
Oak Ridge, TN 37831-6109
Nicholas A. Ashford
Massachusetts Institute of Technology
School of Engineering
Building E40-239
77 Massachusetts Avenue
Cambridge, MA 02139
617-253:1664
Stan Atwood
NC Division of Health Services
306 N. Wilmington St.
Raleigh, NC 27602-2091
919-733-2801
A John Bailer
NIEHS B-302
P.O. Box 12233
111 T.W. Alexander Drive
Research Triangle Park, NC 27709
919-541-4929
Ambika Bathija
729 Ridge Drive
McLean, VA 22101
202-382-4785
Steven Bayard
Carcinogen Assessment Group (RD-689)
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460
202-382-5722
Linda Birnbaum
NIEHS
C3-02
South Campus, Bldg. 101
TW Alexander Drive
Research Triangle Park, NC 27709
919-541-3583
C-2
-------
Salvatore F. Biscardi
Pharmacologist, Oncology Branch
Health & Environmental Review Division
(TS-796)
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460
202-382-4288
Kenneth B. Bischoff
Dept. of Chemical Engineering
University of Delaware
Academy Street
Colburn Laboratory
Newark, DE 19716
302-451-2830
Ann M. Blacker
Rhone-Poulenc AG Company
P.O. Box 12014
Research Triangle Park, NC 27709
919-549-2639
Jerry Blancato
U.S. Environmental Protection Agency
Exposure Assessment Group (RD-689)
401 M Street, S.W.
Washington, DC 20460
202-475-8918
Harold H. Borgstedt
HDI
183 East Main Street
Rochester, NY 14604
716-546-1464
William S. Bosan
Chevron Environmental Health Center, Inc.
15299 San Pablo Avenue
Richmond, CA 94804-0054
Olen R. Brown
University of Missouri
Dalton Research Center
Columbia, MO 65211
314-882-4528
Robert Brown
U.S. Food and Drug Administration (HFF-118)
200 C St., S.W.
Washington, DC 20204
202-485-0080
Ronald P. Brown
Technical Resources, Inc.
3202 Monroe St.
Rockville, MD 20852
301-231-5250
James V. Bruckner
Associate Professor and Director of Toxicology
The University of Georgia College of Pharmacy
Athens, GA 30602
Leo T. Burka
NIEHS
P.O. Box 12233
Research Triangle Park, NC 27709
919-541-4667
Jim Carlisle
California Department of Food and
Agriculture - Medical Toxicology Branch
P. O. Box 942871
Sacramento, CA 94271-0001
James J. Chen
Mathematical Statistician
Department of Health & Human Services
National Center for Toxicological Research
Jefferson, AR 72079
Florence F. Chiao
262 Arlington Ave.
Kensington, CA 94707
415-525-5542
Robert Chinery
Bureau of Toxic Substances Assessment
State of New York
Department of Health
Corning Tower
The Governor Nelson A, Rockefeller Empire
State Plaza
Albany, NY 12237
C-3
-------
M. Ann Clevenger
Environ Corporation
The Flour Mill
1000 Potomac Street, N.W.
Washington, DC 20007
202-337-7444
Harvey Clewell III
AAMRL/TH Building 79 Area B
Wright Patterson AFB :
Ohio 45433-6573
513-255-5740
Murray S. Cohn
Consumer Product Safety Commission
Room 700
5401 Westbard Avenue
Bethesda,MD 20207
301-492-6994
John B. Cologne
University of Washington
Biostatistics, SC-32
Seattle, WA 98195
206-545-2737
Rory B. Conolly
Northrop Services, Inc.
101 Woodman Dr.
Suite 12
Dayton, OH 45431-1482
513-256-3600/258-1150
Kim Grossman
Connecticut Dept. of Environmental
Protection ;
Air Compliance Unit
165 Capitol Avenue
Hartford, CT 06106
(203) 566-2690
Richard F. Cullison
U.S. Food and Drug Administration
Center for Veterinary Medicine
5600 Fishers Lane
Rockville, MD 20857
301-443-5540
Marc H. Davie
S.C. Johnson & Son, Inc.
1525 Howe St., MS 315
Racine, WI 53143
414-631-2710
Barbara Davis
CH2M Hill
P.O. Box 4400
Reston, VA 22090
703-471-1441
Mary E. Davis
West Virginia University Medical Center
Department of Pharmacology and Toxicology
Morgantown, WV 26506
304-293-3414 (office)
304-293-4449 (lab)
Robert L. Dedrick
Division of Research Services
Biomedical Engineering and Instrumentation
Branch
Building 13, Room 3W10C
NIH
Bethesda, MD 20892
301-496-5771
Richard W. D'Souza
Miami Valley Labs
P.O. Box 398707
Cincinnati, OH 45239-8707
513-245-2419
David Dolan
Region 5 (5WD-TUB-9)
U.S. Environmental Protection Agency
230 S. Dearborn Street
Chicago, IL 60604
(312)886-9532
Tom Downs
The University of Texas
School of Public Health
P.O. Box 20186
Houston, TX 77225
713-792-4421
C-4
-------
Elizabeth. A Doyle
Gillette Medical Evaluation Laboratories
1413 Research Blvd.
Rockville, MD 20850
301-424-2000
William Farland
Office of Health and Environmental
Assessment (RD-689)
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460
202-382-7317
Lawrence J. Fischer
Michigan State University
Center for Environmental Toxicology
C-231 Holden Hall
East Lansing, MI 48824
517-353-6469
H.L. Fisher
U.S. Environmental Protection Agency
MD-2
Research Triangle Park, NC 27711
919-541-2631
Nicholas M. Fleischer
U.S. Food and Drug Administration
12 Chancelet Ct.
Rockville, MD 20852
301-443-1223
John M. Frazier
School of Hygiene and Public Health
The Johns Hopkins University
Department of Environmental Health Sciences
615 North Wolfe Street
Baltimore, MD 21205-2179
301-955-4712
Clay B. Frederick
Rohm and Haas Company
Research Laboratories
727 Norristown Road
Spring House, PA 19477
215-641-7000
John R. Froines
UCLA School of Public Health
CHS
10833 LeConte
Los Angeles, CA 90024-1772
213-825-7104
W. Don Galloway
Supervisory Psychologist
Health Sciences Branch (HFZ-112)
Division of Life Sciences
Office of Science and Technology
Center for Devices and Radiological Health
U.S. Food and Drug Administration
12709 Twinbrook Parkway
Rockville, MD 20857
Irene Glowinski
ILSI Risk Science Institute
1126 16th Street, N.W.
Suite 100
Washington, DC 20036
202-659-3306
Robert J. Golden (attending Mon. only)
Karch & Associates, Inc.
1805 Florida Ave., NW
Washington, DC 20009
202-667-3031
Jay Goldring
University of North Carolina/NIEHS
Box 12233, MD C3-03
Research Triangle Park, NC 27709
919 541-3525
Albert F. Gunnison
NYU Medical Center
550 First Avenue
New York, NY 10016
212-340-7300 x 885
Daniel J. Guth
U.S. Environmental Protection Agency
MD-13
Research Triangle Park, NC 27711
919-541-5340
C-5
-------
Dale Hattis
Massachusetts Institute of Technology
Building E40-227
Cambridge, MA 02139
617-253-6468
Bill Hayton
College of Pharmacy
Washington State University
Pullman, WA 99164-6510
509-335-5622 !
James R. Harr
USDA/FSIS
300 12th St., SW
Washington, DC 20250 ;
202-447-2807
Milan J. Hazucha
School of Medicine
The University of North Carolina at Chapel Hill
Trailer #4, Medical Building C, 224H
Chapel Hill, NC 27514
919-962-0126
Jenifer S. Heath
North Carolina Department of Human
Resources
Division of Health Services
P.O. Box 2091
Raleigh, NC 27602-2091
Lyse D. Helsing
Senior Environmental Health Scientist
Risk Science International
1101 30th St., NW ;
Washington, DC 20007 ;
202-342-2206
Carol Henry
ILSI Risk Science Institute
1126 16th Street, N.W.
Suite 100
Washington, DC 20036
202-659-3306
Sara Hale Henry
Department of Health and Human Services
Food and Drug Administration
Washington, DC 20204
Oscar Hernandez
The Mitre Corporation
Metrek Division
7525 Colshire Drive
McLean, VA 22101-3481
Stephen G. Hundley
E.I. DuPont DeNemours & Company
Haskell Laboratory for Toxicology and
Industrial Medicine
P.O. Box 50
Elkton Road
Newark, DE 19714
George J. Ikeda
U.S. Food and Drug Administration
Route 2, Box 277
Laurel, MD 20708
301-344-4004
Richard D. Irons
Molecular Toxicology and
Environmental Health Sciences
Health Sciences Center
University of Colorado
Box C238
4200 East 9th Avenue
Denver, CO 80262
303-270-7170
Annie M. Jarabek
U.S. Environmental Protection Agency
Environmental Criteria and Assessment Office
(MD-52)
Research Triangle Park, NC 27711
Dennis E. Jones
ATSDR
1600 Clifton Rd./F-38
Atlanta, GA 30333
404-488-4835
C.D. Kary
Ashland Oil, Inc.
P.O. Box 391
Ashland, KY 41114
606-329-3062
C-6
-------
R.W. Kilpper
Xerox Corporation
Joseph C. Wilson Center for Technology
Building 0843-16S
Webster, NY 14580
Gary D. Koritz
University of Illinois
College of Veterinary Medicine
Urbana,IL 61801
217-333-7981
Robert I. Krieger
California Dept. of Food and Agriculture
Worker Health & Safety
1220 N Street
Sacramento, CA 95814
916-922-3626
Paul T. LaRocca
Schering Corp.
P.O. Box 32
Lafayette, NJ 07848
201-579-4244
Peter Leung
Schering Corporation
60 Orange St.
Bloomfield, NJ 07003
201-429-4009
Stan L. Lindstedt
Dept. of Zoology and Physiology
University of Wyoming
Laramie, WY 82071
307-766-6100
Ronald J. Lorentzen
U.S. Food and Drug Administration (HFF-104)
200 C Street, S.W.
Washington, DC 20204
202-485-0046
Lawrence K. Low
Pharmacokinetics and Metabolism Section
Environmental and Health Safety Laboratory
Mobil Oil Co.
P.O. Box 1029
Princeton, NJ 08540
609-737-5546
George W. Lucier
NIEHS
P.O. Box 12233
A3-02
Research Triangle Park, NC 27709
919-541-3802
Brian H. Magee
E.C. Jordan Co.
Corporate Place 128
107 Audubon Rd., Suite 301
Wakefield, MA 01880
617-245-6606
Allan H. Marcus
Battelle Memorial Institute
200 Park Drive
P.O. Box 13758
Research Triangle Park, NC 27709
919-549-8970
James M. McKim
U.S. Environmental Protection Agency
Environmental Research Laboratory ~ Duluth
6201 Congdon Blvd.
Duluth, MN 55804
Michele A. Medinsky
Lovelace ITRI
P.O. Box 5890
Albuquerque, NM 87185
505-844-2207
Timothy J. Mohin
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
Michael P. Moorman
Department of Health and Human Services
Public Health Service
National Institute of Environmental Health
Sciences
MD 14-08, Box 12233
Research Triangle Park, NC 27709
919-541-3404
Joyce Mordenti
Genentech
460 Point San Bruno Blvd.
San Francisco, CA 94080
415-266-2771
C-7
-------
David A. Morgott
Eastman Kodak Company
343 State Street
Rochester, NY 14650
Marilyn E. Morris
Department of Pharmaceutics
527 Hochstetter Hall
SUNY-Buffalo <
Amherst, NY 14260
John Overton
Toxicology Branch (MD 82)
Inhalation Toxicology Division
U.S. Environmental Protection Agency
Health Effects Research Laboratory
Research Triangle Park, NC 27711
Linda Penniman !
Occupational Safety and Health Administration
U.S. Department of Labor
Washington, DC 20210
Steven Pike
EnviroMed
2200 E. River Road
Suite 123
Tucson, AZ 85718
602-577-0818
Andrew Pope
National Research Council, NAS r
2101 Constitution Ave., NW
Washington, DC 20418
202-334-2534
Christopher Portier
Statistics and Biomathematics Branch
Division of Biometry and i
Risk Assessment
NIEHS (B3-02)
P.O. Box 12233
Research Triangle Park, NC 27709
919-541-4999
Gloria B. Post
Research Scientist
State of New Jersey
Department of Environmental Protection
Division of Science and Research
CN409 i
Trenton, NJ 08625
609-292-8497
Mary L. Quaife
1506 33rd St., N.W.
Washington, DC 20007
202-333-8391
Martha J. Radike
University of Cincinnati Medical Center
Institute of Environmental Health
Kettering Laboratory
3223 Eden Avenue
Cincinnati, OH 45267-0056
V.C. Ravikumar
NLU School of Pharmacy
700 University Ave.
Monroe, LA 71209
318-343-3683
Ravindra R. Raje
Long Island University
75 DeKalb Avenue
Brooklyn, NY 11201
718-403-1062
Richard H. Reitz
The Dow Chemical Company
Toxicology Research Laboratory
Building 1803
Midland, MI 48640
517-636-5995
S.A. Ridlon
ARCO Chemical Company
3601 West Chester Pike
Newtown Square, PA 19073
Stephen M. Roberts
University of Arkansas for Medical Sciences
4301 W. Markham
Little Rock, AK 72205-7199
Joseph V. Rodricks
ENVIRON Corporation
Terrace Level
1000 Potomac St., N.W.
Washington, DC 20007
202-337-7444
C-8
-------
Brian M. Sadler
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
919-541-6687
Julio A. Salinas
1372 High St.
Westwood, MA 02090
617-769-0021
Jaci Schlachter
U.S. Environmental Protection Agency
726 Minnesota Ave.
Kansas City, KS 66101
913-236-2815
Miriam de Salegui
845 West End Avenue, Apt. 5F
New York, NY 10025
Mary Jean Sawey
School of Public Health
Harvard University
Laboratory of Radiobiology
665 Huntington Avenue
Boston, MA 02115
617-732-1184
Lee R. Schull
Covell Health Center
2043 Anderson Road
Suite F
Davis, CA 95616
916-758-3811
Cheryl Seigel-Scott
U.S. Environmental Protection Agency
Office of Toxic Substances (TS-798)
401 M Street, S.W.
Washington, DC 20460
(202) 382-3895
Danny D. Shen
University of Washington
Department of Pharmaceutics, BG-20
Seattle, WA 98195
206-545-2920
Waheed H. Siddiqui
Dow Corning Corporation
2200 W. Salzburg Rd.
Midland, MI 48640
517-496-4884
I. Glenn Sipes
University of Arizona
Department of Pharmacology and
Toxicology
College of Pharmacy
1703 East Mabel
Tucson, AZ 85721
602-626-7123
Karen J. Skinner
Office of the Commissioner
Department of Health and Human Services
Federal Drug Administration
Rockville, MD 20857
Stu M. Somani
Southern Illinois University
School of Medicine
801 N. Rutledge
Springfield, IL 62708
217-785-2196
Anne Spacie
Department of Forestry and Natural Resources
Purdue University
Forestry Building
West Lafayette, IN 47907
Hugh Spitzer
HESD/7th Floor
American Petroleum Institute
1220 L Street, N.W.
Washington, DC 20005
202-682-8338
John Stoudemire
Genetics Institute
57 Cambridge Park Drive
Cambridge, MA 02140
617-876-1170
C-9
-------
Lester Sultatos
New Jersey Medical School
185 S. Orange Ave.
Newark, NJ 07103-2757
201-456-6612
Angelo Turturro
National Center for Toxicological
Research
Office of the Director
County Road 3
Jefferson, AR 72079
501-541-4517
Curtis Travis
Oak Ridge National Laboratory
P.O. Box 2008
Building 45000 S., MS 109 5204
Bethel Valley Road
Oak Ridge, TN 37831
615-576-2107
John J. Vandenberg
Pollutant Assessment Branch
Emission Standards Division
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
Nicholas E. Weber
Center for Veterinary Medicine
U.S. Food & Drug Administration
5600 Fishers Lane '
Rockville, MD 20857
301-443-6971
Richard J. Wenning
Envirologic Data, Inc.
295 Forest Ave.
Portland, ME 04101
207-773-3020
Gary K. Whitmyre
Versar Inc.
6850 Versar Center
P.O. Box 1549
Springfield, VA 22151
703-750-3000
Alan Wilson
Monsanto Company
645 S. Newstead Ave.
St. Louis, MO 63110
314-694-7986
Carl K. Winter
Division of Toxicology
University of California
Riverside, CA 92521
714-787-5994
Joseph J. Yang
Mobil Oil Company
Environmental Affairs and Toxicology
Department
P.O. Box 1029
Princeton, NJ 08540
609-737-5500
David W. Yesair
BioMolecular Products, Inc.
P.O. Box 347
Byfield, MA 01922
617-462-2224
John F. Young
Reproductive & Developmental Toxicology
N.C.T.R.
Jefferson, AR 72079
501-541-4304
Constantino Zervos
Department of Health & Human Services
U.S. Food and Drug Administration
Washington, DC 20204
*U.S. GOVERNMENT PRINTING OFFICE;! 990 -711B.1S9/ 0 0 to 1
C-10
-------
-------
TJ O
o> .5;
S- oo
-r 3
= CD
m w
cn
CD
C
U!
CD
CO
O
O
rn
-a
cn
o
o
OJ
i
ID
O
£^ ' JIT '
S-< »:
O O i '
T -i 3
lit
Q) (t)
a?
-? 3
3 -<
S3-
~> (U
R> en
O
I
" m
2S
> me
CO 3 D
CD < rt
3 =V CD
CD D>
E «
0)
O
CD
O
3
3' 3 »
D Si * I
- 6'
3
o
X
o>
en
m|
3
3
3
CD
3
CD
0)
T3
m
33
O
C/3
^
> CD
ti
m>
%d
>
o
------- |