UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D,C. 20460
my 26, 1987
SMC-87-030
r F i C t OF
Honorable Lee M. Thomas . T
Administrator
U.S. Environmental Protection Agency
401 M Street, SW
Washington, DC 20460
Dear Mr. Thomas:
The Deputy Administrator and the Assistant Administrator for Research
and Development requested that the Science Advisory Board (SAB) review the
progress made by the Office of Research and Development (OHD) in addressing
EP&'S needs for extrapolation models. The SAB Executive Committee formed
an Extrapolation Models Subconmittee which conducted the review in public
session. The Subcommittee ' s full report, which is attached, describes!
1) the problem that EPA confronts in developing and. using extrapolation
models; 2) ongoing work from the perspective of the '-'type of extrapolation;
3) an analysis of the OKD effort organized according to scientific discipline;
4) a perspective on the overall effort of the Federal government in this
area of research; and 55 the Subcommittee's general ccmments and conclusions*
Models that allow one to extrapolate from one set of scientific
phenomena or observations to another are an important component of the
risk assessment process. The use of extrapolation models is subject to
considerable scientific uncertainty, and many such models lack recent
scientific review. Given EPA's commitment to using risk assessment in
regulatory decision making, it is imperative that the Agency promote
efforts to iitprove extrapolation models.
The subject of extrapolation modeling is ecmplex. The Subcommittee
believes that the field can be described as a multidiseiplinary matrix
so that the work can be viewed frcm different perspectives, such as the
kind of extrapolation process, the stage of model development, the
scientific discipline involved or the general approach to modeling*
Progress in model development can be analyzed from these different
perspectives. The Subcommittee developed the following two principles to
evaluate research plans:
1) If research on an extrapolation model is successful, how
will the Agency be able to better assess risks? TJiiat can
EPA do with an improved model that it cannot do without one?
2) Will successful research on a model establish leadership
for EPA within the scientific community and promote interest
in the model outside of EPA?
-------
-2-
The Subcommittee's major finding is that there is no overall,
conceptually integrated Agency research program on extrapolation modelingr
but a conglomeration of investigator-initiated projects, many of which are
commendable in their design and implementation. The Subcommittee was
impressed with the talent of many of the individuals in the research staff
within the Health Effects Research Laboratory,
The Subeonsnittee' s major recommendation is that EPA should develop
a comprehensive plan for an extrapolation models research program that
should; 1) articulate an overall conceptual objective towards which
individual projects would aijn; 2) enhance EPA's risk assessment-risk
management philosophy; 3) develop a framework that promotes more planning
and resource stability in support of the research? 4) provide a caramon
nomenclature; 5) improve ceramunication among the Agency's organizational
components; and 6) explain to the nonscientist how the research on
extrapolation rodels supports the Agency's regulatory decisions.
EPA must provide leadership within the Federal government to Improve
existing extrapolation models. EPA shares with other regulatory agencies
a great need for better models, and has some resources to perform research
and to stimulate work by the major Federal research organizations. Thus,
extrapolation modeling creates a unique research opportunity and agenda
for EPA.
The Subcommittee appreciates the opportunity to review EPA's ongoing
work in extrapolation modeling. The Science Advisory Board also looks
forward to a continuing involvxnent in the further development and
application of this research. We also request that the Agency formally
respond to our report.
Sincerely,
Ronald Wyzga, Chair
Extrapolation Models Subcommittee
Science Advisory Board
Norton Nelson, Chair
Executive Committee
Science Advisory Board
-------
SAB-EC- 87-030
REVIEW OF RESEARCH IN SUPPORT OF EXTRAPOLATION* MODELS
BY EPA'S OFFICE OF RESEARCH £ND DEVELOPMENT
fay the
Extrapolation Models Subccrattiittee
Science Advisory Board
United States Environmental Protection Agency
May 1987
-------
U.S. ENVIFDNMENTAL PROTECTION AGENCY
NOTICE
This report has been written as a part of the activities of the Science
Advisory Board, a public advisory group providing extramural scientific
information and advice to the Administrator and othei* officials of the
Environmental Protection Agency. The Board is structured to provide a
balanced expert assessment of scientific matters related to problems facing
the Agency. This report has not been reviewed for approval by the Agency,
and hence, the contents of this report do not necessarily represent the views
and policies of the Environmental Protection Agency, nor of other agencies
in the Executives Branch of the Federal government, nor does mention of
trade names or commercial products constitute endorsement of recommendation
for use.
-------
11
U.S. ENVIRONMENTAL PROTECTION AGENCY
SCIENCE ADVISORY BOARD
EXTRAPOLATION MDDSLS SUBCOMMITTEE
Chair
Dr* Ronald Wyzga, Electric Power Research Institute, 3412 Hillview Avenue
P.O. Box 1041, Palo Alto, California 94303
Members
Dr. Janes Bond, Chemistry-Toxicology Group, Inhalation Toxicology Research
Institute, Lovelace Foundation, P.O. Box 5890, Building 9200 Area Y,
Kirkland APB East,. Albuquerque, New Mexico 87113
Dr. Kenneth Brown, Research Triangle Institute, P.O. Box 12194, Research
Triangle Park, North Carolina 27709
Dr. Sergio E. Fabro, DECEASED, Reproductive Toxicology Center, Research
Foundation of the Columbia Hospital for Women, 2425 L Street, M.W.
Washington, D.C. 20037
Dr. James Gibson, Chemical Industry Institute of Toxicology, 16 Davis
Drive, Research Triangle Park, North Carolina 27709
Dr. Michael Cough, Environ Corporation, 1000 Potomac Street, N.W.,
Washington, D.C. 20007
Dr. Marshall Johnson, Professor, Department of Anatomy, Jefferson Medical
College, 1020 Locust Street, Philadelphia, Pennsylvania 19107
Dr. Don E. McMillan, Chairman, Department of Pharmacology, Mail t 638,
University of Arkansas, Medical Sciences, 4301 West Markham St., Little
Rock, Arkansas 72205
Dr, D. Warner North, Principal, Decision Focus Inc. Los Altos Office
Center, Suite 200, 4984 El Gandno Real, Los Altos, California 94022
Dr. Glenn Paulson, Clean Sites Inc., 1199 N. Fairfax St., Suite 400,
Alexandria, Virginia 22314
Dr, Richard Schlesinger, Associate Professor, Institute of Environmental
Medicine, Lanza laboratory, Long Meadow Road, New York University,
Tuxedo, New York 10987
Ex-Of f icio Member
Dr. John Neuhold, Department of Wildlife Science, College of Natural Resources,
Utah State University, Logan Utah 843322
Executive Secretary
Dr. Daniel Byrd III, Science Advisory Board (A-1Q1-F), U.S. Environmental
Protection Agency, Washington, D.C, 20460 (Until March 27, 1987)
Dr. Terry P. Yosie, Director, Science Advisory Board, (A-101), U.S.
Environmental Protection Agency, Washington, DC 20460
-------
Ill
TABLE OF CONTENTS
PAGE
I. EXECUTIVE SUMMARY 1
II, INTPOOCTION " 6
A. Definitions . * 6
B. Agency Uses of Extrapolation Models 7
C. The Complex Nature of Extrapolation Modeling 8
D. Relationships Between Extrapolation Modeling, 11
Pharmacokineties, Statistical Approaches and
RisR Assessment
E. Operational Objectives 11
F. Scope of the Review 12
III. ORD'S APPROACH TO EXTRAPOLATION MODELING 14
A. EPA's Research Committees 14
B. Fonnulation o£ the ORD Plan 14
C, Budgetary Support for ORD's Extrapolation 15
Modeling Research
D. The Parallelogram Approach 15
IV. ORD'S PROGRAM PROM THE PERSPECTIVE OP EXTRMOL&TION PROCESSES 17
A» Extrapolation Between Species 17
B. Extrapolation Between Subpopulations of 18
Differing Sensitivity
C. Extrapolation Between Pathological Endpoints 18
or Organ Systems
D, Interpolation Between Doses 19
E, Extrapolation Between Routes of Administration 20
F. Extrapolation Across Durations of 'Exposure 20
G* Extrapolation Between Times of Effect 20
-------
iv
PAGE
H. Extrapolation Between Developmental Stages 20
I. Extrapolation Between Different Chemical 21
Structures
J, Extrapolation Pron In Vitro Test Data 21
to Whole Animal Effects
V. ORD'S PROGRAM FROM THE PERSPECTIVE OF SCIENTIFIC 23
DISCIPLINES
A. Phamacokinetics 23
B. Carcinogenicity ' 24
C, Mutagenicity 26
P, Non-ionizing Radiation 29
E. Comparative Toxicology 30
P. Neurotoxicology 31
G. Systemic Toxicity 33
H. Inhalation Toxicology 34
I. Reproductive and Developmental Effects 37
VI, THE OVERALL FEDERAL RESEARCH EFFORT ON 40
EXTRAPOLATION MODELING
A. Non-QRD Extrapolation Programs in EPA 40
B, Other Federal Regulatory Agencies 41
C. Other Federal Research Agencies 41
APPENDIX I Office of Research and Development •
Briefing Document Executive Summary:
Status of Extrapolation Modeling Research
Needed To Extrapolate Fran Aninsal Data
To Human Risk, Frcsn High To Low Doses,
And Frcm Acute To Chronic Effects,
September 1985
-------
I. EXECUTIVE SUMMARY
1PA uses risk assessment as a technical basis for developing regulations
and standards. Models that allow one to extrapolate from one set of scientific
phenomena or observations to another are an important component of risk
assessment. Not all participants in the regulatory process outside of EPA
accept the use, or the extent of use, of extrapolation models* Although
the Subcommittee members believe that the use of extrapolation models is
intrinsically acceptable, to others the choice of a particular assumption
or model seems arbitrary. These choices can result in larges differences
in risk estimates and thus, regulatory decisions.
Ml extrapolation models are subject to considerable scientific
uncertainty, and many such models lack recent scientific review. To improve
public acceptance of the regulatory process at EPA, it is imperative that
the Agency promote efforts to improve and validate extrapolation models.
The development of accepted extrapolation models can also inprove the
Agency's use of its resources because these models can be substituted for
more intensive (and often more expensive) collection of directly applicable
data.
The subject of extrapolation modeling is complex. The Subcommittee
believes that the field is multidisciplinary and that the work can be viewed
from different perspectives, such as the kind of extrapolation process, the
stage of model development, the scientific discipline involved or the
general approach to modeling. The Subcommittee developed two principles to
evaluate EPA's research on extrapolation models. These include:
1) If research on an extrapolation tnodel is successful, how will
the Agency be able to better assess risks? What can EPA do
with an improved model that it can not do without one?
2) will successful research on a model establish leadership for
EPA within the scientific community and promote interest in
the model outside of EPA?
EPA's Office of Research and Development (ORD) provided a well
written briefing document for the Subcommittee. However, limitation of the
review to intramural projects within OFD made it difficult for the Subcom-
mittee to evaluate the comprehensiveness and coherence of ongoing work
because key elements may have been externally performed. Scientific
personnel working in EPA's program offices also contribute significantly to
model development. Based on knowledge from other SAB reviews, some Subcom-
mittee members also noted that elements of ORD's ongoing internal work was
also not included, particularly in the areas of ecological and dosimetric
modeIsf or models that define the movement of pollutants from the environment
to a receptor. These omissions, and the time lag in the Subcommittee's
preparation of this report, probably have resulted in recommendations that
parallel more recent ORD work that is underway*
-------
In their presentations, OFD's scientists emphasized specific projects
conducted by various organizational components. Combined with the lack
of a comprehensive strategy to direct them, this emphasis placed the Sub-
committee in something of a dilemma. The Subcommittee had expected more
ofi a synthesis or overview of the GBD work frcra the perspective of
extrapolation processes. Therefore, instead of commenting on a strategic
document and placing more emphasis on reviewing specific models used by the
regulatory offices, the Subcommittee has constructed a synthesis of its own.
*
At present, work on extrapolation models is a small portion of ORD's
overall responsibility. In some cases, small pieces of research are carried
out with no clear relationship to other research projects or to long-tenti
goals. In other cases, whole areas of extrapolation modeling are apparently
ignored. Often, projects are funded for purposes, other than the advancement
of extrapolation modeling. Such "piggyback" funding permits investigators
the time to reorient and refocus their ongoing work onto extrapolation
modeling topics. The overall funding level is low in relation to the
magnitude of the problem facing the Agency. Given these circumstances,
Agency management should not develop unrealistic expectations of ORD*
Support for extrapolation modeling could suffer frctw EPA's approach to
allocating research funds through program-oriented research oonmittees
because the work seldom is program or medium specific. However, there are .
examples which suggest that this is not a uniform problem. The Subcommittee
believes that the interaction between QF0 and Office of Air Quality Planning
and Standards has been productive and could serve as a prototype for other
efforts in EPA. The existing research coramittees and the research initiative
on extrapolation modeling revealed by the briefing document provide a start
toward a program plan that is responsive to overall Agency regulatory needs.
The Subcatmittee identifies ten extrapolation processes that are
important to SPA's regulatory efforts. Each process subsumes many specific
models. The Subcommittee did not possess a detailed description of models
currently used by EPA. It concludes that the scope of ORD's current efforts
is uneven, observing specific weaknesses in research on extrapolation between
times of effect, structure-activity relationships and pathology/organ systems,
The Subcanmittee also reviewed ORD's intramural research on extrapolation
models from the perspective of scientific disciplines represented. Its major
comments includes
1} ORD's carinogenicity program is well-defined, but the various
components of the program are not of equal importance. It was not clear
how the research components were selected. Moreover, modification of the
planned research could result in significant increases in the value of the
work.
2) Extrapolation jnodeling efforts for mutagencity are clearly warranted.
The relevance and validity of the existing research for future risk assessment
efforts were not always apparent in the briefing document, and further
Clarification is desirable. It also appears that this work would be enhanced
by increased statistical analysis. The Subcommittee questions whether the
results from extrapolations based on the parallelogram model will be useful
in predicting adverse human health effects because ORD did not state how
chromosome aberrations or formation of adducts relate to human risk. The study
-------
-3-
of models that extrapolate frcra high to lew exposures is of particular interest
because it is now possible to reasure chemically the formation of adducts
between genetic material and genotoxic chemicals to provide estimates of
internal exposure at much lower doses than previously possible. QRD is partici-
pating in this scientific advance. Measurements of adduct formation in
animals {or humans) in which toxicity occurs will be of general theoretical
importance in extrapolating risks to low doses.
3) The nonionizing radiation research program could contribute in-
fatuation to several extrapolation n»dels that are specific to this source.
The goal of the effort should not, however, be just a model with a certain
number of cells but an insight into adverse physiological effects in humans
as a result of elevated temperature induced by radiofrequency heating. More
attention wight be given to addressing the significance of physiological
effects predicted in humans.
4} The Subcommittee recommends that OFD conduct more of the type of
work reported as comparative toxicology and as structure-activity
relationships among toxicants.
5) The investigators within the neurotoxieology program appear to
coordinate well, and they study the same chemicals under nearly identical
conditions. The quality of the research is uniformly high. Indeed, the
research group at EPA is widely recognized as a leading neurobehavioral
toxicology group in the country. The details of how the developing data
base will be applied to the development of extrapolation models needs to be
articulated in greater detail.
6) There is a need to develop a methodology for risk assessment in
reproductive and developmental toxicology. Some of ORD's work is markedly
out of date, whereas other aspects are abreast of contemporary developmental
biology as it relates to questions of environmental toxicity. The developmental
biology program needs an external, independent source of ongoing guidance
and review from senior scientists in the same field. The plans in the
briefing document to develop new methods for dermal absorption and reproductive
toxicity, although important toxicologically, do not seera to fit with attenpts
to advance risk assessment in these areas,
7) The research of the inhalation toxicology program addresses issues
that are critical to the development of reliable extrapolation models for
pulmonary targets. In general, the program is scientifically sound, and it
systematically attempts to provide those data needed for accurate extrapolation
modeling. The dosiraetry studies comprehensively examine important pollutants
to provide accurate comparative regional dose estimates for several species.
Sensitivity analyses with the developed models can be used to effectively
guide further experimental work. However, the species sensitivity aspect
of the program is not as well focused and appears to be addressing some
important points, while emphasizing some that may not be as critical to
extrapolation models. Some refinement is needed to determine which endpoints
are of health significance.
8} Most of the projects on systemic toxicants are in the formative
stages. The Subcommittee found the lack of integration between this program
and the programs in neurotoxicology, inhalation toxicology and developmental
biology to be particularly frustrating, since the latter subjects are
components of systemic toxicology. Certain organ systems, such as the liver
and kidney, receive no attention in QRD's plan* Work in these areas roay
-------
lead to new risk assessment guidelines. It seems implausible that the
evaluation of test data will generalize much from one organ system to
another. For example, EPA could have one guideline for the assessment of
neurotoxic substances and another for the assessment of substances that
pose developmental risks* This is another reason to establish stronger
linkages between the systemic toxicology program and other programs, and to
provide research coverage of all major organ systems.
Based on its review fron the perspective of scientific disciplines,
the Subcommittee concludes that QRD's program has many sound elements, but
that the effort is uneven and that emissions exist in some important
areas. However, the existing program does provide a good start for a more
conprehensives research effort, and stronger planning should make the
program on extrapolation modeling more effective in meeting the Agency's
regulatory needs.
The Subccmdtte also reviewed some disciplinary efforts that cross-
cut those described above, particularly pharmacokinetics. Pharmacokinetic
approaches within the extrapolation research program range fron non-
existent to quite sophisticated. For example, the reproduction, teratology
and neurotoxicology programs do not discuss pharmocokinetic parameters in the
briefing document, while the carcinogenicity and inhalation toxicology programs
emphasize dosianetry and modeling at a high level of sophistication. The
Subcommittee concludes that certain of the disciplinary programs would benefit
by the inclusion of pharmacokinetic experiments and that EPA should develop
a systematic approach to pharmaeokinetics across all programs of extrapolation
modeling.
In addition to phantiacokinetics, there are other disciplines, such as
statistics, that also integrate information from different organ systems or
extrapolation processes. They also will help to provide coherence to QRD's
extrapolation modeling research effort* Work to refine and improve extrapolation
models is inherently statistical in nature. Laboratory research ainned at this
objective requires the application of statistics for such topics as data analysis,
testing of hypotheses, modeling of dose-reponse curves, experimental design
and interpretation of statistical variation. Risk assessors can use statistical
approaches to analyze large data bases and gain insight into fundamental methods.
Although such work is difficult and often demands a multidisciplinary effort,
statistical approaches provide, for exanple, the default assumptions used by
regulatory agencies when precise data are not available. Given the emphasis
of the briefing document on human endpoints, the absence of epidemiology
research also was notable.
Relative to the available resources, the current research program is
scientifically promising. ORD has developed a number of worthwhile projects
that could iirprove Agency 'risk assessment practices and has recruited a group of
talented investigators. However, ORD lacks a strategic plan* The current plan
brings together independent projects well before the establishment of a compre-
hensive extrapolation modeling program. An overall strategy towards which
the individual scientist can aim does not exist.
-------
-5-
The Subcommittee recommends that EPA initiate work on the plan by
making an inventory of the extrapolation models actually used by the various
regulatory programs and evaluating them. This specific task has merit on
it own, not only as part of the broader planning effort. It could identify
areas in which improved extrapolation models are needed and aid in determining
the implications for research planning. Development of such a plan should
involve all parts of the Agency which make use of extrapolation models,
particularly the program offices.
A comprehensive plan for an extrapolation models research program
shoulds 1) state an overall conceptual objective or framework towards which
individual projects would aim; 2) enhance EPA's risk assessment-risk
management philosophy; 3) develop a framework that pronwtes more planning
and resource stability in support of the research? 4} provide a common
nomenclature? 5) ijtprove communication among the Agency's organizational
components; and 6) explain to the nonscientist how research on extrapolation
tnodels supports the Agency's regulatory decisions.
EPA must provide leadership within the Federal government in
existing extrapolation models. EPA shares with other regulatory agencies a
great need for better models, and EPA has some resources to perform research
and stimulate additional efforts by the Federal research organizations*
Thus, extrapolation modeling creates a unique research opportunity and
agenda for EPA.
-------
II. INTRODUCTION
The charge to the Extrapolation Models Subcommittee was to advise the
Administrator and other senior officials of EPA on the status of research on
extrapolation modeling within EPA's Office of Research and Development (OKD).
The Subcommittee focused,, in particular, on the integration of existing
research efforts and their relevance to EPA's regulatory requirements. It
also addressed the future needs of the research program. Given the importance
of extrapolation models to EPA, other Federal agencies and the scientific
ccmrnunity, the Subcommittee has broadened its charge*to include a survey
of research needs and opportunities that each of these groups, working
individually or collectively can address. In support of this review, the
Subcommittee received a briefing of two days duration and a report with two
appendices. The review is one of a series of SAB efforts intended to
independently evaluate ORD's progress in developing data and methodologies
for use in regulatory decision making.
Dr. Richard Schlesinger was unable to attend the meeting in person
but contributed cements by mail after telephone interviews with appropriate
EPA personnel. Dr. Sergio Fabro attended the meeting, reviewed all materials
related to developmental or reproductive effects and created the structure
of sections v and VI o£ this report. Unfortunately, Dr. Fabro died while the
report was in preparation. Dr. Marshall Johnson, who did not attend the
meeting, volunteered to assume responsibility for completion of the develop-
mental and reproductive effects subsections of section vi.
A. DEFINITIONS
As the Subcommittee views the subject, a "model" is an abstract, con-
ceptual description of an object or process that imitates or describes
essential features of the object or process, often in mathematical or
statistical terms. Models usually are neither well validated nor broadly
applicable, but intend to represent components or examples of a specific
phenomenon, Models often inexactly describe a complicated, poorly understood
object or process. A model can be physical, conceptual or mathematical. For
example, a rodent can be used as a physical model of a human in toxicological
assessment. The idea that a rodent is an analogue of a human can be expressed
diagramatically as a conceptual model. EPA often extrapolates quantitatively
from rodents to humans on the basis of body surface area, estimated as a
function of body weight. This overall concept can be expressed as a
mathematical model, as follows:
2/3 2/3
(rodent weight) = (human weight)
(rodent potency) (human potency)
Although the above is illustrative Of extrapolation model concepts, most
models that EPA uses are more complex than this example.
Extrapolation is the process of projecting beyond the available data
on the basis of the available data. When the model is mathematicalf equations
-------
are developed that are consistent with the phenomenon and other scientific
information, and equations are fitted to the data to develop parameters.
New solutions are then developed with inputs beyond the range of the data.
Following cannon usage in the discussion belcw, reference is made occasion-
ally to "extrapolation" to untested doses outside of the tested range,
which is similar to the other processes discussed. In reality, EPA inter-
polates between doses since an estimate is made within the range of doses,
including an undosed control (or baseline) observation,
B. AGENCY USES OF MTRAPOIATIQN MODELS
Extrapolation models are important for EPA because the Agency rarely
has fully conclusive data on the "cause" of a public health and environmental
problem that is the object of a proposed regulation. Thus, the scientific
assessments, on which regulatory and enforcement actions rest, frequently derive
from one or more extrapolation models. EPA currently uses extrapolation
models in the absence of human data for at least ten sets of scientific
activities. These include extrapolating:
1) Quantitative potency between species,
2) Effects in a "normal" population to subpopulations with different
sensitivity due to a prior disease state that may be genetically
and/or environmentally caused*
3} Qualitative pathology or organ system involvroent.
4) Low dose effects from high dose data (with inherent or empirical
control data).
5} Effects with one route of administration frcw data on another.
6) Effects from different times of exposure (and dose rates).
7) Times of effect.
8) Effects at different developmental stages.
9) Effects of untested chemical structures from data on related
chemical structures.
10) Whole animal effects from test tube or cell culture results.
After their initial development, new extrapolation models are subjected
to technical peer review and sometimes to public ccranent* They may
become widely used if they withstand this scrutiny. The development of
accepted extrapolation models can also result in significant economies
to the Agency because these models can be substituted for more data
intensive (and often more expensive) methods.
In general, the Subcommittee believes that EPA clearly needs to use
extrapolation models to discharge its responsibilities.
-------
—8—
C. THE COMPLEX NATUEE OF EXTRAPOLATION MODELING
As each model develops, work on any of the ten processes of extrapolation
described above proceeds in stages* Hie stages might be described as
fallows;
1) Physical model.
2) Conceptual model.
3) Mathematical 'model.
4) Experimentation and validation.
5) Statistical analysis.
6) Iteration of the above steps.
However, this description of the stages uses terms that are overlapping
and not independent of each other* For example, a physical nwdel need
not be used at all. Statistical analysis sometimes is the first step in
model development. At any stage, an alternative model can be posed for
investigation. For example, scientists may debate whether mice or rats
are a more appropriate physical model to understand sane health effect of
a substance in humans, AS another example, two mathematical models compete
as descriptors of the relationship of carcinogenic potency bet-ween species*
One is body surface area (described above) and the second is body weight,
which differs mathematically, as follows;
(rodent weight) * (human weight)
(rodent potency) (human potency)
EPA currently has a cooperative project with the Department of Defense
that will attenpt to choose between these two (and other) models based
on the available human and animal data. This is a very important project
for the Agency, as the choice between these two models could result in a
re-examination in the level of sane environmental standards. (This work was
not included in the extrapolation modeling program presented to the Subcom-
mittee because the project is externally funded.)
Extrapolation models will differ depending on the biological endpoint
in question. For example, neurotoxic effects and carcinogenic effects are
unlikely to develop in a parallel manner under identical conditions of
chemical exposure. Most of the various toxicologieal disciplines differ
in their techniques, methods and approaches. From the perspective of
laboratory scientists, the technology employed by different disciplines
differs so drastically that discussion of the models across these boundaries
may seem pointless. For example, the neurotoxicologist focuses on the
communication of information by a specialized tissue, using techniques
such as measurement of nerve conduction velocity. The oncologist seeks to
-------
understand the phenomena of uncontrolled cellular growth and metastasis,
perhaps by neasuririg alterations in the structure of genetic material.
Even if the the same event were examined, for example, the induction of a
neuroblastcma, the methods utilized by the two disciplines probably would
differ. The examination of the tissue by each expert could involve differing
conceptual approaches, pathological techniques and ideas of disease
progression, making it unlikely that the disciplines would reach identical
conclusions. This latter outcome poses a great challenge for regulatory
agencies seeking to synthesize scientific data and methods in a risk assess-
ment*
Another way to view the progress of an extrapolation raodel is its
stage of refinement. To some extent, refinement runs parallel to the
stages of development described above. Conceptual novels are approximately
the same as "default assumptions" in the field o£ risk assessment* EPA
uses default assunrptions when specific data on a substance or process are
not available* The default assumption is based on reasonable ideas about
how substances or processes behave in general. At an early stage, the
model may be a vague concept, which later is expressed mathematically.
The mathematical expression is more rigorous. However, as work on a
model proceeds, a larger data base accumulates that can be used to help
evaluate the model for its scientific adequacy. Care must be taken,
however, to insure that the sarae data are not used both to generate and
evaluate the model. Particularly when the model is mathematical, it can be
fitted to different data sets in order to better state the form of the
equation (or its parameters). Discrimination will be gained on hew the
parameters will change with different categories of substances. Further,
confidence in any model will grow as it withstands increasing scrutiny.
The importance of analyzing a model for consistency with the data
and accepted theories of physics, chemistry, medicine and biology has not
received sufficient attention by regulatory agencies. Often, so many
years lapse between proposal of a model and general acceptance that it
contradicts our general scientific understanding and should be discarded.
In seme casest no known or practical way exists to validate a model.
Eventually, as SPA and other organizations collect better data on a
particularly contentious process or substance, it may no longer be necessary
to use extrapolation models. The Agency may have exact data on the phenomenon
of interest and can provide a more accurate, specific assessment with the
result that the uncertainty in risk estimates will decrease in the progres-
sion from educated guesses to validated nodeIs to exact data on the process
in question. This progession contrasts with research on the models them-
selves, where the effort is to validate the equations or parameters of the
model for general use, not to replace it with a description of a single
phenomenon. Because the Subcommittee was asked to describe the progress
that ORD is making in answering the needs of the Agency to extrapolate, an
effort has been made in this report not to confuse work on the models with
data acquisition.
-------
-10-
The mathematical form, or at least the parameters, of a model will
change, depending on the chemical substance involved. For example, the
Metals Subcommittee of SAB's Environmental Health Committee has discussed
with EPA's Carcinogen Assessment Group the idea of basing the extrapolation
between species of carcinogenic potency for inhaled metallic particles
on the processes of deposition and absorption in the lung, rather than
body metabolic rate, The Metals Subconmittee and the Carcinogen Assessment
Group agree that extrapolation based on body metabolic rate can be
appropriate for gases.
Despite the complexity of model development, the scientific goal
for each model of each biological endpoint is the same, namely to reduce
uncertainty to the maximal extent possible. Ultimately, it is desirable
not to extrapolate at all, by aquiring and utilizing information on the
effect in question by direct observation of the target species of concern
and the pollutant of interest at actual environmental exposure levels,
and to have these observations supported by well-validated theories of
the mechanisms involved. Such information will seldom be available to
EPA. Its acquisition is resource intensive for both dollars and equipment
and in the use of scarce personnel with special skills and time. Hence,
this ultimate goal needs to be replaced by another, more feasible one of
having generally accepted extrapolation models with minimal uncertainty
associated with their use. These models would enable EPA and other
regulatory agencies to achieve their goals through a less resource and
data intensive approach.
To better organize the complex task inherent in developing nodels,
the Subcommittee recommends that EPA adopt the idea of a multi-dimensional
matrix. Each dimension of the matrix would indicate one of the following
ways of viewing the subject:
1} Extrapolation processes, as described above.
2) Stages of model development, as described above.
3) Biological effect or toxicological discipline.
4} Substance(s) of concern.
5} Stage of data aquisition.
6) General approach to modeling or to biological phenomena*
If any two of these dimensions were illustrated as two sides of a chart
(or computer spreadsheet), each intersection on the chart would still be
quite complex.
-------
-II-
D« BEIATIONSHIPS BETWEEN EXTRftPOIATION MODELING, PHMMAOOKINETICS,
STATISTICAL APPROACHES AND RISK ASSESSMENT
Pharmacokinetics is the study of the absorption, Metabolism, distri-
bution and elimination of foreign substances (xenobiotics) from the body.
Pharmacokinetic information can contribute to any of the ten processes
(see page 7) for which EPA uses extrapolation models. However,
pharmacc-kinetic information cannot provide a complete basis for a desired
extrapolation process because more factors are involved in each process
than absorptionf metabolism, distribution and elimination of the external
dose {or exposure). For example, in extrapolating between two species,
knowledge of the internal (biologically effective) dose of a substance
can be gained from a knowledge of the pharroacokinetics of the substance.
Indeed, even if the pharmacokinetie data are understood in only one of two
species, scientists can make an informed estimate of the internal dose in
the second species. However, if the internal dose is the same in the two
species, the biological response may differ between them.
All extrapolation models are influenced by pharmaeokinetic data. In
addition, pharmacokinetic data can contribute to an estimate of risk without
the necessary inclusion in an extrapolation model. Therefore, extrapo-
lation modeling and pharraacokinetic analysis overlap each other.
Work to refine and improve extrapolation models is inherently statis-
tical in nature. Laboratory research aimed at this objective requires
the application of statistics for such things as data analysis, testing
of hypotheses, modeling of dose-response curves, experimental design and
interpretation of statistical variation. However, "non-laboratory" research
activities, which the Subcommittee has defined as "statistical approaches,"
also play an important role in extrapolation modeling* Risk assessors
use statistical approaches to analyze large data bases, gain insight into
fundamental methods and develop default assumptions* Such work is difficult,
and it often demands a multidisciplinary effort of skilled investigators*
E. OPERATIONAL OBJECTIVES
The Subcommittee has searched for some principles by which progress in
research on models could be evaluated. Progress could be said to occur if
the research were to:
1) Provide data sufficiently direct and accurate that there is
no need to extrapolate.
2) Generate new physical, conceptual or mathematical nodeIs.
3) Help focus the acquisition of new data on the most crucial
elements'of a model and/or improve the parameters of a model.
4) Provide support information that validates, replaces or
contradicts a model used by the Agency.
-------
-12-
5) Provide critical examples of models (e.g., for a prototype
substance}.
6) Accumulate sufficient empirical data to stimulate statistical
analysis and hypothesis formation.
7) Develop statistical measures of the different classes and
iinpacts of uncertainty, given a set of assumptions for a specific
process, substance or model.
An extrapolation model can be evaluated with respect to its function
according to the following principles.
1) Is the model efficient? Does it lead to improved risk assessments
and aid in rejecting inferior ones? Can it be used by Agency
regulatory staff?
2) Is the model accurate? Does it provide answers that are
numerically close to neasurements in the field?
3) Is the model congruent with (relevant to) the decision at
hand? Is it valid?
4) Is the model flexible? Can it be modified easily to reflect a
different object or process?
5) Is the model transparent? Will a decision maker distrust or
not use a model because it is too far removed from experience?
6) Is the model accepted within the specialized technical
community?
While the two sets of principles stated above generally can assist the
the scientific community in describing the progress achieved in developing
a specific model, they do not easily facilitate the comparison or evaluation
of work conducted by ORD on different models. Instead, the Subcownittee
developed two principles to guide its discussions, evaluate material prepared
by ORD and write this report. They includei
1) If research on an extrapolation model is successful, how will
the Agency be able to improve its assessment of risks? what
can EPA do with an improved model that it cannot do without
one?
2) Will successful research on a model establish leadership for
EPA within the scientific community and promote interest in
the model outside of EPA?
F. SCOPE OF THE REVIEW
At its review meeting,.ORD staff informed the Subcommittee that the
review would not include externally funded work. Thus, extramural grant and
-------
-13-
contract projects supported by OED were not reviewed. Only work done by
ORD employees has been evaluated.
As suggested above, most of risk assessment involves the integration
of hazard and exposure estimates. The Subcommittee found that little of
the research presented at the meeting or in the ORD documents related to
exposure? most enphasized hazard, particularly human health hazard, as the
objective. Most assessments of the hazard to human health presented by
processes or substances do have embedded in them a variety of extrapolation
models. However, the Subcommittee is aware of intramural research done
within ORD on models used in expressing exposure and ecological hazards.
-------
-14-
III. ORD'S APPROACH TO EJCTgMOIAnON MODELING
A. EPA's RESEARCH COMMITTEES
ORD and the regulatory offices jointly establish priorities for research
programs through a process of extensive consultation within six Research
Cctnnrittees. Four of these Committees relate to the-major program offices
air, water, hazardous wastes and pesticides/toxic substances. The remaining
two Research Conmittees focus on multimedia energy issues (including acid
deposition) and interdisciplinary research. These Ccmmittees consist of
Agency officials from both QRD and the program offices and are co-chaired
by research and program managers from both units. Because the Research
Committees serve many clients and address all scientific disciplines that
cross-cut the QRD laboratory structure, they focus with difficulty on
single initiatives, such as extrapolation modeling.
As the Subcommittee understands it, the potential exists for each
Research Committee to give a low priority to research that would significantly
enhance the risk assessment process in all program offices because the
research is not specific to one environmental medium or program. However,
ORD Research Conmittees do provide an opportunity for program offices to
actively develop their own research objectives and introduce scientific
initiatives in the planning process. One example of this intervention, the
interaction between ORD's inhalation toxicology researchers and the Office
of Air Quality Planning and Standards (QAQPS), has led to some of the best
(and meet easily applied) research on extrapolation models covered in this
review* The Subcommittee suggests that EPA as a whole will benefit if it
builds upon the QRD-QAQPS interaction.
B. FORMULATION OF THE QRD PLAN
Improving extrapolation modeling is essential to enhancing EPA's
risk assessment efforts. ORD's research emerged as a result of projects
undertaken by individually creative and ambitious investigators. "Ehese
early efforts were encouraged by other factors, including: 1) OAQPS staff
who understood and supported the relevance of this work to their own
programmatic goals} 2} recent recommendations by the SAB in reviews of
EPA Health Assessment Documents and Criteria Documents? 3) the Agency's
development of new risk assessment guidelines; and 4) the National Academy
of Sciences report on "Risk Assessment in the Federal Government: Managing
the Process."
Research Ccranittee deliberations and interactions among laboratory
scientists and program office staff subsequently led to the identification
of a set of major issues in extrapolation research. The overall goal was
identified as the need to enhance significantly the scientific basis for
risk assessments based on health effects data. The immediate objectives
were identified as conducting the research necessary to produce models for
important extrapolation processes.
-------
The response of ORD's scientists to this set of objectives was to
identify on-going research that appeared to have relevance to extrapolation
modeling and to formulate an initiative to the extrapolation of chemical
and pharmacokinetic properties to health and ecological effects. {See
Appendices to ORD's briefing document). While this effort achieved some
early successes, the initial plan omitted some extrapolation models of
importance to EPA's direct regulatory needs and lacked sufficient detail
to permit evaluation of specific experiments. The Subccnatiittee recommends
that future ORD planning efforts emphasize the specific models currently
used in standard-setting, litigation and enforcement, and their relative
importance.
C. BUDGETARY SUPPORT FOR ORD'S EXTRAPOLATION MODELING RESEARCH
The Subcommittee requested and ORD provided a brief summary of EPA's
funding for research on extrapolation models. The Subcommittee understands
that budget estimates available at the review meeting represent approximations
and that individual investigators partition their effort between different
projects, many of which are not primarily intended to support extrapolation
modeling. This partitioning does not yield precise dollar estimates. The
efforts by individuals apparently are added together, and dollar figures
are based on the projected cumulative effort.
The "ballpark" figure for total support of extrapolation modeling/
about four million dollars for Fiscal Year (FY) 1986, is important in two
regards. First, work on extrapolation models is a small portion of ORD's
overall responsibility. Second, this level of funding is low in relation
to the magnitude of the problem facing the Agency in the area of extrapolation
modeling. Unless additional funding is provided, Agency management should
not develop unrealistic expectations of ORD.
D. THE PARALLELOGRAM APPROACH
During the course of the briefing, ORD staff from more than one discipline
made reference to a "parallelogram" concept which has a certain appeal as
a unifying principle for ORD's effort on extrapolation modeling, if used
with the rodent to human body weight example described in the introduction
(above), a parallelogram would resemble the following;
RODENT BODY HtMAN BODY
SURFACE AREA SURFACE AREA
RODENT TOXIC . HUMAN TOXIC
DOSE DOSE
-------
The Subcommittee believes that the parallelogram approach has both
advantages and disadvantages. When the Agency has a detailed understanding
of the extrapolation process, the parallelogram approach is an excellent
heuristic device, particularly to communicate the work to an audience of
lay persons. It also concisely displays the relationships under discussion.
A parallelogram also oversimplifies the information and tends to
conceal a number of complexities. In the example above, the Agency actually
measures body weight, not body surface area. A parallelogram implies that
all relationships will exhibit simple linear proportionality, which seldom
will be the case. Most importantly, the parallelogram implies that building
extrapolation models is a facile process, leading the uninitiated to believe
that relationships between any variables can be derived easily, which is
wrong* In several instances, the presentations left the Subcommittee with
the feeling that inadequate attention had been given to model formulation
and verification. Nothing intrinsic to the parallelogram approach eeummicates
when it is not applicable.
-------
-17-
IV. ORD'S PROGRAM FROM THE PERSPECTIVE OF EXTRAPOLATION PROCESSES
The Subcommittee does not intend to provide a project by project
evaluation of of ORD's entire program in this section of the report.
Instead, the Subcommittee seeks to discuss the applicability of selected
QRD research to some important issues facing the field of extrapolation
modeling which, in turn, can aid ORD in identifying some future needs of
this program. Ultimately, ORD will have to document the extrapolation
models in use for a comprehensive evaluation to take place.
*
A. EXTOAPQIATION BETWEEN SPECIES
Many of the state-of-the-art advances in extrapolation between species
originated within ORD's Office of Health and Environmental Assessment,
especially in the Carcinogen Assessment Group (C&G) which has pioneered the
use in regulatory decision making of animal data to assess carcinogenic
risks of substances to humans. Host of these accomplishments have not been
funded directly as research projects. Instead,, the research developed
because existing methods did not work during efforts to assess carcinogens
for regulatory purposes. For this reason, much of the creative work of
broader significance has been "bootlegged" from the budget to support-
risk assessments for specific substances.
ORD's Health Effects Research Laboratory is aware of the need to
support extrapolation work, and the Subcommittee concludes that ORD has a
number of worthwhile projects underway that could further improve Agency
practices. The lack of any citation in the briefing document of CAG efforts
that address extrapolation issues is noteworthy and suggests that better
communication is needed among the EPA groups working on extrapolation
models.
SPA basically relies on two approaches to extrapolate between species.
These include: 1) body surface area for carcinogenic risks based on the
incidence of tumors at different exposure levels, and 2) a safety or
uncertainty factor for non-carcinogens, based on an exposure level at which
no adverse effects are observed. Both approaches acquire an additional
degree of safety through an emphasis on the most sensitive species. During
FY'86, ORD has projects underway on the rate of heat loss in response to
radiofrequency radiation, computer simulation of radiofrequency effects,
determination of the ratio of the dose causing adult toxicity to that
causing developmental toxicity, embryo culture, interspecies extrapolation
of genotoxicity (especially for molecular dose to DNA), genetic activity
profiles, comparative toxicity, dermal absorption, auditory and visual
sensory function, male reproduction, uncertainty factors, behavioral/
cognitive studies of animal models for known human effects, studies of
animal measures of behavioral/cognitive effects correlated with human
effects, inhalation toxicology, predictions of ozone absorption and effects,
and pulmonary deposition models for particulates or gases.
Only in a few cases did ORD explain how possible research outcomes
might change existing Agency practices of extrapolation between species.
One such example is the determination of the ratio of the dose causing
-------
-18-
adult toxicity to that causing developmental toxicity. SPA currently
assumes that the ratio does not change across species, but the research in
progress might shew that the average ratio for various substances differs
from one species to another. Such information could enable Agency risk
assessors to evaluate developmental risks more accurately. Overall, GM>
research on extrapolation between species is scientifically adequate and
some EPA scientists working on this process are the leaders in their scientific
f ields.
*
B. EXTRA3EOIATION BETWEEN SUBPOPUIATIONS OF DIFFERING SENSITIVITY
Most EPA risk assessments assume a human population that exhibits a
variable sensitivity to an environmental exposure to a chemical. If the
actual population contains a subpopulation of significantly higher sensitivity,
then supralinearity will occur in dose extrapolation, and EPA's usual
assessment practices will not protect the subpopulation from the effects
of the environmental exposure. It is appropriate for the-Agency to inquire
whether such subpopulations exist or whether the subpopulation's sensitivity
is subsumed within the normal range of sensitivity of the population. If
the subpopulation is not part of the normal population, it also is appropriate
to search for seme means to identify the sensitive subpopulation.
ORD has several projects underway on sensitive subpopulations. Research
on human sensitivity to inhalation of cadmium, phosgene and ozone will help
meet some immediate regulatory needs of the Office of Air Quality Planning
and Standards and will provide detailed data for more generalized efforts
on pulmonary deposition models of gases and particulates, The Environmental
Criteria and Assessment Office in Cincinnati has a project on interindividual
variability of human response to toxic substances.
QRD's overall efforts on extrapolation to sensitive subpopulations are
not integrated and lack focus partly because no definition of sensitive
population or set of objectives apparently exists and partly because most
of the work in this area is very recent. Research planning will especially
benefit from program office input since only a portion of EPA's regulations
require a detailed consideration of sensitive subpopulations. However, OSD
does have available leadership in this area: the inhalation toxicology
program is significantly advancing the state-of-the-art, and efforts of the
Environmental Criteria and Assessment Office in Cincinnati are helping to
focus the issues from a methodological perspective.
C. EXTRAPOLATION BSTWEEN PATHOLOGICAL ENDPQINTS OR ORGAN SYSTEMS
This extrapolation process was omitted from ORD's intial plan. It is,
however, a topic of major importance. For example, EPA currently assumes
that carcinogenesis in rodent species extrapolates quantitatively to humans,
but that the extrapolation is not organ specific across species. This
assumption conflicts with most data from models of specific human disease
processes. EPA's assumption merits further investigation. As another
example, some scientists have shown that, for rodent carcinogens, quantitative
measures of acute toxicity correlate with quantitative measures of carcinogenic
potency. While this finding is controversial, it is of great iitportance
to Agency standard-setting because, if true, it would provide inexpensive
-------
support for the difficult and resource intensive process of evaluating
carcinogens, on which EPA places great emphasis,
While extrapolation across pathological endpoints or organ systems
was emitted fron QTO's plan, the Subcommittee found that seme interesting
work on this subject is underway. In the neurotoxicology program, five
projects (development of a conceptual model of neurotoxicity, development
of animal models for human behavioral effects, correlation of animal measures,
auditory and visual sensory function testing, and molecular neurobiology)
should provide data on the correlation and causal relationships between
different neurotoxic endpoints. In the developmental biology program, work
on male reproduction should show whether different measures of pathology
measure the same or different endpoints. ORD has developed a large data
base which profiles the available data on a specific chemical for many
genetic toxieity endpoints. The Subcommittee suggests that statistical
analysis of these data may determine whether different endpoints are measuring
the same thing, or not. The inhalation toxicology program has developed
physical models that show how different measures of lung toxicity are
associated with each other, and such models should enhance our understanding
of the relationship between damage to the lung and to other organs frcm
different kinds of particulate substances. Ttork in the Environmental
Criteria and Assessment Office in Cincinnati'on the similarity of target
organs and on the severity of effects directly relates to extrapolation
between pathological endpoints. Ml of these projects will produce data
that have the potential of changing and/or improving existing regulatory
practices.
D. INTERPOLATION BETWEEN DOSES
EPA has developed many of the state-of-the-art practices to interpolate
between doses. Old's Carcinogen Assessment Group has pioneered the use of
the so-called "linearized" multi-stage model for the assessment of carcinogenic
risks, Apparently, this accomplishment has not been funded as a research
project per se. Instead, the work occurred, in part, as GAG responded to
ccnroents on proposed guidelines for regulating carcinogens in water. For
this reason, much of the creative work on this model has been "bootlegged"
froro the budget to support regulatory assessments-for specific substances*
ORD's Health Effects Research Laboratory is aware of the need to support
research on interpolation between doses. There is, however, little work
underway in this laboratory to address SPA'S needs in dose interpolation
directly.
More than half of the projects reviewed by the Subcommittee have the
potential to influence Agency practices in this area. The project on the
risk of chemical raotagens at environmental levels has great potential
for the validation of some of EPA's high-to-low dose interpolation roodels
because molecular doses to the genore in humans can be related to the
direct observation of the incidence of cancers in exposed persons. ORD
should clarify whether this research may duplicate other developments
within the Agency addressing the same problem.
-------
-20-
E. BXmPQLATlON BETWEEN ROUTES OF ADMINISTRATION
Although program offices often base their regulatory actions on extrap-
olation between routes of administration, the OfcD program places little
emphasis on this process. It is supported only by work in the inhalation
toxicology program on cadmium and phosgene and by the pharmacokineties program.
F. EXTRAPOLATION ACROSS DURATIONS OF EXPOSURE
Extrapolation between different tiroes of exposure is a fairly constant
feature of EPA'S risk assessments. Usually, the Agency assumes the that
fraction of lifespan is equivalent between species. Testing of this assumption
merits more research support than currently exists. Even in extrapolating
data from the same species to different durations of exposure, the relationship
between the different durations of exposure often is not clear. EPA usually
assumes a ten-fold increase in potency from acute to subehronic exposure
and another ten-fold increase from subschronic to chronic exposure. Additional
research is needed to validate these assumptions or provide new ones.
The radiofrequency program has examined extrapolation across time of
exposure through computer simulation and experiments. This focus is also
a central feature of the pharmacokineties program. The inhalation toxicology
program has projects on pulmonary deposition of particulates and gases that
will provide valuable data. 'The rautational risk project looks at time of
exposure in relation to germ cell progression. The Environmental Criteria
and Assessments Office has a project on dose duration associations. Overall,
research on this topic is wall supported by QKD,
G. EXTRAPOLATION BSTM3EN TIMES OF EFFECT
Extrapolation of the interval between time of exposure and the onset
of effect (latency) has been a particularly difficult aspect of EPA's
carcinogenicity assessments. Often the animal data are inappropriate to
estimate latency since increased tumor prevalence in a study is difficult
to distinguish from reduced latency. Since the Agency's policy is to
extrapolate on the basis of tumor incidence without regard to correspondence
between organ sites or Rind of tumor, latency to the appearance of the
animal tuntors does not necessarily correspond to the latency of human
cancers. The Carcinogen Assessment Group has done sorae work on so-called
"time-to-turoor"'models that have been an important feature of risk assessments
for a few substances, such as ethylene dibrcmide.: The radiofrequency
program and the imtational risk project on germ cell progression also
provide some information on latency of effect, but OKD does not appear to
do rauch research in this area.
H. EXTRAPOLATION BETWEEN DEVELOt^lElSITAL STAGES
EPA typically bases risk assessments on rodent data obtained using
standard toxicological protocols. These data are not informative of whether
some particular developmental stage is more sensitive, yet the regulatory
program offices have to set standards that will protect all developmental
-------
-21™
stages and in some instances, such as adolescents employed in agriculture,'
have to set standards for a specific developmental stage. It often is
not clear hew to extrapolate in a risk assessment from toxicity data on
adults to other developmental stages. Both the neurotoxicology and
developmental biology programs have work underway that is generally
concerned with this important problem. The dermal toxicology project has
already shown that there is no consistent effect of developmental age on
dermal absorption. The Subcommittee concludes that the anphasis in this
area is appropriate to the Agency's needs for new information, in part
because two other research programs have a general emphasis on the
extrapolation problem.
I, EXTf&POLATIQW BETWEEN DIFFERENT CHEMICAL STRUCTURES
EPA has to evaluate potential effects between chemicals of similar
structure in evaluating premanufacture notices, prioritizing lists of
substances for detailed assessments, estimating the effects of certain
mixtures of closely related substances (e.g. petroleum products) and assessing
the weight-ofthe-evidence for toxic effects (e.g. carcinogenicity)* Given
this need, the Subcommittee concludes that the O1D effort on this extrapolation
process is not extensive enough,
QBD described work in progress within the neurotoxicolgy, genetic
toxicology and comparative toxicology programs, but these efforts seem
directed at providing raw data on the effects of various substances which
others could interpret, as did the short-term cancer models project. If
there was any systematic effort to contrast the testing with the generic
chemical structures for which the least data exist, ORD did not articulate
it; neither was an effort apparent to evaluate the data statistically. QRD
did not link the data gathering to the areas in which the regulatory programs
experience greatest uncertainty. Only for the dermal toxicity project was
an explicit effort to build models underway that was coupled with an effort
to improve Kiodel building through aquisition of new data,
J. EXTRAPOLATION FROM IN VITRO TEST t&TA TO WHOLE ANIMAL EFFECTS
IB yitro test systems are rapid, inexpensive and relatively free of
ethical considerations in comparison to whole animal toxicity tests. Since
EPA often has to regulate with limited data, in vitro test systems hold
great promise for carrying out the Agency's mission, OBD seems well aware
of this potential, and the Subcommittee concludes that research of funda-
mental importance is underway for the developing the process of extrapolation
fran in vitro test data to whole animal effects.
The biological markers program can relate biochemical effects (that
potentially can be observed in tissue culture) to the incidence of toxic
effects in humans after certain exposures. Hopefully, these biochemical
effects are also a part of the pathological mechanism of toxicity. The
inpact of such systems *for Agency risk assessments is profound because
causality of a biochemical event in pathogenesis will permit a direct
inference to effects of other substances on the same biochemical marker in
tissue culture. The work on molecular dosimetry (genetic risk of chemical
nwtagens at environmental levels) in the genetic toxicology program promises
to have a similar power in relating the effects of many substances on cell
culture substrates to a few cases in, which the incidence of human cancer
-------
-22-
at known exposures is linked to the levels of DMA. modification in exposed •
persons. The projects on genetic activity profiles and rotational risk to
gem cell stages will inprove the ability of EPA to relate in vitro test
outcomes to whole animal toxicities. The dermal toxicity program has
already shown that human skin does not predict the whole animal absorption
of hydrophobic chemicals, counter to the usual assumption of risk assessors*
The neurotoxicology program has a project underway to validate the predictions
jnade frcm in vitro neurotoxicity tests, and the developmental biology program
has a similar embryo culture effort underway.
The Subcommittee concludes that QRD has some state-of-the-art work
underway on the extrapolation from in vitro test data to whole animal
effects, and that some of the investigators within this broad topic are
the leaders in their scientific specialties.
-------
-23-
V. QRD'S PBOGRftM FROM TSE PERSPECTIVE OF SCIENTIFIC DISCIPLINES
In the following sections, the Subcommittee has reviewed ORD's work on
extrapolation models from the perspective of the scientific disciplines
involved. The review does not always follow along the lines of ORD's
organization. For' example, the Subcommittee preferred to examine several
projects on Genetic Toxicology together, although they are housed in different
ORD offices and were presented separately.
A. PHAMACOKINETICS
Pharmacokinetic approaches in the extrapolation research program range
from non-existent to quite sophisticated. For example, the reproduction,
teratology and neurotoxicology programs do not discuss pharmacokinetic
parameters in the briefing document, while the carcinogenic!ty and inhalation
toxicology programs emphasise dosimetry and modeling at a high level of
sophistication.
The non-ionizing radiation program does not involve chemical administration,
Thus, pharmacokinetic approaches have no role. The reproduction and teratology
programs focus on the renal, immune and cardiovascular systems £or teratology
studies and on gerontology and erKtocrinology for reproductive toxicology
studies. Although this program is involved with work on dermal absorption of
pesticides, the approach is iji vitro and does not involve pharmacokinetics.
The program would profit frcra the availability of pharmacokinetic data.
Similarly, the neurotoxicology program places considerable emphasis on
species comparisons for extrapolation research, For valid extrapolation, it
would seem important to know whether apparent species differences have metabolic
determinants. •
The genetic toxicology program presents a parallelogram method for extrapola-
tion of in vi,yg and in vitro data across species which depends on the development
of dose-effect data. Although the metabolism of cyclqphosphanu.de is mentioned
briefly, and there is a mention of dosimetry under "Future Directions," no system-
atic approach to pharmacokinetics appears to be a part of this project.
The carcinogenicity program gives considerable emphasis to pharmacokinetics
and some of the fruits of this effort could aid other research groups. One
series of experiments is directed toward determining the extent to which the
data from inhalation toxicokinetic studies can be used to make predictions
about the effects of ingested halocarbons. Experiments that attempt to vary
both the route and pattern of chemical administration are in progress to
determine whether the kinetics and toxicity of halocarbons depend on the
route and pattern of exposure. Studies on the toxicokinetics of cadmium
involve both pharmacokinetics and computer modeling. SPA, uses the data to
develop a toxicokinetic model, which is then computer simulated. The
-------
_24-
simulation is used to predict the consequences of changes in the cadmium
level in the food supply. A sophisticated pharmacokinetic approach is the major
emphasis of this group.
The inhalation toxicology program also emphasizes sophisticated phatmaco-
kinetics. As with many inhalation studies, dosimetry across species is a major
concern. The Agency's investigators have developed ozone dosimetry models to
simulate local absorption of ozone in the lower respiratory tract. Thus far,
dose delivery to the lung has been emphasized. The program has considered local
distribution and metabolism of chemicals and plans to combine pulmonary dosimettry
models with pharraacokinetic models for extrapulmonary dosimetry.
Ihe toxic ity mechanisms program attempts to quantify and predict toxic ity
through structure-activity relationships in fish. This project does not emphasize
toxicokinetic parameters. The ccttfsarative toxicology program also concentrates
on fish models, particularly extrapolation from fish to higher vertebrates. The
program compares the pharmacokinetic relationships of different species. Although
specific pharmacokinetic procedures are not presented in any detail, an excellent
opportunity exists for collaborative research between these two groups.
The program on genetic risk of chemical routagens at environmental levels
develops micro-techniques for the detection of trace nutagens. Such work neces-
sarily involves seme consideration of drug metabolism and drug distribution.
Dosimetry methodology will be used to study the sensitivity of developing germ
cells to possible mutagens. Although the briefing document presents few method-
ological details, this program appears to be heavily involved in micro-toxico-
kinetics.
The program on systemic toxicants and chemical mixtures has developed phaona-
cokinetic data in humans, emphasizing the exposure pathology and age of the
subjects. Pharmacokinetic models are being developed for extrapolation purposes,
and their utilization in risk assessment is a major direction of EPA research.
EPA should develop a systematic approach to pharmacokinetics across all
programs of extrapolation modeling. The sophisticated approaches of a group
primarily involved in phanmacckinetic modeling need not be universally applied,
but an apparent lack of conparability exists across programs. If the Agency does
support a program with a direct emphasis on pharraacokinetics, that new program
can provide support to the other programs and lead the coordination effort*
B. CARCINOGENICITY
At the time of the Subcommittee's review QRD's carcinogenicity program has
three objectives related to extrapolation. These include: 1) developing methods
using the results of short-teem tests to detect and determine the relative potency
of carcinogens? 2) examining the distribution of toxic substances and metabolites
(singly or in a mixture) in the human bc% and whether route of administration
influences distribution; and 3) estimating the concentration of cadmium in the
human body for various key organs, given estimates of exposure.
-------
-25-
The value of the proposed research varies with the objective. For the
first objective, current estimates of carcinogenic risk are largely based on
animal bioassays. These experiments, which compare the percentages of animals
demonstrating cancer at two or more exposed levels, are costly and time-
consuming, as they generally continue over most of the lifespan of the test
animal. Short-term tests, which could yield the same information in much
less time for far less cost, would be immensely useful to predict the car-
cinogenicity of previously untested chemicals and mixtures. Different
formulations of mixtures could not practically be tested using long-term
animal bioassays.
Three short-term in vivo bioassays will be studied. The class of genotoxic
chemicals to which each bioassay is sensitive will be determined by literature
review and experiment. The research will investigate the ability of the short-
term bioassays, both individually and as a group, to rank substances according to
their carcinogenic potency. Potency measures will be determined for short-term
tests singly and in combination. These rankings will be compared with rankings
from long-term bioassays.
The Subcommittee recommends that the carcinogenicity program evaluate the
short-term tests in terms of both sensitivity and specificity. This objective
requires a knowledge of both false positive and false negative outcotes, iff A
short-term test is to be useful. Hence, non-carcinogens of various chemical
classes should be tested as well. The tests should be performed in combination
with a larger set of short-term tests than the three under study. Three different
short-term in vivo bioassays are to be used to discriminate carcinogens frent
non-carcinogens and to estimate relative activity. This discrimination can be
made statistically, but it is probably best to test for discrimination with each
bioassay separately.
Similarly, long-term bioassay data fron rats and mice probably should not
be combined because these species often differ in potency for the same substance.
In seme cases, a substance is positive in one species but not the other, or has
only been tested in one species. Comparisons may have to be made among substances
within one species, rather than by combining data from different species* There
is a large data base ("TD50") of tumor incidence data in different species that
might be useful. What organs and tumors will be used to determine potency in
long term bioassays? The Subcommittee suggests that in some situations the
carcinogenicity program will inadvertantly test simultaneously for extrapolation
between organ sites or pathological endpoints* Potency differs substantially
across organs arid tumor types. For example, will the results frem the mouse
lung adenoma bioassay be expected to correlate with cancer in mice at any site?
The Subcommittee did not understand how the potency of complex mixtures will be
determined.
For the second objective, studies of the distribution of toxic materials
are useful because experimental data often exist only for one route of exposure.
It is not clear how to use the results from a feeding study to estimate risks
associated with inhalation exposure. The results from this study could help the
Agency extrapolate better between routes of administration. Given the higher
costs and greater experimental difficulty of inhalation experiments, the carcino-
genicity program might consider substituting feeding bioassays, where possible.
-------
-26-
The Subcommittee recommends that the earcinogenicity program perform selected
experiments on species other than the rat to see if the rat results extrapolate
to other species* The U.S. Air Force has developed physiological pharmacokinetic
nodels for several halocarbon solvents that are well-validated by experimental
data. The carcinogenicity program should consult these models to ensure that no
duplication of effort occurs and that the data are gathered at dose levels that
will provide the greatest amount of information about pharmacokinetic variables.
For the third objective, a risk assessment for cadmium will be more accurate
if it considers the delivered dose of a toxicant at the-site of toxicity rather
than the dose administered in an experiment. There also is a need to examine
the influence of dosing pattern (for example, continuous versus intermittent) and
variability of human response for a given level of exposure. Pharmacokinetic
models and data can help resolve these issues. The SAB Environmental Health
Ccnraittee has commented on the problem of deposition and absorption of cadmium
particles in the lung in a separate report of December 5, 1984.
The carcinogenicity program plans to formulate a physiological pharmacokinetic
nodel for cadmium in humans. Probability distributions for the model input
variables and parameters will be derived empirically. The initial application
of the system will be for cadmium ingestion. Exposure distributions will be
entered into the systems to predict the population frequency distributions of
accumulated cadmium in key organs, such as renal cortex.
ORD's carcinogenicity program isf in general, well-defined. However, specific
elements of the program are not of equal importance and it is unclear how the
elements were selected.
C. MOTM3ENICITY
OK) presented two programs in genetic toxicology, both with several
projects, that are primarily oriented to mutagenicity as an endpoint. Much
of the work in this program also will provide useful results for the carcino-
genic ity program. However, the Subcommittee agrees that mutagerucity is an
appropriate toxicological endpoint of concern for EPA.
The program uses the parallelogram method extensively. In one project,
the investigators extrapolate genotoxicity data from in vitro (rodent and
human cells) to in vivo levels of cellular organization (rodents; humans, if
data are available). This approach will be useful to regulatory programs
when human jln vivo data are not available. This method has been used under a
number of different names, as various investigators have tried to apply the
results of short-term tests to the prediction of genetic arid carcinogenic
hazards. The approach outlined in this proposal is useful and has been used
successfully in other laboratories. For example, the parallelogram method
was applied to use frequency of chromosome aberrations in blood lymphocytes
of raiee and huraans exposed to radiation to establish a usable ratio between
the slopes of the dose-response relationships between the two species*
These types of studies have enabled investigators to predict the frequency of
aberrations that would be produced in humans and the genetic risk for humans
exposed to ionizing radiation.
-------
-27-
The next major step involves using the parallelogram approach with data
for which the dose-response relationships in animals and humans have been
developed, to define the dose at the cellular and molecular level. This is
relatively staple to do in the case of ionizing radiation, where extensive
theoretical work exists on the dose-response relationships, the dose is
well-defined and the response can be readily measured. For chemicals, however,
only the exposure is known, dose-response relationships are poorly understood
and what happens at the cellular and molecular level must be investigated
carefully before the parallelogram approach will be useful. For example, the
concentration to which the intact animal, cell, or human is exposed may have
very little relationship to the actual biological dose to the target tissue,
target cell or target molecule. The investigators need to identify adducts in
the target tissue following chemical exposure. Quantification of adducts will
insure that more appropriate dose-response relationships are utilized in the
parallelogram approach. It is evident from the literature that chemical
exposure will result in many different kinds of adducts and not all of them
may be responsible for the toxic effects of interest. A very important
question that needs to be addressed is what level of adduct formation actually
is harmful. Is there a threshold level of adduct formation below which no
toxic effect will be observed?
The Subcommittee concludes that a major problem in the approach outlined is
that the investigators propose, for the most part, to measure only changes at
the level of the chremosome. Many chemicals are not potent clastogens but do
cause point mutations. In contrast, radiation is a relatively potent clastogen
but a poor inducer of point mutations. Most of the radiation induced mutations
seem to be the result of chromosome deletions and not point mutations. The
investigators should be encouraged to use other endpoints for the approach
to be complete for the other chemicals under study.
The approach used with cyclophosphamide is to use the biological response
(i.e., induction of sister chromatid exchange) as a measure of real dose and
compare the level of exposure needed to double the response in both human
and animal systems. This approach may help in understanding if the exposure
concentration has a simple relationship to the amount of biological damage
observed. However, it has been demonstrated for some chemicals that the
exposure, degree of interaction with DMA and the biological response are not
well-related, especially when dose-rate changes. The investigators seem to
understand these problems and should be encouraged to delve deeper into the
relationship between chemical dosimetry and biological effects.
While it is hot explicitly stated, it should be clear that the investigators
understand that their efforts are directed toward understanding exposure
(dose)-response relationships. A key to this work is to make sure that
dose-to-target-tissue is investigated. While the work with peripheral blood
lymphocytes is appropriate for ionizing radiation, the investigators need to
be careful in using chromosome aberrations in peripheral blood lymphocytes
following exposure to chemicals. Although these data will indicate target
tissue dose better in some situations, in the extreme case damage to peripheral
blood lymphocytes could occur that has no relationship to specific damage in
a different target tissue. The investigators need first to establish, for a
substance, what the relationship is between "dose" as measured in peripheral
blood lymphocytes and "dose" to target tissues.
-------
-28-
One of the .major criticisms the Subcommittee has with the program is
that the investigators are relating their studies to the parallelogram •
concept for human health effects, Wtile the parallelogram approach is
useful for specific lesions or endpoints (e.g., sister chrcmatid exchange or
adduct formation), the Subcommittee questions whether the results from ex-
trapolations based on the parallelogram model will be useful in predicting
adverse human health effects. ORD did not state how chromosome aberrations
or adduct formation relates to human risk. For example, once all the relation-
ships are elucidated between animal and roan in terras of genetic damage, how
will this information be used to predict the toxic potential of a compound
to people?
The parallelogram concept, as developed within this program, needs to
be tested statistically. In particular, the assumption of linearity should
be examined. ORD has scattered data for gamma radiation in vitro in humans,
and fitting a dose-response madel to such data is questionable. The investi-
gators may need to increase the sample size since the number of dicentrics
per cell seemed low. They might also utilize measures of potency that are
more robust than the estimate of the linear coefficient in the model for
gamma radiation, is the applicability of the parallelogram concept for
genetic toxicity being tested with different species of rodents? What measure
of potency is best to use? (The doubling dose?) The investigators sometimes
use the linear coefficient from the multi-stage model instead. It was not
clear what measures of potency and dose were used for studies of peripheral
blood lymphocytes in mice. Repeated experiments may be required to test the
parallelogram concept, and the investigators should determine hew much testing
is required to give estimates within a prescribed degree of accuracy.
The practice of pooling data is open to criticism since conditions are
never constant across different experiments. How would the results compare
if the studies-were used separately to estimate the model? Since extrapolation
constants varied with dose (but were similar for the corresponding sides of
the parallelogram), how will one predict human in vivo response for doses
not -tested?
Even while recognizing the difficulties involved, the Subecnmittee
recommends that the investigators address complex chemical mixtures. While
information on single chemicals or radiation is very useful for the parallelogram
approach to genetic toxicity, the Subcommittee questions whether this approach
will apply to exposure to complex mixtures. Humans are exposed to mixtures of
chemicals, each of which may have toxic potential.
The mutagenicity program has another project that will use-animal data
in which the exposure levels of nutagens are high and the incidence of mitagenic
effects can be observed in small groups o£ animals to extrapolate to the
lower smitagen doses to which humans are typically exposed. The animals will
receive a wide range of exposures. This is an important area of research
on high to low dose extrapolation which, in this program, appears to focus on
genetic risk rather than on carcinogenic risk. Much of the conceptual approach,
however, applies just as well to the problem of carcinogens that act by
chemically modifying DMA. The approach that the investigators aos using is
appropriate and should yield valuable information for use in extrapolation.
-------
-29-
The Subcommittee believes that another important aspect of this program
is research directed toward developing methods to detect very low DNA adduct
levels* Ihe investigators should be encouraged to continue these lines of
research since exposures to low concentrations of a toxicant will probably
result in very low levels of DNA modification, Ihe Subcommittee was pleased
to see that the investigators realize the utility of this approach for inves-
tigations of complex chemical mixtures.
The rttutagenicity prog-am will have to overcome some obstacles. A model
development problem exists in incorporating the low dose data into a dose-
response curve since linear extrapolation does not fit the data well, it
would appear that the embedded'problem of species-to-species extrapolation
remains. The Subcommittee does not understand the extent to which dose rate
will affect the interpretation.
This fflutageriicity program within the Office of Health Research can make
useful contributions to extrapolation modeling, but the Subccnmittee had a
difficult time understanding from the briefing document and oral presentations
where the program is going in the future or how it related to other efforts
elsewhere* Ihe Subcommittee also had difficulity trying to determine if the
approach is going to be sufficiently unique that it will add understanding
to the mechanisms of dainage from chemicals and radiation. Such understanding
would facilitate snaking the large extrapolation jumps between radiation and
chemicals, between in vitro and in vivo measurements of genetic damage and
between animal data and man.
D. NON-IONIZING RADIATION
Non-ionising radiation is discussed on four pages of the QRD briefing
document. Hie Subcommittee also reviewed recent reports by other SAB panels
concerning non-ionizing radiation and interviewed experts in the field,
including the SAB panel chairmen.
The first of these reports (January 31, 1984) is a review of a major
EPA risk assessment source document, Biological Effects of Radiofre
-------
-30-
the heat transfer model to three dimensions and the validation of the thermal
calculations using data from suitable experimental animals, are also delineated
in this document on pages 4-45 and 4-46* While the discussion in EPA's ex-
trapolation models briefing document of the motivation for this research
might be inproved, it essentially addresses research needs identified in the
EPA radiofrequency assessment document and the two SAB letter reports.
The discussion of "Future Directions" on page 7 of the briefing document
could be improved considerably by recognizing that the goal of the effort is
not just a model with a given number of cells but insight into the adverse
physiological effects in humans as a result of elevated tanperature induced
by radiofrequency heating. If the largest energy deposition occurs in the
neck and lower head areas, leading to a temperature increase of approximately
3° C at levels of radiation currently accepted as safe, what physiological
iinpacts does temperature increase iiaply? What organs or sensitive tissues could
be affected? In refining the model, it would seem appropriate to achieve a
finer resolution (by using smaller cells) for the neck, lower head, and other
areas of the body where large temperature increases may occur, and to use
larger cells elsewhere. In this fashion, it may be possible to achieve high
resolution for assessing the physiological effects of potential regulatory
significance without the extensive computational resources needed to use
small cells throughout the body.
In many areas of toxicology, risk assessors estimate human response by
using the results of the most sensitive among small laboratory animal species
that can be tested at low cost, and by scaling the dose from animal to human
using a simple mathematical formula. For non-ionizing radiation this approach
might underestimate the extent of adverse human response. More accurate
methods have been developed based on an understanding of the biological
mechanisms involved and how they differ among species. As our understanding
of biological mechanisms advances, it will be appropriate to apply this
modeling approach to other types of toxic agents as well.
Ihe non-ioniaing radiation extrapolation efforts appear to fit previously
cited needs. More attention, however, might be given to the significance of
physiological effects predicted in humans as well as the validity of these
predictions for humans*
E. COMPARATIVE ICKICOIO3Y
The Subeoratiittee recommends that OBD conduct more of the type of work
reported as comparative toxicology and as structure-activity relationships
among toxicants. It is in these areas of fundamental research where a good
potential exists for discovering answers to the applied questions posed by
extrapolation modeling.
The effort in comparative toxicology is important to the development of
the structure-activity relationship concept at EPA, It is clear that different
species do exhibit different tolerances to a given toxicant. Is it not pos-
sible, then, that some species may have evolved mechanisms for the amelioration
of the effect of a toxicant or group of toxicants? Identifying these mechanisms
among species is a logical step in building the empirical base to 1} test
the structure-activity relationships hypothesis, and 2) initially build
extrapolation models.
-------
- 31 -
Although the research generally is moving in a logical direction, it
is open to seme criticism. The briefing document implies that an understanding
of the underlying control mechanisms and more complete models will somehow
result fron the data to be collected, but OFD needs to explain the logic and
procedures by which this synthesis will be accomplished, 'The 10 by 10 matrix
testing 'regime designed to test the sensitivity between diverse taxoncraic
groups may yield disappointing results if the group relies solely on major
taxonomic groupings (genera, families, orders) as the distinction among
species. The model should depend on the properties of the organic agent and
the species tested. It is not clear that an effective target dose in one
species would predict another species because the target organ may differ by
species. Taxonomic classification is a history of origins and not necessarily
of environmental experience. Sufficient examples of evolutionary divergence
exist within families and genera to cast doubt on a scheme that uses either
families or genera as a category of species classification for the purposes
of determining sensitivity relationships. The family Cyprinidae, for example,
contains species that vary in sensitivity to a toxicant by several orders of
magnitude. The effort would be better served with a matrix that considers
classification of organisms based on environmental experience rather than
taxonomic relationships,
In summary, current efforts are reasonably well conceived, bat might
be improved by placing more emphasis on environmental experience rather than
taxonomic relationships in developing the research agenda.
Neurotoxicology is at a stage as a research field where the emphasis is
on establishing and validating methods for detecting and measuring the
consequences of chemical insults to the nervous system. ORD's neurotoxicology
program has concentrated its efforts on developing rat models for determining
the effects of potential toxins on behavior, neurochemistry and neurcpathology,
'The neurotoxicology program then attempts to validate the animal model by
ccuparing the rat data with that available from humans, with a particular
interest on behavioral measures since behavioral parameters can be measured
non-invasively in man.
The neurotoxicology program has performed an excellent job in developing
methods and procedures for measuring neurobehavioral toxicity. The develop-
ment of the neurotoxic esterase assay as a measure of delayed neurotoxicity
produced by organqphosphate insecticides should have an immediate inpact
on regulatory processes, as it should allow for replacement of the hen test
with the conventional rat model in use for most other types of regulatory
testing of agricultural pesticides.
The approach ot the neurotoxicology program has been to jump directly
from the rat to man Cor model validation. Although man does represent the
ultimate validation, the program nay be relying too heavily on the rat
model. The Subcommittee recommends that the neurotoxicology program place
a greater emphasis on cross-species comparisons (alloroetry). Another area
where the neurotoxicology program could use additional enphasis is pharmaco-
kinetics. The general needs for the development of pharmacokinetic capability
in ORD are documented elsewhere in this report. The use of phasroacokinetic
data as possible explanations for species differences in neurobehavioral
responses to chemical insult could be of great benefit to this research group.
-------
-32-
Members of the SAB Environmental Health Committee have previously
pointed out that some neurotoxic effects clearly do not occur in relation
to blood levels of the neurotoxic agent. (See, for example, comments on the
drinking water health advisory for acrylamide.) For sane substances of
environmental concern, processes in nervous tissue are of greater apparent
importance as determinants of 'human toxicity than tissue dose. Such examples
are of great importance in setting limits on the utility of pharroacokinetic
analysis for risk assessment.
The groups of investigators within the neurotoxicology program appear
to coordinate well, and they study the same chemicals under nearly identical
conditions. The quality of the research is uniformly high. Indeed, the
research group at EPA is widely recognized as being a leading neurobehavioral
toxicology group in the country.
The neurotoxicology program has been concentrating on the effects of
neurotoxins on sensory, motor and cognitive processes and the molecular
mechanisms underlying these effects. In sensory systems, the neurotoxicology
program has taken the approach of developing rapid electrcphysiological and
behavioral msthods for measuring effects of chemicals on the visual and
auditory systems. They have used the pattern-reversal-evoked potential
(PREP) as a model for studying visual acuity in the rat and the brain-stei-
auditory-evoked-response (BSMS) as a model for studying auditory thresholds
in the rat. At a mechanistic level, the effects of specific lesions in
the visual system on the PREP and the effects of cochlear lesions produced
by known ototoxicants on the BSJffiR are being studied with neuropathology
observations made in the same animals.
With respect to cognitive function, the emphasis is on behavior measures.
A microprocessor-based system for use in field studies of human cognitive
function, as well as sensory-motor function, has been developed, although not
yet used to measure cognitive function in toxicant exposed humans- Most of
the other studies involve animal models. These models include place learning,
flavor aversions and operant conditioning procedures. The program has emphasized
comparisons across these behavioral iteasurements, with comparisons of animal
responses with human responses given special attention. Future developments
in this area will focus on the development of additional learning and memory
tasks in aniraals, with special emphasis directed toward those tasks which
can be studied in animals and humans under conparable conditions.
On a mechanistic level, research is conducted relating to the neurochemical
and neurcpatholocjical basis of functional changes produced by toxic chemicals.
One such effort involves determination of the extent to which nervous system-
specific proteins can be used as biochemical markers of neurotoxicity. Animals
are being exposed to known neurotoxins, and the effects on nervous system-
specific proteins are measured by biochemical and radioinrounoassays. Preliminary
data suggest that these proteins predict the cytopathological changes associated
with toxicant exposure and that they may ultimately be sensitive and accurate
predictors of human neurotoxicity.
Another mechanistic approach concentrates on neurotoxic esterase and
its involvment with the delayed neurotoxicity produced by organophosphorcus
compounds* Ihe degree of inhibition of this esterase is highly predictive
of the symptoms of delayed neure-pathology produced by these compounds, and
OFD is suggesting the measurement of the enzyme inhibition as a replacement
-------
-33-
for the hen test currently used within the Office of Pesticides and Ibxic
Substances. The advantage of replacing the hen test includes the opportunity
to incorporate the emyroe inhibition test into conventional rodent toxicity
test protocols*
The future directions of the research program include: 1) a focus
on cross-species extrapolation of sensory test' procedures, particularly to
humans; 2) the use of the micro-processor based system for field studies in
humans; 3) the development of learning and memory tasks in animals that can
be directly compared with human tasks; and 4) further refinement of the
cellular and molecular studies on nervous system-specific proteins.
The briefing document notes that the neurotoxicology program has
concentrated on the effects of chemicals on motor behavior, in addition to
effects on sensory and cognitive processes, but the results of the experiments
on motor behavior are not discussed, nor are the directions of future research
in these areas. Soroe of the issues that may be relevant to discuss include:
1} How is motor function assessed in animal and/or human models? 2) JVre
molecular tests being developed to study the mechanisms underlying changes in
cognitive functions in animal models? 3} What species are used to make
neurotoxicity. ccnparisons? How do laboratories approach the problem of
extrapolation across species with respect to neurotoxicity? 4} Measuring the
inhibition of neurotoxic esterase inhibition and nervous system-specific
proteins are very specific mechanistic tests. Are other types of meehanisns
planned for future study, and if so, which ones? S) Do the cross species
comparisons really involve similar processes or only analagous processes?
6} Is there any attempt to study pharmacokinetic parameters of different
neurotoxins? Do species differences perhaps depend on different metabolic
pathways in different species, or differences in drug delivery? Would dose-
modeling studies be appropriate in making species ccnparisons? 7} Does an
adequate collection of baseline data exist in nontoxicant exposed humans to
validate the use of the micro-processor system in "normal humans" before
studies begin on toxicant exposed populations? 8) Is the neurotoxicology
program examining problems of acute versus chronic exposure, and reversible
versus irreversible changes? 9) If bioassays of nervous system-specific
proteins are to serve as predictors of neurotoxicity in humans, a model for
extrapolation from animals would appear to be necessary. 10) How will the
hypothesis that nervous system-specific proteins are sensitive indicators be
tested? How will prediction be accomplished? 11) How is the test for assessing
visual acuity in rats to be extrapolated to humans?
G* SYSTEMIC TOXICITY
The purpose of the systemic toxicants program is to develop the
assumptions, appropriate modifications and, when necessary, new approaches to
risk assessment for systemic (non-carcinogenic) toxicants including chemical
mixtures. The approach proposed recognizes the need to take into account
both theory and reasonable assumptions, and enphasizes the iimportanee of
understanding the mechanism of toxic action in the test system relative to
the expected outcome in humans. The staff.plans to evaluate the component
parts of the various existing extrapolation models and make revisions or
produce new methods. If new methods evolve, the staff plans to test and
evaluate the newly proposed method. Ationg the tools employed will be
literature searches, data base creation, scientific workshops and symposia.
-------
-34-
The following specific projects are proposed: 1) an assessment of the .
reliability of animal data for predicting human risk; 2} a statistical model
for species extrapolation using categorical response data; 3) estimation of
the impact of human inter-individual variability of humans on response to
toxic substances} 4) phazroacokinetie methods for unproved estimation of
effective dose? 5) development of quantitative methods to assess toxicity of
chemical mixtures? 6) development of a severity-of-the-effacts ranking
scheme', 7) development of reference values for risk assessment; and 8) study of
dose duration associations; extrapolation and interpolation procedures.
Most of these projects are in the formative stages. However, a number of
specific examples of research products are cited to indicate progress.
The Subcommittee found the lack of connections between the excellent
programs in neurotoxicology, inhalation toxicology and developmental biology
and the systemic toxicology program particularly frustrating, since the former
subjects are components of systematic toxicology. Certain organ systems,
such as the liver and kidney, receive no attention in ORD's plan. Ultimately,
work in this area may lead to the development of new risk assessment guidelines,
This is an additional reason to establish stronger linkages between the systemic
toxicology program and other laboratory programs to provide research coverage
of all major organ systems.
The Subcommittee reccmroends that the program use specific chemicals as
examples to explore the proposed techniques. At the time of the Subcommittee's
review meeting, the work was not sufficiently applied nor specific to fully
evaluate research progress or to contribute to needed risk assessments on
important 'problems.
H. INHALATION TOXICQDOGY
The objective of the inhalation toxicology program is to improve the
quantitative extrapolation of inhaled, airborne toxicants, primarily criteria
air pollutants, to pulmonary effects. This objective allows the direct use
of animal inhalation toxieity data in risk assessments by developing quantitative
cross-species interrelationships. To this end, the inhalation toxicology
program' seeks to examine two parameters that are needed to develop such
relationships, namely dpsimetry and species sensitivity, as well as to provide
judgments as to those specific health effects which merit extrapolation.
The goal of the dosiiaetry studies is to determine dose to target sites*
The approach employed is the development of mathematical models, for both
gases and particles, which incorporate parameters of lung structure and
physiology as well as the specific properties Of the toxicant of interest.
These models will be used to predict dose by region within the respiratory
tract.
Current work on gas dosimetry is aimed at predicting the local absorption
of 03 in the lower respiratory tract of experimental animals and humans;
defining the reactions of N02 following, deposition in the lungj refining
knowledge on the composition of mucus in animals and humans so as to improve
estimates of oxidant reactivity; and determining the removal of 03 in the
upper respiratory tract so as to provide more accurate input into the lower
respiratory tract model.
-------
-35-
For particle dosimetry, the program is building mathematical models
to predict the deposition of hygroscopic particles. These will be tested by-
studying the deposition of both hygroscopic and nonhygrosccpic particles in
humans. Deposition is also examined in casts of the respiratory tract of
experimental animals and humans, including children. Lung morphcfnetric
analyses are performed to refine this essential ccnpsnent of a dosimetry
model.
The stated aim of the particle dosiiaetry studies is to assess the regional
deposition of chronically inhaled particles. It is, however, not clear how
the studies outlined in this area address this issue. They appear to be
aimed solely at studying sites of particle deposition in model systems which
will provide important input into the development of empirical models.
It is important that the inhalation toxicology program make full use of
the available data base in the particle deposition area and design studies
that will complement rather than repeat those already preformed. For example,
data are needed on the deposition pattern of ultrafine particles C<0»1 um),
and this could be obtained both in cast systems and in vivo. In addition,
these deposition studies should be conducted in experimental animals to
expand the data base to allow dose extrapolation in this important ambient
particle size range.
If the data are available to perform extrapolation of delivered dose of
insoluble particles from animals to roan, a need exists for a greater modeling
effort. The program assumes a simple linear relationship but should verify
the fit to data statistically. A poor fit will suggest that further efforts
to develop an appropriate iwxlel are needed; data analyse alone will not
suffice.
The goal of the species sensitivity studies is to examine interspecies
differences in sensitivity to equivalent toxicant doses, and to quantitate
these differences. To these ends, various approaches are used, largely
employing three test materials: 03, phosgene, and cadmium. Specific studies
include: examining pulmonary macrophages after both in vitro or in vivo
exposures; in vitro exposures of respiratory tissues for comparison to _in
vivo exposures! assessment of effects of phosgene inhalation in various
species over a range of exposure concentrations? comparison of acute pulmonary
function responses to 03 in various species; determination of the concentration
response relationsip for 03 induced alterations in alveolar epithelial per-
meability? and assessment of the effects of oxidant gases upon collagen
metabolism and turnover.
The projects concerned with determining species sensitivity do not
seem to be as integrated, or as consistently relevant, as those in the dosiinetry
area. EPA staff have chosen various test endpoints, but the Subcommittee
questions the relevance of some in the total picture of the program*
-------
-36-
Another area of critical importance in extrapolation modeling which does
not appear to be addressed in the inhalation toxicology program is analysis of
interspecies clearance rates, both short-term, i.e., mucociliary clearance,
and longer-tern, alveolar clearance. EPA should conduct these analyses in a
methodologically consistent manner to allow direct extrapolation between
species.
It is not clear how the study involving in vitro exposure of human
respiratory tissues will aid in extrapolation modeling* Mthough it is
anticipated that results _in vitro will be compared with"results obtained in
vivo, the importance of this project needs to be clarified. Unlike the
area of macrophage biology where there is a large data base on _in vitro
exposures, which should be scaled to allow extrapolation of in_ vivo effects,
there are few systems using respiratory tissue in culture, and the procedure
is not amenable to widespread use.
Also unclear is the importance of the studies of phosgene sensitivity,
especially since any scaling factors for this material are to be likely
different than those for other gases, such as 03 itself. The briefing
document does clearly state how results with phosgene will help the
inhalation toxicology program in the extrapolation tnodeling of critical
ambient pollutants*
Another study is aimed at assessing collagen turnover and metabolism
in huraans and experimental animals. This is important in assessing the role
of air pollutants in producing fibrotic lung disease, and will allow develop-
went of a scale of the sensitivity of various species to this important
effect of oxidant pollutants.
The construction of an integrated dosimetric biological model for hazard
assessment is an important step in providing accurate, up-to-date and state-
of-the-art extrapolation methods for ambient toxicants. It will, hopefully,
facilitate better use of existing experimental animal toxicologic data and
•new data in the standard-setting process, ,
The inhalation toxicology program has addressed an important issue for
which there are few available data, namely deposition and morphanetry in
children's lungs. This is a critical activity since children may receive a
greater dose for an equivalent exposure level than adults. These studies
should be performed with models of lungs of other sensitive populations—for
example, persons with chronic lung disease.
The inhalation toxicology program is scientifically sound and is addres-
sing critical issues in extrapolation modeling. lh« dosimetzy studies are
systemically examining important pollutants to provide accurate interspecies
regional dose estimates. Sensitivity analyses with the models developed can
be used to guide further experimental work, effectively. However, the species
sensitivity aspect of the program is not as well focused, and appears not to
be addressing some important points, while emphasizing some that may not be
critical to extrapolation models. Some refinement Is needed on determining
those endpoints which are of health significance.
-------
-37-
I. REPRODUCTIVE MID DEVELOPMENTAL EFFECTS
The objectives of the developmental biology program are to establish
means for risk extrapolation to humans from data obtained under experimental
laboratory conditions. The program addresses a wide range of potentially
adverse effects on human reproduction and development including effects on
fertility, pregnancy outcome, and long-tern postnatal functional effects.
The program seeks to increase the sensitivity and specificity of reproductive
toxicity testing. Plans are incorporated in the program to develop more
sensitive methodologies able to detect lesions that cannot be identified
with the presently available testing techniques. There is a need to develop
a methodology for risk assessment in reproductive and developmental toxicology.
However, plans for developing rethods for dermal .absorption and reproductive
toxicity, although important toxicologically, do not seem to fit with the
plan that attempts to advance our knowledge in these two areas of risk asses-
sment.
In teratology response studies, the problem of general maternal toxicity
in the formation of birth defects is addressed. The issue is fundamentally
imjportant for identifying agents that produce developmental effects at
maternally toxic exposure levels. In any proposed methodology for quantitative
risk assessment, there must be an evaluation of the dose-effect relationship
irrespective of the mechanism(s) by which these effects occur. To assess
the role of maternal toxicity in the formation of birth defects is an important
objective, but it does not directly contribute to the development of a quanti-
tative risk assessment methodology.
In presently available teratologic testing systems, a number of problems
are recognized; 1) a variety of organ systems are not evaluated because of
technical difficulties? and 2) for several effects with high background
incidence it is difficult to assess the exact toxicologic -importance. The
developmental biology program has approached these problems by assessing
any long-term significance of "non-teratogenic fetal toxicity" and by
the postnatal evaluation of organ systems (e.g. renal, immune, and cardiac
functions) in the neonate. Beyond a few isolated studies of diverse
organ systsras, the only substantial literature of potential manifestations.
of perinatal insult is of effect on parameters somewhat related to central
nervous system function. Good, clear examples of effects produced
in functional capacities at exposure levels below those able to produce
other signs of altered in utero development in Segment II evaluation are
not available. This topic merits further investigation.
The program supplies three approaches to the problem of interspecies
risk evaluation. The first approach is to develop in vitro sytems that
possess the metabolizing functions characteristic of different species,
including the human. The program plans to observe rodent invitro embryo
develops&ent in the presence of metabolizing systems which possess different
capacities characteristic of the different species. The aim is to enable
examination of the effects of human metabolites on the developing rodent
embryo. This may be useful for a general understanding of the toxicologic
importance of sane class of chemicals to the rodent embryo developing
in vitro and for increasing our knowledge of the relationship, if any,
-------
-38-
between chemical structure and reproductive toxicity* The likelihood
that this type of approach will significantly contribute to the establishment
of a quantitative reproductive risk assessment methoaology is meager.
By testing the effect of various metabolizing systems from different
species on the development of one species, this approach addresses the
important problem of species difference in teratogenicityf but only
indirectly.
The second approach is to evaluate whether the A/D*ratio (defined as
the ratio of the dose of a chemical exerting adult toxicity to the dose that
causes developmental toxicity) is constant in different species. The A/D
ratio is an important observation for quantitative risk assessment methodology.
The usual assumption is that the A/D ratio does not vary across species.
However, A/D ratio appears rot to be constant across species, based on this
group's preliminary studies. The ratio may need to be defined statistically,
as has been done elsewhere. For example, A is the dose that is toxic to
percent x of the adults, and D is the dose that is toxic to percent y of the
embryos. Perhaps for the right choice of x and y, the ratio is constant
across species. The development of mathematical models for dose response in
animal studies is difficult because of the ccn|>lexity of the maternal-fetal
system and litter effects. This group is proceeding with the project
with outside consulting support, which is commendable.
Currently, the, no-observed-effeet-level plus margin-of-safety method is
used for species extrapolation. The value of the A/D ratio work depends on
the accuracy with which the biochemical lesions used are predictive of
teratogenes is.
The third approach to interspecies risk evaluation is to use molecular
markers of teratogenic action (such as the formation of DNA adducts by
alkylating agents, effects on microtubular function, or changes in biochemical
pathways) in order to extend the lower measurable bounds of the dose-response
curve* The developmental biology program hopes that, by quantifying biochemical
lesions which putatively precede teratogenic effects, it will be possible to
define the shape of the dose-response curve with actual data. This approach
is interesting, but it implies that the detected biochemical abnormalities
are causally related or linked in sane way to the teratogenic action. This
may not be a general rule since biochemical lesions unrelated to teratologic
action are likely to be detected. This also assumes that the wide spectrum
of teratogenic effects will have a ccrrroonality of biochemical mechanisms,
which is an unlikely proposition.
In research on reproductive toxicology, the developmental biology division
has three projects. The first is general reproductive effects extrapolation.
In an attempt to increase the ability to extrapolate between species, rats
and hamsters exposed to a selected agent are followed with morphological and
behavioral tests frcro weaning through puberty, breeding and gestation, up to
the Fl generation.
The approach to endocrine and aging effects is to inclement specific
neurcendocrine measures necessary to identify the mechanisms and/or the
sequence of events mediating the disruptive effects of toxic substances on
reproductive function in the young-adult-geriatric animal. At the same time,
-------
attempts will be made to 1) identify mechanisms reponsible for reproductive
aging, and 2) determine how age-related changes alter the organism's risk
following exposure to xenobiotics.
The effort on male reproductive function involves assessing testicular
function in animal models and humans in an attempt to determine whether
changes in the structure and function of the rodent reproductive tract
predict impaired reproductive function in humans. Ihese experiments include
morphological evaluations and an in .vitro assessment of the reproductive
functions and how exposure to various xenobiotics modify there.
The developmental biology program evaluates the difficulties inherent
in the extrapolation of animal data on skin absorption into the human situation,
and has developed a number of interesting in vitro techniques.
The Subcommittee has a mixed evaluation of the status of extrapolation
modeling for reproductive and developmental effects. Some of the work
is out of date, whereas other aspects are highly germaine and abreast of the
contemporary developments in developmental biology as it relates to questions
of toxicity. 'Jhe description in the briefing document consists of a series
of questions that apply toxologic questions to on-going research interests.
Ibis emphasis is unfortunate and should be changed to address more relevant
questions and techniques needed to answer the more important questions. The
developmental biology group needs an external, independent source of on-going
guidance and review from senior scientists in the same field. The individual
scientists involved in the developmental biology group tend to be of high
caliber and motivation. The program merits this attention, so that its
projects will become less diffuse and not distracted from, developing
extrapolation models. The group has adequate resources and, if directed
rather than diffused, could have a significant innpact.
-------
-40-
VI, THE OVERALL FEDERAL RESEARCH EFFORT ON BmUTOLATlON MODELING
Without extrapolation, testing of chanieals in laboratory aniuials is
pointless. Given the critical importance of extrapolation, and the millions
of dollars spent by the Federal research and regulatory agencies on tojdeity
testing, it should be expected that major efforts are underway to develop and
examine extrapolation methods. EPA does not have a program that focuses directly
on extrapolation method development and evaluation. In its place are exciting
research projects on toxicology tests and other efforts directed at extrapolation*
This conclusion was also stated in a recent review of U.S. research directed
at examining and improving risk assessment for carcinogens. The review^, prepared
by the Environ Corporation for the ILSI-Risk Science Institute, concluded that less
than 5 to 10 percent of the research budgets of institutions involved in risk
assessment is directed at inproving methods, including extrapolation. The bulk
of the latter research is supported by EPA.
Little overlap exists in the material surveyed by the ILSI-Risk Science
Institute and the present Subcommittee report. The former studied extramurally
funded research on extrapolation o£ carcinogenic effects by key institutions
throughout the entire U.S. , whereas the latter reviewed extrapolation of all
health effects only in intratsurally funded work in EPA'S ORD. However, both
groups' findings have sane remarkable similarities. Both reviews conclude that
extrapolation efforts are insufficiently funded and uneven with respect to the
particular scientific issues addressed. The ILSI-Risk Science Institute study
also highlights the importance .of coordinating research efforts among Federal
agencies.
The National Academy of Science Committee on Institutional Means for
Assessment of Risks to the Public Health listed fifty-nine "components" of risk
assessment that might be improved. Of these, the ILSI-Risk science Institute
survey identified twelve studies that examine the relationship between
administered dose and target tissue dose, and seven that seek to identify
biological markers of human exposure. Those are important extrapolation
processes, and Section IV of the Subcommittee's review has discussed more extrap-
olation processes that are also important* The ILSI-Risk Science Institute found
that no research was funded for twenty-seven components, and only one study was
underway for the remainder.
A. NON-ORD EXTRAPOLATION PROGRAMS IN SPA
Ihile research at EPA is focused in GRD, extrapolation modeling also occurs
in many of the regulatory offices. These include: 1) the Hazard Evaluation
Division within the Office of Pesticide Programs? 2) the Health and Environmental
Review Division within the Office of Toxic Substances » and 3) the Criteria and
Standards Division within the Office of Drinking Water. Each group uses such
models frequently to carry out their respective missions. That ORD requested a
review of its own effort is laudable, but the omission of the non-QRD scientific
assessment activities from the current plan limits the usefulness of the plan.
Beyond the extrarourally funded work that is managed by ORD, the Office of Toxic
Substances has housed the "Gene-Tox" program, which collates world-wide data on
J-V. EQDRICKS and C. ST* HIL&IRE, Review of Current Research Activities
to Improve Risk Assessment and Identification of Major Gaps* Prepared for
the ILSI-Risk Science Institute by Environ Corporation, November 6, 1985.
-------
-41-
bioassay methods for many genetic endpoints and has work underway on structure-
activity relationships that has the potential to iaprove the Agency's practice of
extrapolation between similar chemical structures. The Office of Solid Waste and
Emeigeney Response has a support contract with the Centers for Disease Control.
Other examples exist that would further demonstrate the necessity to have an
Agency-wide plan.
B. OTHER FEDERAL REGULATORY AGENCIES
the need for extrapolation rnodels is felt most strongly in regulatory
agencies, Therefore, they should provide the core leadership, direction and
support Cor the B'ederal effort. At present, EPA appears to carry most of the
responsibility, although the Food and Drug Administration has the support of the
Center for Food Safety and Applied Nutrition and the National Center for Toxico-
logical Research, To avoid unnecessary duplication and to utilize scare resources
optimally, EPA needs to coordinate its research planning with the other Federal
regulatory agencies.
C. OTHER FEDERAL RESEARCH AGENCIES
The Department of Energy and the Department of Health and Human Services
support research on extrapolation modeling. Within the Department of Health and
Human Services a number of organizations are involved, including the Centers for
Disease Control {the Center for Environmental Health, the Agency for Tbxic
Substances and Disease Registry) and the National institute for Occupational
Safety and Health. The National Institutes of Health has several organizations
involved, especially the National Cancer Institute and the National Institute of
Environmental Health Sciences, EPA's research plans should explicitly take into
consideration the contributions of these agencies.
-------
- 42 -
JffiHEMHX I
Sf&TUS OF EXTRAPOLATION
RESEARCH HEEDED TO SHWJQIATE
FSCM ANIMAL CfttA TO HDMM RISK,
FBOM HIGH TO LOW DOSE3, -
FPCM M3IIE TO G3SONIC 2PFSCTS
VOLUME I
BRIEFING DOCUMENT
Prepared By
Office of Research and Developrnent Staff
' TO assist the Science advisory Beard panel
convened to ptovicla the Administrator of EFA
with a prcgrarmatic review of extrapolation
related research conducted in CRD.
FOR REVI3W PURPOSES ONLI
DO $QT CITE OS SEP8CDUCS
SEPTEMBER 1935
-------
43 -
EXECUTIVE SUMMARY
This briefing document was prepared by Office of Besearch and Devel-
opment (QRO) staff to assist the EPA Science Advisory Beard (SAB) in their
review of research in progress dealing with extrapolation of nonhuxsan labora-
tory data to roan* Various components of OBD research related to extrapola-
tion are discussed. Hie findings and reconraendations of the SAB will be
transmitted to the EPA, Administrator and Assistant Administrator.
The overall extrapolation program considers the needs of the various
program offices and research comittees. This document provides an overview
of the extrapolation research projects in the Office of Health Research
(OHR)r Office of Environmental Processes and Effects Research (DEFER), and
Office of Health and Environmental Assessment (OHEA). Here, we describe
those elements of the program that relate to major issues in extrapolation
research:
1. Extrapolation from in yitro techiques to whole animals.
2. Extrapolation of laboratory animal data to humans.
3. Extrapolation of results fron high dose exposure to low dose
(ambient) exposure.
4. Extrapolation of results £mn acute or subchronic exposure to
continuous exposure/chronic effects.
The overall goal of the extrapolation research program is to provide a
significant enhancement of the scientific basis for risk assessments based on
health affects data. Extrapolation of effects using data from ecosystems
species is also included in the research program* With unproved extrapola-
tion methods, major uncertainties in the health data bases- can be. better
resolved, leading to more precise risk assessments, thereby improving risk
management judgments. To these ends, ORD has developed a research plan
consistent with program office needs, research connittee priorities, avail^
able expertise and resources, and state-of-the-art science. In further
support of this program the Assistant Administrator for ORO has recotmended
an increase of 1.3 million dollars and 4,9 positions in the budget request
for FY 87. this represents partial funding of a large research initiative on
advanced methods for extrapolation; a full description of the initiative may
be found in Volume II, Appendix 1-8. the increase is designed to strengthen
ongoing efforts and to focus on areas which are in the forefront of scien-
tific knowledge.
The research effort has built upon the recanraendations of the National
Academy of Sciences and needs defined in EPA criteria documentsr proposed
risk assessment guidelines, and such reports as the NSC "Risk Assessment
in the Federal Government: Managing the Process," and EPA's "Risk Assess-
ment and Management: Framework for Decision Making," as well as frcm various
program reviews. Interactions among laboratory scientists, program offices,
and research carol e tees also have been important to the development of the
overall program. The program is designed to reduce uncertainties and iisprove
the accuracy and precision of risk assessments when sufficient human clinical
and epidemiolcgical data are not available.
-------
- 44 -
Efeurotoxieology - Techniques for extrapolation between sensory,
motor, and cognitive effects of high to low dose, acute to sub-
chronic r and cross species extrapolation.
Geneticjltoxioology - Mutagenicity test battery for human hazard
estimation; molecular dosiroetry for comparative mutagenesis, ear-
cincgenesis, and risk assessment.
Carcinogenicity (Mammalian) - Statistical -oethcds for estimating
carinogenic potency of organicsj utility of route to route extrapo-
lation in risk assessment; predicted probability distributions of
kidney cortex and urine cadmium levels; mathematical simulations
of pharmacokinetics of drinking water contaminants.
Inhalation Toxicology - Combining dosimetry and species sensitivity
data for quantitative extrapolation of animal toxicological results
to man. Efforts include developing theoretical models for gaseous
and particulate deposition in man and animals, model validation and
mechanistic studies, experimental dosinetry studies, comparisons of
species sensitivity to oxidants, studies providing improved input
data for interspecies comparisons .of delivered dose, etc.
DEFER: Toxieity Mechanisms * This effort attenpts to predict tcadcity of a
chemical to fish on the basis of molecular descriptors and chemi-
cal properties. To do so a sequence of measureable histologic,
biochemical, physiological, pharmacokinetic, and behavioral respon-
ses are measured to define the acute rode of toxic actions,
Comparative; Toxicology - The objective of the program is to provide
the necessary toxicological data to extrapolate dose responses
between invertebrates and lower vertebrates and between lower and
higher vertebrates (including man).
CUBA; Genetic Risk of Chemical Mutagens - This research program is de-
signed to provide a scientific basis for risk estimates calculated
by using- extrapolations from the relatively high inutagen doses used
in animal mutation studies to the lower nutagen doses associated
with human exposures,
Systanic Toxicants and Chemical Mixtures - This program validates
risk assessment assumptions, develops appropriate theoretical mod-
ifications, and when necessary, develops new approaches to risk
assessments for systemic (non-carcinogenic) toxicants and for mix-
tures of various chemicals presenting either carcinogenic or ncn-
carcinqgenic risks.
-------
- 45 -
INTRODUCTION
A major goal of QRD is to improve the scientific basis for* extrapolation
which will enhance the precision of risk assessments. Valid extrapolation
methods are essential if EPA is to optimally utilize a highly diverse and
conplex health data base in making risk assessment decisions. The existing
health data bases for most chemicals have cannon problens: in vitro or-animal
data often strongly suggest potential human hazards, but it is hunans who
must be protected, most often by regulation of human exposure levels? health
data often exist for high doses, but toxicity resulting from ambient exposure
is not sufficiently quantified7 and many experiments indicate a hazard fron
acute exposure, but humans may be exposed chronically and experience a dif-
ferent degree or type of effect. Making such extrapolations, as outlined
above, especially in a quantitative or semi -quantitative manner, is exceed-
ingly conplex and not yet precise.
Multifaceted research approaches must be applied to account for the
inherent complexities of the issues. Each key issue mist be addressed at a
pace consistent with the state-of-the-art of a given issue. For example,
animal to man dosiaetric extrapolation of inhaled chemicals needs to be
approached initially for simple cases of chemicals not undergoing biotrans—
formation to gain basic understanding needed to solve more difficult profalans
for many inhaled organic chemicals. There are other extrapolation issues
needing more elementary approaches, such as cases for which animal models of
developmental toxicity and neurotoxicity need to be refined and mechanisms
understood in relation to human mechanisms before such models can be applied
to collect data for ultimate extrapolations.
Thus, areas of erophasis for ORD's extrapolation program are:
1. Improving the scientific basis for extrapolation.
2. Decreasing uncertainties in , risk assessments by" improving the
precision of extrapolations.
3. Responsiveness to the extrapolation needs of the program offices.
These considerations, issues, and goals were incorporated into the devel-
opment of the extrapolation research projects to be described. 'The scope of
the overall program is broad, given the expertise, resources, and specific
missions of the research groups involved. The chapters of this document are
organized by research groups to facilitate the presentation and are as fol-
lows;
Non-ionizing radiation - Scaling physiologic effects of radio
frequency radiation exposure and mathematical modeling of thetao-
regulatory systens.
Reproduction and teratology - Mult vs developing embryo minimal
dose extrapolation; maternal toxicity in teratogertesis, the role of
metabolic regulation during differentiation; reproductive toxico—
logical testing co improve extrapolation of effects.
-------
_ 46 -
TABLE OP CONTENTS
Page No.
introduction .......... 1
Extrapolation Related Research in the
Office of Health Research (CUR) ..... 3
Non-ionizing Radiation ,,.,.»,,.,..».... 4
Reproduction and Teratology 8
Neurotoxicology ............... 16
Genetic Toxicology ,,.»....» . . 21
Carcinogenic! ty (Mansnalian) . . , **..*. 27
inhalation Toxicology ................. 34
Extrapolation Related Research in the
Office of Environmental Processes
and Effects Research (OEPER) .........*..... 44
Ibxicity Mechanisms .......... .45
Conparative Toxicology ........... .48
Extrapolation Belated Research in the
Office of Health and Environmental Assessment ....... 51
Genetic Risk of Chemical Mutagens
at Environmental Levels .......... 52
Systemic Toxicants and Chemical Mixtures ........ 54
------- |