United States Risk EPA/625/3-89/015
Environmental Protection Assessment March 1989
Agency Forum
k*EPA Workshop Report on
EPA Guidelines for
Carcinogen Risk
Assessment
-------
EPA/625/3-89/015
March 1989
Workshop Report on EPA Guidelines for
Carcinogen Risk Assessment
Assembled by:
Eastern Research Group, Inc.
6 Whittemore Street
Arlington, MA 02174
EPA Contract
for the
Risk Assessment Forum
Technical Panel on Carcinogen Guidelines
U.S. Environmental Protection Agency
Washington, DC 20460
uooa 1G7Q
-------
NOTICE
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
This document was assembled by Eastern Research Group, Inc., Arlington,
Massachusetts, for the EPA Risk Assessment Forum. Sections from individual
contributors were reproduced as received without retyping in the interest of
time. Relevant portions were reviewed by each workshop chairperson and
speaker. Their time and contributions are gratefully acknowledged.
-ii-
-------
CONTENTS
PAGE
INTRODUCTION 1
MEETING AGENDA 5
COLLECTED WORKSHOP MATERIALS
Qualitative Evidence Issues 9
o Relevance of Tumors in Animals to Human
Carci nogeni ci ty 11
o Weight-of-Evidence Classification Scheme 21
Quantitative Evidence Issues , 27
o Dose Scaling Across Species 29
o Incorporation of Mechanistic Data into
Quantitative Risk Assessment 37
o Additivity/Independence of Mechanism to
Background Processes 45
Summary and Concluding Remarks 59
APPENDICES
Appendix A EPA Risk Assessment forum Technical Panel
and Associates A-l
Appendix B List of Participants B-l
Appendix C List of Observers C-l
Appendix D Introductory Plenary Session D-l
iii
-------
WORKSHOP ON CARCINOGEN RISK ASSESSMENT
January 11-13, 1989
Virginia Beach, Virginia
.*
INTRODUCTION
On September 24, 1986, the U.S. Environmental Protection Agency (EPA)
issued guidelines for assessing human risk from exposure to environmental
carcinogens (51 Federal Register 33992-34003). The guidelines set forth
principles and procedures to guide EPA scientists in the conduct of Agency
risk assessments, to promote high scientific quality and Agency-wide
consistency, and to inform Agency decision-makers and the public about these
scientific procedures. In publishing this guidance, EPA emphasized that one
purpose of the guidelines was to "encourage research and analysis that will
lead to new risk assessment methods and data," which in turn would be used to
revise and improve the guidelines. Thus, the guidelines were developed and
published with the understanding that risk assessment is an evolving
scientific undertaking and that continued study would lead to changes.
As expected, new information and thinking in several areas of carcinogen
risk assessment, as well as accumulated experience in using the guidelines,
has led to an EPA review to assess the need for revisions in the guidelines,
particularly in the areas of classification of carcinogens and the decision
logic for when and how to apply quantitative risk estimation methods. On
August 26, 1988, EPA asked the public to provide information to assist this
review (53 Federal Register 52656-52658) and on December 12, 1988, the Agency
announced -that a workshop for analysis and review of these issues would be
held in Virginia Beach, Virginia on January 11-13, 1989 (53 Federal Register
49919-20).
The Workshop was part of a three-stage process for reviewing and, as
appropriate, revising EPA's cancer risk assessment guidelines. The first
-------
stage began with several information gathering activities to identify and
define scientific issues relating to the guidelines. For example, EPA
scientists and program offices were invited to comment on their experiences
with the cancer guidelines. Also, the August 1988 Federal Register notice
asked for public information on use of these guidelines. Other information
was obtained in meetings with individual scientists who regularly use these
guidelines.
The Virginia Beach Workshop completed this information-gathering stage of
EFA's preliminary review of the guidelines. For the Workshop, EPA assembled
experts in various aspects of carcinogen risk assessment to study and comment
on the scientific foundation for two general aspects of the guidelines. As
outlined in the Workshop agenda, workgroups studied "qualitative" issues
bearing on the classification of chemicals as potential carcinogens and
"quantitative" questions on extrapolating data from test animals to human
populations.
EPA emphasized that the agenda had been deliberately limited in two
important ways. First, although the Agency recognized that many issues were
ripe for discussion, the Virginia Beach Workshop was limited to the specific
subject matter areas outlined in the agenda. EPA plans to study other issues
in later stages of its review. Second, EPA stressed that the Agency was not
expecting consensus on, or resolution of, all issues. Rather, the Workshop
was an information-gathering exercise --a scientific forum for objective
discussion and analysis among the invited panelists. Background issues papers
were developed to help focus discussion on technical questions, and the
Chairman of each workgroup session prepared a brief summary of each discussion
for presentation at the closing plenary session (Section 3 of this Report).
-------
In Che second stage of Che three-scage guidelines review, EPA will analyze
Che information described above Co make decisions about changing Che
guidelines, Co determine Che nature of any such changes and, if appropriaCe,
Co develop a formal proposal for peer review and public comment. EPA's
.»
analysis of Che information collected so far suggests several possible
ouccomes, ranging from no changes at this time Co substantial changes for
cercain aspecCs of Che guidelines.
In Che Chird stage of chis Agency review, any proposed changes would be
submitted Co scientific experts for preliminary peer review, and then Co Che
general public, other federal agencies, and EPA's Science Advisory Board for
commenC. All of chese comments would be evaluated in developing final
guidance.
-------
WORKSHOP Oil CARCIHOCBM RISK ASSESSMENT
Virginia B*ach, Virginia
January 11-13, 1999
TUESDAY, JAMDARY 10. 1989
7:30PM - 9:30PM
Early Registration/Check-in
WEDNESDAY, JANUARY 11. 1919
7: 30AM - 8:30AM
8:30AM - 11:30AM
8:30AM
8:4 1
10:00AM
10:10AM
10:30AM
11:15AM
Registration/check-in
Continental Breakfast served in the
Horizon Lounge
PLUMY
Welcome and Introduction
Perspective on EPA'a Carcinogen Guidelines
- Environmental Protection Agency
- Public Interest
- Industry
- European
•
Logistic Announcements
BREAK (20 minutes)
Participants Discussion
Workshop Overview; Directions
to the Work Groups
Dorothy Patton
John Moore
Prederica Perera
James Wilson
Kees Van der Heijden
Kate Schalk
Moore, Perera,
Wilson, Van der Heijden
William Parland
11:30AM - 1:OOPM
LUNCH
-------
WEDNESDAY, JANDARY 11, 1989 (continued)
WORKSHOP SESSIONS
IrOOPM - 5:40PM
3:00PM
5:00PM
5:20PM
6:OOPM - 7:30PM
Qualitative Issues - Virginia Room
Two Sessions: Animal Tumors,
Height-of-Evidence
Participants
Quantitative Issues - Chesapeake Boon
Three Sessions: Scaling, Mechanisms,
Additivity/Independence
Participants
Roy Albert
Margaret Chu
David Clayson
Marilyn Fingerhut
Gary Flaw
John Grahaa
Richard Hill
Kin Hopper
Eugene McConnell
Colin Park
Janes Popp
Peter Preuss
Ellen Silbergeld
Thomas Slaga
Robert Squire
Raymond Tennant
Kees Van der Heijden
Kelvin Andersen
Carl Barrett
Linda Birnbaum
Murray Conn
Robert Dedrick
William Farland
Michael Gallo
David Gaylor
James Gillette
Daniel Krewski
Arnold Kuzmack
Frederica Perera
Christopher For tier
Richard Reitz
Lorenz Rhomberg
Stephen Safe
Robert Scheuplein
Thomas Starr
James Swenberg
Curtis Travis
James Wilson
GENERAL: John Moore, John Ashby
Animal Tumor Work Group*
Chair: Eugene McConnell
*
BREAK (20 minutes)
*This group continues on Thursday.
RECEPTION
Cash Bar and Hors d'oeuvres
in the Horizon Lounge
Scaling Work Group
Chair: Melvin Andersen
BREAK (20 minutes)
Observer Questions/Comments
Chair's Summary
-------
THURSDAY, JANUARY 12, 1989
8:00AM - 12:30PM Animal Tumor Work Group (continued)
9:20AM -------- Observer Questions/Comments
9:40AM --------- Chair's Summary
10;00AM - BREAK (20 minutes)
10:20AM Wetqht-of-Bvtdence Work Group
Chair: Gary Plaram
12:00 NOON
12:20PM
12:30PM - 1:45PM
1:45PM - 6:00PM
3:30PM
5:20PM
5:40PM
Mechanisms Work Group
Chair: Michael Gallo
BREAK (20 minutes)
LUNCH
Wetght-of-Evidence Work Group (continued)
BREAK (20 minutes)
Observer Questions/Comments
Chair's Summary
Observer Questions/Comments
Chair's Summary
LUNCH
Additivlty/lndependence Work Group
Chair: Daniel Krewski
BREAK (20 minutes)
Observer Questions/Comments
Chair's Summary
-------
FRIDAY, JANUARY 13, 1989
8:30AM - 12:00 NOON
8:30AM
8:40AM
10:15AM
10:30AM
11:30AM
12:00 NOON
PLENARY
Introduction
Work Group Reports
- Animal Tumor
- Weight-of-Evidence
- Scaling
- Mechanisms
- Additivity/Independence
BREAK (IS minutes)
Participants' Discussion
WRAP-UP
ADJOURNMENT
Richard Hill
Eugene McConnell
Gary Plamm
Melvin Andersen
Michael Gallo
Daniel Krewski
McConnell, Plamm
Andersen, Gallo, Krewski
John Ashby
Dorothy Patton
-------
QTLZTATZVB BVDMCB ISUS
R«l«vanc« of Tumors in Animal* to Human Carcinogsnieity
Pr«-m««ting Zssua Papar
work Oroup summary
W«ight-of-Bvid«nc« Classification
Pr«-m««ting zasua Papsr
Work Qroup summary
-------
PRE-MEETING ISSUE PAPER
HQRK5H3P TOPIC
BELEVBNOB OF TCKKS IN ANIMMS TO tOftN CARCTNOGSaCTIY
GDAIS;
1. To further develop the list of factors (Attachment 1) that should be
considered in making evaluations about the significance of tupior findings
in animals to human carcinogenicity.
2. To analyze in detail sane of the identified factors as to whether various
outcomes tend to strengthen, weaken, or void the presumption of human
carcinogenicity based on animal tumor data.
Tumors in chemically fo>Mtgd animals generally signal that a chemical may
be carcinogenic to humans. However, a careful review of all relevant
information can strengthen or weaken a final judgment about carcinogenicity in
In the late 1970s both IARC (1977) and NCI's National Cancer Advisory
Board (NCAB, 1977) stated that carcinogenic responses in animals signaled
potential effects in exposed humans. later, an OSTP (1985) cancer principle
gave deference to the IAPC language, "that in the absence of adequate data in
humans it is reasonable, for practical purposes, to regard chemicals for which
there is sufficient evidence of carcinogenicity in animals as if they
presented a carcinogenic risk to humans." for the use of animal tumor data to
predict human carcinogenic risks. Even given this, Sir Richard Doll stated in
a symposium organized through IARC (1985; p. 5-6) that
. . . chemicals with 'sufficient' evidence for carcinogenicity [in
animals] have generally been accepted as posing a potential hazard
to man, even in the absence of detailed knowledge of the mechanisms
by which they exert their effect. . . . More recently, however, it
has come to be realized that carcinogens with* sufficient* evidence
for carcinogenicity [in animals] may act by different mechanisms and
that even for these substance reasons may be found that make
extrapolation from one species to another inappropriate. It seems,
therefore, that experimental toxicological data on carcinogenesis
can forewarn us of potential hazard to man, but that cur final
decision must rest on a full assessment of our total knowledge. . .
To this end, the OSTP report stated that the presumption should be reviewed in
light of other information in reaching a final position on human
carcinogenicity. The EPA guidelines (EPA, 1986) also expressed this position
in its weight-of-evidence approach.
Therefore, the question is how to determine the degree of human relevance
of animal tumor observations and how to determine what are the factors and
analyses chat can be used in answering the question.
11
-------
The yieuicnt workgroup is to perform tuo tasks. The first is to identify
those generic factors (attributes) that one sifts through in evaluating the
relevance of tumors in animals to human carcinogenicity. A short list of
potential factors (Attachment 1) is attached for workgroup consideration and
modification. The second task is to analyze some of the identified factors it
more detail as they, alone or in combination, apply to the determination of
human carcinogenicity. The workgroup discussions should not be distracted by
debates on issues such as "What is a 'promoter'?" but rather should focus on
the use of information on promotion characteristics in making a judgment of
human carcinogenicity.
Two sample data configurations are included to help identify factors for
Attachment 1 and to begin the discussion on the relevance of the various
factors. Data configurations in one case (Attachment 2) suggest increased
confidence and in the other case (Attachment 3) suggest decreased confidence
of the relevance of animal findings to humans. Some of the attributes are
components of a given or of several long-term animal studies; others are
mechanistic, pharmacokinetic, or other biological information on the
substance; still others reflect structure-activity relationships (SAR). Some
of these factors may stand alone in dUininishing/increasing the significance of
animal tiimnr data to human cancer hazard but many need to be considered with
additional factors to determine the impact on the question of human relevance.
12
-------
XFDICBMEMT X
SAMPIE
a. Route of administration to animals significantly different/similar
to human exposure*
b. High/low background incidence of tumors at specific organ sites.
c. Presence/absence and extent of cellular toxicity where tumors occur.
d. Presence of evidence for or against the progression of preneoplastic
and early neoplastic lesions to malignancy.
e. Degree of consistency of timrnr outcome in repeat bioassays.
f. Number and sex of animal species affected.
g. Dose-response characteristics.
h. Organ or tissue sites of tumors and existence of analogs in humans.
a. Differences/similarities in tcxification/detoxification pathways
(comparative metabolism).
b. Differences/similarities in absorption, distribution, excretion, as
well as rates of metabolism (qualitative and quantitative
pharmacokinetics).
c. Ability of parent compound or metabolites to bind covently with
cellular macromolecules.
d. Outcome in genotoxicity tests for a range of end points.
e. Presence/lack of promotion activity.
f. Findings on the influence of the chemical or metabolites on
physiological adaptive mechanisms (e.g., presence or lack of
hormcxM* disturbance, oxidative stress, alutathione denlefciaro
3. SMI
~~***y^* m • • %MBMg**»^v9 *u^^«JUAtJbaBU0 ^9««^» f ŁALC0«IUG W.L J.CU.JV Wl.
ie disturbance, oxidative stress, glutathione depletion)
a. Nature of evidence of carcinogenicity of compounds chemically
related to parent chemical/metabolites.
b. Electrophilicity/biological alkylating ability of
analogs/metabolites.
c. Knowledge about the toxicological effects of analogs/metabolites.
d. Mechanistic insights on analogs/metabolites.
13
-------
JQEl Of
is a
data tfrat strongly
tested ff
1.
2.
3.
4.
A clear dose-response effect.
The induction of tumors is seen at
subchronic, or chronic toxicity.
below observed acute,
Tumors are induced in multiple animal species and/or at multiple sites
a sneei<
Preneoplastic lesions and benign tumors are induced early and rapidly
progress to malignancy in a dose-related fashion.
1. Tumors occur at doses with no evidence for disruption of honeostasis.
2. The chemical and/or its metabolites are shown to induce HA damage/repa:
and form covalent adducts with «*»linT^i* macrcnolecules.
3. The chemical and/or its metabolites are found to induce gene mutations
and/or structural chromosome aberrations.
4. HiarinacoJdnetic data help explain v*y certain organ/tissue sites develoj
and others did not.
III. SftR
1. The parent chemical is expected to be metabolized the same way in humans
as in the test species.
2. The parent chemical and/or metabolites belong to a class of human
-------
ATTACHMENT 3
Configuration of eoqperinental data that suggests the aniaal findings nay not
be relevant to toman carcinogenicity.
1. Induced tumors in - animals found only at doses where substantial tissue
and cellular injury occurs.
i) Such injury is observable soon after administration of the chemical
ii) Nature of the injury is known or found to lead to cellular
proliferation and hyperplasia.
2. Induced tumors show little progression to malignancy.
3. Induced tumors arise from tissue for which there is no human analog.
4. Tumor tissue/cell type occurs with high and variable incidence in the
test species while it is very rarely seen in humans.
TT » PHPfT3%hrti iff. r\ffiP^ QQt i ^frt ff _
1. Chemical appears not to induce mutations and CNA damage/repair.
2. Significant differences exist in pharmacokinetics and metabolism between
test species and humans.
3. Doses that induce tumors are either non-physiological or clearly disrupt
homeostasis.
III. SAR factors;
1. The parent compound and/or its metabolites are not members of a class of
human carcinogens.
15
-------
*EPA (1986) U.S. Environmental Protection Agency, Guidelines for carcinogen
risk MHnnffmrmt. Fed. Reg. 51:33992-4003.
IARC (1977) IARC Technical Report 77/002, Preamble. International Agency for
Research on Cancer, lyon.
IARC (1985) Interpretation of negative epidemLological evidence for
carcinogenicity (Wald, N.J. & Doll, R., Eds.). IARC Scientific
Publication No. 65. International Agency for Research on Cancer, Lyon.
**NCAB (1977) General criteria for assessing the evidence for carcinogenic!^
of chemical substances: Report of the Subcommittee on Environmental
Carcinogenesis, National Cancer Advisory Board. J. Natl. Cancer Inst.
58:461-5.
OSTP (1985) Office of Science and Technology Policy, Chemical carcinogens: a
review of the science and its asBnciated principles. Fed. Reg. 50:10371-
442. (*Reprinted, Env. Health Perspectives 50:201-207 only, 1986.)
* Enclosed for all participants.
** Enclosed for qualitative workgroup.
16
-------
Chair summary of Work Group suasion on
the) R«lejvanc« of Tumors in Animals to Hunan Carcinoganicity
iOIK 6BOOP IHOBI H01UB AID ABTMAL TUMDt DAZ4
January 11-12, 1989
Chair: Ernest E. McConnell
The animal tumor workgroup met Wednesday afternoon, January 11. and
Thursday morning, January 12. The workgroup was divided into two parts. Dr.
McConnell lead the discussion on various factors that impact on the relevance
of the animal bioassay to human hazard evaluation. Dr. Popp followed by
leading a discussion of metabolic, toxicologic, and mechanistic factors as
well as structure activity relationships.
The bioassay factors* that were discussed were the relevance/importance of:
1. Tumors with a high versus low background incidence
2. Toxicity in the target organ, i.e. the organ showing chemically
related neoplasms
3. Route of administration
4. Consistency between studies
S. Consistency between species/strains/sexes
6. Evidence of progression/regression of tumors
7. Induction of single types of tumors versus multiple types of tumors
8. Induction of benign versus malignant tumors
9. Importance of latency in tumor induction
10. Dose response - tumors only at the Maximum Tolerated Dose (MTD) or at
doses below the MTD
11. Tumor site in animals compared to site in humans
The strengths and weaknesses of each factor were discussed.
The consensus of the workgroup was that all of the above factors are
important and are relevant in determining human health hazards. Importantly
*It was assumed that the bioassay was conducted in an adequate manner and that
the tumor response was clearly related to the test chemical.
17
-------
it was agre«d that it would not be appropriate to focus too ouch on a single
factor, but that the various factors should be integrated to determine the
relevance of the bioassay for hazard identification.
Dr. Popp then lead 4 discussion on the following toxicologic,
mechanistic, metabolic and SAR factors.
1. Genetic Toxicology
- some discussion of the definition of a positive/negative
- application of gene tox data as it applies to weight of evidence and
mechanism. Concluded that it was an important risk factor and was
important in mechanism decisions.
2. Promotion
- the discussion became bogged down to some degree on a definition of
promotion
- consensus that it was important to determine if a chemical would fall
into that category t
- however, no consensus on how the information would be used for human
hazard evaluation. The issue probably needs to be revisited in the
future.
3. Cell Proliferation
- discussed the mechanisms involved in cell proliferation, i.e.,
mitogen versus regeneration
- important to establish if proliferation occurred early, late or
throughout the study, magnitude of proliferation, and whether it is
in target organ (only or in other organs)
- needs to be looked at on case-by-case basis. No general rule on
relevance.
4. Metabolism and tissue binding
- important to know but some skepticism on value for human hazard
evaluation.
- interspecies comparisons are important considerations. Important to
see if there are similarities, especially in the animal target organ
compared to the human target organ.
- the presence and type of binding can be important in understanding a
possible mechanism. However, need to be careful not to over
interpret the data.
- Understanding of biological half-life (persistence) of chemical is
important
18
-------
5. Physiologic adaptation
by itself may be of little value for human hazard evaluation, but
with information could be useful.
may be of more use for other toxic endpoints such as reproductive and
fetal development studies.
6. SAR
• probably of importance within specific groups/classes of chemicals.
• Important to remember that SAR is not infallible.
There was a consensus that all of the above factors are important in
evaluating the potential carcinogenic hazard of a chemical. It was stressed
that the information should be pooled with the bioassay data for an in-depth
human hazard evaluation.
SUMMARY
It was agreed that this type of information can make a fairly simple
discussion more difficult but the results are worth the effort and the current
state of the science demands that we do it. However, the totality of the data
should be "clear" and "strong" before it is used to enhance or detract the
bioassay results.
Moore's original assumption was that "Positive animal studies are
presumptive evidence of human hazard identification." It may be appropriate
to add the clause "in the absence of other relevant information." In other
words, the above factors can add or subtract the weight of evidence of this
assumption.
19
-------
PRE-MEBTING ISSUE PAPER
TOPIC
1. To analyze issues relevant for developing weight-of-evidence
classification
2. To propose weight-of-evidence classification(s) that embody discussions
from this and the previous workshop.
How can one integrate all infornation into an overall weight-of-evidence
that reflects the judgment of potential human carcinogenicity of an
agent?
Both the OSTP (1985) cancer principles and the EPA (1986) cancer risk
assessment guidelines have stressed a weight-of-evidence determination in
evaluating potential carcinogenic effects from exposure to chemicals in
humans. This entails consideration of all relevant human and long-term
animal studies along with metabolic/pharmacokinetic information, various
mechanistic considerations, and structure-activity relationships (SAR).
Little guidance is given in the EPA guidelines as to how this should be done.
The EPA classification is an adaptation of the approach developed by IARC
(1982).
LKLJBKttti'J.CMS AND
This workgroup is to perform two tasks. The first is to discuss a number
of issues (questions) relevant to making weight-of-evidence determinations.
The second task is to review certain existing classifications which can serve
as models for identifying attributes for ccnsideration and develop/propose
weight-of-evidence classifications.
1. Should a classification scheme emphasize whether an agent has
carcinogenic properties in a general sense ( i.e. , intrinsic
carcinogenic activity without regard to species)?
2. Should a classification scheme emphasize the determination of human
carcinogenic potential of an agent?
3. Given that the EPA weight-of-evidence determination goes beyond the
analysis conducted by IARC in support of human carcinogenicity,
should EPA continue to use an lARC-like classification?
4. Should the use of judgmental descriptors like "probable" and
~ '•possible" human carcinogens be continued?
21
-------
5. How can the confidence in the information and level of concern in
support of human carcinogenicity be incorporated into a weight-of-
evidence determination?
6. Should we expand/condense the number of groups in the EPA
classification system?
7. Should the potential for human exposure and carcinogenic "potency"
of the chemical be factored into the determination of whether a
substance poses a carcinogenic hazard to humans?
8. How should adequacy of testing be added as a component in a
classification scheme, since data bases on chemicals vary
significantly?
9. Given the deliberations on the relevance of tumors in animals to
human carcinogenicity and the current workshop discussions, what
might a classification scheme(s) look like?
10. What guidance coming from the hazard identification step might be
given as to the means of estimating human carcinogenic risk?
1. EPA (1986) classification, (attached)
2. IARC (1987) classification, (attached)
3. "Tripartite1* scheme, (attached)
1. Retain the current EPA classification but refine criteria for
categorizing the level of evidence.
2. Modify the current EPA classification to harmonize with IARC
(1987).
3. Modify the current EPA classification scheme along other lines that
incorporate insights obtained from this workshop and elsewhere.
22
-------
*EEA (1986) U.S. Eradronmental Protection Agency, Guidelines for carcinogen
risk Msinrnnmt. Fed. Reg. 51:33992-34003.
IARC (1982) International Agency for Peiienuh on Cancer, IARC Monographs on
the evaluation of the carcinogenic risk of chemicals to humans.
Supplement 4, pp 11-14. Lyon.
OSTP (1985) Office of Science and Technology Policy, Chemical carcinogens: a
review of the science and its associated principles. Fed. Rag.
50:10371-10442. (*Reprinted, Environ. Health Perspect. 50:201-207, only,
1986.)
* Enclosed for all participants.
23
-------
Chair's Summary of Veight of Evidanea
Classification Sassion
Chair: Gary Flama
Tha session began with an introduction from Richard Hill who described
what EPA hoped to accomplish from the "weight-of-evidence classification"
session. Or. Hill explained that it has bean apparent to the Agency through
comments received as a result of the Federal Register Notice of Intent to
review its cancer assessment guidelines and through other mechanisms that
EPA's current cancer classification system is subject to challenge. A major
issue that has arisen in the course of the public comment period is the
difference between a "strength-of•the-evidence" approach to classification as
opposed to a "weight-of-the-evidence" approach to classification. The
argument being that the application of EPA's current classification approach
is more like a "strength-of-the-evldence" approach which focuses only on how
good the animal cancer data are without adequate regard for the significance
of the animal cancer data to humans. The "weight-of-evidence" approach to
classification is assumed to focus on the meaning and significance of animal
findings to humans. Dr. Hill expressed the view that it is important for the
Agency to have the benefit of the discussion and the expression of views
flowing from the discussion of this issue by the various experts gathered
around the table. He emphasized that it is unnecessary for the group to come
to consensus, only to have an indepth and penetrating discussion of the
issues.
\
As a related exercise, a list of IS chemical carcinogens was circulated to
the members of the panel, and the members were asked to evaluate each on a
scale of 1 to 10 in which 1 is most relevant to humans and 10 least relevant.
The exact charge to the panel was "to the extent you are acquainted with the
data on these substances, give your personal evaluation of the likelihood each
substance is a human carcinogen under historical exposure. Don't guess or
resort to the conclusions reached by IARC, EPA, NTP and others. (Key: 1
[human] to 10 [not human]; N - don't know)." The panel was given
approximately 10 minutes to complete the questionnaire.
-------
One important conclusion that can be drawn from this exercise is that for
certain carcinogens (i.e., dimethyInitrosamine, 1,2-dibromo-3-chloropropane)
there was strong agreement on the significance and relevance of the animal
cancer findings to humans. For instance, panelists, with the exception of
,»
one, scored dimethyInitrosamine as 2 or less. Similarly, there was good
agreement that d-limonene and nitrilotriacetic acid were not likely to be
human carcinogens under historical exposure conditions. On the other hand,
for many other known animal carcinogens, such as carbon tetrachloride, there
was little agreement among panelists with scores at both ends and in the
middle of the scale.
Of the 9 chemicals from the list that have been evaluated by EPA (FDA
compounds excluded), all but two were designated as B2 (probable human
carcinogen) based on sufficient animal tumor findings. Vinyl chloride and
formaldehyde were classed higher due to combined human and animal evidence.
This raises the question of whether the B2 (i.e., sufficient animal)
category should be subdivided into two or more categories. It was emphasized
by some that no single factor could serve to offset the general presumption of
relevance of animal data to humans. In all cases a combination of many
factors would be required to provide convincing evidence that animal data were
not relevant to humans.
It was argued by some panelists that the rationale used by panel members -
whether wittingly or unwittingly - to classify certain carcinogens as more or
less relevant to humans was predicated on mechanistic considerations and that
such considerations were key to any further division of the B2 category.
Certain examples were pointed to as supporting this position, for instance,
the binding of Or^-globulin and its relationship with renal tumors, certain
hormonal disturbances or imbalances, and thyroid cancer and the relationship
between urinary stones and bladder tumors.
25
-------
Son* panelists opined that a sliding scale should be used to indicate the
degree of relevance of animal data to humans while others were opposed to this
approach. One panelist expressed the view that only when the scientific
evidence was clear, should any judgment be made about the lack of human
relevance; until that tine the presumption that the animal data were relevant
should hold. Several panelists expressed the view that in virtually all
circumstances the Judgment that animal data are not directly relevant to
humans must involve some consideration of human exposure. One panelist argued
that the assurance that BHA is not carcinogenic to humans is strongly
dependent on the conditions of human exposure as opposed to the notion that
animals are fundamentally and qualitatively different from humans in regard to
the induction of cancer by BHA. Similar arguments were made in regard to
other agents like those that produce urinary stones.
The panel, after several hours of lively discussion, recognized that there
are probably animal carcinogens that would not under specific conditions exert
their effects in humans. The panel discussed the level of evidence required
to establish that such minimal carcinogens are unlikely to affect humans and
discussed various methods of introducing this concept into the EPA risk
assessment system.
26
-------
QUANTITATIVE EVZDBNCB ISSUES
Oo«« Sealing Across Species
Pre-meeting Issue Paper
Work flroup Summary
Znoorporation of Mechanistio Data into Quantitative
Risk Assessment
Pre-meeting Issue Paper
Work Group summary
Additivity/Zndependenoe of Mechanism to Background
Processes
Pre-meeting Issue Paper
Work Oroup summary
27
-------
PRE-MEETINQ ISSUE PAPER
1UW.C
DOSE SCALING ACROSS
QQAIS;
1. To H<«rai«» the considerations that enter into the choice of a cross-
species scaling factor, including identification of the species
differences the scaling is intended to adjust.
2. To evaluate arguments in support of various scaling methods that have
boon proposed.
3. To reconmend default dose-scaling methods for use when few chemical
specific data are available, along with criteria for use of alternatives
when appropriate data are available.
In view of the developing understanding of comparative pharmacokinetics,
how should ctacoa be scaled so as to be expected to be of equivalent
carcinogenic effect across species vten (a) no rftnrinal -specific data are
available, and (b) **y»* mat-at-ini in or p*v»i i»tmit"in are available?
Arguments have been made for various dose-scaling factors to be used in
the absence of chemical specific data. The OSTP (1985) document recognized
there are no clear choices among the factors but stated that using body weight
or surface area as conversion factors may be reasonable. The EPA (1986)
Guidelines for Carcinogen Risk Assessment choose scaling daily amount by body
surface area as a default position. Although the guidelines recommend the use
of chemical-specific information on pnarmanokinetics and metabolism for
scaling, no guidance is given as to how to use them. Progress on this
question would be enhanced by a clearly articulated rationale for the choice
of any particular scaling method and an enumeration of the specific species
differences that scaling is intended to accomodate.
i >:. .1
AND
The most important focus will be on examining the various elements that
contribute to species differences in carcinogenic potency and questioning how
and whether a cross-species scaling of dose can serve to correct for them.
This will include a discussion of both pharmaookinetic and pharmacodynamic
factors, as well as an examination of the various ways in which target-
tissue doses can be expressed. Once the demands being placed on a scaling
factor are articulated, one can djgraiga the utility of various data and
theoretical formulations (e.g. , allometzy) in lending support to one or
another scaling method. Finally, one can consider how the appropriate scaling
method may vary as a function of different amounts and patterns of
phazmacokinetics and metabolism, as well as different modes of carcinogenic
action.
29
-------
The deliberations will be guided by focusing on the following set of
questions, for which a sense-of-the-neeting position will be sought:
1. Is dose scaling applied in order to correct for phamacokinetic
differences only (i.e., rendering tissue-level exposures equal) or
should it also correct for species differences in phaxnacodynamics
(the effect on elements of the carcinogenic process)?
2. Tissue-level exposures are inherently multidimensional,
consisting of changing concentration over tine. What considera-
tions enter into choosing an appropriate one-dimensional sumnary
measure (e.g., AUC, peak concentration, mg-eq metabolized)?
3. What relative tissue-level exposures can be expected to produce
equal risks across species in view of different numbers of cellular
targets, different rates of cell turnover and ENA repair, and
different lifespans? How does this expectation on the presumed
mechanism of carcinogenicity?
4. What is the utility of alloaetric predictions of relative tissue
exposures in animals and humans following equal administered doses?
Is the cumuli of "physiological time" helpful in equating
pharmacokinetic and/or pharmacodynamic processes across species?
5. Can appropriate "default11 assumptions about dose scaling be
constructed for use in the absence of adequate pharmacokinetic and
pharmacodynamic data on a compound? Do such considerations suggest
that the current Guidelines recommendation for scaling applied doses
by surface area (body weight to the 2/3 power) should be changed?
6. Should the "default" scaling factor depend on whether the putative
carcinogenic species is believed to be the parent compound, a stable
metabolite, or a reactive intermediate of metabolism?
7. Does the use of target tissue exposures in quantitative risk
assessment necessarily entail the predictionof site concordance of
induced cancers across species?
30
-------
Chair Suoaary of Work Oroup Session on
Oo«« sealing Across Species
Chair: Melvin Anderson
1. To discus* the considerations that enter into the choice of a
cross species scaling factor, including identification of the
species differences the scaling is intended to adjust.
2. To evaluate the^ arguments in support of various scaling
methods that have been proposed.
3. To recommend default dose-scaling methods for use when few
chemical specific data are available, along with criteria for use
of alternatives when appropriate data are available.
The panel wrestled with the issue of the scientific
considerations that enter into development of an appropriate
cross species scaling factor. It was recognized that the scaling
factor consists of at least two parts. One relates to species
differences in pharmacokinetics. This gives rise to differences
in tissue dose in various species even though the exposure
situations are equivalent. The second relates to tissue
sensitivity ,that is, different outcomes in various species even
though they receive identical tissue exposures. It was
recognized that the current use of a scaling factor is intended
in some poorly articulated way to account for both of these
factors. Despite the ability to define these two contributing
elements in the interspecies scaling factor, in practice it
proved virtually impossible for the working group to analyze them
independently. This was likely due to their long association in
regulatory policy.
Considerable discussion focused on the arguments which
support the current default scaling methods. The data of
Freireich et al. have been referred to as the basis for selection
of the surface area scaling presently used by the US EPA. These
data show that the toxicity of various chemotherapeutic agents
across species is approximately a function of surface area and
not a function of body weight. These particular d ta do not
address cancer outcome in various species, but are measures of
acute toxicity. In fact they relate different measures of acute
toxicity in "the two different species - mouse and humans. The
rationale expressed by agency (US EPA) representatives for
reliance on this data set was that many relationships scale as
body surface area, this acute toxicity of chemotherapeutic agents
is just one of them, and that it is perhaps to be expected that
cancer potency should also follow this kind of relationship.
This represents a recapitulation of the historical use of the
factor, but not a scientific justification for its use.
31
-------
In contrast, scientific arguments for interspecies
sensitivity of tissues to equivalent exposures might be surmised
from analysis of the data on solid tumors caused by radium
ingestion in several species. The results of Raabe et al. Mere
discussed and it was suggested that they support a scaling factor
related to total absorbed dose per lifetime instead of absorbed
dose per day. While these results may be misleading because they
represent analysis of radiation carcinogenesis and not chemical
carcinogenesis directly, no other compelling information on
interspecies tissue sensitivity for cancer causation was provided
to analyze the issue with chemicals. There does exist a body of
data on chemical carcinogenesis from people subjected to
chemotherapy who develop second tumors. These data are somewhat
obscured by the very different dosing scenarios in the patients
and in the exposed animal populations, but could be reviewed to
see if they illuminate the issue of tissue sensitivity to
equivalent doses.
There were numerous comments that the position of equal
sensitivity of tissues to cancer across species was not
appropriate for certain chemicals (the dioxins, for example) and
these chemicals that act through receptor mechanisms needed to be
looked at carefully. No specific recommendation followed,
however, for what would be an appropriate 'way to express tissue
sensitivity (although there was some mention of receptor number
as a predictor of response from dioxin type chemials from Dr.
Safe).
The default condition, scaling to surface area, as it now
stands can be defined as representing either both correction
terms - tissue sensitivity and delivered dose - or only one of
them. A comment made was made by one participant that
pharmacokinetics cannot address the interspecies sensitivity.
This really was a restatement of a belief that the scaling factor
as historically used and currently defined is for tissue
sensitivity alone. This position, as noted below, was not a
universally held position in the Federal sector regulatory
agencies.
The US EPA uses an adjustment factor related to body surface
area and assumes that larger species are at greater risk. The
data used by Freiereich et al. to support this factor has been
reanalyzed by.Travis, who gave an overview of his analysis. By
conducting a more complete statistical analysis, a slightly
different slop* (0.75 vs 0.67) was obtained. The initial
published analysis only showed that the slope of 0.67 was
consistent with the data and did not determine it exactly. There
was discussion about whether the 0.75 power would be a better
default position and whether the two agencies (US EPA and FDA)
might use the O.75 power as a common value for their interspecies
dose scaling. Not unexpectedly, there was little support of such
a move by any agency representatives. Each agency seems
comfortable with its own tradition in arriving at the
interspecies scaling factor. Various panel members did comment
on the fact that conformity in approach to a common problem would
32
-------
be desirable. A frequently cited study by Crump and colleagues
was also discussed. This work correlated the calculated animal
and human potencies of various carcinogens and, apparently, found
a somewhat better correlation with body weight than surface area.
It was felt by the panel and by an observer (Chao Chen, US
EPA) that the Crump et al. study does not provide irrefutable
support of a body weight correlation.
The topic addressed next was dose scaling. The discussion
was still burdened with an inexact definition of what the panel
intended when addressing this part of the total cross species
scaling problem. Does dose scaling correct for all interspecies
differences or only for part of them? Dr. Reitz presented a brief
summary of the use of pharmacokinetic modeling to calculate
particular measures of tissue dose. He stressed that there were
different measures of tissue dose that are expected to be
associaited with particular presumed mechanisms of toxicity.
These 'mechanisms' are not biological distinction at a fine
molecular level, but gross distinctions of the nature of the
chemical that causes toxicity. Discussed were cas'es where the
parent chemical is the genotoxic species, where a stable
metabolite is the gsnotoxic species, and where a highly reactive
chemical is the form reacting with DNA. Again, it was stressed
that some knowledge of the chemistry and biology of the system
was required to make these distinctions. The issue of just how
much 'mechanistic' information is required before pharmacokinetic
data can be used was a lively topic of conversation. There was
some limited feeling that the system had to be completely defined
in its total complexity before novel data could be used to alter
the risk assessment process. The rationale is that there is a
process in place now that provides some protection, it should not
be lightly changed because an error in the new approach might be
to increase risk and should be implemented only after virtually
universal acknowledgement that the new approach was a proper
method of calculating tissue response.
Despite differences in opinion about when the body of
evidence would be sufficient, there seemed to be general
agreement that pharmacokinetics could be used to estimate the
appropriate measure of tissue dose to aide in dose scaling and to
determine relative adjustment factors for interspecies scaling.
The issue of the relevant measure of dose to be calculated from
pharmacokinetic models was not very thoroughly addressed except
noting that the proper choice of tissue dose must be related to
the presumed mechanism of carcinogenicity. Similarly, while
allometric relationships are valid for many physiological
processes, biochemical processes involved in metabolism may not
behave nearly so coherently and need to be measured directly to
support such dose calculations. At least one comment was
directed toward developing a data base to test the expected crass
species extrapolation of tissue dose as predicted based by
pharmacokinetic arguments for chemicals with various mechanisms
of toxicity. It was noted by several participants that dose
scaling for reactive metabolites might best be represented as
related to an inverse of surface area.
33
-------
Several comments were taken from the observers. One issue
raised was whether it was necessary to build a conservative
default position into the decision tree or whether some other
process might be applied that considered the entire body of data
including the uncertainty with respect to mechanism. The same
person (Dr Stevenson, Shell) asked whether there shouldn't be
some incentive to encourage work to increase the data base for
resolving some of the scientific issues in dose scaling.
The efforts of the panel to successfully address the points
in its charge was severely hampered by its inability to attain
concensus on what the interspecies scaling factor is really being
used to correct for. The question remained heavily mired in
policy, in discussions about what the overall scaling factor
should be, in recapitulating the rationalization of its present
value, and in maintaining a general aversion to creating any new
default position. It was extremely difficult to seriously
examine the problem in its component parts. The panel, however,
is not unique in its inability to define the role played in
quantitative risk assessment by factors intended to adjust for
'dose-scaling across species'. There does not seem within the
agency any clearly articulated idea of what the factor should
address. In fact, among individuals from the three agencies - US
EPA, FDA, and CPSC - represented on the panel, there were three
very different ideas espoused as to how the correction should be
used and as to what was included in the totality of dose-scaling
across species.
One said that dose scaling was not appropriate. No
correction should be made to the animal dose to account for the
species differences and the tissue dose in experimental animals
should further be corrected based on a surface area relationship
(CPSC). A second espoused the position that the pharmacokinetics
could be used to account for dose delivery in humans, and after
taking delivered dose in humans into account, a surface area
correction should be applied. In this case the surface area
correction becomes the interspecies tissue sensitivity scaling
factor (US EPA). The third said that the correction was entirely
for dose, not tissue sensitivity, and when data on the
appropriate delivered dose metric were convincing, they could be
used directly. This position suggests that the correction is
entirely for delivered dose (FDA). There -really is a very
remarkable difference of opinion within the federal regulatory
agencies on the nature of this interspecies dose scaling factor.
It is no wonder that the panel was confused about the problem.
-------
Perhaps this final not* should be regarded as simply a personal
comment of the chairman, but it seems that the differences among
the agencies on interspecies dose scaling are growing further and
further apart as more scientific information becomes available.
Some common ground needs to be struck to avoid an impression of
arbitrary regulation of chemicals by the various regulatory
agencies in the federal sector.
35
-------
PUB-MEETING ISSUE PAPER
HQRRSBQP TQFIC
BULKFCBXEECN OP MEX2BNISTXC DAX& HOD QUANTITATIVE RISK ASSESSMENT
QQMfi;
1. To identify types'of mechanistic and biological data that support or
suggest alternative approaches to the current EPA default methods for
lew-dose extrapolation of dose-response relationships.
2. To evaluate the assumptions and data requirements of proposed mechanistic
approaches to quantitative risk assessment.
3. To develop considerations to be applied in choosing approaches to low-
dose extrapolation, and in evaluating their reliability.
How does one apply knowledge of carcinogenic mechanisms to the
development and use of low-dose extrapolation techniques?
The current EPA (1986) Guidelines state: "When pharmacckinetic or
metabolism data are available, or when other substantial evidence on the
mechanistic aspects of the carcinogenic process exists, a low-dose
extrapolation model other than the linearized multistage procedure might be
considered on biological grounds. When a different model is chosen, the risk
assessment should clearly disniss the nature and weight of evidence which led
to the choice." The linearized TtnlHstaqe procedures makes little use of what
is known about the process of chemical carcinogenesis, and aims only at
setting an upper bound below which the true dose-response relationship is
expected to lie.
AND
The initial deliberation will examine various proposed approaches to the
use of mechanistic data in dose-response analysis. The examination will focus
on questions relating to the choice among mathematical extrapolation models,
including an analysis of the specific mechanistic interpretations of car-
cinogenesis invoked by each method. The data required to implement each
method, and the ease and reliability with which they may be obtained, will be
considered. Finally, the group will be asked to suggest criteria for when
particular methods may be used. These criteria should reflect the uncertainty
in identifying key mechanisms and the difficulty of reliably measuring key
model inputs.
The deliberations will be guided by focusing in the following set of
questions, for which a sense-of-the-meeting will be sought:
1. How have data concerning mechanism of carcinogenesis in animals been
used to argue that the assumption of linearity of response at low
doses may be untenable?
-------
2. What data (other than tumor response) support this conclusion? Are
the data useful for quantitative imm'iuimint of response?
3. What underlying hypotheses about carcinogenesis support the various
approaches to dose-response modeling? Are these hypotheses
generally accepted in the scientific oonoounity in general terms? for
specific chemical responses or Ti*^0^ of action?
4. For key model parameters, what is known of the range of response
under normal and Hi*»n
-------
CHAIR SUMMARY OF WORK GROUP SESSION ON INCORPORATION
OF MECHANISTIC DATA INTO QUANTITATIVE RISK ASSESSMENT
Chair: William Farland
INTRODUCTION
^
This session began with an abbreviated discussion of the elements of dose
response analysis in risk assessment: the- selection of a data set for use,
determination of equivalent exposure units between species, and choice of an
extrapolation model. In particular, the question of how to extend the
information obtained from high dose studies into the range of interest,
several orders of magnitude lower, was discussed. The issue was framed as
follows: How does one apply knowledge of carcinogenic mechanism(s) to the
development and use of low dose extrapolation techniques?
The session can be summarized by focussing on three main questions which
were addressed in the discussion, highlighting the points which were made and
relating conclusions which were reached by the group. Finally a set of
general conclusions is presented in an attempt to further convey the thinking
of the workgroup.
QUESTION 1. HOW HAVE DATA CONCERNING TUMOR RESPONSE AND/OR MECHANISM OF
ACTION BEEN USED TO EVALUATE THE ASSUMPTION OF LINEARITY OF RESPONSE IN THE
RANGE OF "INFERENCE"?
This question was addressed first from the perspective of the analysis of
data from exceptionally large cancer studies. Examples which were presented
and discussed included the NCTR study, termed the* "EDoi Study", using 2-AAF,
the BIBRA study with nitrosamines and the IRDC study of saccharin. Despite
the size of these studies, the point was made that they covered relatively
narrow ranges of exposure. In addition, while responses were either linear or
superlinear in the observed range, the data do not allow the differentiation
of linearity from non-linearity at the low end of the observed range and into
the range of inference. These studies also show that for practical reasons,
it is not possible to collect data at much lower than a 1% response rate so
39
-------
Chat our ability to evaluate the shape of the dose-response curve is limited
to this range.
Further discussion led to consideration of theoretical issues related to
linearity /non-linearity. •> These included a brief discussion of such issues as
the impact of pharmacokinetics as well as choice of a dose-response model on
the shape of the dose-response curve. Saturation of activation or elimination
pathways were cited as reasons for some perceived non-linearities of response.
In addition, the choice of cross-species factors affecting metabolism and
distribution may have profound effects for site-concordance of expected
response.
Several points were made about the choice of dose response models. With
regard to the linearized multistage approach, the EPA's current "default"
model, discussion focussed on the decreasing support for its "biological"
underpinnings. Also, its ability to accommodate only tumor response data
limits its ability to capture much of the data considered in the weight-of-
the-evidence determination. The point was made that it is one among several
models which incorporate low dose linearity but that the important issue was a.
determination of where linearity of response began and what the slope was.
This approach was contrasted with so-called "two-stage models" such as
that developed by Moolgavkar and others. These models allow the incorporation
of more of the data and, hence, more of an understanding of cancer biology.
Use of such models on a routine basis will require an increased understanding
of sensitive parameters in the model and how to address them, e.g., data or
inference. While the discussion of this general approach was favorable,
caution was voiced that more data than is currently collected would be needed
to implement their use. In addition the difficulty in collecting needed data,
e.g., proliferation rates for initiated cells, reflects limited systems and
expensive studies. Such data collection should not be viewed as something
which would be routine.
The following conclusions were reached regarding the stated question:
40
-------
1) High dose tumor data are not very useful in evaluating the low dose shape
of the dose-response curve but they are generally all we can expect to have;
2) There are various reasons to expect or at least to hypothesize non-
linearity in the high dose region, but there are fewer for the low dose
region, i.e., many processes can be expected to become linear at lower doses;
3) Biologically based models can incorporate theories of carcinogenesis but
they must be flexible enough to be modified with increases in knowledge.
Don't over simplify!
QUESTION 2. WHAT DATA OTHER THAN TUMOR RESPONSE CAN BE USED TO EVALUATE THE
ISSUES RAISED IN RESPONSE TO QUESTION 1?
Major topics discussed in response to this question included the role of
DNA adducts and cell proliferation in predicting tumor response. With regard
to DNA adducts, points were raised about uncertainty in the direct role that
certain adducts were playing in the carcinogenic process. Support for their
role is based on correlations in the shape of the dose-response curve,
efficiency of formation, persistence, site-specificity and other factors.
These data allow the dose-response curve to be extended to lower doses if the
inference for the role of adducts in causation of tumors is strong enough.
Adducts should generally be considered to be linear at low doses although not
necessarily at high doses. Examples of these concepts involved discussion of
some data sets illustrating linearity such as those for DEN, 2-AAF, and ETO.
On the other hand, chemicals such as vinyl chloride, formaldehyde, NNK, and
BaP were used to illustrate non-linearity. Data sets from NNK with saturated
activation and BaP with saturated detoxication were discussed as examples
illustrating the basis for sublinearity and superlinearity of high dose tumor
responses.
Cell proliferation data can also be helpful in understanding or predicting
tumor respojnse. In addition to allowing for the detection of background
-------
initiation of potential tumors, cell proliferation as an expression of high
dose toxicity can lead to enhanced induction of tumor response for relatively
weak initiators. This is in contrast to the situation where cytotoxicity
actually depresses a tumor response due to either random or selective cell
killing. Such enhancement responses.can be a response to genotoxicity, e.g.,
with oncogene activation, but it can also be non-genotoxic in nature, e.g.,
with Offc-globulin and male rat kidney response. Nevertheless, the cell
proliferation response seems most often to represent a high dose phenomenon
and needs to be taken into account when extrapolating to lower doses.
Conclusions reached regarding this question were as follows:
1) Data are accumulating to support correlation between certain adducts and
tumor response but causation remains an uncertainty;
2) Linearity of adduct formation, although with differing efficiency depending
on specific adducts, is expected and seen at low doses, but not necessarily at
high doses;
3) Cell proliferation can enhance response of adducts or background processes,
but should be regarded as necessary but not sufficient for most tumor
responses;
4) Data on adduct formation and cell proliferation can be collected to feed
into biological dose-response models but these techniques are limited and may
be prohibitively expensive.
QUESTION 3. WHAT ARE THE UNDERLYING BIOLOGICAL PROCESSES WHICH SUPPORT THE
USE OF BIOLOGICALLY-BASED APPROACHES TO DOSE RESPONSE?
Discussion of this question stimulated further discussion on several
topics described in above sections. It reflected the general belief that
cancer is a multistage process which is impacted by such things as
pharmacokinetics, mutagenicity, mitogenicity, and cytotoxicity. In addition,
-------
the discussion focussed on hormonal modulation of tumor response. Chemical
impacts on hormonal control of cellular processes represent yet another point
in the multistage process where chemical impacts can be seen. Examples of
this effect are perceived as receptor-based processes, so special
consideration must be given to understanding high versus low dose
interactions.
The point was made in this part of the discussion that there are
techniques available for the incorporation of inexact data into models and fo#
understanding the impact of certain data choices. These techniques must be
further explored if we are to be able to use data as have been discussed in
the face of uncertainty.
Conclusions reached in this section of the discussion include:
1) All of the data can be built into cancer models if the models are given the
necessary flexibility;
2) The question of effects of a chemical on multiple points of the cancer
process must be better understood from both the biological and from the
modeling perspectives;
3) There is still a long way to go in understanding what the most important"
biological data are for evaluating tumor responses or predicting tumor
response at low doses.
In addition to the general conclusions which summarize each of the
discussions on individual questions, the following represent areas of general
consensus coning out of the overall discussion:
1) We have to start somewhere to incorporate more data into quantitative risk
assessment. The use of pharmacokinetic data to explain the shape of the dose-
response curve, i.e., the development of the biologically effective dose
concept, appears to be most promising;
-------
2) Bioassay tumor data will continue to be used as the basis for quantitative
assessment, but other information is useful and necessary, both qualitatively
and quantitatively, to put the tumor data in proper perspective;
*
3) Novel risk assessment approaches should be encouraged both inside and
outside of EPA. Scientific opinion (general consensus) should be used as a
criterion for use of these techniques since the process will continue to rely
on many inferences;
4) Biological models under development should be flexible and incorporate as
much of the biology and data as possible. We should not be satisfied with
oversimplifications if we are to improve on existing techniques rather than
just replacing them with approaches of equal or greater uncertainty,
5) Research must be encouraged to address areas which have been identified and
to develop methods for providing the data necessary to be able to replace
assumptions in low dose extrapolation models.
-------
PRE-MEETING ISSUE PAPER
NQRK5HGP TOPIC
M1IITI VHY/MEfagMglCE GP IBCBMOSM TO HM3UAUM3 PROCESSES
GDKLS:
1. To examine, for several proposed mechanisms of chemical carcinogenesis,
whether the biological action of carcinogenic agents is independent from
or additive to the processes responsible for the genesis of background
tumors.
2. To H<«rai«« biological processes that support the theoretical arguments
for the additivity or independence assumptions.
3. To recommend procedures for estimating risks at low doses that take
proper account of mechanistic aspects of the additivity/ independence
issue, both in cases where appropriate biological information is
available and when it is not.
one combine biological and H'MJJLW^'"8'*! arep •'• • ii H into guidance
on the existence and magnitude of risk from low-dose exposures to chemical
Questions have arisen as to the means of evaluating cancer risks from
chemical exposures when "spontaneous" tumors of the same type occur in
unexposed individuals. Some have argued from theoretical grounds that if a
chemical acts upon cells in ways similar to the causes of "background" cancer,
then even small exposures may marginally accelerate the tumorigenic process,
leading to a linear, no-threshold elevation of risk over background at low
doses (Crump et al., 1976; Peto, 1978). Alternatively, if a chemical acts in
other ways, there would be independence in action between the chemically-
induced effects and the background processes, which could possibly lead to
excess risks fxon low doses that are markedly sub-linear. For example, some
mechanisms of carcinogenesis may elevate risk over background only when
organisms are perturbed out of a "normal physiological range" by high doses.
Both the OSTP (1985) document and the EPA (1986) Guidelines for Carcinogen
Risk Assessment espouse the use of linear extrapolation as default science-
policy decisions but stress the importance of mechanistic, metabolic, and
pharmacokinetic information, when it exists, in deriving high-to-low-dose
extrapolations.
The theoretical arguments about additivity and independence are couched
in very general terms, while the biological investigation of carcinogenic
mechanisms addresses the action of specific elements of the process. One now
needs to seek the specific biological meaning behind the theoretical concepts
of "additivity" and "independence" in order to understand how the theoretical
and biological arguments can be brought to bear on the question of estimating
risks fron low doses of carcinogens. It should be noted that the question
here is not one of extrapolating the shape of the dose-response curve between
high and low doses, but rather of the behavior that should be expected of any
45
-------
such extrapolation at doses low enough that the interaction (if any) with the
background tumoriqenic proneseog predominates.
The initial deliberation win focus on the theoretical arguments about
the low-dose consequences of carcinogenic processes that are either additive
to or independent from background. The notions of additivity and independence
will be examined in the light of various specific proposed mechanisms of
carcinogenesis to try to determine what "additivity1* and "independence" mean
in biological rather than statistical terms. The means of distinguishing
between additive and independent effects and of measuring them at low levels
will be examined. The final focus is on recommendations about appropriate
treatment of possible low-dose risks when some mechanistic knowledge is
available.
The deliberations will be guided by focusing on the following set of
questions, for which a sense-of-the-meeting position will be sought:
1. Is the etfditive-to-background position an assumption, or are there
data to suggest that it describes the underlying biological truth?
2. How does the statistical argument that low-dose linearity is to be
expected when mechanism is additive to background fare in view of
knowledge of various mechanisms of carcinogenesis?
3. How can we distinguish cases of independent and additive background
in practice? What biological data can help in trying to make this
distinction?
4. What is known about the low-dose properties of done effect curves
for elements of proposed mechanisms of carcinogenesis (e.g.,
mutation, cytotcodcity, receptor binding)?
5. Practically speaking, are we able to measure very small elevations
in these processes over background (so small that they imply trivial
cancer consequences) in order to detect a virtual (or practical)
threshold?
6. For quantitative purposes, should a putative epigenetic carcinogen
be treated as acting independently from or additively to low levels
of other such agents in the human environment? of genotoxic agents
in the human environment?
7. In view of the above issues, under what circumstances might it be
appropriate to assume that carcinogenesis has or does not have a
dose threshold? What criteria must be satisfied to treat a
carcinogen as acting independently from background, and how should
exposures to these substances be viewed vis-a-vis exposures to
substances that may be additive to background?
-------
Crump, K.S., O.G. Hoel, C.H. Langley, and R. Peto (1976) Fundanental
carcinogenic processes and their duplications for low dose risk
Research, 36:2973-9.
*EPA (1986) U.S. Environmental Protection Agency, Guidelines for Carcinogen
Risk Assessment. Federal Register, 51:33992-4003.
OSTP (1985) Office of Science and Technology Policy, Chemical carcinogens: A
review of the science and its associated principles. Federal Register,
50:10371-442. (*Reprinted, Environmental Health Perspectives, 50:201-
207, only 1986)
Pato, R. (1978) Carcinogenic effects of chronic exposure to very low levels of
toxic substances. Environmental Health Perspectives, 22:155-9.
* Enclosed for all participants.
-------
Chair: Daniel Krevski
Hie quantitative estimation of risks airaociattd with low levels of exposure to
ji
carcinogens present in the environment is an important pan of carcinogen regulation.
Presently, there is a strong tendency to employ risk estimation methods which assume
that the dose response curve for carcinogenesis is linear in the low dose region. This
position is reflected in the principles proposed by the OSTP (1986), who stated that
"When data and information are limited...and when much .uncertainty exists
regarding the mechanism of carcinogenic action, models or procedures which
incorporate low-dose linearity are preferred."
This position is also reflected in EPA's current Carcinogen Risk Assessment Guidelines
(EPA, 1986).
While emphasizing risk estimates derived using some form of linear extrapolation,
the Agency Guidelines recognize that such estimates may be more appropriately viewed
is plausible upper limits on risk, and that the lower limit may well be effectively zero.
While the Guidelines also state that procedures for obtaining a best estimate lying
somewhere these two extremes generally do not exist at the present time, it may be
possible to move away from the upper limit in some circumstances. The Agency has
recently taken a concrete step in this direction in its consideration of the possibility of
a threshold for the induction of thyroid tumors (Paynter ei aj. 1988).
Current practice within the Environmental Protection Agency is to use the
linearized multi-stage model for low dose cancer risk flniitifHUH* The most important
aspect of this practice is not the choice of the multi-stage model itself for risk
estimation purposes, but rather the linearized form of the model Because the model is
-------
constrained to be low dose linear, it may be expected to yield risk estimates comparable
to those based on other linear extrapolation procedures, including those proposed by
Gaylor ei al (19_) and Krewsiri fit al (1986).
*Q Backround and Low Dose Linearit
Additiviry to background is often cited in support of the assumption of low dose
linearity in carcinogenic risk assessment In the additive background model proposed by
Crump ei al (1976) and Peto (1978), spontaneous tumors are associated with an effective
background dose, with exposure to carcinogens present in the environment adding to
this background dose. In this regard, Crump fit al (1976) stated that
"if carcinogenesis by an external agent acts additively with an already
ongoing process, then under almost any model the response will be linear at
low doses".
Hoel (1980) subsequently demonstrated that this result also holds even in the case of
partial additivity. This result is reexpressed as follows in the current Guidelines:
"If a carcinogenic agent acts by accelerating the same carcinogenic process
that leads to the background occurrence of cancer, the added effect of the
carcinogen at low doses is expected to be virtually linear."
The basic idea behind the additive background model is illustrated in Figure 1.
•
Here, the spontaneous response rate is considered to arise as a consequence of an
effective ^background dose, with the effects of the test chemical acting additively to
background in a dose-wise fashion. The face that the excess risk over background
P(d+S) - P(5) is linear at low doses follows from the fact that the secant between doses
of 5 and (5+d) converges to the tangent to the dose response curve as the dose d of
the test compound becomes small.
-------
Within the framework of this model, the only condition required for this result to
hold is that the probability of tumor occurrence be a smooth strictly increasing function
of dose. No further assumptions are required concerning either the mathematical form
«•
of the dose response relationship or the lexicological mechanism by which tumors are
induced.
It is worth noting that low dose linearity implied by this model refers to the slope
of the dose response curve at an applied dose of zero. Without further assumptions, no
further general statements can be made about the magnitude of this slope, nor about
the range of low doses over which this linear approximation will hold reasonably well.
For the multi-stage model, however, the linear approximation holds very well even at
doses which double the background tumor rate, provided that the spontaneous response
rate is not exceedingly small (Crump el ai., 1976).
Mnnlinearity at High Doses
The existence of linearity of low doses does not imply that the dose-response curve
will also be linear at high doses. In particular, curvature at high doses can occur due
to factors such as saturation of absorption or elimination pathways or pharmacokinetic
processes involved in metabolic activation. Nonlinearity at high doses can also occur
due to re*nrnfoi* of DNA repair systems or the induction of cellular proliferation.
Dose-response curves for chemicals which can both cause DNA damage and induce
cellular proliferation can be subject to a high degree of upward curvature, as with the
hockey stick shaped dose-response curves for 2-AAF induced tumors of the urinary
bladder. Similarly, secondary carcinogens which act as a consequence of high dose
toxicity may be effective only at relatively high doses.
50
-------
It was noted that nonlinearity due to saturation of elimination pathways results in
upward curvature (as is the case with methylene chloride), whereas saturation of
activation processes leads to downward curvature (as, for example, with vinyl chloride).
However, for those processes which saturate in accordance with Michaelis-Menten
kinetics, the amount of the proximate carcinogen formed at low doses will be directly
proportional to the administered dose since such processes are essentially first order at
low doses (Murdoch tl a]., 1987).
If the pharmacokinetics of metabolic activation are known, dose-response may be
assessed in terms of the dose delivered to the target tissue. This may result in a more
nearly linear dose response curve, which greatly facilitates statistical extrapolation to
low doses (Hoel ej al, 1983; Krewski ej a].. 1986).
Often, the only data available on which to base estimates of low dose risk are the
tumor occurrence rates at two or three dose levels used in carcinogen bioassay. It was
noted that linear extrapolation procedures such as the linearized multi-stage model
applied to bioassay data acquired at high doses may lead to overestimates of risk when
the linear component of the dose response curve does not become manifest until much
low doses (see Figure 2). It was further noted that linearizicd estimates of low dose
risk are also highly insensitive to the experimental data in the sense that even large
perturbations in the data may not have much impact on the estimates of low dose risk.
This insensitiviry is further demonstrated by the strong association between carcinogenic
potency and the maximum tolerated dose shown by Bernstein d al (1985). For these
reasons, estimates of risk based on linear extrapolation of bioassay data provide
somewhat crude upper bounds on the risks associated v ith low levels of exposure.
51
-------
It was suggested that improved estimates of low dose risk might be attained if
additional data on pre-neoplastic effects were exploited. For example, consideration
could be given to the use of DNA adducts which may be shown to be related to tumor
.*
induction in specific cases. In such cases, it should be possible to monitor adduct
formation at doses far below those at which reliable estimates of tumor occurrence
rates can be obtained. This information could then be used in conjunction with the
bioassay results to obtain more accurate estimates of low dose risks.
Molecular D^yn^try
One area which offers considerable promise for improving our understanding of
how neoplastic changes occur at low doses is molecular dosimetry. A forthcoming
report by the National Academy of Sciences indicates that different chemicals may
induce different kinds of DNA lesions, involving anywhere from two to thirteen sites on
molecular DNA. At the same time, there is evidence that spontaneously occurring DNA
lesions can be different from those caused by exposure to alkylating agents. This
suggests that fingerprinting of DNA damage in exposed and unexposed individuals may
provide a means of distinguishing between additive and independent background in
practice.
In the case of a genotoxic agent which acts completely independently of
background, the dose-response curve for tumor induction can be linear or nonlinear. If
neoplastic conversion can occur as the result of a single pro-mutagenic DNA lesion, the
linearity of adduct formation at low doses implies linearity with respect to tumor
induction. If two or more pro-mutagenic lesions are required to create a malignant
cancer cell, however, the dose-response curve will be nonlinear even at low doses. This
52
-------
essentially represents a multi-hit independent background model in which the dose-
response curve is proportional to dose raised to a power equal to the number of DNA
lesions required for neoplastic conversion.
,»
Dose-Responsje. Modeling
Dose-response modeling is an integral pan of the quantitation of cancer risk. The
biological basis for the multi-stage model currently used by the Agency for this purpose
is incomplete in that no cognizance is taken of tissue growth or cell kinetics. A
moderately large number of stages (up to six or seven) may also be required to describe
dose-response curves exhibiting high upward curvature, raising questions of biological
interpretation.
For these reasons, more biologically based dose-response models have received
considerable attention in recent years. Perhaps the most widely discussed model is the
two-stage birth-death-mutation model developed by Moolgavkar, Venzen and Knudson,
hereafter referred to as the M-V-K model. This model presumes that initiated cells
arise from normal stem cells following the occurrence of a single mutational event.
Initiated cells may undergo clonal expansion, and possibly sustain a second mutation to
form a cancerous celL The model also provides for growth of the target
tissue over time.
Although this model currently enjoys considerable biological appeal, and can
describe a variety of dose-response curves, it may not provide a complete description of
the events involved in tumor induction in all cases. For example, more than two
mutations may be required for neoplastic conversion in some cases. Similarly,
53
-------
nongenotoxic factors may be required to foster development of a malignant tissue mass.
As additional components are incorporated, the tractability of the model will be
reduced.
.»
Despite this potential for further elaboration, die two-stage M-V-K model is
viewed as biologically more meaningful than the Armitage-Doll multi-stage model.
Application of this model will however require supplementary data on tissue growth and
cell kinetics, in addition to bioassay data on tumor occurrence. Attempts to estimate
all of the parameters involved in the model without such supplementary data will likely
result in unstable estimates (Portier, 1987). Separate estimates of the parameters
governing the two mutation rates characterizing the genotoxic processes in the model
will also require more elaborate bioassay protocols than those currently in use (EPA,
1987). Nonetheless, applications of this model should be encouraged in order to gain a
better appreciation of its practical utility.
General Mechanistic (Vmsideratipna
A number of mechanistic considerations relevant to low dose risk assessment were
identified. First, the examination of non-neoplastic changes involved in carcinogenesis
may provide some insight as to the magnitude of low dose risks. For example, data on
DNA damage and cellular proliferation may provide a means of obtaining more accurate
estimates of low dose risks. This tnight be done within the context of the M-V-K
model or through the use of more sensitive biomarkers such as DNA adducts which can
be measured at doses well below those at which tumor occurrence rates can be measured
directly.
-------
Whether or not hormonaily mediated tumors fall within the additive background
framework is somewhat unclear. Such tumors can occur following depression of
circulating thyroid produced hormones, with compensatory thyroid hyperplasia induced by
thyroid regulating hormones secreted by the pituitary gland (Clayson, D.B., 1989). As
noted previously, it has been suggested that this process may demonstrate a threshold
dose below which tumors may not occur. On the other hand, if thyroid tumors can occur in
nonexposed individuals as the result of natural fluctuation in hormone levels, an additive
background model may apply.
The potential risks posed by exposure to low doses of tumor promoting agents are
somwhat unclear. If spontaneous promotion can occur, the possibility that an additive
background model may apply exists, along with the attendant implication of low dose
linearity.
Finally, it was noted that multiple mechanisms may be involved in chemical
carcinogenesis, including both genotoxk effects and nongenotoxic effects such as
cellular proliferation. In general, existence of multiple pathways of carcinogenesis
renders the elimination of partial additivity to background more difficult
55
-------
References
Bernstein, L., Gold. L.S., Ames, B.N., Pike, M.C. & Hoel, D.G. (1985). Some tautologous
aspects of the comparison of carcinogenic potency in rats and mice. FinndflTnfintnl it
Applied Toxicology 5,79-86.
.*
Clayson, D.B. (1989). Can a mechanistic rationale be provided for non-genotoxic
carcinogens identified in rodent bioassays? In preparation.
Crump, K.S., Hoel, D.G., Langley, C.H. & Peto, R. (1976). Fundamental carcinogenic
processes and their implications for low dose risk assessment. Cancer Research 26,
2973-2979.
Environmental Protection Agency (1986). Guidelines for carcinogen risk assessment.
Federal Register 51. 33992-34003.
Environmental Protection Agency (1987). Report of the EPA Workshop on the
Development of Risk Assessment Methodologies for Tumor Promoters. Report No.
EPA/600/9-87/03. EPA, Washington, D.C.
Gaylor, D.W.
Hoel, D.G. (1980). Incorporation of background in dose-response models. Federation
39.73-75.
Hoel, D.G., Kaplan, N.L. & Anderson, M.W. (1983). Implication of nonlinear kinetics on
risk estimation in carcinogenesis. Science 212,1032-1037.
Krewski, D., Murdoch, D. & Dewanji, A (1986). Statistical modeling and extrapolation of
carcinogenesis ^^tn In: Modern Statistical Methods fa Chronic Disease Epjdc.inifllo.gy
(S.H. Moolgavkar & RX. Prentice, eds.). Wiley, New York, pp. 259-282.
Murdoch, D., Krewski, D. & Crump, K.S. (1987). Mathematical models of carcinogenesis.
In: Cancer Modeling (J.R. Thompson & B. Brown, eds.). Marcel Dekker, New York,
pp. 61-89.
Office of Science and Technology Policy (1985). Chemical carcinogens: a review of the
science and its associated principles. Federal Register SQT10371-10442.
Paynter, OJL, Burin, G.C, Jaeger, R.B. & Gregorio, C,A. (1988). Goitrogens and thyroid
follicular cell neoplasia: evidence for a threshold process. Regulatory Toxicology &
Pharmacology 8. 102-119.
Peto, R. (1978). Carcinogenic effects of chronic exposure to very low levels of toxic
substances. Environi[nfn|[pl Health Perspectives 22.155-159.
Portier, C. (1987). Statistical properties of a two-stage model of carcinogenesis.
Environmental Hpalth Perspectives 76. 125-131.
56
-------
Probability
of
Response
P(d + S)
V/l
-vl
Dose d
Effective
Background
Dose (S)
Background Dose
plus
Applied Dose(d + 8)
Figure 1.
The additive background model.
-------
Probability
of
Response
P(d)
Observed Response
Linear Extrapolation
from
Lowest Dose
/
I
Actual Dose Response Curve
Dose d
Figure 2. Linear extrapolation of bioassay data.
-------
SUMMARY 3^ID CONCLUDING PBMARKS
by John Ashby
Virginia Beach; February 1989
This has been a highly successful meeting. When one reads the work group
remits that were prepared prior to the meeting one realizes how much
thought and work had preceeded this gathering. In particular, by their
design and with the questions they pose, they define the major areas of
uncertainty 1n carcinogen classification and risk assessment modelling.
The EPA have therefore essentially called the bluff of those who think
there are simple answers to these problems, and 1n so doing, they provided
the right atmosphere for the productive meeting we have just taken part
In.
The thing that strikes one most forcibly 1s that this meeting would have
been Inconceivable a few years ago. Things are changing rapidly in this
area, and the EPA judged the moment correctly to hold this meeting - one no
longer has to use the^terms non-genotoxic carcinogen or acceptable risk in
a lowered voice - they are now subjects for serious study. It 1s
interesting that risk assessment 1s already a widely discussed topic well
beyond scientific arenas - this has been enforced by the sensationalizing
of certain putative human hazards. Quite by chance, this topic was aired
1n a British Newspaper the day I left for this meeting. The topic was the
'toxldty' of potato skins, and It followed closely on the egg/Salmonella
debate that 1s still taking place in Britain. John Akass wrote the
following:
59
-------
'What weight of potato skins does the average man have to eat
before he calls an ambulance? We can be trusted with these details
because most mature adults are reconciled to the fact that the
ordinary run of life contains unpleasant, but acceptable hazards.
Institutes that exist to give us the jitters should also tell us how
much we should tremble'.
The final sentence of that quotation describes in common words what we have
been attempting to achieve for carcinogens over the past days.
It will perhaps help to give an example of the problem we face. I have
chosen two rodent carcinogens defined by the US NTP. Both are formally
carcinogenic, yet scientific instinct leads us to expect a greater
potential hazard for man from the first carcinogen of the two (Slide 1).
It 1s when one attempts to justify that feeling that problems are
encountered, because we have not yet collectively agreed the factors that
contribute to high or low human hazard ratings for an animal carcinogen.
This example could be repeated many times from the large database of the
NTP alone (this and related examples are summarized In Mutation Research,
224, 17-115, 1988). Other examples of animal carcinogens with different
implicit hazard ratings for humans are shown in Slide 2; it is the
scientific formalIzation of the assumptions made on this slide that is
currently required.
An interesting topic has surfaced several times at this meeting - the
difference between 'strength of evidence' and 'weight of evidence'. The
first is a description of the scientific validity of an observation, the
60
-------
en
I Code No | CAS No.
NIPTwfcftaportNo (yaw)
ShuchMortoMMlnMw
I 151 136-20-6
M0< 16761
Cuptoon
1 IOJ 1 83446-30-6
306(1866)
CMoriMtadpMMm*
(Cn:43ftCU
•fart
•^•^•i^™— •
•f
-
Saknoneto
»1*!!7
*
-
1 T<
*••
CS
HQ
L
8
2Q
MB
Ct
ItatoCI
c| I !„
1 *• 1 "
0 76 60
O 17 6
0 40 46
MKMOQEMOr
fcTBAl*
II
0 62 78
2 66 23
0 33 61
2 It 6
rOATA
NTPTcdmc^R
MK»(
' I . 1
C|L |H
2 7 16
0 7 M
12 24 32
•part
%TBA>
«
C | t | M
2 11 13
0 4 13
4 20 26
004
I- Card nog enicity data for cupferon (NTP TR 100, 1978) and chlorinated paraffins (NTP TR 305, 1986),
, ,
as abstracted in Nutation Research, 2Qi, 17, 1987. The syabols used In this Table are described In that paper,
but the key ones are TBA « tumour bearing anlnals, C, L and H are control, low and high dose test groups. The
tissue codes are CS - Circulatory system, HG - Harder Ian gland, L - Liver, S - Stoaach, ZG • Zynbal's gland,
HS - Haenopoeltic systea. The first compound, cupferon, was classified as positive (P) In all four test
groups. The second compound was classified using current levels of evidence and gave clear evidence (CE) of
card nogenl city only as a leukaemgen In the nale Mouse. These two cheaicals are therefore representative of
putative geno toxic and non-genotoxic carcinogens. The caret nogenl city of the chlorinated paraffin was
observed at relatively high dose-levels (top dose of 5g/kg in the aouse) and haeaopoeitlc tumours appeared in
the male mice controls (12% TBA). This slide captures the purpose of the present Meeting by highlighting the
need to agree a method for ranking carcinogens in terms of possible human hazard.
-------
For Human Risk Assessment
NaS
DEHP
LIM
BHA
NaCI
HCHO
TNM
*
*
*
*
*
*
*
OBN
OMN
DEN
MNU
BP
S mustard
BCME
bladder
liver
kidney
stomach
skin
respiratory
lung
Why ? MECHANISM/NATURE OF EFFECT
Slide 2. Examples to illustrate the implicit purpose of .this meeting.
NaS - sodium saccharin, DBN * dibutylnitrosamlne, OEHP -'2-diethyl-
hexylphthalate, OMN * dimethylnitrosamine, LIM • limonene,
DEN « diethylnitrosamine, BHA - butylated hydroxyanlsole,
MNU * N-methyl-N-n1trosurea, NaCI - common salt, BP • benzo[a]pyrene,
HCHO - formaldehyde, BCME - bis(cMoromethyl)ether, TNM - tetranitromethane
S-mustard - b1s(2-chloroethyl)sulphide.
62
-------
assumptions can only yield soft predictions. The fact that 1t 1s possible
to Integrate many different datasets in order to derive a perception of
overall risk was roost convincingly Illustrated by the '10-carcinogen
questionnaire' completed by a work group yesterday (elsewhere in this
document). The fact that the whole of the scale (1-10) was used (1.7-9.2
Infact) confirms that we generally agree that the weight of evidence
suggests that sodium saccharin does not present the same human hazard as
does vinyl chloride, y|t each are clearly carcinogenic in adequate animal
studies. Turning now to the main consensus points derived from this
meeting.
It seems that there is common agreement that chemicals can elicit a
carcinogenic response In rodents by more than one mechanism. The most
obvious of these Is by direct Interaction of the chemical with ONA-
genotoxlc carcinogens. A range of other possible mechanisms are now being
entertained as plausible, but often all we have are empirical markers or
hazy hypotheses of how the tumour Incidence Is Increased. Nonetheless,
these Indications are valuable for they encourage further research, and
usually that strengthens belief In an alternative (non-genotox1c)
mechanism of carclnogenldty. This leads to probably the major conclusion
of the meeting, as follows • If. In the fullness of time, we come to
recognize 1 variety of distinct mechanisms bv which chemicals can cause
cancer 1n rodents, then 1t will follow that each Qf these mechanism* mav
require the development of different risk assessment model* for
extrapolation to hum^n, carcinogenic hazard,. Further, some of these
iTOChinlsms mav be shown to be of low or zero relevance to man. The meeting
accepted this as a working hypothesis and then discussed Us Implication if
established experimentally as true.
63
-------
The electrophllic or genotoxic carcinogens are perhaps the easiest to
recognize, and they probably present the greatest individual potential
human hazard. The key points when approaching their definition and
assessing their likely human hazard were discussed as follows:
a) Useful SAR data exist already.
b) Genotoxicity assessment is possible using in vitro and in vivo assays.
c) ONA adducts in vivo can provide useful indications of exposure, but
such data should not be over-interpretated in terms of degree of
carcinogenic initiation.
d) Non-genotoxic effects (toxicitles) such as hormonal changes,
hyperplasia, necrotic effects, etc, Induced by the same chemical may
play a critical role in the eventual carcinogenic outcome of a
bioassay. These may also influence the observed dose-response
relationships, itself affecting risk assessment.
e) Conservative low-dose extrapolation of data may be in order, albeit
such extrapolation may be modified by data indicating
metabol1c/spec1es/DNA-repair (etc) differences, or the critical
involvement of non-genotoxic toxicitles of the chemical.
The current problem faced is that this profile fails to fit an increasing
number
-------
non-genotox1c carcinogens, rather, it was considered necessary to
synthesize a variety of data leading to a confident prediction of a
mechanise Independent of Induced primary DNA damage. Both positive and
negative sources of data are theoretically available for Integration, as
follows:
Negative: a) Non-alerting chemical structure.
b) Conclusion of non-genotoxicity, particularly in vivo.
c) Absence of DNA lesions in vivo.
d) Evidence indicating the potential for bioaccumulation,
ie, a failure to metabolize and excrete the agent
rapidly.
Positive: e) Evidence for an alternative mechanism of carcinogenic
action.
f) Weak or highly specific carcinogenic effect in rodents,
especially if in a tissue with a significant
spontaneous tumour incidence. Factors such as long
latency period and the Induction of benign as opposed
to malignant tumours also figure here.
Other Factors: g) Knowledge of metabolic factors that cannot operate in
man, of carcinogenic dose-levels that could not be
achieved in man, or of pharmacokinetic factors unique
to rodent species.
h) Specific knowledge that the mechanism of carcinogenic
action could not apply to man.
65
-------
When all of these factors are considered it may be possible to classify
animal carcinogens according to the scheme shown In Slide 3. Examples of
chemicals that could possibly fill these separate boxes (Slide 3) are shown
*
beneath each box and are discussed briefly 1n the legend to the slide. Of
course, the majority of animal carcinogens are currently in the centre box,
and only experimental data (in addition to the bioassay data) can lead to a
chemical being upgraded or downgraded in terms of potential human hazard.
Much discussion took place regarding how such studies could be encouraged
by the 'rewarding' of new data by a movement in classification, but the
matter was not resolved. Use of either a sliding scale or sub-sections
within the central box were discussed (1e, near to a presumed human hazard,
or nearer to an assumed rodent specific/high dose carcinogen).
Substantial data were presented during the meeting Indicating that
butylated hydroxyanisole (BHA) 1s probably the best candidate to date for
placing in the category 'not relevant to humans at expected exposure
level'. This was because the rodent forestomach tumour data indicate the
absence of cardnogenicity below -2% in diet. Likewise, the several male
rat kidney-specific carcinogens that lead to retention of aZM-globulin in
the kidney are the nearest to being established as of no relevance to man
•
[chemicals such as Hmonene (lim) and trimethylpentane (IMP)].
At the other extreme, all present at the meeting were agreed that agents
such as dibromochloropropane (DBCP), the flre-retardant TRIS, the several
nitrosamines such as DEN and benzo[a]pyrene can probably be regarded as
likely human carcinogens, given appropriate exposure - even in the absence
of positive human epidemiological data [all examples in this text were used
66
-------
Animal Carcinogens
Human
BZD
VC
etc.
:•:
•: Presumed
:j Human
•:
•:•
I I I I I I I I I I
Not Relevant
Below X ppm
1 1 1 1 1 1 1
Not
Relevant
i i i i i i i
NC
DBCP
DEN
BP
The
Majority
BHA
NaS
NTA
TU
IMP?
LIM?
DHEP?
CAP
etc.
Slide 3. A classification scheme that allows for animal carcinogens to be segregated Into different levels of
potential human hazard. This scheme represents a synthesis of the several such schemes proposed at the
meeting. The assignment of chemicals to particular categories is speculative at this stage. Note how this
classification scheme is based as weight of evidence, not just strength of evidence for individual datasets.
Chemical abbreviations not identified in the legend to Slide 2 are as follows: BZD = benzidine, VC = vinyl
chloride monomer, DBCP = dibromochlorpropane, NTA = nitrilotriacetic acid, TU = thiourea, TMP =
trimethylpentane and CAP = caprolactam.
-------
as Illustrations of principles during the meeting; they should not be
regarded as precedents for the principles being discussed In the absence of
a more detailed review of the literature].
A critical aspect of assigning a tentative mechanism of action to a
carcinogen 1s that 1t can Influence the type of model used for the
extrapolation of the animal cancer data to man. Thus, evidence 1s building
up to Indicate that liver-specific carcinogens that Induce peroxisome
proliferation 1n the liver (agents such as DEHP) will be non-carcinogenic
as dose-levels that do not lead to perturbations 1n I1p1d metabolism in the
liver. Further, the magnitude and type of the hyperplastlc wave that
usually follows peroxisome proliferation can Influence the eventual tumour
Incidence, as Is thought to apply also to the longevity of the test species
and Its spontaneous liver tumour Incidence. Such complex and linked
requirements for cardnogenlcity are most efficiently handled by multi-
stage models, perhaps based on that described by Moolgavkar and his
colleagues. Further, there are some Indications that this particular
mechanism of cardnogenlcity (peroxisome proliferation) may not be relevant
to humans, in which case the risk assessment 1s essentially achieved with
that knowledge (but fact 4 assumption, see earlier).
Risk assessment models were discussed in great detail, and the following
general conclusions were drawn:
a) There is an urgent need to stop talking about modelling in the
abstract, and to start to use them. Data should be published and
. methods compared in an open manner.
68
-------
b) Particular attention should be paid to developing multi-stage models.
c) Attention should be afforded to deriving empirical 'markers' of
cardnogenidty for use In low-dose studies on agents reported to be
carcinogenic at higher dose-levels (eg, Swenberg's use of kidney foci
as markers of TNP-induced kidney cancer in male rats).
d) Attempts should be made to isolate those stages in a multi-stage model
that are directly chemically dependent. This will open the door to
consideration of additive effects (enhancement of spontaneous tumour
.Incidence) and inductive effects (the initiation of novel tumours by
the test agent).
In order that any of the above hopes can be realized, the meeting
recognized that certain disciplines must be exercised by all involved;
three being discussed 1n detail:
1) That terminology should be tightened up. Thus, it is of-little value
to use a loose word such as 'hyperplasia'. If this word 1s considered
relevant to a particular carcinogen or putative carcinogenic mechanism,
the following facts (and probably more) should be clearly identified.
•
a) Was the hyperplasia acute or chronic, focal or general?
b) Was Its observation acute or extended?
c) Was a single measurement or an integrated measurement made?
d) How was hyperplasia monitored -
i) S-phase cells
ii) mitotic figures
iii) tissue weight
69
-------
t) Uhtn used for risk assessment, will hyperplasla in the whole tissue
or for a sub-population of cells be used?
2) What 1s the specificity for cardnogenlcity of the marker event being
used. For example, 1s a2M-globul1n also accumulated In tissues other
than the male rat kidney during exposure to Hmonene, and 1f so, why
are tumours not Induced In those tissues?
3) The Issue of 'scaling dose' requires urgent classification. At present
It 1s used in two ways. First, via surface area of a species, to
derive trans-species dose extrapolation. Second, via pharmacokinetic
results, to rationalize differences in carcinogenic outcome between
species or routes of administration.
In summary, in order to make progress in this highly complex area we need
to agree a common language based on sound scientific observations.
Two final considerations seem to be important. First, It is clear that
molecular studies of proto-oncogene activation should'soon resolve many of
the fundamental aspects of chemical cardnogenlcity. These should impinge
directly on mechanisms of action, and as such they should Influence risk-
assessment models 1n a fundamental manner. Second, the Issues at stake
here are fundamental to a sound policy for the categorization and risk
assessment of chemical carcinogens. As such, progress should be attempted
at a fundamental level and on a broad front. In particular, these
principles should not be evaluated and developed in an adversarial
70
-------
atmosphere with a major chemical of commerce. If that happens, this
critical Issue will become clouded by political factors of no true
relevance to chemical carcinogenidty.
.*
In closing, I would congratulate Or. Hill and his colleagues for organizing
such a stimulating and open-minded meeting. Progress in this area 1s
urgently required, but it can only be based upon sound science.
71
-------
APPENDIX A
EPA RISK ASSESSMENT FORUM TECHNICAL PANEL
AMD ASSOCIATES
A-l
-------
BPA RISK AJ
CARCINOGBM RISK ASSBSSMBHT
FORUM TBCHHICAL PAMBL AMD ASSOCIATES
Richard Hill, William parland, Co-chairs
Donald Barnes
Margaret Chu
Arnold Kuzmack
Peter Preuss
Lorenz Rhomberg
John Moore, Chairman, Risk Assessment Council
Dorothy Patton, Chair, Risk Assessment Forum
Linda Tuxen, Technical Liaison, Risk Assessment Forum
IHVITB0 NORKSHOP PARTICIPANTS:
Qualitative laauea
Roy Albert
David Clayson
Marilyn Fingerhut
Gary Flamm (Chair, WOE)
John Graham
Kim Hopper
Eugene Mcconnell (Chair, ANIMAL DATA)
Colin Park
James Popp
Ellen Silbergeld
Robert Squire
Ray Tennant
Kees Van der Heijden
Quantitative lasueo
Melvin Andersen (Chair, SCALING)
Carl Barrett
Linda Birnbaum
Murray cohn
Robert Dedrick
William Parland (chair, MECHANISMS)
David Gaylor
James Gillette
Daniel Krewski (Chair, BACKGROUND)
Frederica Perera
Christopher Portier
Richard Reitz
Stephen safe
Robert Scheuplein
Thomas Starr
James Swenberg
Curtis Travis
Robert Scheuplein
James Wilson
John Ashby
CONTRACTOR ASSOCIATES:
Kate Schalk, Conference Services, Eastern Research Group
Trisha Hasch, Conference Services, Eastern Research Group
Herbert Page, Scientific Consultant, Eastern Research Group
Elaine Krueger, Environmental Health Research, Eastern Research Group
A-2
-------
APPENDIX B
LIST OP PARTICIPANTS
B-l
-------
O.S.
. PROTOCTIOH AGBMCY
OH CABCIPOCni RISK AS8BSSMBHT
January 11-13, 1989
Virginia Beach, VA
IBVITTO PARTICIPANTS
Roy Albert
Department of Environmental Health
University of Cincinnati - Medical
Center
3223 Eden Avenue
Cincinnati, OH 45267
513-558-5701
Melvin Andersen
AAMRL/TH
Wright Patterson APB
Ohio 45433
513-255-3423
John Ashby
Imperial Chemical Industries, Ltd.
Central Toxicology Laboratory
Alderly Park
Nacclesfield SK10 4TJ
Cheshire, England
011-44-625-512833
Carl Barrett
National institute of Environmental
Health Sciences
P.O. Box 12233
Research Triangle Park, NC 27709
919-541-2992
Linda Birnbaum
National institute of Environmental
Health Science*
South Campus, Bldg. 101, C3-02
TW Alexander Drive
Research Triangle Park, NC 27709
919-541-3583
David B. Clayson
Toxicology Research Division
Bureau of Chemical safety
Pood Directorate
Health Protection Branch
Health and Welfare Canada
Tunney's pasture
Ottawa, Ontario KlA OL2
Canada
Murray S. Conn
Consumer Product Safety Commission
Room 700
5401 Westbard Avenue
Bethesda, MD 20207
301-492-6994
Robert L. Dedrick
Division of Research Services
Biomedical Engineering and
Instrumentation Branch
Building 13, Room 3W13
National institutes of
Health
9000 Rockville Pike
Bethesda, MD 20892
301-496-5771
Marilyn Pingerhut
National Institute of
Occupational Health and
Safety
4676 Columbia Parkway (R-13)
Cincinnati, OH 45226
513-841-4203
Gary Plamm
Chair, Weight of Evidence
Classification Scheme
11760 Indian Ridge Road
Reston, VA 22090
703-620-9249
David Gaylor
National Center for lexicological
Research
Highway 365 North County RD 13
Jefferson, AR 72079
501-541-4000
B-2
-------
James R. Gillette
Laboratory of Chemical Pharmacology
National Heart, Lung, and Blood
institute
Building 10, ROOM 8N117
National institutes of Health
Bethesda, Mo 20892
301-496-2593
John Graham
Harvard School of public Health
677 Huntington Avenue
Boston, MA 02115
617-732-1090
Kim Hooper
California Department of Health
Services
2151 Berkeley Way, Room 504
Berkeley, CA 94704
415-540-2012
Daniel Krewski
Chair, Additivity Independence
to Background
Environmental Health Directorate
Health Protection Branch, Room 115
Health and Welfare Canada
Environmental Health Centre
Tunney's Pasture
Ottawa, Ontario K2A 06Z
Canada
613-954-0163
Gene McConnell
Chair, Confidence in Application of
Animal Tumor Data to Humans
P.O. Box 348-25
Route 6
Raleigh, NC 27613
919-848-1576
Colin Park
Health and Environmental Sciences
Dow Chemical Company
Building 1803
Midland, MI 48674
517-636-1159
Feederica Perera
Division of Environmental Health
Sciences
Columbia School of Public Health
60 Raven Avenue, B-109
New York, NY 10032
212-305-3465
James Popp
Chemical Industry Institute
of Toxicology
P.O. Box 12137
Research Triangle Park, NC 27709
919-541-2070
Christopher Portier
Statistics and Biomathematics
Branch (B3-02)
Division of Biometry and
Risk Assessment
National Institute of
Environmental Health Sciences
P.O. Box 12233
Research Triangle Park, NC 27709
919-541-4999
Richard H. Reitz
Health and Environmental Sciences
Dow Chemical Company
Building 1803
Midland, MI 48674
517-636-5995
Stephen H. Safe
Texas A&M university
Veterinary Physiology and Pharmacology
Agronomy Road at University Drive
Room 107, VMA Building
College Station, TX 77843
409-845-7261
Robert Scheuplein
Pood and Drug Administration
CFSAN/OTS HFP-100, Room 2025
200 C Street, S.W.
Washington, DC 20204
202-245-1372
Ellen Silbergeld
Environmental Defense Fund
1616 P Street, N.W. .
Washington, DC 20036
202-387-3500
Robert A. Squire
Robert A. Squire Associates, Inc.
1515 Labelle Avenue
Ruxton, MD 21204
301-821-0054 or
301-955-3273
B-3
-------
Thomas a. Starr
Chemical industry Institute of
Toxicology
P.O. Box 12137
Research Triangle park, NC 27709
919-541-2070
James Swenberg
Chemical industry institute of
Toxicology
P.O. Box 12137
Research Triangle Park, NC 27709
919-541-2070
Raymond Tennant
National Toxicology Program (MD-B4-02)
Cellular and Genetic Toxicology Branch
National institute of
Environmental Health Sciences
P.O. Box 12233
Research Triangle park, NC 27709
919-541-4141
Curtis Travis
Oak Ridge National Laboratory
P.O. Box 2008
Building 4500 S., MS 109 5204
Bethel Valley Road
Oak Ridge, TN 37831
615-576-2107
Rees Van Der Heijden
Director of Toxicology
National institute of Public Health
and Environmental Protection (RIVM)
P.O. Box 1
3720 BA Bilthoven, The Netherlands
011-31-30-74-2080
James Wilson
Monsanto Company (32NE)
800 N. Lindbergh Blvd.
St. Louis, NO 63167
314-694-8879
HPA PAKKCIPABW
Margaret Chu (RD-689)
Office of Environmental Health
and Assessment
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, B.C. 20460
202-382-7335
William Parland (RD-689)
Office of Health and Environmental
Assessment
Office of Research and Development
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
202-382-7315
Richard Hill
Office of Pesticides and Toxic
Substances (TS-788)
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
202-382-2897
Arnold Kuzmack
Program Development and Evaluation
Division (WH-550B)
Office of Drinking Water
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, DC 20460
202-392-5515
John Moore
Acting Deputy Administrator (A-101)
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
202-382-4711
Dorothy Patton (RD-689)
Risk Assessment Forurn
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
202-475-6743
Lorenz Rhomberg (RD-689)
Office of Health and Environmental
Protection
Office of Research and Development
U.S. Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
202-382-5723
-------
APPENDIX C
LIST OF OBSERVERS
C-l
-------
U.S. ranmtONMBMTAI. PROTECTION AGENCY
WORKSHOP OH CARCIROGBM RISK ASSESSMENT
Virginia Beach, VA
January 11-13, 1989
OBSERVERS
Nary Argus
Office of Toxic substances
U.S. Environmental Protection
Agency (TS-796)
401 N Street, S.W.
Washington, DC 20460
202-382-4250
Karl Baetcke
U.S. Environmental Protection
Agency (TS-796)
401 N Street, S.W.
Washington, DC 20460
202-382-4302
Tim Barry
U.S. Environmental Protection
Agency (PM-220)
401 M Street, S.W.
Washington, DC 20460
202-382-2038
Steve Bayard
U.S. Environmental Protection
Agency (RD-689)
401 M Street, S.W.
Washington, DC 20460
202-382-5722
Diane Beal
Office of Toxic Substances
U.S. Environmental Protection
Agency (TS-796)
401 M Street, S.W.
Washington, DC 20460
202-382-4259
Jerry Blancato
U.S. Environmental Protection
Agency (RD-689)
401 N Street, S.W.
Washington, DC 20460
202-475-8918
John J. Boland
Johns Hopkins university
513 Ames Hall
Baltimore, MD 21213
301-338-7103
W. Ray Brown
Research Pathology Services, Inc.
438 B. Butler Avenue
New Britain, PA 18901
215-345-7070
Richard Brunker
U.S. Environmental Protection
Agency (3HW26)
841 Chestnut Street
Philadelphia, PA 19107
215-597-9800
Betsy Carlton
National Sanitation Foundation
P.O. 1468
Ann Arbor, MI 48106
313-769-8010
Peter Chan
Rohm and Haas
Toxicology Department
727 Norristown Road
Spring House, PA 19477
215-641-7489
Gail Charnley
Meta Systems, Inc.
Suite 700
370 L1 Enfant Promenade
Washington, DC 20024
202-488-7494
C-2
-------
Chao Chen
O.S. Environmental Protection
Agency (RD-689)
401 M Street, S.W.
Washington, DC 20460
202-382-5719
Jerome Cole
International Lead Zinc Research
Organization
2525 Meridian Parkway
P.O. Box 12036
Research Triangle Park, NC 27709
919-367-4647
Marion Copley
U.S. Environmental Protection
Agency (TS-769C)
401 M Street, S.W.
Washington, DC 20460
202-357-7491
Thomas A. Cortina
Halogenated Solvents Industry
Alliance (HSIA)
1225-19th Street, N.W. Suite 300
Washington, DC 20036
202-223-5890
Ila Cote
OAQPS (MD-12)
U.S. Environmental Protection
Agency
Research Triangle Park, NC 27711
919-541-5342
Richard A. Davis
American Cyanamid Company
Toxicology & Product Safety
Department
1 Cyanamid Plaza
Wayne, NJ 07470
201-831-4004
Arnold Den
Office of Regional Administration
U.S. Environmental Protection
Agency
215 Fremont Street
San Francisco, CA 94105
415-454-0906
Reto Engler
U.S. Environmental Protection
Agency (TS-769C)
401 M Street, S.W.
Washington, DC 20460
703-557-7491
Dr. John L. Festa
American paper institute
1250 Connecticut Ave., N.W.
Washington, D.C. 20036
202-463-2587
Bob Fegley
U.S. Environmental Protection
Agency (PM-221)
401 M Street, S.W.
Washington, DC 20460
202-382-5490
D. Douglas Fratz
Chemical Specialties Manufacturers
Association
Suite 1120
1001 Connecticut Avenue, N.W.
Washington, DC 20036
202-872-8110
Tom Fuhremann
Monsanto company
800 N. Lindberg Boulevard C2SE
St. Louis, MO 63167
314-694-8819
Janice Garrett
Oak Ridge National Lab
Building 4500S, MS 109 5204
P.O. Box 2008
Oak Ridge, TN 37831
615-576-2107
Paul J. Garvin
Amoco Corp.
MC4901
200 B. Randolph Drive
Chicago, IL 60601
312-856-5991
John G. Hadley
Owens corning Fiber Glass Corp.
Route 16, P.O. Box 415
Granville, OH 43023
614-587-7228
C-3
-------
Carol Henry
Risk science institute
1126 16th Street, N.W.
Washington, DC 20036
202-659-3306
Leslie King
Chemical Manufacturers Association
2501 M Street, N.W.
Washington, DC 20037
202-887-1323
Sara Henry
Pood & Drug Administration-*
HPP156
200 C Street, S.W.
Washington, DC 20204
202-472-5707
Charalingayya Hiremath
U.S. Environmental Protection
Agency (RD-689)
401 N street, S.W.
Washington, DC 20460
202-382-5725
Cheryl Hogue
Chemical Regulation Reporter
1231 25th Street, N.W.
Washington, DC 20037
202-452-4584
Donald Hughes
Procter & Gamble
ivorydale Tech Center
Cincinnati, OH 45217
513-627-5254
Albert Hugh
Texas Eastern Corp.
P.O. Box 2521
Houston, TX 77252
713-759-7203
Edwin Jeszenka
Tennessee Eastman Chemicals
Division
Eastman Kodak Company
P.O. Box 511
Kingsport, TN 37662
615-229-4884
Renata Kimbrough
Office of Regional Operations
U.S. Environmental Protection
Agency (A-101)
401 M Street, S.W.
Washington, D.C. 20460
202-382-4727
Neil Krivanek
Dupont
P.O. Box 50 Elton Rd.
Newark, NJ 19714
302-366-5274
Lisa Y. Lefferts
Center for Science in the
Public Interest
1501 16th St., NW
Washington, D.C. 20010
202-328-7334
Steve Lewis
Exxon Biomedical Sciences
Nettlers Road - CN 2350
East Millstone, NJ 08875-2350
201-873-6063
Bertram Litt
Litt Associates
3612 Veazey Street, N.W.
Washington, DC 20008
202-686-0191
Ronald Lorentzen
Pood & Drug Administration
200 C Street, SW
Washington, DC 20204
202-485-0046
Donald Lynam
Ethyl Corporation
451 Florida Street
Baton Rouge, LA 70801
504-388-8008
William Lynch
Rohm & Haas
Toxicology Department
727 Norristown Road
Spring House, PA 19477
215-641-7489
-------
Robert NcGaughy
U.S. EnvirotuMntal Protection
Agency (RD-689)
401 N Street, S.W.
Washington, DC 20460
202-382-5889
John C. Niddleton
U.S. Borax Company
412 Crescent Way
Anaheim, CA 92801
714-490-6000
Robert Moolenaar
American Industrial Health Council
1330 Connecticut N.W., suite 300
Washington, DC 20036
202-659-0060
Steve Nesnow
Health Effects Research
Laboratory (MD-68)
U.S. Environmental Protection
Agency
Research Triangle Park, NC 27711
919-541-3847
Jerome Puskin
U.S. Environmental Protection
Agency (ANR46)
401 M Street
Washington, DC 20460
202-475-9640
Robert D. Putnam
Putnam Environmental services
P.O. Box 12763
2525 Meridian Parkway
Research Triangle Park,
N.C. 27709
919-361-4657
Gerhard Raabe
Mobil Corporation
Corporate Medical Department
150 East 42nd Street
New York, NY 10017
212-883-5368
Reva Rubenstein
U.S. Environmental Protection
Agency (WH-562B)
401 M Street, S.W.
Washington, DC 20460
202-382-4798
David sandier
American industrial Health Council
1330 Connecticut N.W., Suite 300
Washington, DC 20036
202-659-0060
Mitchell sauerhoff
CIGA-GBIGY
400 Parmington Road
Farmington, CT 06032
203-674-6338
Robert Scala
Exxon Biomedical Sciences
Mettlers Road-CN2350
East Millstone, NJ 08875-2350
201-873-6061
Rita Schoeny
Environmental criteria and
Assessment Office
U.S. Environmental Protection
Agency
26 W. Martin Luther King Drive
Cincinnati, OH 45268
513-569-7814
Cheryl Siegel Scott
U.S. Environmental Protection
Agency (TS-798)
401 M Street, S.W.
Washington, DC 20460
202-382-2282
Robert Sielken, Jr.
Sielken, Inc.
Suite 410
3833 Texas Avenue
Bryan, TX 77802
409-846-5175
C-5
-------
Peter sherectz
Virginia Department of Health
Bureau of Toxic Substances
109 Governor Street
Room 918
Richmond, VA 23219
804-786-1763
Helen Shu
Syntex Corp.
3401 Hillview Aver.ue
Palo Alto, CA 94304
415-852-1184
Donna Sivvlka
Nickel Producers Environmental
Research Association (NIPERA)
Alston Tech Park, Suite 104
100 Capitola Drive
Durham, NC 27713
919-544-7722
Dwight Smith
.Covington & Burling
1201 Pennsylvania Ave., MW
P.O. Box 7566
Washington, D.C. 20007
202-662-5528
Jerome Smith
Lead industries Associates
292 Madison Avenue
New York, NY 10017
212-578-4750
Jerry Smith
Rohm and Haas Company
Independence Nail West
Philadelphia, PA 18105
215-592-2986
Roy Smith
U.S. Environmental Protection
Agency (3HW16)
841 Chestnut Street
Philadelphia, PA 19107
215-597-6682
E.J. Sowinski
ICI Americas, inc.
Concord Pike & Murphy Road
Wilmington, DE 19897
302-575-4503
Hugh Spitzer
American Petroleum institute
1220 L Street, N.w.
Washington, DC 20005
202-682-8338
Donald Stevenson
American industrial Health council
1330 Connecticut N.W., suite 300
Washington, DC 20036
202-659-0060
Michael Terraso
Texas Eastern Corp.
P.O. Box 2521
Houston, TX 77252
713-759-7203
Rosalind Volpe
international Lead Zinc Research
Organization
2525 Meridian Parkway
P.O. Box 12036
Research Triangle Park, NC 27709
919-367-4647
Michael Waters
U.S. Environmental Protection
Agency
Health Effects Research Laboratory
Research Triangle Park, NC 27711
919-541-2537
Michael Watson
U.S. Environmental Protection
Agency
1200 6th Avenue
Seattle, WA 98101
206-442-1072
Larry Wetzel
CIBA-GEIGY
Agricultural Division
P.O. Box 18300
Greensboro, NC 27419-8300
919-292-7100
John Yoder
Lead industries Associates
292 Madison Avenue
New York, NY 10017
212-578-4750
C-6
-------
Lauren zeise
California Department of Health
Services (RCHAS)
2151 Berkeley Way, Annex 2
Berkeley, CA 94704
415-540-3221
C-7
-------
APPENDIX D
INTRODUCTORY PLENARY SESSION1
'This section includes the text and/or summary of four presentations given
during the Introductory Plenary Session on Wednesday, January 11. Because EPA
had not asked the speakers to prepare formal papers, the following texts are
based on tape'recordings of each speaker's presentation, edited for clarity.
Each speaker has reviewed and approved the material presented here.
D-l
-------
INTRODUCTORY PLENARY SESSION
John A. Moore. An EPA View
In 1986, some 3-odd years ago, EPA published in final form its
guidelines for carcinogen risk assessment. However, as I think
almost everybody in this room knows, the Agency began using them in
1984 while it was drawing comment on its draft guidelines. I think
we made it clear then, and if we didn't, we'll make it clear today,
that we see those guidelines as needing to have a use in all seasons,
both at the time they were published, as well as into the future. In
order for them to maintain some degree of utility, it's incumbent
upon us to make sure that they reflect advances or current beliefs as
they relate to scientific knowledge, as well as to take stock of the
experience that we've gained from their use these last several years.
I think on net the Agency feels that the guidelines have served them
reasonably well. They are in the main, we believe, reasonable
principles for the consistent review of carcinogenic data.
At the same time, I think there is within the Agency a consensus that
they can and probably should be improved. Improvement can occur in
two broad ways. One can make basic changes in the guidelines
themselves, and/or one can make changes in the implementation
practices within the framework of the guidelines as they currently
exist. A number of months ago, a technical group within the Agency
reviewed the guidelines with an eye towards identifying areas that
may be ripe for more intensive investigation. They recommended that
D-2
-------
wholesale revision of the guidelines probably was not appropriate.
They also identified, however, two broad topics that merited further
scrutiny. First, the weight-of-the-evidence process that's used for
evaluating whether or not a chemical process may pose a carcinogenic
risk to man and* secondly, the issues associated with quantitative
dose response assessment.
There were four activities that were planned and have been pursued.
First, we solicited Agency-wide the various offices at headquarters,
as well as regions, for comments on what they thought of the
guidelines. Next, we conducted a limited number of interviews of
certain people within the Agency and also outside of the Agency; for
example, people who had served in our science advisory groups and who
had reviewed products of the use of the cancer guidelines, as well as
State government people. These guidelines, although they're
developed to assist the Agency in getting on with its business,
certainly end up having implications far beyond the borders of EPA.
We're keenly aware of that and would like to make sure that we
understand how.they're being used outside of the Agency, with the
idea of their either being used properly from our perspective or
possibly being misused. Not that there's necessarily much we can do
about it, but we certainly should be aware of it. Thirdly, we
identified through the Federal Register the fact that we were
considering guideline review, and fourthly, we began planning for the
workshop that begins today.
The first three of those activities have been completed. Upon
completion of this workshop, we will synthesize what we think we've
heard and plan some course of action. A preliminary evaluation of
the first three things identified three themes.
(1) Some people felt chat the guidelines in their current application
were just fine; we don't need to change a thing. If you have to
change something, tinker with it on the margin at best; for instance,
D-3
-------
maybe change some language to reflect the changes in language that
IARC guidelines had undergone since the Agency's guidelines had come
out.
(2) A number of people said the guidelines are fine. The problems
are with how you" people use the guidelines. In this case, the claim
is that we don't always use the breadth or totality of the
information that's available on a given chemical, even though the
guidelines clearly allow that and maybe even suggest that that's what
should be done on a case-by-case basis. They assert that we too
often are rote, if you will, in our application of default positions
and, indeed, maybe need to analyze a little bit more in some areas.
An example that somebody might give is that we always seem to pick
the most sensitive indicator, as opposed to the most appropriate
indicator in that data set. They make a distinction between the two.
Why do you always do this? Why don't you show a little bit more
selectivity?
(3) Finally, there are those that are on the opposite end of the
spectrua, in that no matter how you try to tinker with the
guidelines, they really require a reanalysis. Two main areas were
commonly cited that need to be looked at in order to make an
improvement on the current product. One is this identification of
human carcinogenic hazards, while the other pertains to quantitative
questions associated with a risk extrapolation. Obviously the
workshop that we're headed into here tries to draw upon these two
aspects. What is the process that you need to go through to make
that initial qualitative determination? How do you express it?
Secondly, how do you get into some quantitative expression of the
risk?
We identified under those two broad areas five topics that we think
are of priority interest, based on our own beliefs and experiences,
as well as comments that were received. They are reflected in the
-------
five breakout activities for the meeting. Let's go through the two
qualitative topics.
There are obviously a number of issues surrounding the question of
whether or not * chemical poses a carcinogenic hazard to humans.
Questions have arisen about the current classification scheme built
into the guidelines. For example, our Science Advisory Board more
than once has pointed out the difficulty in interpreting experimental
data and placing it along that continuum between B2 and C. I don't
get many comments and letters talking about the A or the E part of
the guidelines. They all focus .around B2, C; B2-C. This range is
broad, and everyone's focusing right there. Are the guidelines truly
doing everything that they should be doing? Is the problem inherent
in the classification scheme or is it a flaw in how we use the
classification scheme? I think we need to take a look at that.
What is the relevance in the finding of tumors in animals and of
other experimental data to human carcinogenicity? We're asking this
group to hopefully consider the bases for judging the relevance of
the experimental data. What can you do and what can't you do? How
far can you go and still be basically following a logical application
of science, as opposed to doing something else that represents a
decision, but not necessarily the logical step or a comfortable step
based on where we are with existing knowledge.
The guidelines state, and the Agency believes, that positive animal
studies are presumptive evidence of potential human carcinogenic
hazard. But at the same time, I don't believe the presumption is
infallible. I think the challenge is to better capture the fervor of
our belief associated with any particular chemical. Can we, or as
some people say, must we admit that in certain cases animal tumors or
experimental findings may be of less relevance to humans? How do, or
should, we distinguish between a ore species' response and a multiple
species' response? How do we do a better job if we/ve got two
D-5
-------
responses, rat and mouse, and the nature of the second tumor response
is different from a histogenic standpoint? We don't express that
very well. Can we do better?
Certainly, concordance across species is used to strengthen one's
,<
belief a chemical is a human carcinogen. If it's positive in
elephants, giraffes, rats, and mice and they all produce the same
type of tumors, similar dose-response, etc., we're sure it is one.
We all use it that way, I think, fairly effectively. But how about
trying the case when the data don't quite fit together. We don't do
it very well or don't agree on how to do it, and when we're done, I
don't think we communicate these cases very well.
Let me get out a pet peeve. The fact that dimethyInitrosamine is a
flaming genotoxic agent and a carcinogen in dozens of species
probably doesn't have a lot of relevance to what we do in our day-to-
day practice as a regulatory agency. I don't see too many DEN's. I
think the things that we struggle with are the paradichlorobenzenes
or the diethylhexophthalates or the ethylene thioureas.
Paradichlorobenzene: mouse liver tumors by gavage; male rat kidney
tumors; two species. What do you do with it? DEHP is nongenotoxic
but produces clearly reproducible effects in bioassays. What does
that mean? Ethylene thiourea produces thyroid tumors. What does one
do with it, in that these tumors might be secondary to a disruption
in physiologic responses that are hormonally derived. These are the
types of things that the Agency struggles with, not the boring
negatives or the flaming positives.
You've got to assemble all experimental and other information into a
weight-of-evidence expression. 1 think we've got to realize by
weight-of-evidence statement, we're not simply trying to amass the
evidence in support of carcinogenicity, but to assess thoroughly in
that expression whatever may be the overall implications of the data
D-6
-------
for humans. And whatever is done, how do you get that across in
something less than a 32-chapter book?
Let's now go to quantitative issues. Agency experience and reviewed
comments suggest that we need to evaluate carefully several topics
relative to quantification of risk. For example, the guidelines give
guidance on interspecies scaling in the absence of any relevant
information. They implied at the time they were written, that if one
had some pharmacokinetic data, it probably would help to bridge the
gap. Let's take methylene chloride, in that we have that
information. Everybody said, okay, now, let's do it! Then came the
discussions, and we found we don't have agreement on how or what to
do. Do you apply it only across human doses? Do you scale doses
across species. How does one account for possible differences in
pharmacodynamics? Hopefully, we can get some sense out of that
session at the workshop.
Another question is when to quantitate cancer risk, and some claim we
often seem to be oblivious to the qualitative data. Should we always
proceed on the same path? In one case, you might have a strong,
robust data set that makes everybody comfortable. Both sexes, both
species, dose-response, etc. In another case, you have little
evidence of careinogenieity; you believe it's reproducible, but the
response is fairly modest; one sex, one site. Yet we treat the two
cases the same because the paradigm we use does not consider the
qualitative evidence that goes into it. Should we always take the
same route or should we be doing something different? If you forget
what the Agency says in response to that question, look at what the
Agency does and it probably gives a clearer insight as to what it's
all about. I think the Agency struggles with the fact of having to
do the same thing all the time or feeling compelled that it's got
little choice but to do the same thing at all times.
D-7
-------
You've got two parts of the Agency, the Office of Water, as well as
the Pesticide Program, that often say, given a C categorization,
we're not going to do a quantitative cancer risk estimate. Why are
they doing that? I think part of the reason is they don't believe
the outcome if quantitative procedures are rotely applied. There's a
glimmer there of something maybe one should look at. It's fine to
say you're not going to extrapolate risk, but the next question is,
okay, what are you going to do? After all, there certainly was some
type of positive response; it wasn't negative.
Maybe this will be the last meeting that we'll have to discuss
scaling doses across species. What do we do and how do we do it?
Can we finally come to an agreement on what needs to be done rather
than continue to say, well, there are two ways we do it and both make
sense. Frequently, we argue both sides and then go different ways.
The public sits there and wonders what's going on. Can we do better?
I think there's been a more serious insight on that issue in the last
year that might allow us to move off the dime a little bit one way or
another. Hopefully, you'll look at that.
In what way can biological data, bearing on potential carcinogenic
mechanisms, be used in selecting the means for extrapolating from
high doses to low doses. We don't mean something different
necessarily. We.can accept some alternative if it's clearly better,
but difference for difference sake isn't going to necessarily get us
anywhere. I'm uncomfortable about always using the upper 95%
confidence limit on the linear component as a means of expressing
what it is that might happen. Isn't there something else that could
be included as part of the equation, like central tendency, or MLE?
I don't know what it is, but expressing the worst all the time, which
is what we do, doesn't give the true picture. How can we do it
better?
D-8
-------
Lastly, what are the consequences of whether a chemical induces
tumors by actions that are similar to or different from those
accounting for background tumors. Do you treat a mouse liver tumor
or a mouse lung tumor response differently than some other type which
is rare? If the*answer is yes, what's the decision logic that goes
along with doing that rather than having it be a black box that comes
out with any particular chemical?
Believe it or not, we assembled the participants of this workshop in
the belief that they'd be able to work on the five topics that we
identified. They have something special to lend to those five
topics. With that in mind, we urge you to stay on those five topics.
I think there's certainly a fair amount of grist for the mill in
those five topics. It isn't as if you won't have something to do.
Identify those issues that you think might be right for some
intensive evaluation or re-evaluation now, as contrasted to those
that need further work and might be ripe for consideration 3 or 5
years from now or 10 years from now. For those that are amenable to
possible change, identify the range of options. I'm not sure a
consensus will jump out on all of these r'ssues, so what we really
want is a very good casting of the pros and the cons associated with
those. Let's be candid as to what speaks for it and what speaks
against it. On the other hand, if consensus doesn't emerge, don't
run away from the issue.
The other thing, let's not discuss some pending regulatory decision
in the guise of this meeting, whatever it may be. Let's check our
baggage at the door. We all carry it, we all have our biases. I
gave you a couple hints of mine in this talk. Let's try to make it
as open as possible. Remember, the Agency is at the beginning of a
fairly long path deciding what should be revised and how to do it.
If there is revision, chare will be ample opportunity to comment on
D-9
-------
and
not
further massage whatever those changes may be. This meeting is
going to be the basis for coming up with something.
My last comment would be that, as you'll hear this morning and I hope
continue to hear throughout the conference, there is more than one
way of doing some'thing. The way the Agency's cancer risk assessmencs
are being done in some instances with certain types of chemicals is
different from the way others are doing it. Cognizant of the fact
that these differences exist, I'm not at all happy that the
differences do exist. I think this world get? smaller and smaller
and I think we always need .to see if one can reconcile differences
rather than continue to watch differences grow one way or another.
To the degree that there are differences, the key thing is at least
to try to understand why the differences exist. Then maybs from that
point you might try to reconcile some of these differences. For
instance, certainly the B2-C approach is addressed differently
internationally than is currently done by EPA, or possibly elsewhere
in this country.
Thank you.
D-10
-------
Frederica Perera, A Public Health View
I'd like to start by giving .a broadbrush context to this whole
discussion. The context of this workshop is that we have a very
major problem in this country today. That is to say, there are
460,000 deaths each year in the U.S. from cancer and the great
majority of these are preventable. Preventable because they're
attributable to environmental factors; not only lifestyle factors,
(such as smoking and diet) which involve voluntary behavioral
choices, but involuntary exposures to industrial and manmade
carcinogens in the workplace, in the ambient air, the drinking water,
and the food supply. While it's really not possible to estimate the
exact contribution of any one of these to the total burden of human
cancer, I think we all recognize that exposure to these industrial or
manmade carcinogens is pervasive and significant. This is the
special province and responsibility of EPA, and, we need to deal more
effectively with these problems.
Therefore, I would suggest that the standard for review for these
guidelines is whether they provide a workable, scientifically
supportable system for the timely assessment of carcinogens and for
their regulation in order to prevent cancer. That's my jumping-off
spot.
Turning to the 1986 guidelines, as I discussed in my written comments
to EPA, I think that the basic principles are justified. These
principles are the reliance on experimental data in the absence of
epidemiology and the general assumption that carcinogenesis is a non-
threshold phenomenon. In part, this latter assumption is derived
from the inability to define population thresholds.
D-ll
-------
I had specific comments and criticisms about certain aspects of the
'86 guidelines, including the classification scheme. I'm concerned
about the fact that this scheme could easily become a kind of pigeon-
holing system with automatic regulatory implications that are not
biologically founded. For example, category C, or "possible
.»
carcinogens" includes certain chemicals with at least one clear
positive test result, such as vinylidene dichloride, styrene,
lindane, para-dichlorobenzene, all significant human exposure.
Certain offices in EPA have proposed to regulate those chemicals
because they fell into the C rather than the B ("probable") category.
I think there is no biological basis for this approach. I would
recommend that the revised guidelines state clearly that a chemical
with a clear positive bioassay result and significant human exposure
will be a candidate for regulation as a carcinogen.
Moving on now to the second area that I think needs strengthening;
that is the area of the criteria for reviewing pharmacokinetic-based
risk assessments and mechanistic models. Of course, the guidelines
now state that EPA will consider such data, but as Jack Moore said,
they're not sure how to do it. I think there's a need for consistent
minimum criteria to be met by each model and assessment that is
presented to the Agency for its consideration.
On my next slide, I've listed the type of information one would
ideally want to have. This is not a trivial research agenda here, as
many of you know. The list includes: identity of the active
species' or critical metabolite for carcinogenesis; an understanding
of whether pharmacokinetic processes arŁ linear or saturable; and
whether the carcinogen itself can affect or modulate any of these
processes. (We know, for example, that formaldehyde can induce cell
proliferation, bind to DNA, cause mutation, and inhibit DNA repair.)
Also, one would want to identify chemical interactions that are
likely in the human exposure situation, as well as differences
between chronic and acute exposure; whether there are interspecies
D-12
-------
differences; and, very importantly, the range of interindividual
variation in the human population. I will be discussing this last
point in a moment. To summarize, proposed models should be
accompanied by a clear discussion of uncertainties and assumptions in
the model and by results of sensitivity analysis.
Similarly, with mechanistic models or theories, there should be a
requirement for adverse supporting evidence. For example, in the
last 5 or 10 years, we have repeacedly seen proposals to split up the
world of chemical carcinogens into those which directly damage DNA,
the genotoxic carcinogens, and those which appear to act by other
mechanisms. However, it just isn't that simple: there is no bright
line between these two groups. Indeed, data from the EPA GeneTox
program, from Dr. Mike Waters and his colleagues, show that many so-
called classical nongenotoxic carcinogens, are indeed positive in
several, not just one, but several short-term tests for genetic
toxicity. These include DOT, asbestos, diethylstilbestrol,
chloroform, trichlorethylene, perchlorethylene, phenobarbital, ethyl
alcohol, and sodium saccharin. In addition, recently, researchers at
NIHS, Drs. Reynolds and Marshall, have shown that two so-called non-
genotoxic liver carcinogens, furan and furfural, in fact, induce a
novel mutation in rat oncogene and liver tumors.
Let us turn to the central question here: "Is low-dose linearity a
valid assumption?" Low-dose linearity follows from either the notion
that individuals are not homogeneous in their responses or that the
effect of any single carcinogen of concern has the ability to add on
to the effect of ongoing processes and background exposures.
Let's examine whether there are new data that shed light on this
question. Results of human studies on metabolism, DNA binding, and
DNA repair show that indeed in humans there is considerable
variability between individuals in terms of the metabolism of
aromatic amines, polycyclic aromatic hydrocarbons, and other
D-13
-------
environmental chemicals. For example, there is a range of 3 to 160
for specific aspects of human metabolism. In vitro studies with the
carcinogens benzo(a)pyrene, aflatoxin Bl, and dimethylnitrosamine
show a 150- to 200-fold variation in DNA binding. With respect to
enzymes involved in DNA repair, human studies show a 2- to 65-fold
range.
Indeed, human studies also suggest very wide variation. We have
recently studied a group of foundry workers with exposure to
polycyclic aromatic hydrocarbons and have seen in the exposed group a
range in adduct levels 0 to 2.8 femtomoles. of adduct per microgram of
DNA. Similarly, in smokers of about a pack and a half a day, DNA
adduct levels differed widely. The same variability was seen by
other researchers in a study of 0«-methylguanosine adducts in
patients from China, presumably exposed to dietary nitrosamines.
Łia-platinum is a chemotherapy agent that is given to patients in
treatment for epithelial cancers. The doses are Jtandardized to body
surface area. Here, too, researchers have seen wide variation, as
was true for ethylene oxide-hemoglobin and 4-aminobiphenyl-
hemoglobin adducts.
On the question of background levels, in each of the studies just
discussed, with the exception of cis-platinum. the so-called
unexposed or control group also had a mean adduct level significantly
greater than 0 and again there was a range in results. While these
phenomena certainly need to be investigated further, this information
is supportive of the assumption of low-dose linearity.
Finally, I want to turn to an issue which is not adequately addressed
in the guidelines. One often hears that the linearized multistage
model is a very conservative model and is always going to give an
upper bound estimate of the risk. But neither this model nor other
models that are available are fully addressing the problem that there
are certain segments of the population that are likely to be more
D-14
-------
susceptible to the effects of carcinogenic exposures. These include
the young, people with pre-existing disease, and the elderly, for
example. Of great concern are young children, of whom in this
country today we have 18 million between the ages of 1 and S years,
and 22 million,.,0 to 5 years. At this age, children can be assumed
to be more vulnerable to the effects of environmental carcinogens
than adults. First of all, they have a greater Intake on a kilogram-
of-body-weight basis of drinking water (>3.5 times); and a two- to
six-fold greater consumption of food. Thus, their intake of
carcinogens in these media is correspondingly greater than for
adults.
Secondly, certain physiological factors can increase their
susceptibility to exposure: greater retention of dose, decreased
detoxification, less effective DNA repair systems, higher rate of
proliferation, and immature immune systems. Finally, this young
population has a longer future lifetime over which cancer can develop
as a result of their early exposures. In other words, their future
lifetimes will exceed the latency of cancer. For these three
reasons, EPA should routinely do risk assessments for children as a
separate population.
D-15
-------
James Wilson, An Industry View
The job change is still in the future by a month...I've been
afflicted with Potomac fever, I guess. My assignment is the industry
perspective, and I'm a little hesitant to advertise what I have to
say as representative of the views of industry, whatever industry may
mean. Even among my colleagues at Monsanto, there's a healthy
difference of opinion on a number of the issues that we will talk
about. Among my colleagues active at the American Industrial Health
Council, there are perhaps even larger, or at least equal differences
of opinion, and outside of these groups, there exists an even greater
diversity of views. So what you'll get this morning is a personal
perspective strongly colored by my 20-odd years of experience as a
research chemist in the chemical industry. The question is, is it
appropriate now to consider some changes in the way the carcinogen
risk assessment guidelines are drawn?
Let me first talk about the purpose of these and other such
guidelines. Recall that they're intended primarily for use by
professionals, frequently fairly inexperienced professionals.
Professionals, whether they're toxicologists, physicians, architects,
engineers, or whatever, share a few key traits, and for us this
morning, the most important of these is their orientation towards
solving particular problems. Each sick person to a doctor, each
bridge to a structural engineer, presents a unique case, with unique
problems. The problem facing the professional is usually fairly well
defined: to make the sick person well, to design a bridge to cross a
particular river at a particular point. In developing a solution co
each of these particular problems, the professional relies on both
general and specific knowledge, and upon judgment built up over time
as different problems are addressed and solved. Professionals
necessarily have to rely on an incomplete base of information for
solving the problems. Whenever possible, they rely on scientific
D-16
-------
information because they, like the rest of us, find such information
reliable. But when that's not available, they have to make do with
what's at hand. Each profession develops its own rules of thumb,
lessons that are learned over years by practitioners and passed on •
from one to another, that help them in this process of solving
particular problems. Much of what can be found in the carcinogen
risk assessment guidelines turns out to be this kind of preset,
codified rule. Most of these come from toxicology, and have evolved
in response to the profession's need to set adequately safe limits
for exposure to toxic substances.
Toxicology is a relatively young profession, one whose underlying
science buse is not so well developed, perhaps, as those of
engineering and medicine. Some of its rules of thumb have a basis in
science, but they're purely practical. One example is the use of the
most sensitive species or test as a basis for setting standards.
This can be understood in terms of the demands on the profession, but
is patently inconsistent with the way scientists analyze data. Host
people believe that science-based procedures give better results than
those based solely on practice. They do this, at least in part,
because we now recognize science as a process for developing reliable
knowledge. To borrow a phrase from the English physicist John Zyman,
who published a little book on the subject about 10 years ago,
"science provides the best basis for making decisions in a material
world, when relevant scientific information is available." This
being the case, it's obviously in the interest of the Agency to adopt
and use the most scientific methods that it can, and in the interest
of all of us for the Agency to do so. In my opinion, in the roughly
5 years that have elapsed since serious work was done on these
guidelines, there have been significant changes in the science
underlying some parts of the guidelines. I suggest that these parts
should be revisited.
D-17
-------
One chlng chat has become abundantly clear in the last few years is
the importance of mitotic rate to cancer risk. Scientists in the
field of biochemical genetics recognized more than 2 decades ago that
mutations essentially don't occur unless a cell with damaged DMA
undergoes division. The implications of this for carcinogenicists
j
were not recognized by scientists in that field until about a decade
ago, when, independently, two groups of scientists, Moolgavkar and
Knudson, and Greenfield, Ellwein, and Cohen, found they had to take
the age distribution and mitotic count to explain the quantitative
incidence of cancer. Todd Thorslund, then at the Agency, was one of
the first people in risk assessment to recognize the potential
importance of this finding of the Moolgavkar-Knudson theory to cancer
risk assessment, and he started working on his ideas sometime about
1985. We are still working on the implications this theory poses for
cancer risk assessment, and how to take them into account. One of
the critical implications is that any treatment that increases
mitotic rate over background also increases risk. This means, among
other things, that those people who are larger are more highly at
risk than those who are smaller, as a recent paper by Albanes pointed
out. More cells means more mitoses over a lifetime, and thus a
greater chance that a cell suffering a mutation in a critical locus
will occur. Both overfed people and overfed rats have higher cancer
rates.
Now, more importantly for our purposes here today, is that data on
dose response of mitotic rate assumed a much larger importance than
any of us had recognized in the past. Results coming from Cohen and
Ellwein's work suggest that acutely elevated mitotic rate and
treatment in which a burst of increased cell division occurs, chang
risk only to a very small degree. They found further that the risk
.from chronic elevation is very nonlinear. The guidelines need to
recognize the fact that this data is important and that it will be
becoming available over the next few years.
D-18
-------
Another implication of the theory, which Moolgavkar has forcefully
pointed out recently, is that the approximation used by Crump and his
coworkers in devising the linearized multistage extrapolation
procedure fails at high incidence. That is, it fails under precisely
the conditions where it is most likely to be employed by the Agency
and others, where the tumor incidence is something over .3, or over
30 percent. Under these conditions, the separate contributions of
mitogenic cell division stimulation and mutagenic direct attack on
DNA begin to interact in a synergistic fasnion. A hockey stick-
shaped dose response curve is then the result.
Under these conditions, even with a strongly genotoxic agent such as
benzopyrene or 2-acetylaminofluorene, the dose response is not
necessarily linear, as the work of Gaylor and his colleagues on the
ED-01 study have pointed out. Another indication of that is in a
paper by Zeise and Crouch on benzopyrene in the rat forestomach.
Thus, the linearized multistage procedure can be claimed to be a
plausible upper bound of the risk only if the mitogenic contribution
to the incidence can be distinguished. The guidelines need to take
that into account.
The discussion of interspecies extrapolation has generated much more
heat than light over the past few years. It's time now for this to
be addressed. Within the last few years, some real scientific
understanding of this process has begun to emerge. Personally, I am
convinced by the work of Travis and his associates that the default
scaling procedure should be that of the three-quarters power of body
weight, but not everyone agrees. More important is that our
understanding of the factors that control dose scaling have been
expanding much more rapidly. For instance, the difference of opinion
between the Dow Chemical Company and the EPA over low-exposure
extrapolation on the effects of dichloromethane may turn on whether a
detoxification product, such as a formaldehyde derivative from the
glutathione pathway, is formed spontaneously, by a direct chemical
D-19
-------
reaction, or by an enzymatic reaction of some kind. That kind of
point becomes the determining factor in how an extrapolation is to be
performed. We've gone beyond what is consistent with the default
procedures that are called for in the guidelines. The guidelines
need to recognize that more sophisticated information can be, and is
being obtained.
Another new development is the accumulation of data that allow us to
draw conclusions about the ability of animal models to predict human
cancer hazard. Some consensus now exists on predictability in a
number of cases. For example, the rat bladder is known not to be a
good model for human cancer hazard, concerning exposure to
aminobiphenyl or other polynuclear aromatic amines, most likely due
to a pharmocokinetic difference between man and the rat. On the
other hand, recent human evidence confirms that inert implants that
cause chronic inflammatory response increase the chance that a tumor
will form at the site of inflammation. Further, it's very likely
that a repeated injection of any substance into the same spot in a
human will cause a tumor at that spot, from the same mechanism as
that of a solid-state carcinogen. However, such tumors by themselves
in animals are not predictive of a human cancer hazard for the
material that's injected. The physiology of the rat urinary tract,
especially in the young male rat, is different enough from that of
humans that substances that induce bladder calculi only under
conditions found in rat urine are not predictive of human cancer
hazard. The sane thing can be said of the nephrotoxins that complex
with Qtjuglobin and retard its decomposition leading to the sequence
of events described by Swenberg and his coworkers. Even though the
evidence of carcinogenic response to inhaled polynuclear aromatic
hydrocarbons is surprisingly sparse, nearly everyone believes that
they pose a serious threat. Most of us assume that the hazard comes
from the genotoxic properties of the metabolites of benzopyrene and
so on, yet there's good evidence of action through a mitogenic
mechanism as well. The angiosarcomas introduced in rats by inhaled
D--20
-------
vinyl chloride do predict human response; we don'C know if chemically
similar compounds do the same. Finally, we come to the interesting
case of rodent goitrogens. Clearly, there are significant
quantitative differences between humans and rats in the way they
respond to ethylene thiourea or sulfamethazine. McLuin believes
there are also qualitative differences, and that the rodent tumors
are not predictive of human response. There is some evidence to
support this view, but there is not a consensus on that conclusion.
In summary, the first question to be asked of any animal experiment
is, does the model allow us to.predict a human response? The
guidelines need to recognize this fact.
The fields of DNA adducts and oncogenes have exploded, or the two
fields have exploded and now they're growing together. We can say a
few things for certain about the information that's derived from
those kinds of experiments. The one thing that is clear, however, at
least in theory, is that DNA adduct information can be used as an
indicator of integrated chronic exposure. Information much beyond
that has still to be deciphered.
Let me close with a reminder that people outside the Agency will use
the results of the classification scheme, if not the scheme itself,
for purposes outside of the Agency's control. Many will use the
classification as an indication of the danger posed by the substance.
Chemicals that are "known carcinogens" will be regarded, probably
correctly, as more dangerous than those that are probable
carcinogens, and so on. I suspect that, conscious or not, within the
Agency the classification scheme serves a purpose similar to that
used by people outside of the Agency. I would suggests two possible
alternatives: either make the classification "possible carcinogen" a
very large one, by including virtually all compounds for which there
are inadequate data; or discard from the classification scheme
altogether all substances except those regarded by the Agency as
serious threats to human health. For instance, Monsanto commented on
D-21
-------
one example of a competitive situation where the classification of
one compound as a category C and another compound as not classified
was used to effect the sale of a particular product to customers. If
that were the only example of this kind of distortion of commerce
that ever occurred, the subject would not be worth further
.»
discussion. However, that's very unlikely. What happened once is
probably happening at other times, and should be kept in mind as the
Agency considers revisions to that part of the guidelines. Thank you
very much.
D-22
-------
Kees Van Der Heiden, A European View1
Or. Van Der Heiden described che European approach to carcinogen
evaluation, noting that while the operational definitions are similar,
strict guidelines do-not exist and most compounds are evaluated on a case-
by-case basis. Use of mechanistic data is important in classifying, and
it is likely that carcinogens act by various important mechanisms. Some
distinctions can be made; for example, one can have clearcut genetic
mechanisms and clearcut nongenetic mechanisms. Indirect genetic
mechanisms are more difficult to deal with, however, and attempting to
separate genetic from nongenetic mechanisms is not simple in practice.
The general approach in both Europe and the United States is similar in
that it starts with an evaluation of weight of evidence. However, in
Europe, weight of evidence is based on whether a compound is a carcinogen
in experimental animals rather than a human carcinogen. The basic weight-
of-evidence evaluation in regard to experimental animal evidence is about
the same internationally.
Beyond the initial animal bioassay evaluation, additional information is
usually requested. However, if good positive results are found in a
bioassay, then other information, e.g., biokinetics, pharmacokinetics,
scaling factors, is not necessary. A number of mechanistic considerations
are usually pursued with the results of short-term tests of ONA
interaction; mechanistic studies are preferred over multiple bioassays.
Information on nutritional imbalance or physical/chemical properties of
the compound are then utilized in evaluating the results.
In emphasizing a mechanistic approach to classification, most chemicals
are found to fall somewhere in between genotoxic and nongenotoxic. For
these substances, a choice must be made. For the genotoxic compounds, a
clear, nonthreshold approach is used to calculate risk: a simple, linear
'This is a summary of remarks based on the taped presentation. It has
been reviewed and approved by the speaker.
D-23
-------
nonthreshold model which draws a straight line from the lowest positive
point to zero. For nongenotoxic compounds, risk assessments utilize a
threshold approach based on mechanistic considerations and the assumption
of the reversibility of the effect.
.*
Carcinogens fall into basically three groups: genotoxins, nongenotoxins,
and uncertain mechanisms. While compounds can be placed in the latter
group if there is limited evidence from animal bioassays, e.g., tumors
only at high dose levels, these are compounds which are in some way
questionable.
The regulatory approach for clear, genotoxic carcinogens has been to avoid
exposure to the extent possible. When there is an alternative, a
genotoxic carcinogen is not allowed. When there is a profound
technological need, a limit value is derived employing a simple
calculation to adjust body weight differences between the test species and
humans. The limit value is thus different from an acceptable daily limit.
Scaling factors are not used. Since there has been no agreement despite
extensive discussion, this method was considered the best choice.
* US. OOVCNNMEMT PWNTHW OffKf: 1*88- 648" 010;-0000 7
------- |