November 1992
Draft Working Paper
           Working Paper for Considering Draft Revisions

      to the U.S. EPA Guidelines for Cancer Risk Assessment

THIS DOCUMENT IS A PRELIMINARY DRAFT.  Until formal announcement by the U.S.
Environmental Protection Agency is made in the Federal Register, the policies set forth in the 1986
Guidelines for Carcinogen Risk Assessment, as they are now interpreted, remain in effect. This
working paper does not represent the policy of the U.S. Environmental Protection Agency with respect
to carcinogen risk assessment.
                      Office of Health and Environmental Assessment
                          Office of Research and Development
                          U.S. Environmental Protection Agency
                                  Washington, D.C.
                                                               Printed on Recycled Paper

                           DRAFT-DO NOT QUOTE OR CITE


       This document is a draft working paper for review purposes only and does not constitute
Agency policy. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.

                         DRAFT-DO NOT QUOTE OR CITE
List of Figures ...................... . . . ... ... .....'..... ... ....	  vi
Authors and Contributors	 vii
PREAMBLE ........ ....	 . .	..........:.......	. . . .	2
1.   INTRODUCTION . ..... . . . ....... ..	;.../... . . ... ........ .. . .	9
    1.1.  PURPOSE AND SCOPE OF THE GUIDELINES...... . . . 1 . . . . ..... .  . . . ... .'. . . 9
    1.3.  ORGANIZATION OF THE GUIDELINES .. .....'.... ... ...... .... ...... 11
    1.4.  APPLICATION OF THE GUIDELINES  . . . . . . ... . . ..............	... 11
2.   HAZARD ASSESSMENT .	: '. . .... .v.  . . . ... .	.... ..... ?	12
    2.1.  INTRODUCTION . . ... . . ..... . . . . .!. .L .'...  .......		12
    2.2.  INTEGRATING DATA FOR HAZARD ASSESSMENT  . ........	. . . .	13
    2,3.  ANALYSIS OF HUMAN DATA . . . . . .... .......... . . . . . . , . .  ... .... ... 14
         2.3.1. Epidemiologic Studies ......... . . , ... .  ... . . . . .. ... . . ... .	14
        Exposure Focus ...................:................. 14
        Types of Epidemiology Studies	15
         2.3.2. Elements of Critical Analysis  .................. . .	15
        Exposure	.....!.../.................. 15
        Population Selection Criteria .........;.................. 16
               2.3.2:3.   Confounding Factors . . . . . . . .... . . . .... . . . .	17
        Sensitivity ....... .... ......;	17
        Criteria for Causality  . ...... ... . . .........".	 19
         2A1. Category!	;...	;.................;... 21
         2.4.2. Category 2	..............21
         2.4.3. Category 3	 21
         2.4.4. Category 4	 21
         2.5.1. Significance of Response .......... ... ..... .V	'.-.22
         2.5.2. Historical Control Data   ............ .....	23
         2.5.3. High Background Tumor Incidence	 24

                          DRAFT-DO NOT QUOTE OR CITE

                               CONTENTS (continued)

          2.5.4.  Dose Issues  . . ..	  24
          2.5.5.  Human Relevance	•......'	25
          2.6.1.  Physical-Chemical Properties  	• — .........  25
          2.6.2.  Structure-Activity Relationships (SAR)	  26
          2.6.3.   Metabolism and Pharmacokinetics	27
          2.6.4.  Mechanistic Information	r ................ r. . . . .	29
        Genetic Toxicity Tests	 .... .............  29
        Other short-Term Tests		.......  32
         Short Assays for Carcinogenicity	  33
        Evaluation of Mechanistic Studies		 .  33
    2.7.   SUMMARY OF EXPERIMENTAL EVIDENCE .....	. . . . • • • • •  34
          2.7.1.  Category 1			, .	.	..-.-.'	, - —	. .  36
          2.7.2.  Category 2				,...,..., 37
          2.7.3.  Category 3 				, ....		- - - •		••••• • - 3-7
          2V.7.4.  Category 4			 . ...... .	.  38
    2.8.   HUMAN HAZARD CHARACTERIZATION 		 .... .	  38
          2.8.1.  Purpose and Content of Characterization	 40
          2.8.2.  Weight of Evidence; for Human Careinogenicity ......................  40'
        Descriptors  					42
                2.8.22.  Examples of Narrative Statements			4:3

          3.2.1.  Response Data  . —	 — .,. . .  46
          3.2.2.  Dose Data ... —			 .— . -- • - - -	• • • 47
        Base Case-Few Data			 48
        PharmacoMnetic Analyses  .............. —	 . . 49

                         DRAFT-DO NOT QUOTE OR CITE

                              CONTENTS (continued)

         Additional Considerations for Dose in Human Studies .......... 50
          3.3.1.  Analysis in the Range of Observation		 . . .	51
          3.3.2.  Extrapolation . .	...	 53
          3.3.3.  Issues for Analysis of Human Studies	.54
          3.3.4.  Use of Toxicity Equivalence Factors (TEF)	 55
    3.4.   DOSE-RESPONSE CHARACTERIZATION ........	.	....	.... 55


    5.1.   PURPOSE			 .	. . 57
    5.2.   APPLICATION . . . . .		 . .	. . . . .	. .	58
          5.3.   CONTENT	  . . .		• • • • •	   - 58
                5.3.1.    Presentation and Descriptors	 58
                5.3.2.    Strengths and Weaknesses	 . . .	58

6.   REFERENCES .	. . .			 -	 60

                          DRAFT-DO NOT QUOTE OR CITE
                                 LIST OF FIGURES
Figure 1 . ...	 —,„.,.;........ — ..............	................... 39

                           DRAFT-DO NOT QUOTE OR CITE

                           AUTHORS AND CONTRIBUTORS

      This draft working paper was prepared by an intra-Agency EPA working group chaired by
Jeanette Wiltse of the Office of Health and Environmental Assessment.


                             DRAFT-DO NOT QUOTE OR CITE

                           WORKING PAPER FOR CONSIDERING
                                DRAFT REVISIONS TO THE

       This working paper identifies cancer risk assessment issues that some Agency scientists have
been discussing as a basis' for possible proposed revisions to EPA's 1986 Guidelines for Carcinogen
Risk Assessment.  The working paper is being given to other scientists to obtain early comment on the
many issues that remain undeveloped or are still under discussion.  The working paper is not a
proposal. It has not been reviewed or approved by any EPA official, and the proposal that is
eventually approved is likely to be very different in many respects from this working paper. When
proposed revisions are ready, EPA will publish them in the Federal Register for public comment.

       Until formal announcement by the U.S. Environmental Protection Agency is made in the
Federal Register, the policies set forth in the 1986 Guidelines for Carcinogen Risk Assessment, as they
are now interpreted, remain in effect. This working paper does not represent the policy of the U.S.
Environmental Protection Agency with respect to carcinogen risk assessment.

                              DRAFT-DO NOT QUOTE OR CITE

       The United States Environmental Protection Agency (EPA) 1986 guidelines on carcinogenic
risk assessment (51 FR 33992, September 24, 1986)) stated that, "...[a]t present, mechanisms of the
carcinogenesis process are largely unknown..." .  This is no longer true. The last several years have
brought research results at an explosive pace to elucidate the molecular biology of cancer.  This new
knowledge is only beginning to be applied in generating data about environmental agents.  Guideline
revisions are intended to be flexible and open to the use of such new kinds of data even though the
guidelines cannot fully anticipate the future forms that carcinogenicity testing and research may take.
At the same time, the guidelines address assessment of the kinds of data that are the current basis of
carcinogenicity assessment as a result of the past two decades of development of the science of risk
assessment. Because methods and knowledge are expected to change more rapidly than guidelines can
practicably be revised, most of the Agency's development of procedures for cancer risk assessment
will henceforth be accomplished through publication of technical work performed under the aegis of
the Agency's Risk Assessment Forum.  The technical documents of the Forum are developed by a
process that engages the general scientific community with EPA scientists.  The documents are made
available for public examination as well as for scientific peer review through the EPA Science
Advisory Board and other groups. The Forum sponsored two workshops in which areas of potential
revision to the guidelines were discussed by scientists from public and  private groups. (USEPA,
1989a; USEPA, 1991a).

Major Changes from 1986 Guidelines
       Revisions in this working paper differ in many respects from the Agency's 1986 guidelines.
The reasons for change arise from new research results, particularly about the molecular biology of
cancer, and from experience using the  1986  guidelines.
       One area of change is increased emphasis on providing characterization discussions for each
part of a risk assessment (hazard, dose-response, exposure, and risk assessments). These serve  to
summarize the assessments with emphasis on explaining the extent and weight of evidence, major
points of interpretation and rationale, strengths and weaknesses of the evidence and analysis, and
alternative conclusions that deserve serious consideration.
       Two other areas of major change are in:

                              DRAFT-DO NOT QUOTE OR CITE

        (1)     the way the weight of evidence about an agent's1
               hazard potential is expressed; and
        (2)     approaches to dose-response assessment.
 1.  To express the weight of evidence for carcinogenic hazard potential, the 1986 guidelines provided
 tiered summary rankings for human studies and for animal bioassays.  These summary rankings of
 evidence were integrated to place the overall evidence in alphanumerically designated classification
 groups A through E, Group A being associated with the greatest probability of carcinogenicity.  Other
 experimental evidence played a modulating role for ranking. Considerations such as route of exposure
 (e.g., oral versus  inhalation) and mechanism of action were not explicitly captured in a
        These working revisions take a different approach.  The idea of summary ranking of individual
 kinds of evidence is retained and expanded, but these are integrated differently and expressed in a
 narrative weight of evidence characterization statement. {Whether an alphanumerical rating will be
 a part of this statement is an unresolved issue still under discussion at EPA.)
        The narrative statement is preceded by summary rankings of human observational evidence
 and of all experimental evidence. The summary ranking for experimental evidence is composed of
 long-term animal bioassay evidence and all other experimental evidence on biological and chemical
 attributes relevant to carcinogenicity. This stepwise approach anticipates marshalling evidence  and
 organizing conclusions as analysis proceeds, for convenience of consideration.  It also gives explicit
 weight to certain  kinds of experimental evidence that previously were considered in a "modulating"
        The narrative statement provides a place to describe evidence by route of exposure and to
 describe the hazard assessment and dose-response implications of mechanism of action data in
 characterizing the overall weight of evidence about human carcinogenicity.
 2.  The approach to dose-response assessment is another area of major change.  It calls for a stepwise
 analysis that follows the conclusions reached in the hazard assessment as to potential mechanism of
 action.  Two steps divide the analysis into modeling in the range of observed data and analysis of
 dose-response below the range of observed data.
!The term "agent" is used throughout (unless otherwise noted) for a chemical substance, mixture, or
physical or biological entity that is being assessed.

                              DRAFT-DO NOT QUOTE OR CITE
(The process for combining all the findings relevant to human carcinogenic potential is a matter
of continuing discussion at EPA. This working paper presents one of a number of suggested
approaches.  The objective is to be integrative and holistic in judging evidence while at the same
time giving guidance to .junior scientists in various disciplines about how to marshal and present
(How to use mechanistic information in dose-response assessment is incompletely developed in
these working paper. Specific issues are pointed out in later sections.}
Perspectives for Carcinogenicity Assessment
       The following paragraphs summarize part of the current picture of the events in the process of
carcinogenesis. Most of the research cited was conducted with experimental approaches not
commonly used to study environmental agents. Nevertheless, as this picture is elaborated, more
experimental approaches will become available for testing specific mechanisms of action of
environmental agents. Even before this happens as a general forward step, information currently
available for some agents can be interpreted in light of this picture to make informed inferences about
the role the agent may play at the molecular level.
       Normally, cell growth in tissues is controlled by a complex and incompletely understood
process governing the occurrence and frequency of mitosis (cell division) and cellular differentiation.
Adult tissues, even those composed of rapidly replicating cells, maintain a constant size and cell
number (Nunez et al., 1991).  This appears to involve a balance among three cell fates:  (1)  continued
replication or loss of ability to replicate, followed by (2) differentiation to take on a specialized
function or (3) programmed cell death (Raff, 1992; Mailer, 1991; Naeve et al., 1991; Schneider et al.,
1991; Harris, 1990).  As a consequence of either the inactivation of processes that lead to
differentiation or cell death, replicating cells may have a competitive growth advantage over other
cells, and neoplastic growth clonal expansion can result (Sidransky et al., 1992; Nowell, 1976).
       The path a cell takes is determined by a timed sequence of biochemical signals. Signal
transduction pathways, or "circuits" in the cell, involve chemical signals that bind to receptors,
generating further signals in a pathway whose target in many cases is control of transcription of a
specific set of genes (Hunter,  1991; Cantley et al., 1991; Collum and Alt, 1990).  A cell produces its
own constituent receptors, signal transducers, and signals,  and is subject to signals produced by other
cells, either neighboring ones or distant ones, for instance, in endocrine tissues (Schuller, 1991).  In
addition to hormones produced by endocrine tissues, numerous soluble polypeptide growth factors
have been identified that control  normal growth and differentiation (Cross and Dexter, 1991; Wellstein

                              DRAFT-DO NOT QUOTE OR CITE

et al., 1990). The cells responsive to a particular growth factor are those that express transmembrane
receptors that specifically bind the growth factor.
       One can postulate many ways to disrupt this kind of growth control circuit, including
increasing or decreasing the number of signals, receptors, or transducers, or increasing or decreasing
their individual efficiencies. In fact, human genetic diseases that make individuals cancer-prone
involve mutations that appear to have some of these effects (Hsu et al., 1991; Srivastava, 1990;
Kakizuka et al., 1991). Tumor cells found in individuals who do not have genetic disease have also
been shown to have mutations with these consequences (Salomon et al.,  1990; Bottaro et al., 1991;
Kaplan et al., 1991; Sidransky et al., 1991).  For example, neoplastic cells  of individuals with acute
promyelocytic leukemia (APL) have a mutation that blocks cell differentiation in myeloblasts that
normally give rise to certain white cells in blood. The mutation  apparently alters a receptor that
normally responds positively to a differentiation signal.  Patients with APL involving this mutation
have been successfully treated by oral administration of retinoic acid, which functions as a chemical
signal that apparently overrides the effect of the mutation,  and drives the neoplastic cells to stop
replicating and differentiate. This "differentiation therapy" demonstrates the power conveyed by
understanding the growth control  signals of these cells (Kakizuka et al.,  1991; deThe et al., 1991).
       Several kinds of gene mutations2 have been found  in human and animal cancers. Among
these are mutations in genes termed tumor susceptibility genes. One kind, mutations that amplify
positive signals to replicate or avert differentiation, are termed oncogenes (proto-oncogenes in their
normal state).  Another kind are mutations in genes involved in generating negative growth signals,
termed tumor suppressor genes  (Sager, 1989).  Damage to these two kinds of genes has been found in
cells of tumors in many animal and human tissues including the sites of the most frequent human
cancers (Bishop, 1991; Malken et al., 1990; Srivastava et al., 1990; Hunter, 1991).   The functions and
deoxyribonucleic acid (DNA) base sequences of the genes are highly conserved across species in
evolution (Auger et al., 1989a, b; Kaplan,  1991; Hollstein  et al.,  1991; Herschman, 1991; Strausfeld et
al., 1991; Forsburg and nurse, 1991).  Some 100 oncogenes and several tumor suppressor genes have
thus far been identified; specific functions are known for only a few.
       The growth control circuit can also be altered without permanent genetic change by, for
example, affecting the responsiveness of signal receptors, the concentration of signals, or the level of
2 The term "mutations" includes the following permanent structural changes to DNA: single base-pair
changes, deletions, insertions, transversions, translocations, amplifications, and duplications.
                                                5               -                        11/05/92

                              DRAFT-DO NOT QUOTE OR CITE

gene transcription (Holliday, 1991; Gross and Dexter, 1991; Lewin, 1991). These can come about
through mimicry or inhibition of a signal or through physiological changes such as alteration of
hormone levels that influence cell growth generally in some tissues.
        Current reasoning holds that cell proliferation which results from changes at the level of DNA
sequence or DNA transcription, from changes at the level of growth control signal transduction, or
from cell replication to compensate for toxic injury to tissue can begin a process ,of neoplastic change
by increasing the number of cells that are susceptible to further events that may lead to uncontrolled
growth. Such further events may include, for instance, errors in DNA replication that occur normally
at a low background rate or effects of exposure to mutagenic agents.  Effects on elements of the
growth control circuit, both permanent and transient, probably occur continuously in virtually all
animals due to endogenous causes. Exogenous agents (e.g., radiation, chemicals, viruses) also are
known to influence this process in a variety of ways.                                    ..   '
        Endogenous events and exogenous  causes such as chemical exposure appear to increase the
probability of occurrence of cancer by increasing the probability of occurrence of effects  on one or
more parts of the growth control circuit.  The specific effect of one exogenous chemical,  aflatoxin Bl,
on a tumor suppressor gene has been postulated on the basis of molecular epidemiology.  Mutations in
the tumor suppressor gene p53 are commonly found in the more prevalent human cancers, e.g., colon
carcinomas, lung cancer, brain and breast tumors (Levine et al., 1991; Malkin et al, 1990).
Populations with high exposure to aflatoxin Bl have a high incidence of hepatocellular carcinoma
showing a base change at a specific codon  in the p53 gene (Hollstein et al., 1991). However, the
patterns of base changes in this gene  that are found in virus-associated hepatocellular carcinomas and
at other sites  of sporadic tumors showing p53 gene mutation are different from the pattern found in
aflatoxin Bl-exposed populations, supporting the postulate that the specific codon change is a marker
of the effect of aflatoxin Bl  (Hayward et al., 1991).
        Research continues to reveal more  and more details about the cell growth cycle and to shed
light on the events in carcinogenesis at the molecular level.  As molecular biology research progresses,
it will become possible to better understand the potential mechanisms of action of environmental
carcinogens.  It has long been known that many agents that are carcinogenic are also mutagenic.
Recognition of the role of oncogenes and mutations of tumor suppressor genes has provided specific
ideas about the linkage of chemical mutagenesis to the cell growth cycle.  Other agents that are not
mutagenic, such as hormones and other chemicals that are stimulants to cell replication (mitogens), can

                             DRAFT-DO NOT QUOTE OR CITE

be postulated to play their role by acting directly on signal pathways, for example as growth signals or
by disrupting signal transduction (Raf£ 1992; McCormick and Campisi, 1991; Schuller, 1991).
       While much has been revealed about likely mechanisms of action at the molecular level, much
remains to be understood about tumorigenesis.  A cell that has been transformed, acquiring the
potential to establish a line of cells that grow to a tumor, will probably realize that potential only
rarely. The process of tumorigenesis in animals and humans is a multistep one (Bouk, 1990; Fearon
and Vogelstein, 1990; Hunter, 1991; Kumar et al.,  1990; Sukumar, 1989; Sukumar,  1990), and normal
physiological processes appear to be heavily arrayed against uncontrolled growth of a transformed cell
(Weinberg, 1989).  Powerful inhibition by signals from contact with neighboring normal cells is one
known barrier (Zhang et al.,  1992).  Another is the immune system (at least for Viral infection). How
a cell with tumorigenic potential acquires additional properties that are necessary to enable it to
overcome these and other inhibitory processes is unknown.  For known human carcinogens studied
thus far, there is an often decades-long latency between exposure to carcinogenic agents and
development of tumors, which may suggest a process of evolution (Fidler and Radinsky, 1990; Tanaka
et al., 1991; Thompson etal., 1989).
       The events in experimental tumorigenesis have been described as involving three stages:
initiation, promotion, and progression.  The initiation stage has been used to describe a point at which
a cell has acquired tumorigenic potential. Promotion is a stage of further changes, including cell
proliferation,  and progression is the final stage of further events in the evolution of malignancy (Pilot
and Dragan, 1991).  The  entire process involves a combination of endogenous and exogenous causes
and influences.  The individual human's  susceptibility is likely to be determined by a combination of
genetic factors and medical history (Harris, 1989; Nebreda et al., 1991), lifestyle, diet, and exposure to
chemical and physical agents in the environment.
       A number of key questions about carcinogenesis have no generic answers—questions such as:
How many events are required? Is there a necessary sequence of events? The answers to these
questions may vary for different tissues and species even though the nature of the overall process
appears to be the same.  The fact that the nature of the process appears empirically to be the same
across species is the basis for using assumptions that come from general knowledge about the process
to fill gaps in empirical data on a particular chemical. Knowledge of the mechanisms that may be
operating in a particular case must be inferred from the whole of the data and from  principles on
which there is some consensus in the scientific community.

                              .DRAFT-DO NOT QUOTE OR CITE

        Information from studies that support inferences about mechanism of action can have several
applications in risk assessment. -For human studies, analysis of ONAjlesions in tumor cells .talcen from
humans, together with information about the lesions that a putative tumorigenic agent causes :in
experimental systems, can provide support for or contradict a causal inference about the agent and the
human effect (Vahakangas et al.,  1992; Hollstein et al., 1991; Hayward et al., 1991).
        An agent that is observed to cause mutations experimentally may be inferred to have potential
for carcinogenic activity (U.S. EPA, 1991a). If such an agent is shown to be carcinogenic in animals
the inference that its mechanism of action is through mutagenicity is strong.  A carcinogenic agent that
is not mutagenic in experimental systems, but is mitogenic or affects hormonal levels or causes toxic
injury  followed by compensatory growth may be inferred to have effects on growth signal transduction
or to have secondary carcinogenic effects.  The  strength of these inferences depends in each case on
the nature and extent of all the available data.
        These differing mechanisms of action at the molecular level have different dose-response
implications for the activity of agents. The carcinogenic activity of a direct-acting mutagen should be
a function of the probability of its reaching and  reacting with DNA. The activity of an agent that
interferes at the level of signal pathways with many potential receptor targets should be a function of
multiple reactions. The activity of an agent that acts by causing toxicity followed by compensatory
growth should be a function of the toxicity.

                               DRAFT-DO NOT QUOTE OR CITE
                                      1. INTRODUCTION
        The new guidelines will revise and replace EPA Guidelines for Carcinogen Risk Assessment
 published in 51 FR 33992, September 24, 1986. Through guidelines, EPA provides its staff and
 decisionmakers with guidance and perspectives necessary to their performing and using risk
 assessments.  Publication of EPA's guidelines also provides basic information about the Agency's
 approaches to risk assessment for those who participate in Agency  proceedings, or in basic research or
 scientific commentary on the subjects the guidelines cover.
        As  the National Research Council pointed out in  1983 that there are many questions
 encountered in the risk assessment process that are unanswerable based on scientific knowledge (NRC,
 1983). To  bridge the uncertainty that exists in areas where there is no scientific consensus, inferences
 must be made to ensure that progress continues in the assessment process.  While the application of
 scientific inferences is both necessary and useful, the bases for these inferences must be continually
 reviewed to assure that they remain consistent with predominating scientific thought.
        The guidelines incorporate basic principles and science policies based on evaluation of the
. currently available information.  Certain general assumptions are described that are to be used when
 data are incomplete.  Standard, default assumptions are described in order to maintain consistency and
 comparability from one assessment to the next.  However, these guidelines explain that such
 assumptions are to be displaced by facts or better reasoning when appropriate data are available. Short
 of displacement, an analysis of any promising alternatives is expected to be presented alongside
 default assumptions.
        These guidelines serve two policy goals that must be balanced:  first, to maintain consistency
 of procedures that will support regularity in Agency decisionmaking and, second, to be adaptable to
 advances in science. Each risk assessment must balance these goals.  To assist in balancing these and
 other science policies, the Agency will rely on input from the general scientific community through the
 Agency's established scientific peer review processes.  The Agency will continually adapt its practices
 to new developments in the science of environmental carcinogenesis, and restate or revise, where
 appropriate, the principles, procedures, and operating assumptions of the risk assessment process.
 Changes will be made through either revisions to these guidelines or, more frequently, issuance of
 documents on scientific perspectives and procedures and science policies that are developed under the
 aegis of the EPA Risk Assessment Forum.

                                               9                                       11/05/92

                              DRAFT-DO NOT QUOTE OR CITE

       Under these guidelines all available direct and indirect evidence is considered to assess
whether the weight of the combined evidence supports a conclusion about potential human
carcinogenicity.  Direct evidence for carcinogenicity in humans comes from epidemiolpgical studies of
cancer or, in a few instances, from case reports.  Other data providing direct evidence can come from
long-term animal cancer bioassays. Indirect evidence comes from a variety of information about
lexicological and biochemical effects related to carcinogenicity.
       The most direct evidence for identifying and  characterizing an agent's human cancer hazard
potential is from human epidemiologic studies in which cancer is attributed to exposure to a specific
agent.  These studies are rarely available because the identification and follow up of populations of
sufficient size and sufficient exposure to detect underlying risk is rarely feasible. Moreover, exposure
to many potential but unidentifiable causative factors is frequent, making statistical attribution of
incidence of a cancer to  a single agent difficult.  Much of the human  evidence comes from
occupational studies in which workplace exposure to  an agent has been high, and the increased
incidence of a cancer attributed to the agent has  been distinguishable from other potential causes.
Studies that are statistically not powerful enough to discern an association between environmental
exposure and tumor incidence or to distinguish among potential  causative factors are unable to show
that an agent is not carcinogenic.  Such studies, if well conducted, may nevertheless be used to
estimate a "ceiling" on an agent's carcinogenic potency.
       Long-term animal cancer bioassays are more  frequently  available for more agents than are
epidemiologic studies. Approximately 400 of these have been conducted by the National Cancer
Institute and National Toxicology Program (NTP)(Huff et al., 1988; NTP, 1992) and many additional
ones have been conducted by others.  The correspondence between positive results in human studies
and long-term animal cancer bioassays is high (Tomatis et al., 1989; Rail, 1991) in the limited number
of cases in which comparison is possible. In the absence of epidemiologic information, tumor
induction in animal assays remains the best single piece of direct evidence on which to evaluate
potential human carcinogenic hazard (OSTP,  1985).  Results of  animal studies have to be carefully
analyzed along with other relevant data (such as metabolism and pharmacokinetic data used to
compare animals and humans) to evaluate biological  significance, causation, and reproducibility of
results, and to determine the reasonable inferences  about human hazard they support (Allen et al.,
1988; Ames and Gold, 1990).

                              DRAFT-DO NOT QUOTE OR CITE

        Data on physicochemical characteristics and biological effects of an agent that make it more or
less likely to affect processes involved in producing neoplasia provide important evidence supporting
influences about carcinogenic potential.  These include, for example, the ability to alter genetic
information, influences on cell growth, differentiation, and death, and structural and functional
analogies to other compounds that are carcinogenic.

        These  guidelines follow and should be read with two other publications that provide basic
information and general principles.  These are:  Office of Science and Technology Policy (OSTP,
1985) Chemical Carcinogens: A Review of the Science and its Associated Principles (50 FR 10371),
and National Research Council (NRC, 1983), Risk Assessment in the Federal Government: Managing
the Process (Washington, DC, National Academy Press).  The 1983 NRC document provided the 1986
guidelines with a thematic organization of risk assessment into hazard identification, dose-response
assessment, exposure assessment, and risk characterization.  This thematic organization has been
slightly revised in these guidelines to focus attention on the importance of characterization in each part
of the assessment.  Nonetheless, the four questions addressed in these four areas remain the same; they
are: Can the agent present a carcinogenic hazard to humans?  At what levels of exposure?  What are
the conditions  of human exposure?  What is the overall character of the risk, and how well do data
support conclusions about the nature and extent of the risk?

        The guidelines are to be used within the policy framework already provided by applicable EPA
statutes and do not alter such policies. The Guidelines provide general directions for analyzing and
organizing available data.  They do not imply that one kind of data or another is prerequisite for
regulatory action to control, prohibit, or allow the use of a carcinogen.
        Regulatory decision making involves  two  components:  risk assessment  and risk management.
Risk assessment defines the adverse health consequences of exposure to toxic agents. The risk
assessments will be carried out independently from considerations of the consequences of regulatory
action.  Risk management combines the risk assessment with directives of regulatory legislation,
together with socioeconomic, technical, political, and other considerations, to reach a decision as to
whether or how much to control future exposure to the suspected toxic agents.

                              DRAFT-DO NOT QUOTE OR CITE
                                 2. HAZARD ASSESSMENT
       Hazard assessment covers a wide variety of data relevant to the question, can an agent pose a
human carcinogenic hazard? Available data may include: long term animal cancer bioassays and
human studies, physical-chemical properties of the agent and its structural relationship to other
carcinogens, studies of cellular and molecular interactions and mechanisms of action, and results from
lexicological tests and experiments on the bioavailability and transformation of an agent in
experimental animals and humans. Hazard assessment results are summarized in a hazard
characterization that conveys the nature and impact of available data and appropriate scientific
inferences about human  carcinogenic hazard.
       Experience shows that the nature and extent of information available on  each agent is different
and can vary from a wealth of epidemiologic data to only physical-chemical properties. Frequently,
results from a  long-term animal carcinogenesis bioassay are the only direct evidence available for the
evaluation. These guidelines follow the assumption that chemicals with evidence to demonstrate
careinogenicity in animal studies are likely to present a carcinogenic hazard to humans under some
conditions of exposure (OSTP, 1985).  At the same time, there may be mechanistic, physiological,
biochemical, or route-of-entry differences which alter the lexicological consequences in humans from
those observed in ihe particular animals tested. When the results of animal testing are extrapolated to
humans, effects observed at high continuous exposures are often projected to low or intermittent
exposures and results from one route of exposure are often extrapolated to  other routes of exposure.
The risk analysis must examine each assumption and extrapolation for mechanistic and biological
plausibility. The elements of hazard assessment described below are the foundation for these
       The characterization of an agent's carcinogenic human hazard potential depends on the weight
of all the relevant evidence.  Studies are evaluated according to accepted criteria for study quality,
sensitivity, and specificity.  These have been described in several publications (Interagency Regulatory
Liaison Group, 1979; OSTP, 1985; Peto et al., 1980; Mantel, 1980; Mantel and Haenszel, 1959;
Interdisciplinary Panel on Careinogenicity, 1984; National Center for Toxicological Research, 1981;
National Toxicology Program, 1984; U.S. EPA, 1983a, b, c; Baseman, 1984). The hazard
characterization describes how likely the agent is to be carcinogenic to humans,  including the
judgment whether or not the hazard is considered to be contingent on certain conditions of exposure

                               DRAFT-DO NOT QUOTE OR CITE

(e.g., oral versus dermal exposure). The characterization summarizes the basis of, and confidence in,
inferences drawn from data and the rationale for conclusions about weight-of-evidence; these are
accompanied by judgments on issues and uncertainties that cannot be resolved with available
        The characterization of potential hazard is qualitative.  It does not address the magnitude or
extent of effects under actual exposure conditions. However, observations and conclusions from the
hazard characterization that are relevant to quantitative dose-response analysis are carried forward to
the section on quantitative dose-response analysis, and those that are relevant to actual exposure
conditions are discussed in the risk characterization.

        The assessment of potential carcinogenic hazard to humans is a process in which many kinds
of data are integrated to  examine the inferences and conclusions they support. The process is
conducted as an interdisciplinary effort.
        While the discussion that follows explores data analyses along separate disciplinary lines and
provides for making intermediate summaries of human observational data and experimental data, it
must be recognized that this is done simply for convenience of organization and marshalling of
thought, and the individual analyses are interdependent not separate.  Each kind of analysis, from
evaluation of human studies to structure-activity relationship analysis, looks to the others for
interpretive  alliance and perspective.  Confidence in conclusions is built upon the  overall coherence of
inferences from different kinds of data as well as  confidence in individual data sets.
        For  example, in examining the issue of causation as part of human studies analyses, one uses
knowledge of the biological activity of the agent in animal systems and of pertinent features of its
structure, metabolism and other properties to address issues of biological plausibility of a causal
hypothesis.  Likewise, where there are no epidemiplogic studies and one is examining relevance of
animal responses to human hazard potential, one uses human data to  address comparative biology of
animals and humans with respect to, for instance,  metabolism,  pharmacokinetics, physiology, and
disease history.

                             DRAFT-DO NOT QUOTE OR CITE

23.1. Epidemiologic Studies
       Epidemiology is the study of the distribution of a disease in a human population and the
determinants that may influence disease occurrences.  Epidemiologic studies provide direct information
about the response of humans who have been exposed to suspect carcinogens and avoids the need for
interspecies extrapolation of animal lexicological data.
23.1.1. Exposure Focus
       An identification of hazard in a human population depends critically on the exposure
assessment, which consists of two components: (a) the qualitative determination of the presence of an
agent in the environment and (b) the quantitative assessment.  An exposure assessment which includes
an attribution of quantified exposure to an individual is considered more precise and will carry more
weight in an evaluation of human hazard. In many epidemiologic studies, the populations are selected
and studied retrospectively,  and the time  between exposure and observation of effects is very long
because of the latency of cancer. The past exposure is a critical determinant. In an environmental
situation, quantitative exposure assessment is usually difficult to achieve due to lack of measures of
past exposure. This is one reason why occupational studies where exposure is based on job
classification are often used for identifying environmental hazard.  Past occupational exposures are
usually considered to be at higher levels  than those encountered environmentally; therefore, the
question whether  any identified hazard is pertinent at lower exposure levels needs to be addressed.
       Exposure  assessment becomes more complicated when the exposure is to a complex mixture of
incompletely identified chemicals. In addition, human exposures to agents can occur by more than
one route as compared to the controlled exposure regimens used in the animal carcinogenicity studies
(e.g., occupational exposure to  solvents can occur through inhalation and dermal absorption). The
characterization of the patterns of exposure to identify exposure-effect relationships is another
consideration.  Important exposure measurements in epidemiologic studies include cumulative exposure
(sometimes time-weighted), duration of exposure, peak exposure, exposure frequency or intensity, and
"dose" rate. Some insight on which measurement of exposure will be the best predictor of a cancer
can come from an understanding of the disease process itself.
       In epidemiological studies, "biological markers", usually the reaction products of an agent or
its metabolite with DNA or a protein or other markers of exposure such as excretion of metabolites in
urine have been increasingly considered as reliable measures of exposure. More rarely a marker of

                              DRAFT-DO NOT QUOTE OR CITE

effect specific to an agent may be found (Vahakangas et al., 1992).  Information on the relationship
between exposure or effect and markers is often derived from metabolism and kinetic studies in
animals. Validation of the relationship with comparative human data is needed to support confidence
in use of such markers.

(The generic issue of use biomarkers of exposure and effect is still under consideration.) Types of Epidemiology Studies
        Various types of epidemiologic studies or reports can provide useful information for
identifying hazards.  An important consideration is the validity and representativeness of the studied
population with respect to the larger population of interest  Study designs include cohort, case-control,
proportionate ratio, clinical trials, and correlational studies.  In addition, cluster investigations and case
reports, while not constituting studies, may yield useful information under certain situations (e.g.,
reports associated with exposure to vinyl chloride and diethylstilbestrol).  The above designs have
well-defined strengths  and limitations (Breslow et al., 1980; 1987; Kelsey et al., 1986; Lilienfeld and
Lilienfeld,  1979; Mausner and Kramer., 1985; Romman, 1986).

2.3.2.  Elements of Critical Analysis
        Aspects of the available human data, which are described in this section, are evaluated to
determine whether there is a causal relationship between exposure to the agent and an increase in
cancer incidence. Certain elements of analysis are brought to bear on the criteria for causality, which
are listed and discussed in Section  In general, these elements address the study design and
conduct; the ability to  sort out the potential role of the agent in question as opposed to other risk
factors; assessment of exposure of the study and referent populations to the agent and to other risk
factors; and, given all of the above, the statistical power of the study or studies. Exposure
       Exposure is the foundation upon which any exposure-effect relationship is evaluated.  Often,
the exposure is not to a single agent, but to a combination of agents (e.g., exposure to
chloromethylmethyl ether and its ever-present contaminant bischloromethyl ether). When exposures
occur simultaneously, it is generally assumed that each chemical exposure contributes to the exposure-
or exposures-effect relationship.

                              DRAFT-DO NOT QUOTE OR CITE

       Exposure can be defined in hierarchical levels.  Greater weight will be given to studies where
exposures are more precisely defined and can be quantified. The broadest definition of exposure is
that inferred for a group of individuals living in a geographic area.  At this level, it is not known
whether all individuals are exposed to the agent, and if exposed, the patterns and lengths  of exposure.
The result is a mixture of individuals with higher exposure and those with little or no exposure. This
leads to exposure misclassification, which, if random, may result in a study's reduced ability to detect
underlying elevations in risk. For the same reasons, exposure as defined by assignment to a broad
occupational category in the absence of qualitative or quantitative data yields less useful information
on an individual's exposure.
       A more recent application in epidemiologic studies is the use of job-exposure matrices to infer
semi-quantitative and quantitative levels of exposure to specific agents (Stewart and Herrick, 1991).
The job-exposure matrix has been applied to occupational scenarios where at least  some current and
historical monitoring data exist.  In examining exposure levels inferred from a job-exposure matrix, the
basis of the monitoring data must be considered—whether data are from routine monitoring or  reflect
accidental (i.e., higher than average) releases.
       Biological markers are indicators of processing within a biological system.  Using such a
marker as a measure of exposure is potentially the most reliable level of data since the quantity
measured is thought to more precisely characterize a biologically available dose, rather than exposure
that is the amount of material presented to the individual and is usually  inferred from a measurement
of atmospheric concentrations (NAS, 1989).  Validated markers are the most desirable, i.e., markers
which are highly  specific to the exposure and those which are highly predictive of disease (Blancato,
OHR Biomarker Strategy, cite published paper;  Hulka and Margolin, 1992) (e.g., urinary arsenic
(Entertine et al., 1987), and alkylated hemoglobin (hemoglobin adducts) from exposure to ethylene
oxide (Callemen et  al., 1986; van Sittert et al., 1985).
233.3,. Population Selection Criteria
       The study population and the comparison or referent population are identified and examined to
decide whether or not comparisons between populations are appropriate and to determine the extent of
any bias resulting from their selection. The ideal referent population would be similar to the study
population in all respects except exposure to the agent in question.  Potential biases (e.g., healthy
worker effect, recall bias, selection bias, and diagnostic bias) and the representativeness of the studied
population for a much larger population are addressed.

                              DRAFT-DO NOT QUOTE OR CITE

        Generally, the referent population in cohort studies consists of mortality or incidence rates of a
larger population (e.g., the U.S. population). The healthy worker bias is specific to occupational
cohort studies, and it asserts that an employed population is healthier than the general population
(McMichael, 1976).  The influence of the healthy worker effect is toward a more favorable mortality
in the exposed population; this influence is thought to decrease with increasing age and to have less
influence on site-specific cancer rates.  The influence of the healthy worker effect is thought to be
minimized by the use of an internal comparison group (e.g., incidence or mortality rates of employees
who are from the same company, but not among the employees in the study .population).
        In case-control designs, the potential for differences in recalling past events (recall bias)
between the case and control series needs to be evaluated. The characteristics of the control series
also need to be discussed. Hospital controls have associated limitations with respect to possible
associations with the exposure of interest.  Randomly-selected population or community controls are
thought to be more like cases in the case series; however, response rates are often lower.  Confounding Factors
        A confounding variable is a risk factor for the disease under study that  is distributed unequally
among the exposed and unexposed populations. Adjustment for possibly confounding factors can
occur either in the design of the study (e.g., matching on critical factors) or in the statistical analysis
of the results. If adjustment within the study data is not possible due to the presentation of the data or
because needed information was not collected during the study, indirect comparisons may be made
(e.g., in the absence of direct smoking data from the study population, an examination of the possible
contribution of cigarette smoking  to increased lung cancer risk and to the exposure in question may
include information from other sources such as the American Cancer Society's longitudinal studies
(Hammond, 1966; Garfmkel and Silverburg, 1991).
        In a collection of heterogenous studies possible  confounding factors are usually randomly
distributed  across studies.  If consistent increases in cancer risk are observed across the collection of
studies, greater weight is given to the agent under investigation as the etiologic factor even though the
individual studies may not have completely adjusted for confounding factor. Sensitivity
       Epidemiologic studies which consist of a large number of individuals with sufficient exposure
to a putative cancer-causing agent and adequate length of time for cancer development or detection are

                              DRAFT-DO NOT QUOTE OR CITE

considered to have a greater ability to detect cancer risk. Studies for review, however, do not always
fulfill these criteria.  In addition, the ability to detect increases in relative risk associated with
environmental exposure is very difficult due to heterogeneous exposure regarding both pattern and
levels and which potentially bias risk toward the null hypothesis of no effect.
        If the underlying risk is actually increased, examination of persons considered at higher risk
increases the detection ability of a study.  Such examination may include an evaluation of risk among
individuals with higher or peak exposure,  with greater duration of exposure, or with the longest time
since first exposure (to allow for latency of effect), and those of older age, and those with long
        A study in which no increases in risk were observed may be useful for inferring an upper limit
on possible human risk.  Statistical reanalysis is another approach for examining the sensitivity of
results from an individual study (e.g., the dose-response relationship reported in one formaldehyde-
exposed cohort (Blair et al., 1986) has been examined by several investigators (Blair et al., 1987;
Sterling and Weinkam, 1987; Collins et al., 1988; Marsh, 19920).  These further analyses are a
reaggregation of exposure groups or an examination of the influence of a subgroup on the disease
incidence of the much larger group.                                                     ,
        Statistical methods for examining several studies together are  frequently applied to the
collection of data.  These methods, commonly referred to as meta-analysis, are used to contrast and
combine results of different studies with the goal of increasing sensitivity. In meta-analysis, study
results are evaluated as whether they differ randomly from the null hypothesis of no effect (Mann,
1990); meta-analysis presumes that observed results are not biased.  If an underlying effect is not
present, the observed results should appear randomly distributed and cancel each other when studies
are combined (Mann, 1990).  Several important issues are pertinent to meta-analysis. These are
controlling  for bias and confounding prior to combining studies, criteria for study inclusion,
assignment of weights to individual studies, and possible publication and aggregation bias.  Greenland,
1987 discusses  may of these issues in addition to identifying methodologic approaches.
{Participants at the December 4, 1992, Society for Risk Analysis on cancer risk assessment issues
were asked to look at meta-analysis.}

                               DRAFT-DO NOT QUOTE OR CITE  Criteria for Causality                        '
        A causal interpretation is enhanced for studies to the extent that they meet the criteria
described below.  None of the criteria, with the exception of a temporal relationship, should be
considered as either necessary or sufficient in itself to establish causality.  These criteria are modelled
after those developed by Hill in the examination of cigarette smoking and lung cancer (Rothman,
        a.      Temporal relationship:  This is the single absolute requirement, which itself does not
               prove causality, but which must be present if causality is  to be considered.  The
               disease occurs within a biologically reasonable time frame after the initial exposure.
               The initial period of exposure to the agent is the accepted starting point in most
               epidemiologic studies.
        b.      Consistency: Associations are observed in several independent studies  of a similar
               exposure in different populations.  This criterion also applies if the association occurs
               consistently for different subgroups in the same study.
        c.      Magnitude of the association:; A causal relationship is  more credible when the risk
               estimate is large and precise   (narrow confidence intervals).
        d.      Biological gradient: The risk ratio is correlated positively with increasing exposure or
               dose. A strong dose-response relationship across several  categories of exposure,
               latency, and duration is supportive although not conclusive for causality given that
               confounding is unlikely to be correlated with exposure. The absence of a dose-
               response relationship, however, should not be construed by itself as evidence of a lack
               of a causal relationship.
        e.      Specificity of the association: The likelihood of a causal interpretation is increased if
               a single exposure produces a unique effect (one or more cancers also found in other
               studies) or if a given effect has a unique exposure.
        f.      Biological plausibility:  The association makes sense in terms of biological  knowledge.
               Information from animal toxicology, phannacokinetics, structure-activity relationship
               analysis and short-term studies of the agent's influence on events in the carcinogenic
               process are considered.
        g.      Coherence:  The cause-and-effect interpretation is in logical agreement with what is
            ;   known about the natural history and biology of the disease, i.e., the entire body of
               knowledge about the agent.

                              DRAFT-DO NOT QUOTE OR CITE


(The process in combining all findings relevant to human carcinogenic potential is an issue for
further development. The need for this summarization step for human evidence and the one in
Section 25 for experimental evidence are open questions at EPA.)

       Each epidemiological study is critically evaluated for its relevance with respect to the
exposure-effect relationship, exposure assessment such as intensity, duration, time since first exposure,
and methodological issues such as study design, selection and characterization of comparison group,
sample size, handling of latency, confounders, and bias.
       Following critical evaluation, the totality of the weight-of-evidence for human careinogenicity
is assessed and summarized according to one of the following four categories, which are meant to
represent a judgment regarding the weight of all of the human evidence even if only one study exists
on the subject  Rarely, the judgment can be based on a series of case reports. More likely, the
evaluation will involve several studies.  Inferences from summary analyses such as meta-analysis can
provide support for placement into these categories.  In addition, evidence that the agent  in question is
metabolized to a compound, for which independent human evidence exists, is supportive  of the
       The weight a particular study or analysis is given in the evaluation depends on its design,
conduct, and avoidance of bias (selection, confounding, and measurement) (OSTP, 1984). Results,
both positive and null, are considered in light of the study's rigor. The weight of evidence is based on
the plausibility of the association and the conclusiveness of observed findings. Greater plausibility and
conclusiveness can be ascribed to an exposure-effect relationship when it can be explained in terms of
adherence to the criteria for causality, including coherence with other evidence such as animal
toxicology.  The plausibility of exposure-effect relationship also can be bolstered or mitigated by
evidence of structure-activity relationship analysis with well characterized agents, studies of
mechanism of action, understanding of metabolic pathways, and other indirect evidence relevant to
human effects.  A mixture (e.g., cigarette smoke, coke oven emissions) may be categorized as an agent
when causation is ascribed to the mixture, but not to necessarily to its individual components.

                              DRAFT-DO NOT QUOTE OR CITE

2.4.1.  Category 1
        Plausible evidence exists, and from this evidence a conclusive causal association can be
judged. Cause and effect relationships are supported with results from well-designed and conducted
studies in which random or nonrandbm error can be reasonably excluded.

2.4.2.  Category 2
        Evidence exists to suggest that causal association is   plausible; however, such evidence is not
conclusive due to a number of reasons which may include lack of consistency, wide confidence
intervals which may or may not include a risk, or absence of an observed dose-response relationship.  ,
The effect of random or nonrandom error in individual studies which could influence the risk ratio
away from the null is considered  minimal.  This category covers a broad range of possible weights of
evidence.  At the top of the category are highly suggestive, but short of convincing data. At the
bottom of the category are suggestive but weak data.  A statement of the relative position of data in
this continuum accompanies the description of the data as Category 2.

2.4.3.  Category 3
        The body of evidence is inconclusive. The assertion of a causal association is not plausible
from the available data in which studies of equal quality have contradictory results in which random or
nonrandom error is a more likely explanation for observations of increased risk.  This category also
applies when no epidemiologic data are available.

2.4.4.  Category 4
        The available studies are  designed with defined ability to detect increases in risk, and resultant
risk ratios are precise with tight confidence intervals.  Evidence derived from the studies consistently
show no positive association between the suspect agent and cancer.  The evidence is described as
showing no cause and effect  relationship at the exposure levels studied. It is not considered to show
that the agent is non-carcinogenic under all circumstance unless the evidence is so  complete that
potential for human carcinogenicity can be eliminated.

                              DRAFT-DO NOT QUOTE OR CITE

       Long-term animal studies are evaluated to decide whether biologically significant responses
have occurred and whether responses are statistically significantly increased in treated versus control
animals.  The unit of comparison is an experiment of one sex, in one species.

2.5.1.  Significance of Response
       Evidence for carcinogenicity is based on the observation of biologically and statistically
significant tumor responses in specific organs or tissues.  Criteria for categorizing the strength of
evidence of animal carcinogenicity  in bioassays have been established by the National Toxicology
Program  (NTP, 1987). Animal study results are evaluated for adequacy of design and conduct (40
CFR Part 798).  The results are described and biological  significance of observed toxicity is evaluated
(non-neoplastic endpoints included).

(For EPA's purposes, the criteria for evaluating animal cancer bioassays are still under review,
and could be somewhat different  from those of NTP.  Nevertheless, much of the animal cancer
data available to EPA carries the NTP designations of "clear, some, equivocal, or none".)

       Interpretation of animal studies is aided by the review of target organ toxicity and other non-
neoplastic effects (e.g., changes in the immune and endocrine systems) that may be noted in
prechronic or other lexicological studies.  Time and dose-related changes in the incidence of
preneoplastic and neoplastic lesions may also be helpful in interpreting responses in long-term animal
       It is recognized that chemicals that induce benign tumors also frequently induce malignant
tumors, and that certain benign tumors may progress to malignant tumors.  Benign and malignant
tumor incidence are combined for analysis of carcinogenic hazard when scientifically defensible
(OSTP, 1985; Principle 8).  The Agency follows the National Toxicology Program framework for
combining benign and malignant tumor incidence of a particular site (McConnell, 1986).
       Elevated tumor incidences hi adequate experiments are analyzed for biological and statistical
significance.  Generally, a statistical test that  shows a positive trend in dose-response at a level of
significance of five percent (i.e., the likelihood of false positive results is less than five percent)
supports  a conclusion that the experiment is positive. If false positive outcomes are a serious concern,
the use of a formal multiple comparison adjustment procedure should be considered.  No rigid decision

                              DRAFT-DO NOT QUOTE OR CITE

rule should be used as substitute for scientific judgment.  Other statistical tests may be applied if the
trend test is not statistically significant or, for some reason, not applicable for a given experiment.
The significance level should be adjusted if multiple comparisons of the same data are made, in order
to avoid raising the overall likelihood of false positives (Haseman 1983, 1990; U.S. FDA, 1987).
        Data from all long-term animal studies, positive and negative, are to be considered in the
evaluation of carcinogenicity.  Different results according to species, sex, or strain, or by route of
administration,  duration of study or site of effect are not unexpected.  The issues are how different
results affect the weight of evidence and whether the differences suggest the operation of any
particular mechanisms of action or tissue  sensitivity that may assist in judging human relevance.

2.5.2.  Historical Control Data
fNOTE TO THE READER: The issues of how to consider historical control data and high
background tumors are knotty ones. For high background tumors there are varying views, some
question relevance, but usually there are insufficient data about the mechanism of action to
question its relevance. Others point to the fact that both humans and animals have tissues with
high background rates.}
       Historical control data often add valuable perspective in the evaluation of carcinogenic
responses (Haseman et al., 1984). For the evaluation of rare tumors, even small increases in tumor
response over that of the concurrent controls may be significant compared to historical data. Historical
data can also identify sites with high spontaneous background in the test strain.  Nevertheless,
historical control data have limitations as compared to concurrent control data.  One limitation is the
potential for genetic drift in laboratory strains over time that makes historical data less useful beyond a
few years.  Other limitations are the differences in pathological examinations at different times and in
different laboratories; these are due to changes over time in criteria for evaluating lesions and to
variations in preparation techniques and reading of tissue samples between laboratories.  Other
differences may include biological and health differences in animal strains from different suppliers.
Concurrent controls are, for these reasons, more valuable comparison for judging whether observed
effects in dosed animals are treatment related.
       Comparison of an observed response that appears to be treatment related with historical  control
data may call the response into question if the observed response is well within the range of historical
control data.  Whenever historical control data are compared with the current data the reasons should

                              DRAFT-DO NOT QUOTE OR CITE

be given for judging the historical control data to be adequately representative of the current expected
response background.

233.  High Background Tumor Incidence
       Tumor data at sites with high spontaneous background requires special consideration (OSTP,
1985; Principle 9).  Questions raised about high background tumors in animals (and humans) are
whether they are due to particular genetic predispositions or ongoing proliferative processes that are
species-specific prerequisites to a neoplastic response or, on the other hand, represent sensitivities due
to biological processes that are alike among species. Answering these questions requires a body of
research data beyond the data obtained in standard animal studies.  Unless there are research data to
establish that such tumor data at a site occur because of a mechanism-of-action that is unique to the
species, strain, and sex with the high background, the tumor data are considered, as are other tumor
data, in the overall weight of evidence.  These data may receive  relatively less weight than other tumor

2.5.4.  Dose Issues
       Long-term animal studies at or near the maximum tolerated dose level (MTD) are used to
ensure an adequate power for the detection of carcinogenic activity of an  agent (NTP, 1984; IARC,
1982). The MTD is a dose which is estimated to produce some  minimal  toxic effects in a long term
study (e.g., a small reduction in body weight), but should not shorten an animal's life span or unduly
compromise normal well-being except for chemically induced carcinogenicity. (International Life
Sciences Institute, 1984; Haseman, 1985).  Assays in which the MTD may have been exceeded or may
not have been reached require special scrutiny.
       Exceedance of the MTD in a study may result in tumorigenesis that is secondary to tissue
damage or physiological damage and is more a function of this damage than of the carcinogenic
influence of the particular agent tested.  Inferences drawn from the study  must consider observed non-
neoplastic toxicity and the tissues affected, as well as the existence of carcinogenic effects in tissues,
or at doses, not affected by the exceedance.  Study results at doses that exceed the MTD can be
rejected if toxic damage is so severe as to compromise interpretation.
       Null results in long-term animal studies at exposure levels above the MTD may not be
acceptable if animal survival is so impaired that the sensitivity of the study is significantly reduced

                              DRAFT-DO NOT QUOTE OR CITE

below that of a conventional chronic animal study at the MTD.  The import of non-positive studies at
exposure levels below the MTD may be compromised by lack of power to detect effects.

2.5.5. Human Relevance
       Relevance of tumor responses to human hazard is a judgment that is integral to analysis of
bioassay results. The assumption is made under these guidelines that observation of tumors at any
animal tissue site supports an inference that humans may respond at some site. This assumption is
reexamined as data on the issue become available for specific responses. The Agency will undertake
analyses of relevance issues as needed in reports to be published from time to time (e.g., USEPA,
       If information on the mechanism of tumorigenesis supports the conclusion that a response seen
in an animal study is unique to that species or strain, the response is considered to provide no
evidence for human hazard potential (U.S. EPA, 199la).  Agency decisions of this kind  about
particular animal responses are made and published under the aegis of the EPA Risk Assessment
Forum.  Such mechanistic uniqueness is be differentiated from quantitative differences in dose-
response which are not, £er_se, issues of relevance.

       Certain structural, chemical, and biological attributes of an agent provide key information
about its potential to cause or influence carcinogenic events. These attributes  and comparative studies
between species provide information to support carcinogenic hazard identification and compare
potential activity across species. The following sections provide guidance for inclusion of analyses of
these kinds of evidence in hazard identification.

2.6.1. Physical-Chemical Properties
       Physical-chemical properties that can affect the agent's absorption,  tissue distribution
(bioavailability), biotransformation, or chemical degradation in the body are analyzed as  part of the
overall weight of evidence on hazard potential.  These include, but are not  limited to: molecular
weight, size, and shape; physical state (gas, liquid, solid); water or lipid solubility that can influence
retention and tissue distribution; and potential for chemical degradation or stabilization in the body.

                              DRAFT-DO NOT QUOTE OR CITE

       Interaction with cellular components and reactivity with macromolecules is a second major
area covered.  Factors such as molecular size and shape, electrophilicity, and charge distribution are
analyzed to decide whether they would facilitate such reactions by the agent.

2.6.2.  Structure-Activity Relationships
       The role of structure-activity relationship (SAR) analysis in the  assessment of the carcinogenic
risk of an agent in question is dependent upon the availability and the quality of the lexicological data
on the agent.  For chemicals  with data from  reasonably conducted studies, SAR analysis is useful in
providing input to determine the probable mechanism of action, which is important for hazard
identification and for decisions on the appropriate methodology for quantitative risk assessment. For
chemicals with either unsatisfactory or inadequate carcinpgenicity data,  SAR analysis may be used to
generate, bolster, or mitigate the carcinogenic concern for the chemical, depending on the strength of
and confidence in the SAR analysis.  In addition, SAR analysis can also serve as a guide to evaluate
carcinogenic potential of untested chemicals.
       Currently, SAR analysis is most useful for chemicals that are believed to produce
carcinogenesis, at least initially, through covalent interaction with DNA (i.e., DNA-reactive mutagenic
electrophilic or proelectrophilic chemicals) (Ashby and Tennant, 1991; Woo and Arcos., 1989).  In
analyzing the SAR of DNA-reactive mutagenic chemicals, the following parameters should be
considered (Woo and Arcos,  1989):
       a.      the nature and reactivity of the electrophilic moiety or moieties present;
       b.      the potential  to form electrophilic reactive intermediate^) through chemical,
               photochemical; or metabolic activation;
       c.      the contribution of the carrier molecule to which the electrophilic moiety(ies) is
       d.      physicochemical properties (e.g., physical state, solubility, octanol-water partition
               coefficient, half-life in aqueous solution);
       e.      structural and substructural features (e.g., electronic, stearic, molecular geometric);
       f.      metabolic pattern (e.g., metabolic pathways and activation and detoxification ratio);
       g.      the possible exposure route(s) of the subject chemical.

                              DRAFT-DO NOT QUOTE OR CITE

       Following compilation of a carcinogenicity database for structural analogs, the above
parameters are used to compare and  place the subject chemical as to its carcinogenic potential among
its analogs or congeners. In addition, the analysis is supplemented with any available information on
the pertinent toxic effects of the compound, its potential metabolites, and its structural analogs.  The
pertinent toxic effects are those known to contribute to carcinogenesis such as immune suppression or
       Suitable SAR analysis of non-DNA-reactive chemicals and of DNA-reactive chemicals that do
not appear to bind covalenfly to DNA requires knowledge or postulation of the most probable
causative mechanism(s)  of action (e.g., receptor-mediated, cytotoxicity related) of closely related
carcinogenic structural analogs. Examination of the physicochemical and biochemical properties of the
subject chemical may then allow one to assess the likelihood that such a mechanism also may be
applicable to the chemical in question and to determine the feasibility of conducting SAR analysis
based on the mechanism.

2.6.3. Metabolism and Pharmacokinetics
       Studies of the absorption, distribution, biotransfbrmation and excretion of agents are used to
make comparisons among species to assist in determining the implications of animal responses for
human hazard assessment, to support identification of lexicologically active metabolites, to  identify
changes in distribution and metabolic pathway or pathways over a dose  range  and between species,
and to make comparisons among different routes of exposure.
       In the absence of data to compare species, it is necessary to assume that pharmacokinetic and
metabolic processes are  qualitatively comparable.  If data are available (e.g., blood/ tissue partition
coefficients and pertinent physiological parameters of the species of interest), physiologically based
pharmacokinetic models can be constructed to assist in determination of tissue dosimetry, species-to-
species extrapolation of  dose, and route-to-route extrapolation (Connolly and Andersen, 1991).
       Analyses  of adequate metabolism and pharmacokinetic data can be applied toward the
following as data  permit. Confidence in conclusions is greatest when in vivo data are available.
       a.      Identifying metabolites and reactive intermediates of metabolism and determining
               whether one or more  of these intermediates are likely to be responsible for the
               observed effects. This information on the reactive intermediates will support and
               appropriately focus SAR analysis, analysis of potential mechanisms of action, and, in

                              DRAFT-DO NOT QUOTE OR CITE

               conjunction with physiologically based pharmacokinetic models, estimation of tissue
               dose in risk assessment (D'Souza et al., 1987; Krewski el al., 1987).
       b.      Identifying and comparing the relative activities of relevant metabolic pathways in
               animals with those in humans. This analysis can give insight on whether extrapolation
               of results of animal studies to humans will produce useful results.
       c.      Describing anticipated distribution within the body, and possibly identifying target
               organs.  Use of water solubility, molecular weight, and structure analysis can support
               inferences  about anticipated qualitative distribution and excretion. In addition,
               describing whether the agent or metabolite of concern will be excreted rapidly or
               slowly or will be stored in a particular tissue or tissues to be mobilized later can
               identify issues in comparing species and formulating dose-response assessment
       d.      Identifying changes in pharmacokinetics and a metabolic pathway or pathways with
               increases in dose. These changes may result in the formation and  accumulation of
               toxic products following saturation of detoxification enzymes.  These studies have an
               important role in providing a rationale for dose selection in carcinogenicity studies.  In
               addition, these studies may be important in estimating a dose over a  range of high to
               low exposure for the purpose of dose-response  assessment.
       e.      Determining the bioavailability of different routes  of entry by analyzing uptake
               processes under various exposure conditions. This analysis supports  identification of
               hazard for untested routes of entry.  In addition, use of physicochemical data (e.g.,
               octanol-water partition coefficient information)  can support an inference about the
               likelihood  of dermal absorption (Flynn, 1990).
       In all of the above-listed areas of inquiry,  attempts are made to clarify and describe as much as
possible the variability to be expected because of differences in species, sex, age, and route of entry.
Utilization of pharmacokinetic information takes into account that there may be subpopulations of
individuals who are particularly vulnerable to the effects of an agent because of
metabolic deficits or pharmacokinetic or metabolic differences (genetically or environmentally
determined) from the rest of the population.

                              DRAFT-DO NOT QUOTE OR CITE

2.6.4.  Mechanistic Information
       {The material in this section is only a start. Substance-specific risk assessments may
       have little or no data in this category.  Even when data are available, there is no
       standard for what is acceptable or what to expect.  If there are no data, we will have to
       use default assumptions. How much information is enough is difficult to say until testing
       in this area is more regular.)

       "Knowledge of carcinogenic mechanisms is incomplete in all cases.  Information on how
particular agents are likely to cause cancer may, however, be useful for appreciating more accurately
the hazard that such agents pose to humans" (IARC, 1991). Results from short-term lexicological tests
and molecular and cellular mechanistic studies are also useful in the interpretation of epidemiological
and rodent chronic bioassay data used in hazard identification and characterization.  These data may
provide guidance for dose-response modelling.
       Testing for tumorigenicity is usually done in long-term assays that involve exposure for much
of an animal's lifespan.
       Data from the long-term animal studies and the toxicity studies preceding them (e.g., evidence
of lesion progression, or lack of progression, and hyperplasia at the same site as the neoplasia) may
suggest a line of inquiry for further study. Cell necrosis is often an early finding (e.g., 20-90 days)
and provides indirect  evidence for subsequent tissue regeneration and compensatory growth
mechanisms when these events are not directly observed.  Other early changes observed during pre-
chronic studies range  from biochemical changes to altered hormone levels to organ enlargement
(hyperplasia) to specific and marked histopathological changes (Hildebrand et al., 1991).
       Conventional animal cancer bioassays provide little information on mechanism of action.
Short-term animal assays generally have more defined study designs to provide information about
potential mechanisms of action. A large number of short-term assays examine biological activities
relevant to the carcinogenic process (e.g., mutagenesis, tumor promotion, aberrant intercellular
communication, increased cell proliferation, malignant conversion, immunosuppression). In the future,
mechanistic-based end points should play an increasing, and perhaps major, role in the assessment of
cancer risk. Genetic Toxicity Tests
       Information on genetic damaging events induced by an agent is revealing about the possible
mechanism of action of a carcinogen.  Although the effectiveness of genetic toxicology tests in

                              DRAFT-DO NOT QUOTE OR CITE

predicting cancer has been questioned (Brockman and DeMarini, 1988), the ability of these tests to
detect mutagenic carcinogens has not been seriously challenged (Brockman and DeMarini, 1988; Prival
and Dunkel, 1989; Tennant and Zeiger, 1992; Shelby et al., 1992; Jackson et al.,  1992).
       Recent studies on oncogenes provide evidence for the linkage between mutation and cancer
(Bishop,  1991); activation of protooncogenes to oncogenes can be triggered, for example, by point
mutations, DNA insertions, or chromosomal translocation (Bishop, 1991).  In addition, the inactivation
of tumor suppressor genes  (anti-oncogenes) can occur by chromosomal deletion or aneuploidy
(chromosome loss), and mitotic recombination (Bishop,  1989; Varmus, 1989; Stanbridge and Vavenee,
       Genetic toxicology tests have been described  in various reviews (Brusick, 1990; Hoffman,
1991). The EPA has published various testing requirements and guidelines for detection of
mutagenicity (USEPA, 199la).  A useful method to "portray" data graphically, and which provides a
reasonable starting point for analysis, is the genetic activity profile (GAP) methodology developed by
the USEPA (Garrett et al.,  1984; Waters et al., 1988).
       Many test systems  have been developed to assay agents for their mutagenic potential3. These
include assays for changes in DNA base pairs of a gene (i.e., gene mutations) and microscopically
visible changes in chromosome structure or number.  Structural aberrations include deficiencies,
duplications, insertions, inversions, and translocation.  Other assays that do not measure gene
mutations or chromosomal aberrations per se provide  some information on an agent's DNA damaging
potential  (e.g., tests for DNA adducts, strand breaks, repair, or recombination).
       Distinguishing a carcinogenic agent as a mutagen or nonmutagen is an  important decision
point in defining the mechanism of action. To designate a putative carcinogen as a mutagen,  there
should be confidence that the primary target is DNA.  Mutagenic end points that involve stable
changes in DNA structure are emphasized because of their relevance  to carcinogenesis. These include
gene mutations and chromosomal aberrations.
       To be of value in cancer risk assessment, genetic toxicology data must meet the demands of
scientific scrutiny.  A higher level of confidence that  a carcinogen is  a mutagen is assigned to agents
that consistently induce direct structural changes in DNA in a number of test systems.  Although
important information can be gained from in vitro assays, a higher level of confidence is given to a
data set that includes in vivo evidence. In vivo data is emphasized because many agents require
    3 Ability to induce heritable or stable alterations in DNA structure and content.
                                             30                                      11/05/92

                              DRAFT-DO NOT QUOTE OR CITE

 metabolic conversion to an active intermediate for biological activity.  Metabolic activation systems
 can be incorporated into in vitro assay; however, they do not always mimic mammalian metabolism
 perfectly. If available, human genetic toxicity end points relevant to carcinogenesis are important in
 vivo data.                                                                      ,
        It is not possible to illustrate all potential combinations of evidence, and considerable judgment
 must be exercised in reaching conclusion.  Certain responses in tests that measure DNA damaging
 potential (e.g., DNA repair activity, adducts or strand breakage in DNA) other than gene mutations and
 chromosomal aberrations may provide a basis for raising the level of confidence in designating a
 carcinogen as mutagenic.
        There are many other mechanisms by which agents cause genetic damage secondary to other
 effects. For example, an agent might interfere with DNA repair or possibly increase DNA damage
 through an increase in oxidative radical production (Ceratti et al., 1990). Reliance on evidence for
 induced gene mutations or chromosomal aberrations to define a mutagenic  carcinogen is not meant to
 downplay the importance of these secondary mechanisms or other genetic end points.
        Aneuploidy (i.e, a change in chromosome number) may play an important role in the
 development of some tumors  (Kondo et al., 1984; Cavenee et al., 1983; Barrett et al., 1985), but it
 may result from  interactions with cellular components (e.g., mitotic apparatus) other than with DNA.
 For this reason, aneuploidy is not considered evidence for designating a carcinogen as mutagenic.
 Aneuploidy is important information regarding potential carcinogenicity by other genetic mechanisms
 and should be factored into the evaluation concerning mechanisms of action.
       Because  mutagenic carcinogens have been observed to  induce tumors across species and at
 multiple ;sites, evidence of both mutagenicity and tumor responses in multiple species or sexes
 significantly increases concern for the human carcinogenic potential of an agent.  Absence of
 mutagenicity in multiple test systems gives insight into alternative mechanisms by which non-
 mutagenic carcinogens may act. The consideration of alternative non-mutagenic mechanisms does not
 necessarily provide a basis for discounting positive results in the animal cancer bioassay and thus
 does not negate the concern for human risk.  On the other hand, evidence for non-mutagenicity and
 the lack of responses in a chronic rodent bioassay increases the confidence  that an agent is not a
human hazard.

                              DRAFT-DO NOT QUOTE OR CITE
2.6.43. Other Short-Term Tests
       In addition to genetic toxicity tests, information on increased cell proliferation, cell
transformation, aberrant intercellular communication, receptor mediated effects, changes in gene
transcription (i.e., events that involve a change in the function of the genome) can provide useful
information in the evaluation of mechanism of action and insight into the carcinogenic potential of an
agent. It is not possible to describe all the data that might be encountered in a substance-specific
assessment. Thus, the most conventional ones or those that are currently emphasized are mentioned as
       Cell proliferation plays a key role at each stage in the carcinogenic process and it is well
established that increased rates of cell proliferation are associated with increased cancer risk.  This
increased risk is due to  the increased susceptibility of proliferating cells to both spontaneous genetic
damage as  well as that induced by mutagens. Therefore, mitogenic activity in a mutagenic agent could
be expected to further increase the probability of mutagehesis and, therefore,  carcinogenesis.  Cell
proliferation or mutation alone are insufficient to cause neoplasia; further events are required for cells
to escape from growth control, to attain the ability  to grow independently, and to acquire invasiveness.
       Evidence for the increased rate of cell division may be determined by measuring the mitotic
index, or by supplying a specific DNA precursor to the cell (e.g., 3H-thymidine or bromodeoxyuridine)
and counting the percentage of cells that have incorporated the precursor into the replicating DNA, or
by immunodetection of  proliferation-specific antigens.  These analyses are carried out in vitro, during
pre-chronic studies, or as part of the long-term animal cancer bioassay.
       Non-mutagenic  carcinogens are more likely than mutagenic carcinogens to affect a specific sex
or organ.  Stable cell populations with  a potential for a high rate of cell replication are more often
affected than cell populations with a naturally high rate of replication. These properties have been
used to develop  two stage initiation-promotion studies  based on  preneoplastic lesions or tumors of the
mammary gland, urinary bladder, forestomach, thyroid, kidney, and liver.  Such tests provide
mechanistic insight as well as supportive evidence for carcinogenicity (Drinkwater, 1990).
       Several short-term tests respond to both mutagenic and non-mutagenic carcinogens.  Assays
for measuring perturbation of gap-junctional intercellular communication may provide and indication
of carcinogenicity, especially promotional activity,  and provide mechanistic information (Yamasaki,
1990). Cell transformation assays have been widely used for studying mechanistic aspects of chemical
carcinogenesis because in vitro cell transformation  is considered to be relevant to the in vivo
carcinogenic process.

                               DRAFT-DO NOT QUOTE OR CITE

 2.6.43.  Short-Term Assays for Carcinogenesis
        In addition to more conventional long-term animal studies, other shorter-term animal models
 can yield useful information about the carcinogenicity of agents. Some of the more common tests
 include mouse skin (Ingram and Grasso, 1991), transplacental and neonatal carcinogenesis (Ito,  1989),
 mammary gland tumor studies and preneOplastic lesions or altered cell foci (e.g., in liver, kidney,
 pancreas). Currently, increased research emphasis is being put on alternative approaches to the  chronic
 rodent cancer bioassay.  As an example, significant progress is being made using fish models (Bailey
 et al., 1984; Couch and Harshbarger,  1985).  Evaluation of Mechanistic Studies
        The entire range of data about an agent's physical-chemical properties, structure-activity
relationships to carcinogenic agents, and biological activity in vitro and in vivo is reviewed for
mechanistic insights.  The weight and significance of the observation of carcinogenic activity of the
agent in vivo can be greatly influenced by the available data in several areas, all of which should be
considered. Discussion should summarize available data  on the agent's effects on DNA structure or
expression and its effects on the cell cycle. Types of information to be considered include: whether
the agent is a mutagenic or a non-mutagenic carcinogen,  specific effects on proto-oncogenes or tumor
suppressor genes and DNA transcription, and structural or functional analogies to agents with the
above effects.
        Information demonstrating effects on the cell cycle would include: mitogenesis, effects on
differentiation, effects on cell death (apoptosis), tissue damage resulting in compensatory cell
proliferation, receptor-mediated effects on growth-signal transduction, and structural or functional
analogies to agents with  the above effects.
        Information demonstrating effects on cell interaction might include: effects on contact
inhibition of growth, intracellular communication, or immune reactions, and structural or functional
analogies to agents with  these effects.
        These are not intended to be exclusive of other pertinent data not specifically listed.  In
addition, available data on the comparative pharmacokinetics and metabolism of the agent in animals
and humans is assessed to consider whether similar mechanisms of action may be operating in humans
and animals.  (A similar summarization of evidence has been reported by IARC,  1991).
        In evaluating carcinogenic potential and mechanism of action, analyses and conclusions based
on short-term tests are accompanied by a discussion of the level of confidence that can be applied to

                              DRAFT-DO NOT QUOTE OR CITE

all the data. The level of confidence is based on the following (not necessarily exclusive) factors: (a)
the spectrum of endpoints relevant to carcinogenesis and the number of studies used for detecting each
end point and consistency of the results obtained in different test systems and different species, (b) in
vivo as well as in vitro observations, (c) the consistency and concordance of test results, (d)
reproducibility of the results within a test system, (e) existence of a dose-response relationship, and (f)
whether the tests are conducted in accordance with appropriate protocols agreed upon by experts in the
field. For, example, a high level of confidence in describing the potential influence of an agent on
carcinogenic events is based on results covering a number of events relevant to stages of
carcinogenesis, a number of studies including in vivo tests showing consistent trends and good
concordance. A low confidence data set is one that was sparse or has incongruous results and no clear
data trends.
       The strength of an hypothesis about mechanism of action generated by analysis of data in the
above areas should be described by the following criteria:
       a.     The operation of the mechanism in carcinogenesis must have been explained by a body
              of research data and have been generally accepted in the scientific community as a
              mechanism of carcinogenesis;
       b.     There must be a body of experimental data that show how the agent in question
              participates in the mechanism of action. In the absence of data about the mechanism
              of action of an agent, decisions are made using default assumptions:
       c.     That animal effects are relevant to human effects; and
       e.     That the agent affects carcinogenesis with dose and response relating linearly at low
       Both of  these science policy assumptions are supported by current knowledge of carcinogenic
processes, in the absence of better data.  Each assumption must be examined in substance-specific risk
assessments and replaced or joined by alternative analysis when adequate scientific data exist.

(Criteria and examples for categorization of experimental evidence are major issues, particularly
the weight of evidence contribution of research data of new kinds of genes and signal
transduction pathways of growth control.}

                              DRAFT-DO NOT QUOTE OR CITE

        A summary is made of all the experimental evidence that is relevant to human carcinogenic
        The confidence of an agent is potentially carcinogenic for humans increases as the number of
animal species, strains, or number of experiments and doses showing a carcinogenic response
increases. It also increases as the number of tissue sites affected by the agent increases and as the time
to tumor occurrence or time to death with tumor decreases in dose-related fashion. Confidence also
increases as the proportion of tumors that are malignant increases with dose and if the observed tumor
types are historically rare in the species.                                          •
{The appropriate use of molecular biological data in the overall weight of evidence is a question.
The strength of inferences to be drawn from data such as tumor susceptibility or gene effects is
an unsettled issue.}
       The weight of other experimental evidence increases or decreases the weight of findings
relevant to human hazard in the following ways listed below.  Findings in vivo add to the weight of
evidence more rapidly than in vitro findings.
       •       physical-chemical properties and structural or functional analogies can support
               inferences of potential carcinogenicity;
       •       results in a number of short-term studies that are consistent can support inferences
               about potential human effects;
       •       evidence of mutagenic effects on proto-oncogenes or tumor suppressor genes;
       •   -   - evidence of effects on cell growth  signal transduction affecting cell division,
               differentiation; or cell death; and
       •       induction of neoplastic behavioral characteristics in cells in culture or in vivo.
       The summarization of experimental evidence refers only to the weight of evidence that an
agent may or may not be carcinogenic in humans,  not the  dose-response relationship, which is the
subject of a separate analysis.
       The following four categories are used to summarize all of the experimental  data relevant to
inferences about human  carcinogenic potential of an agent. Tumor responses that the Agency has
found to be not relevant for inferring human hazard are not given weight.  Other responses whose
relevance is unresolved are noted in  the categorization of evidence.  Categorization is a matter of

                              DRAFT-DO NOT QUOTE OR CITE

scientific judgment, and the descriptions below are to be used as guidance in making that judgment,
not as absolute criteria.

2.7.1.  Category 1
       The following examples illustrate persuasive evidence of carcinogenic potential. Other
combinations of data also may be persuasive. In prospect, continued research on the role of agents in
mutations of proto-oncogenes and tumor suppressor genes and related research on receptor-mediated
effects on growth control genes also may provide persuasive data.

1.  Long-term animal experiments showing increased malignant and benign tumors
       a.  when the increased incidence of tumors is in more than one species or in more than one
       experiment (i.e., results are  complicated with different routes of administration, or affect a
       range of dose levels)
                       at multiple sites, or
                       at a limited number of sites with a supporting weight of evidence from
                       structure-activity analysis, or available short-term tests;
       b.  when there is a response to an unusual degree in a single experiment with regard to high
       incidence of a low-incidence background tumor, unusual site or type of tumor, or early age at
                       with a dose-related increase in a highly malignant tumor or in early death with
                       cancer, or
                       with a supporting  weight of evidence from structure activity analysis or from
                       available short-term studies; or
        c.  in more than one experiment, at a single site
                       with a highly supportive weight of evidence from SAR analysis and numerous
                       consistent findings of effects on carcinogenic processes in short-term studies,
                       with a dose-related increase in tumor malignancy.

                              DRAFT-DO NOT QUOTE OR CITE

2. Evidence that an agent is readily converted to a metabolite for which independent human or animal
evidence is categorized as Group 1 and data are supportive of like pharmacokinetic disposition, or
short-term studies of the agent are comparable in result with those of the metabolite.

3. Short-term experiments that demonstrate an agent's influence on carcinogenic processes in vivo
consistent with in vitro studies, SAR, and physical-chemical properties that are highly supportive of
carcinogen activity. These are supported by studies showing comparable metabolism and
pharmacokinetics between study species and humans.

2.7.2.  Category 2
       Examples for this category include:
1. A long-term animal experiment or experiments showing increased incidence of malignant tumors or
combined malignant and benign tumors that falls short of the weight for categorization as Category 1.

2. Evidence that an agent is readily converted to a metabolite for which independent human or animal
evidence is Category 2 and data are supportive of like pharmacokinetic disposition, or short-term
studies of the agent are comparable in result with those of the metabolite.

3. Short-term studies and other evidence as described in together with data supporting the
likelihood of comparability in metabolism and pharmacokinetics between species.

2.7.3.  Category 3
       The experimental evidence does not support a conclusion either way about potential
carcinogenicity because:
               too few data are available;
               evidence is limited to tumorigenicity and is found solely in studies in which the
               manner of administration (e.g., injection) or other aspects of study protocol present
               difficulties of interpretation; or
       •       evidence of carcinogenicity is found at a single animal site in one species and sex in
               one or more experiments; the  response is weak and without characteristics that give
               weight to a conclusion about potential human carcinogenicity.

                             DRAFT-DO NOT QUOTE OR CITE

       For example, data are inconclusive if experimental data apart from the animal response do not
support any positive inference about the agent's carcinogenic potential and if the animal response has a
consistent pattern of most of the following characteristics:
              At least two species have been tested, and the tumor response is seen only at the
              highest dose, in one sex, and one species.
              The tumor incidence is predominantly "benign and is seen only in one target .organ.
       •      The tumor is recognized as a common tumor type in that species, strain, and  sex. In
              addition, the observed tumor rate, although statistically significant in the experiment, is
              at or near the upper range of the historical control incidence.
              The tumors do  not cause death in the affected animals during the duration of the study
              and do not appear sooner in the treated animals than in the controls.
       Such evidence may add some weight to results of the human studies.

2.7.4.  Category 4
       This summarization would apply when no increased incidence of neoplasms has been observed
in at least two well-designed and well-conducted animal studies in different species including both
sexes.  The exposures are specified and the implication is that either the agent is not carcinogenic or
the studies had insufficient power to detect an effect.

       Evidence from all of the elements of hazard assessment are drawn together for an overall
characterization of potential human hazard as indicated in Figure 1.

                        DRAFT-DO NOT QUOTE OR CITE
Figure 1

                             DRAFT-DO NOT QUOTE OR CITE

2.8.1.  Purpose and Content of Characterization
       The major lines of observational human evidence and experimental evidence and reasoning are
clearly described. Major judgments made in the face of conflicting data are particularly highlighted
and explained, as are the assumptions or inferences made to address gaps in information. The
strengths and weaknesses of the available data are described and related to resulting confidence in the
characterization.  The hazard characterization addresses not only the question of carcinogenic
properties, but also, as data permit, the question of the conditions (dose, duration, route) under which
these properties may be expressed.
       To provide a basis for combining hazard and environmental exposure data in the final risk
characterization,  the hazard characterization points to differences expected according to route of
exposure, if such differences can be determined.  The assumption is made that the hazard is  not route-
specific, if this is reasonable and not contradicted by existing data. Information about the plausible
mechanism or mechanisms of action is characterized and its implications for dose-response assessment
are explained, including  conditions of dose and duration.

2.8.2.  Weight of Evidence for Human Carcinogenicity

(NOTE TO THE READER:  The question as to whether to abandon our alphanumerical system
entirely or merge it with a narrative statement has not been decided. We may retain labels of A,
B, C, etc., labels for weight of evidence groups.}
       A brief narrative statement is used to summarize the weight of evidence.  It incorporates
judgment about data from all elements of hazard assessment  A summary statement cannot resolve
data interpretation issues; it can only focus judgments and help convey them.  The purpose is to give
the risk manager a sense of the evidence and of the risk assessor's confidence in the data and their
interpretation for the assessment of human carcinogenicity potential and to allow comparison of weight
of evidence judgments from case to case. A weight of evidence conclusion incorporates judgments
both about overall confidence in a set of data as a basis for drawing conclusions and about the
consistency and congruence of inferences supported by the set of data.
       A weight of evidence conclusion is based both observational data from human studies and
experimental data.  All of the elements of analysis included in hazard assessment form the basis of
judgment  The summarizations of experimental evidence and human evidence are ingredients for a
weight of evidence statement. Note that animal tumor responses that the Agency considers not

                              DRAFT-DO NOT QUOTE OR CITE
relevant for inferring human hazard are not weighed.  However, unresolved questions about relevance
are all noted and considered in the statement.
       As the first step, a decision is made on whether the evidence is adequate or not adequate for
characterization. "Not adequate" means that the existing data are inadequate overall to support a
conclusion because either there are too few data or the data are flawed due to experimental design or
conduct, or because findings are not substantial enough to support inferences either way about
potential human carcinogenicity.  Typically, human or experimental data that are in Category 3 would
be considered as not adequate for characterization.
       If the evidence is adequate for a weight-of-eviderice determination, it is described within a
narrative statement. The narrative statement explains  the weight of evidence by summarizing the
content and contribution of individual lines of evidence and explaining how they combine to form the
overall weight of evidence. The statement highlights  the quality and extent of data and the
congruence, or lack of congruence, of inferences they support. The statement also highlights default
assumptions used to address gaps in knowledge.
       The statement gives the weight of evidence by route of exposure, pointing out the basis of
anticipated differences and whether the default assumption supporting extrapolation of hazard potential
between routes has been used and is appropriate. Anticipated potency differences by route  are pointed
out, based on comparatively poor to  ready absorption  by.different routes (see § 2.6.3. Metabolism and
Pharmacokinetics).                                              ,
       The statement discusses the data implications  for mechanism  of action^ It recommends a
general approach or approaches for dose-response assessment in accordance with what the hazard data
imply about the nature of dose-response below the range of observation of ^available studies.  A weight
of evidence for hazard by any mechanism is characterized.  Thus, for example, an agent that is
estrogenic and not  likely to cause permanent genetic changes is characterized as a carcinogenic hazard,
with any limitations of dose being explained in the narrative statement.  The quantitative dose-response
estimation or shape of the dose-response curve does not affect the weight of evidence for hazard.
       The statement notes whether its source is an individual EPA office or an EPA consensus.  The
overall conclusion  is noted by use of one of the following descriptors: "known,"  "highly likely," or
"likely" to be a human carcinogen; "some evidence" or "not likely to be a human carcinogen at
exposure levels studied or alternately under conditions of environmental exposure." These
descriptors fall along a continuum of likelihood that an agent has human carcinogenic potential.  More
than one descriptor may apply to a single agent if the weight of evidence differs by route of

                             DRAFT-DO NOT QUOTE OR CITE

administration.  Also, two descriptors may be applied if the evidence for a route is judged to fall
between two descriptors. These standard descriptors are provided for the purpose of maintaining
consistency of expression of conclusions from case to case. The text of the narrative statement as a
whole is the primary means of conveying information on the weight of evidence.

2.83.1.  Descriptors

{The number of descriptor categories for total weight  of evidence is a continuing issue. The
evidence is along a continuum. How many descriptors are needed to represent the continuum?
What are the criteria for establishing them?}

       Explanations of the general levels of evidence associated with descriptors  in terms of the
summarizations of evidence made in the course of a hazard assessment are as follows:
       "Known" to be carcinogenic in humans is a statement that evidence is convincing (Category
1) that the agent has observed carcinogenic effects in humans by a specified route or routes of
       "Highly likely" is a statement that:
       1.      there is persuasive experimental evidence of carcinogenicity (Category 1) and
               suggestive  human evidence (Category 2), or
       2.      there is persuasive experimental evidence (Category  1) showing a very strong animal
               response (multiple tumor sites in more than one species), or
       3.      an agent is known to be a carcinogen in humans by one route of exposure (known) is
               also absorbed by another route, making carcinogenic effects " highly likely" by the
               second route.

       "Likely"  is a statement that:
       1.      there is persuasive experimental evidence (Category  1),  or
       2.      there is suggestive evidence from human data (Category 2) with experimental evidence
               (Category 2) that supports the likelihood  that the human effects seen were due to  the
               agent in question.

                              DRAFT-DO NOT QUOTE OR CITE

       "Some evidence" is a statement that:
       1.     there is experimental evidence (Category 2), or
       2.     suggestive human evidence (Category 2).
       However, the totality of the evidence is weak because findings are inconsistent, or there are
many gaps in the data.

       "Not likely to be a human carcinogenic at exposure levels studied or alternately, under
       conditions of environmental exposure" is a statement that:
       1.     human evidence has been summarized as no evidence at exposure levels studied
              (Category 4), and there are no positive animal findings, or
       2.     experimental evidence has been summarized as no evidence at exposure levels studied
              (Category 4), and there are no positive human findings, or
       3.     the occurrence of carcinogenic effects is not expected for a particular route of human
              environmental exposure (oral, dermal, inhalation) because the agent is not absorbed by
              that route, or
       4.     the mechanism of carcinogenicity of an agent operates only at doses above  the range
              of plausible environmental exposure, e.g., carcinogenesis as a secondary effect of
              another effect that occurs only at high doses, or
       5.     the occurrence of carcinogenic effects depends on administration of the agent in a
              manner that has no parallel with plausible environmental exposure, e.g., injection of
              polymers.   .
       This descriptor is explained in the narrative statement as being applicable only to the specific
exposure levels studied or environmental exposure conditions which are given in the statement. Examples of Narrative Statements
                                         Compound X
       Following review of all available data relevant to the potential human carcinogenic hazard of
       X (CAS # 000001), the ....  Office of EPA concludes that X is not likely to be carcinogenic
       to humans by any route of exposure at environmental levels.  This determination is based on
       experimental evidence. No human studies on X are available for evaluation. The evidence
       supporting this  finding is the animal response.
       With dietary administration, X caused a statistically significant increase in the incidence of
       urinary bladder hyperplasia and tumors (urinary bladder transitional cell papillomas and

                       DRAFT-DO NOT QUOTE OR CITE
carcinomas) in male but not in female Charles River CD rats at high dose levels (>30,000
ppm).  The tumors were seen only at dose levels producing calculi in the kidneys, ureters and
the urinary bladder. The presence of the urinary bladder calculi was associated with a decrease
in the urinary pH.  The urinary bladder calculi were almost always associated with urinary
bladder hyperplasia (>90%). A major metabolite of X did  not cause any increase in tumor
incidence in another bioassay in rats. X was not carcinogenic in mice in well-conducted

The in vivo (mouse micronucleus test) and in vitro (in bacteria and yeast) short term- studies
on X indicate with medium confidence that X is not genotoxic. Structure-activity-relationship
analysis reveals no chemicals which are related to X and also induce tumors.     It is
concluded that the tumor response in male rats was secondary to stone formation at high
doses, and may be a phenomenon unique to the male rats.  No dose-response analysis is
recommended unless a high-dose environmental exposure to humans is discovered.

                                  Compound Y

Following review of all available data relevant to the potential human carcinogenic hazard of
Y (CAS # 000002), EPA concludes that Y is likely to be carcinogenic to humans by all routes
of exposure.  This determination is based on experimental evidence. No  human studies are
available for evaluation. The strongest lines of evidence supporting findings on Y are animal
experiments and structure-activity relationships.

Rodent studies showed statistically significant increases in the incidence of liver tumors
(hepatocellular adenomas and carcinomas combined) in two strains of mice, in two
independent and adequately conducted studies.  The increases of liver tumors occurred at high
and low doses. Y also produced a statistically significant increase in stomach tumors
(papillomas) in both male and female mice at a dose also producing significant mortality and
reduced body weight (-18% to -23% throughout the study)  and the presence of white foci and
ulcers in the stomach of occasional animals.

Y, administered orally, did not induce tumors in F344 rats in an adequately conducted study.
Data from acute inhalation toxicity and dermal absorption studies  show that Y  is absorbed by
both dermal and inhalation exposure.

Y caused gene mutations and chromosome aberrations in JJ. melanogaster and  DNA damage in
yeast, but it did not induce mutagenic effects in either in vitro or in vivo mammalian systems.
The mutagenicity data set is of low confidence, and it neither supports nor contradicts
inferences about carcinogenicity.  In addition, it does not suggest a mechanism of action.

Structure-activity relationship analysis shows that Y is very closely related in structure to eight
other chemicals, all of which produce liver tumors in mice, rats, or both.

Based upon the above analysis, it is suggested that the dose-response analysis employ a default
assumption of linearity at low dose and consider the liver tumor in mice as an  appropriate

                              DRAFT-DO NOT QUOTE OR CITE

                             3.  DOSE-RESPONSE ASSESSMENT

        Dose-response assessment tests the hypothesis that an agent has produced an effect and
portrays the relationship between the agent and the response elicited.  In risk assessments, dose and
response observations from experimental or epidemiological studies are often projected to much lower
exposure levels encountered in the environment.4  In addition, the mathematical models used for
extrapolation are based on general assumptions about the nature of the carcinogenic process.  These
assumptions may be untested for the particular agent being evaluated (Kodell, in press).  If the dose-
response relationship is developed from an experimental animal study, it also must  be extrapolated
from animals to humans. Because of these inherent uncertainties, projections well outside the range of
the observed data are treated as bounding estimates, not as true values.
Information that shows a comparable pharmacokinetic and metabolic response to an agent in humans
and animals greatly increases confidence in the dose-response analysis. Data suggesting  that an agent
works through a common mechanism of action in humans and animals also greatly increases
confidence in the low dose extrapolation.  In the absence of such data, default approaches provide
upper-bound estimates of response at low doses, with a lower limit as small as zero at very low doses.
In the absence of dose-response data on members of a class of agents, it may be possible to construct
a set of toxicity equivalence factors (TEF) to be used to quantify dose-response by  reference to an
already-characterized member of the class.

3.2. Elements  of Dose-Response Assessment
       The elements of dose response analysis include selection of response data and dose data,
followed by a stepwise dose-response analysis. The first step in the dose-response  analysis is fitting of
the data in the range of study observation; the second step, if needed, is extrapolation of  the dose-
response relationship  to the range of the human exposure of interest.
    4 For this discussion, "exposure" means contact of an agent with the outer boundary of an
organism. "Applied dose" means the amount of an agent presented to an absorption barrier and
available for absorption; "internal dose" means the amount crossing an absorption barrier (e.g., the
exchange boundaries of skin, lung, and digestive tract) through uptake processes; and the amount
available for interaction with an organ or cell is the "delivered dose" for that organ or cell. For more
detailed discussion see Exposure Assessment Guidelines	FR	__:(1992).

                              DRAFT-DO NOT QUOTE OR CITE

A dose-response assessment should take advantage of available data to support a more confident
analysis. When data gaps exist, assumptions based on current knowledge about the biological events
in carcinogenesis and pharmacokinetic processes are used.
3.2.1.  Response Data
       Appropriate response data, as well as mechanistic information from the hazard characterization,
are applied in the dose-response assessment The quality of the data and their relevance to human
exposure are important selection considerations.
       If adequate positive human epidemiologic data are available, they are usually the preferred
basis for analysis.  Positive data are analyzed to estimate response to environmental exposure in the
observed range. (USEPA, 1992a).  Extrapolation to lower environmental exposure ranges is carried
out, as needed. If adequate exposure data exist in a well-designed and well-conducted epidemiologic
study that detects no effects, it may be possible to obtain an upper-bound estimate of the potential risk.
Animal-based estimates, if available, are also presented, and the animal results are compared with the
upper-bound estimate from human data for consistency.
       When animal studies are used, response data from a species that responds most like humans
should be used, if information to this effect exists.  When an agent was tested in several experiments
involving different animal species, strains, and sexes at several doses and different routes of exposure,
the following approach to selecting the data sets is generally used:
       a.      The tumor incidence data are separated into data sets according to organ site and
               tumor type.
       b.      All biologically and statistically acceptable data sets are examined.
       c.      Data sets are analyzed with regard to route of exposure.
       d.      A judgment is reached based on biological criteria as to which set or sets best
               represents the body of data for the purpose of estimating human response.  This
               judgment is augmented with judgment as to the statistical suitability of the data for
               modeling in  the experimental data range.  The hazard characterization is the point of
               reference for the initial judgment. The following characteristics of a data set favor its
                      •  high quality of study protocol  and execution;
                      •  malignant neoplasms;
                      •  earlier onset of neoplasm;

                              DRAFT-DO NOT QUOTE OR CITE

                      * greater number of data points to define the relationship of dose and
                      response;                                        -
                      • background incidence in test animal is not unusually high;
                      • most sensitive-responding species are used; or
                      • data on a related effect (e.g., DNA adduct formation) or mechanistic data to
                      augment the tumor.
        Appropriate options for presenting results include use of a single data set, combining data from
different experiments (Stilfler etal., 1992), showing a range of results from more than one data set,
representing total response in a single experiment by combining animals with tumors or a
combination of these options.  The rationale for selecting an approach is presented, including the
biological and statistical considerations involved; The objective is to  provide a best judgment of how
to represent the observed data.
        Benign tumors are usually combined with malignant tumors for risk estimation if the benign
tumors are considered to have the potential to progress to associated malignancies of the same
histogenic origin. (McConnell, 1986). When tumors are thus combined, the contribution to the total
risk of benign tumors is indicated.  The issue of how to consider the contribution of the benign tumors
should be discussed in the dose-response characterization and risk characterization.
        Data on certain endpoints related to tumor induction may be used to extend  dose-response
analysis below the relatively high dose range in which tumors are observable.  These data permit
extension of the curve-fitting analysis (S wenberg et al., 1987) and may provide parameters for
applying a mechanism-based model (US EPA DioxinAssessment, 1992c).  Data might include
information on receptor binding, DNA adduct formation, physiological effects such as disruption of
hormone activity, or agent-specific alterations in cell division rates. In considering whether such
endpoints can  be applied, key issues are confidence that the data reflect carcinogenic effects of the
agent and that these have been well measured with a dose-effect trend.

3.2.2. Dose Data
       Regardless of the source, animal experiments or epidemiologic studies, several questions need
to be addressed in arriving at an appropriate measure of dose. One question is whether data are
sufficient to estimate internal dose or delivered dose. Part of this question is whether the parent
compound, a metabolite, or both agents are closer in a metabolic pathway to a carcinogenic form.

                              DRAFT-DO NOT QUOTE OR CITE

       The delivered dose to target is the preferred measure of dose.  In practice, there may be little
or no information on the concentration or identity of the active agent at a site of action; thus, being
able to compare the applied and delivered doses between routes and species is an ideal that is rarely
attained. Even so, incorporating data to the extent possible is desirable.
       Even if pharmacokinetic and metabolic data are sufficient to derive a measure of delivered
dose to the target, the dose-response relationship is also affected by kinetics of reactions at the target
(pharmacodynamics) and by other steps in the development of neoplasia.  With few exceptions, these
processes are currently undefined.
       The following discussion assumes that the analyst will have data of varying detail in different
cases about pharmacokinetics and metabolism.  Approaches to limited data are outlined as well as
approaches and judgments for more sophisticated analysis based on additional data. Base Case — Few Data
        Where there are insufficient data available to define the equivalent delivered dose between
species, it is assumed that delivered doses at target tissues are directly proportional to applied doses.
This assumption rests on the similarities of mammalian anatomy, physiology, and biochemistry
generally observed across species. This assumption is more appropriate at low applied dose
concentrations where sources of nonlinearity, such as saturation or induction of enzyme activity, are
less likely to occur.
        The default procedure is to scale daily applied doses experienced for a lifetime in proportion to
body weight raised to the 3/4 power (W3/4).  Equating exposure concentrations in parts per million
units for air, food, or water is an alternative version of the same default procedure because daily
intakes of these are in proportion to W3/4. The rationale for this factor rests on the empirical
observation that rates of physiological processes consistently tend to maintain proportionality with
W3M. A more extensive discussion of the rationale and data supporting the Agency's adoption of this
scaling factor can be found in (USEPA, 1992b).
        The differences in biological processes among routes  of exposure (oral, inhalation, dermal) can
be great, due to, for example,  first pass effects and differing results from different exposure patterns.
There is no generally applicable method for accounting for these differences in uptake processes in
quantitative route-to-route extrapolation of dose-response data in the absence of good data on the  agent
of interest.  Therefore, route-to-route extrapolation of dose data will be based on a case-by-case
analysis of available data.  When good data on the agent itself are limited, an extrapolation analysis

                              DRAFT-DO NOT QUOTE OR CITE

 can be based on expectations from physical chemical properties of the agent, properties and route-
 specific data on structurally analogous compounds, or in vitro or in vivo uptake data on the agent.
 Route-to-route uptake models may be applied if model parameters are suitable for the compound of
 interest.  Such models are currently considered interim methods; further modeldevelbpment and
 validation is awaiting the development of more extensive data  (see generally, Gerrity and Henny,
 1990). Pharmacokinetic Analyses
       Physiologically based mathematical models are potentially the most comprehensive way to
 account for pharmacokinetic processes affecting dose.  Models build on physiological compartmental
 modeling and attempt to incorporate the dynamics of tissue perfusion and the kinetics of enzymes
 involved in metabolism of an administered compound.
       A comprehensive model requires the availability of empirical data on the carcinogenic activity
 contributed by parent compound and metabolite or metabolites and data by which to compare kinetics
 of metabolism and elimination between species. A discussion of issues of confidence accompanies
 presentation of model results (Monro, 1991).  This includes considerations of model validation and
 sensitivity analysis that stress the predictive performance of the model. Another assumption made
 when a delivered dose measure is used in animal-to-human extrapolation of dose-response data is that
 the pharmacodynamics of the target tissue(s) will be the same in both species. This assumption should
 be discussed, and confidence in accepting it should be considered in presenting results.
       Pharmacokinetic data can improve dose-response assessment by accounting for sources of
 change in proportionality of applied-to- internal dose or to delivered dose at various levels of applied
 dose.  Many of the sources of potential nonlinearity involve saturation or induction of enzymatic
 processes at high doses. An analysis that accounts for nonlinearity (for instance, due to enzyme
 saturation kinetics) can assist in avoiding over estimation or under estimation of low dose if
 extrapolation is from a sublinear or supralinear part of the experimental dose-response curve. (Gillette,
 1983). Pharmacokinetic processes tend to become linear at low doses, an expectation that is more
 robust than low-dose linearity of response (Hattis,  1990). Thus, accounting for nonlinearities allows
better description of the shape of the curve at higher levels of dose, but cannot determine  linearity or
nonlinearity of response at low dose levels (Lutz. 1990: Swenberg et al.,' 1987).

                             DRAFT-DO NOT QUOTE OR CITE Additional Considerations for Dose in Human Studies
       The applied dose in a human study has uncertainties because of the exposure fluctuations that
humans experience compared with the controlled exposures received by animals on test. In a
prospective cohort study, there is opportunity to monitor exposure and human activity patterns for a
period of time that supports estimation of applied dose (USEPA, 1992a). In a retrospective cohort
study, exposure is based on human activity patterns and levels reconstructed from historical data,
contemporary data, or a combination of the two.  Such reconstruction is accompanied by analysis of
uncertainties considered with sensitivity analysis in the estimation of dose (Wyzga, 1988; USEPA,
1986). These uncertainties can also be assessed for any confounding factor, for which a quantitative
adjustment of dose-response data is made  (USEPA, 1984).
       Exposure levels of groups  of people in the study population often are represented by an
average when they are actually in a range. The full range of data are analyzed and portrayed in the
dose-response analysis when possible (USEPA, 1986).
       The cumulative dose of an agent is commonly used when modeling human data. This can be
done, as in animal studies, with a default assumption in the absence of data that support a different
dose surrogate. Given data of sufficient quality,  dose rate or peak exposure can be used as an
alternative surrogate to  cumulative dose.

        Because risks at relatively low exposure levels generally cannot be measured directly either by
animal experiments or by epidemiologic studies of reasonable sample size, a number of mathematical
models have been developed to extrapolate from high to low dose. Different extrapolation models
may fit the observed  data reasonably well but may lead to large differences in the projected risk at
lower doses.  As was pointed out by OSTP (1985 see Principle 26), no single mathematical procedure
is recognized as the most appropriate for low-dose extrapolation in carcinogenesis. Low-dose
extrapolation procedures use either mechanistic or empirical models.  When sufficient biological
information exists to  identify and describe a mechanism of action, low-dose extrapolation may be
based on a mathematical representation of the mechanism.  When the mechanism is unknown or
information is limited, low-dose is derived from an empirical fit of a curve compatible  with the
available information.
        If a carcinogenic agent acts by accelerating the same carcinogenic process that  leads to the
background occurrence of cancer,  the added effect on the population at low doses marginally above

                              DRAFT-DO NOT QUOTE OR CITE

background level is expected to be linear.  Above background level, the population response may
continue to be linear in the case of an agent acting directly on DNA, or the population response may
be influenced by individual variability in sensitivity to phenomena such as disruption of hormone
homeostasis or receptor-mediated activity.  If the agent acts by a mechanism with no endogenous
counterpart, a population response threshold may exist (Crump et al., 1976; Peto, 1978; Hoel, 1980;
Lutz, 1990). The Agency reviews each assessment as to the evidence on carcinogenesis mechanisms
and other biological or statistical evidence that indicates the suitability of a particular extrapolation
model.  When longitudinal data on tumor development are available, time-to-tumor or survival models
may be used and are preferred. In all cases, a rationale is included to justify the use of the chosen
        The goal in choosing an approach is to achieve the closest possible correspondence between
the approach and the view of the agent's mechanism of action developed in the hazard  assessment. If
the hazard assessment describes more than one mechanism as plausible and persuasive given the data
available, corresponding alternative approaches for dose-response analysis are considered.

3.3.1. Analysis in the Range of Observation
        In portraying dose response in the range of observed  data, analyses incorporate as much
reliable information as possible.  Pharmacokinetie data or interspecies scaling is used to derive human-
equivalent measures of the animal-administered dose. The empirical response data analyzed include
tumor incidence data augmented, if possible, by incidence data on effects leading to the tumor
response, e.g., DNA adduct or other effect-marker data (Swenberg, 1987).
        Dose-response models span a hierarchy that reflects an ability to incorporate different kinds of
information. If data to support it are available, a mechanism-based procedure is the preferred approach
for modeling.  A mechanism-based procedure is explicitly devised to reflect biological  processes.
Theoretical values for parameters, e.g., theoretical cell proliferation rates, are not used to enable
application of a mechanism-based model  (Portier, 1987).  If such data are absent, a mechanism-based
model is not used. An example of a mechanism-based model is the  receptor mediated  toxicity model
for dioxin, under development at  EPA (U.S. EPA, 1992c).
        Dose-response models based on general concepts of a mechanism of action are next in amount
of information required. For a specific agent, model  parameters are  obtained from laboratory studies.
Examples are the two-stage models of initiation, clonal expansion, and progression developed by

                             DRAFT-DO NOT QUOTE OR CITE

Moolgavkar et al. (1981) and Chen et al. (1991).  Such models require extensive data to build the
form of the model as well as to estimate how well it conforms with the observed carcinogenicity data.
       Empirical models, which do not incorporate information about mechanism  of action, form the
rest of the hierarchy. Among these, time-to-tumor models incorporate longitudinal information on
tumor development.  Simple quanta! models use only the final incidence at each dose level.  The
linearized multistage procedure is an example of an empirical model.
       If a mechanism-based model is judged to be not suitable, the analysis uses an empirical model
whose  underlying parameters correspond to the putative mechanism of action identified in the hazard
characterization. A multistage model (Zeise et al., 1987) structured with time to response as the
random variable is appropriate when time is the dominant factor for probability of response.  This is
the approach when available information described in the hazard characterization is consistent with an
assumption that there is no threshold of response for individuals.  When the probability of effect is due
to the distribution of thresholds for individuals in the  population, a model considering dose as the
random variable may be used.  This may be considered an appropriate approach when the mechanism
has been identified as one such as disruption of hormone homeostasis.

(The issue of appropriate dose-response models is still under discussion at EPA.)

       Ordinarily, models are expected to provide an adequate fit to the observed dose-response
information. The outcome of most tests of goodness of fit to the observations is not an effective
means  of discriminating among models that all provide an adequate fit. Although  a model may
adequately fit the observed dose-response information, all models have limitations  in their ability to
describe the underlying processes and make projections outside the observed information.  A  prime
consideration is the potential for model error, that is the possibility that a model might appear to fit the
observed data but be based on an inadequate mathematical description of the true underlying
mechanism.  This is especially crucial when making inferences outside the range of observation, as
alternative models may provide an adequate fit to  the observed information but have substantially
different implications outside the range of observation.
       Sometimes an inadequate fit might be improved by incorporating more information.  For
example, data in which there is high mortality may be poorly fit unless competing risks of death by
toxicity are taken into consideration with time-to-tumor information and survival adjustments.  If an

                              DRAFT-DO NOT QUOTE OR CITE

adequate fit cannot be obtained, it may be necessary to give less weight to the observations most
removed from low-dose risk., e.g,, from the highest dose level in a study with several dose levels.
        Statistical considerations can affect the precision of model estimates. These include the
number and spacing of dose levels, sample sizes, and the precision and accuracy of dose
measurements.  Sensitivity analysis can be performed to describe the sensitivity  of the model to slight
variations in the observed data. A large divergence between upper and lower confidence bounds
indicates that the model cannot make precise projections in that range.  All of these considerations are
important in determining the range in which a model is supported by data.  :
        With the recent expansion of readily available computing capacity, computer-intensive methods
are being adapted to create simulated biological data that are comparable with the observed
information.  These simulations.....data as  a whole are idiosyncratic rather than typical of the true state
of risks           ;                                                           ,
        The lowest reliable area of a curve is  identified as  a result of the data modeling.  This point is
generally at the level of not less than a 1.0 percent response if only animal tumor response data are
available. (This 1.0 percent response level is  about an order of magnitude below the potential power
of a standard rodent study to detect effects.)  The lowest reliable area may be extended below a 1.0
percent response if based on a more powerful study, on  combined studies, or on joining the analysis
of tumor response  data with data on other markers of effect. This lowest reliable area provides an
estimate that can be used for comparision  with similar analyses of the observed range of noncancer
effects of an agent (USEPA, 1991f).
3.3.2.  Extrapolation
       Using the lowest reliable point from the first step of analysis as a point of departure, the
preferred approach for this second step of analysis still is a mechanism-based model, if data support it.
If a mechanism-based model has been used to portray  the observed data, the question in this step is
whether confidence in the model extends to using it for extrapolation.  If data are insufficient to
support a mechanism-based model, extrapolation is done by a default procedure whose parameters
reflect the general mechanism or mechanisms of action considered to be supported by the available
biological information.
       If the mechanism of action being considered leads to an expected linear dose-response
relationship, the linearized multistage model or a model-free approach may be appropriate (Gaylor  and
Kodell, 1980; Krewski.. 1984; Flamm and Winbush, 1984).

                                 DRAFT-DO NOT QUOTE OR CITE

          The mechanism of action being considered may project that the dose-response relationship in
   the population is most influenced by the differences in sensitivities.  In this case, a model including
   tolerance distribution parameters may be used to provide estimates of the proportion of the population
   at risk for specific doses of interest, e.g., 1/1000, 1/10000 lifetime risk levels. This approach requires
   data for a mathematical portrayal of the distribution.

   (NOTE: The appropriate empirical modeling approaches for extrapolation are an undecided
   issue when a putative mechanism of action has been recognized but data are not supportive of a
   mechanism-based model. Further technical analysis and discussion are necessary before this
   section can be completed.}

          Alternatively, the mechanism  may be one that involves a population threshold. In these cases,
   extrapolation is not made. Instead, a  "margin of exposure" presentation is made in the risk
   characterization.  The margin of exposure in this context is the lowest reliable dose-response area from
   observed data divided by the environmental dose level of interest.
   3.33. Issues for Analysis of Human Studies
          Issues and uncertainties arising in dose-response assessment based on epidemiological studies
   are analyzed in each case.   Several sources of uncertainty need to be addressed in the dose-response
   analysis.  Consideration needs to be given to the data on the exposure and mortality experience of the
   study population and of the population that will represent the background incidence of the neoplasm(s)
   involved. In this area, there are potentials for mistakes or uncertainty in the data or adjustments to the
   data concerning the occurrence or level of exposure of the population members, mortality experience
   of a population, incomplete follow-up of individuals, exposure (or not) of individuals to confounding
   causes, or consideration of latency of response.  These are assessed by analyzing the sensitivity of
   dose-response study results to errors where data permit.  Other kinds of uncertainty can occur because
   of small sample size which can magnify the effects of misclassification or change assumptions about
   statistical distribution that underlie tests of statistical significance (Wyzga, 1988). These uncertainties
   are discussed. Where possible, analyses of the sensitivity of results to the potential variability in the
   data in these areas are performed.
          The suitability of various available mathematical procedures for quantifying  risk attributed to
   exposure to the study agent is discussed.  These methods (e.g.; absolute risk, relative risk, excess
'** additive risk) account differently for duration of exposure and background risk, and one or more can


                              DRAFT-DO NOT QUOTE OR CITE

be used in the analysis as data permit. The use of several of these methods is encouraged when they
can be used appropriately in order to gain perspectives on study results.

3.3.4.  Use of Toxicity Equivalence Factors
       A toxicity equivalence factor (TEF) procedure is one used to derive quantitative dose-response
estimates for agents that are members of a category or class of agents.  TEFs are based on shared
characteristics that can be used to order the class members by carcinogenic potency when cancer
bioassay data are inadequate for this purpose (USEPA, 1991c). The ordering is by reference to the
characteristics and potency of a well-studied member or members of the class. Other class members
are indexed to the reference agent(s) by one or more shared characteristic to generate their TEFs.  The
TEFs are usually indexed at increments of a factor of 10. Very good data may permit a smaller
increment to be used. Shared characteristics that may be used are, for example, receptor- binding
characteristics, results of assays of biological activity  related to carcinogenicity, or structure-activity
       TEFs are generated and used for the limited purpose of assessment of agents or mixtures of
agents in environmental media when better data are not available.  When better data become available
for an agent, its TEF should be replaced or revised.
       Guiding  criteria for the successful application of TEFs are (USEPA,  1991c):
        1.     A demonstrated need. A TEF procedure should not be used unless there is a clear
              need to do so.
       2.     A well-defined group of chemicals.
       3.     A broad base of toxicological data.
       4.     Consistency in relative toxicity across lexicological endpoints.
       5.     Demonstrated additivity between toxicities of group members for assessment of
       6.     A mechanistic  rationale.
       7.     Consensus among scientists.
       The conclusions of dose-response analysis are presented in a characterization section.  Because
alternative approaches may be plausible and persuasive in selecting dose data, response data, or
extrapolation procedures, the characterization presenls the judgments made in such selections. The

                              DRAFT-DO NOT QUOTE OR CITE

results for the approach or approaches chosen are presented with a rationale for the one(s) that is
considered to best represent the available data and best correspond to the view of the mechanism of
action developed in the hazard assessment
       The exploration of significant uncertainties in data for dose and response and in extrapolation
procedures is part of the characterization.  They are described quantitatively if possible through
sensitivity analysis and statistical uncertainty analysis.  If quantitative analysis is not possible,
significant uncertainties are described qualitatively. Dose-response estimates are  appropriately
presented in ranges or as alternatives when equally persuasive approaches have been found.
       Numerical dose-response estimates are presented to one significant figure and qualified as to
whether they represent central tendency or plausible upper-bounds on risk or, in general, as to whether
the direction of error is to overestimate or under estimate risk. For example, the straight line
extrapolation used as a default is typically considered to place a plausible upper- bound on risk at low
doses. On the other hand, a tolerance distribution model used as a default to portray risk-specific
response distribution of the population may greatly underestimate risks if the mechanism is in fact a
linear, nonthreshold one. (Krewski, 1984).
       In cases, where a mechanism has been identified that has special implications for early-life
exposure, differential effects by sex, or other concerns for sensitive subpopulations, these are
explained.  Similarly, any expectations that high dose-rate exposures may alter the risk picture for
some portion of the population are described. These and other perspectives are recorded to guide
exposure assessment and risk characterization.

                                4. EXPOSURE ASSESSMENT

       Guidelines for exposure assessment of carcinogenic and other agents are published in USEPA,
1992a. The exposure characterization is a key part of the exposure assessment; it is the summary
explanation of the exposure assessment. The exposure characterization
       a.      provides a statement of purpose, scope, level of detail, and approach used  in the
               assessment;                                                              .
       b.      presents the estimates of exposure and dose by pathway and route for individuals,
               population' segments, and populations in a manner appropriate for the intended risk

                             DRAFT-DO NOT QUOTE OR CITE

       c.     provides an evaluation of the overall quality of the assessment and the degree of
              confidence the authors have in the estimates of exposure and dose and the conclusions
              drawn; and
       d.     communicates the results of exposure assessment to the risk assessor, who can then use
              the exposure characterization, along with the characterization of the other risk
              assessment elements, to develop a risk characterization.
       In general, the magnitude, duration, and frequency of exposure provide fundamental
information for estimating the concentration of the carcinogen to which the organism is exposed.
These data are generated from monitoring information, modeling results, and or reasoned estimates.
An appropriate treatment of exposure should consider the potential for exposure via ingestion,
inhalation, and dermal penetration from relevant sources of exposures, including multiple avenues of
intake from the same source.
       Special problems arise when the human exposure situation of concern suggests exposure
regimens, e.g., route and dosing schedule that are substantially different from those used in the
relevant animal studies.  The cumulative dose received over a lifetime, expressed  as average daily
exposure prorated over a lifetime, is an appropriate measure of exposure to a carcinogen particularly
for an agent that acts by damaging DNA.  The assumption is made that a high dose of a carcinogen
received  over a short period of time is equivalent to a corresponding low dose spread over a lifetime.
This approach becomes more problematic as the exposures in question become  more intense but less
frequent, especially when there is evidence that the agent acts by a mechanism  involving dose-rate

                        5.  CHARACTERIZATION OF HUMAN RISK

       The risk characterization is prepared for the purpose of communicating results of the risk
assessment to the risk manager.  Its objective is to be an appraisal of the science  that the risk manager
can use,  along with other decisionmaking resources, to make public health decisions.  A complete
characterization presents the  risk assessment as an integrated picture of the analysis of the hazard, dose
response, and exposure. It is the risk analyst's obligation to communicate not only summaries of the
evidence and results, but also perspectives on the quality of available data and the degree of

                             DRAFT-DO NOT QUOTE OR CITE

confidence to be placed in the risk estimates. These perspectives include explaining the constraints of
available data and the state of knowledge about the phenomena studied.

       A risk characterization is a necessary part of any Agency report on risk, whether the report is a
preliminary one prepared to support allocation of resources toward further study or a comprehensive
one prepared to support regulatory decisions. Even if only parts of a risk assessment (hazard and
dose-response analyses for instance) are covered in a document, the risk characterization will carry the
characterization to the limits of the document's coverage.

       Each of the following subjects should be covered in the risk characterization.
5.3.1.  Presentation and Descriptors
       The presentation of the results of the assessment should fulfill the aims as outlined in the
purpose section above.  The summary draws from the key points of the individual characterizations of
hazard, dose response, and exposure analysis performed separately under these guidelines. The
summary integrates these characterizations into an overall risk characterization (AIHC, 1989).
       The presentation of results clearly explains the descriptors of risk selected to portray the
numerical estimates. For example, when estimates of individual risk are used or population risk
(incidence) is estimated, there are several features of such estimates that risk managers need  to
understand.  They include, for instance, whether the numbers represent average exposure circumstances
or maximum potential exposure. The size of the population considered to be at risk and the
distribution  of individuals' risks within the population should be given.  When risks to a sensitive
subpopulation have been identified and characterized, the explanation covers the special
characterization of this population.

5.3.2.  Strengths and Weaknesses
        The risk characterization summarizes the kinds of data brought  together in the Analysis and the
reasoning upon which the assessment rests.  The description conveys the major strengths and
weaknesses of the assessment that arise from availability of data and the current limits of
understanding of the process of cancer causation. Health risk is a function of the three elements of

                               DRAFT-DO NOT QUOTE OR CITE

hazard, dose response, and exposure.  Confidence in the results of a risk assessment is, thus, a function
of confidence in the results of the analyses of each element.  The important issues and interpretations
of data are explained, and the risk manager is giveif a clear picture of consensus or lack of consensus
that exists about significant aspects of the assessment.  Whenever more than one view of the weight of
evidence or dose-response characterization is supported by the data and the policies of these
guidelines, and when choosing between them is difficult, the views are presented together.  If one has
been selected over another, the rationale is given; if not, both are presented as plausible alternative
results. If a quantitative uncertainty analysis of data is appropriate, it is presented in the risk
characterization; in any case, qualitative discussion of important uncertainties is appropriate.

                             DRAFT-DO NOT QUOTE OR CITE
                                      6.  REFERENCES
Allen, B. C.; Crump, K. S.; Shipp, A. M. (1988>Correlation between carcinogenic potency of
chemicals in animals and humans. Risk Anal. 8: 531-544.

American Industrial Health Council, (1989)  Presentation of risk assessment of carcinogens. Report of
an Washington, D.C. ad hoc study group on risk assessment presentation.

Ames, B. N.; Gold, L. S. (1990) Too many rodent carcinogens: mitogenesis increases mutagenesis.
Science 249: 970-971.

Ashby, J.; DeSerres, F. J.;  Shelby, M. D.; Margolin, B. H.; Isihidate, M. et al., eds. (1988) Evaluation
of short-term tests for carcinogens: report of the International Programme on Chemical Safety's
collaborative study on in vivo assays.  Cambridge, United Kingdom: Cambridge University, v. 1,2;
431 pp., 372 pp.

Ashby, J.; Tennant, R. W.  (1991) Definitive relationship among chemical structure, earcinogenicity
and mutagenicity for 301 chemicals tested by the U.S. NTP. Mutat. Res. 257: 229-306.

Auger, K. R.; Carpenter, C. L.; Cantley, L. C.; Varticovski, L. (1989a) Phosphatidylinositol 3-kinase
and its novel product, phosphatidylinositol 3-phosphate, are present in Saccharomyces cerevisiae. J.
Biol. Chem. 264: 20181-20184.

Auger, K. R.; Sarunian, L. A.; Soltoff, S; P.; Libby, P.; Cantley, L. C. (1989b) PDGF-dependent
tyrosine phosphorylation stimulates production of novel polyphosphoinositides in intact cells. Cell 57:

Aust, A. E. (1991) Mutations and Cancer.  In: Li, A. P. and Heflich, R. H., eds., Genetic Toxicology.
Boca Raton, Ann Arbor, and Boston: CRC Press, p. 93-117.

Bailey et al. (1984)

Barrett, J. C; Oshimura, C. M.; Tanka, N.; and Tsutsui, T. (1985) Role of aneuploidy early and late
stages of neoplastic progression of Syrian hamster embryo cell in culture. In: Dellarco, V. L.; Voytek,
P. E. and Hollaender, eds., Aneuploidy: etiology and mechanisms, New York: Plenum, in press.

Bimer et al., (1990) Biomonitoring of aromatic amines. Ill:  Hemoglobin binding and benzidine and
some benzidine congeners.  Arch. Toxicol. 64(2) :97-102

Bishop, J. M. (1991) Molecular themes in oncogenesis. Cell 64:235-248.

Bishop, J. M. (1989) Oncogens and clinical cancer. In: Weinberg, R.A. (ed). Oncogens and the
molecular origins of cancer. Cold Spring Laboratory Press, pp.327-358.

                              DRAFT-DO NOT QUOTE OR CITE
Blair, A.; Stewart, P.; O'Berg, M.; Gaffey, W.; Walrath, J.; Ward, J.; Bales, R.; Baplan, S.; Cubit, D.
(1986) Mortality among industrial workers exposed to formaldehyde. J. NatLCancer Inst. 76: 1071-

Blair A.; Stewart, P. A.; Hoover, R. N.; Fraumeni, J. F.; Walrath, J.; O'Berg, M.; Gaffey, W. (1987)
Cancers of the nasopharynx and oropharynx and formaldehyde exposure. (Letter to the Editor.) J.
NatLCancer InsL 78: 191-192.

Blair, A.; Stewart, P. A. (1990) Correlation between different measures and occupational exposure to
formaldehyde. Am. J. Epidemiol. 131:570-516.

Blancato, J. (   )

Bottaro, D. P.; Rubin, J. S.; Falettb, D. L.; Chan, A. M. L.; Kmieck, T. E.; Vande Woude, G. F.;
Aaronson, S. A.  (1991) Identification of the hepatocyte growth factor receptor as the c-met proto-
oncogene product.

Bouk, N. (1990) Tumor angiogenesis: the role of oncogenes and tumor suppressor genes. Cancer Cells
2: 179-183.

Breslow, N. E.; and Day, N.E. (1980) Statistical methods in cancer research. Vol.1 - The analysis of
case-control studies.  IARC Scientific Publication No. 32. Lyon, France: International Agency for
Research on Cancer.

Bressac, B.; Kew, M.; Wands, J., Oztuk, M. (1991) Selective G to T mutations of p53 in
hepatocellular carcinoma from southern Africa. Nature 429-430.

Brockman,  H. E.; DeMarini, D. M. (1988) Utility of short-term test for genetic toxicity in the
aftermathof the NTP's analysis of 73 chemicals. Environ. Molecular Mutagen. 11: 121-435.

Brusick, _. (1990)

Callemen, C.J.; Ehrenberg, L.; Jansson, B.; Osterman-Golkar, S.; Segerback, D.; Svensson, K;
Wachtmeister, C.A.  (1978) Monitoring and risk assessment by means of alkyl groups in  hemoglobin in
persons occupationally exposed to ethylene oxide.  J. Environ. Pathol. Toxicol. 2:427-442.

Callemen et al. (1986)                               .                                      \

Canfley, L. C;; Auger, K. R.; Carpenter, C; Duckworth, B.; Graziani, A.; Kapeller, R.;  Soltoff, S.
(1991) Oncogenes and signal transduction. Cell 64: 281-302.

Castagna, M.; Takai, Y.; Kaibuchi, K.; Sano, K.; Kikkawa, U.; Nishizuka, Y. (1982)  Direct activation
of calcium-activated, phospholipid-dependent protein kinase by tumor-promoting phorbol esters. J.
Biol. Chem. 257: 7847-7851.

                              DRAFT-DO NOT QUOTE OR CITE
Cavcnee, W. K.; Dryja, T. P.; Phillips, R. A.; Benedict, W. F.; Godbout, R.; Gallic, B. L.; Murphree,
A. L.; Strong, L.C.; White, R. L. (1983) Expression of recessive alleles by chromosome mechanisms
in reteiroblastoma. Nature 303: 779-784.

Cerutti, P.; Larsson, R.; and Krupitza, G. (1990) Mechanisms of carcinogens and tumor progression.
Harris, C.C.; Liiotta, L. A., eds. New York: Wiley-Liss. pp. 69-82.

Chen, C.; Farland, W. (1991) Incorporating cell proliferation in quantitative cancer risk assessment:
approaches, issues, and uncertainties. In: Butterworth, B.; Slaga, T.; Farland, W.; McClain, M. (eds.)
Chemical induced cell proliferation:  Implication for risk assessment. Wiley-Liss.

Collum, R. G.; Alt, F. W. (1990) Are myc proteins transcription factors? Cancer Cells 2: 69-73.

Connolly, R. B.; Andersen, M. E. (1991) Biological based pharmacodynamic models: tools for
lexicological research and risk assessment. Ann. Rev. Pharmacol. Toxicol. 31: 503-523.

Coouch and Harsbarger  (1985)

Cross, M.; Dexter, T. (1991) Growth factors in development, transformation, and tumorigenesis.  Cell

Crump, K. S.; Hoel, D. G.; Langley, C. H.; Peto, R. (1976) Fundamental carcinogenic processes and
their implications for low dose risk assessment. Cancer Res. 36: 2973-2979.

deThe, H.; Lavau, C.; Marchio, A.; Chomienne, C; Degos, L.; Dejean, A. (1991)  The PML-RAR°=
fusion mRNA generated by the t (15;17) translocation in acute promyelocytic leukemia encodes  a
functionally altered RAR.  Cell 66: 675-684.

Drinkwater, N. R. (1990) Experimental models and biological mechanisms for tumor promotion.
Cancer Cells,,  2(1):8.

D'Souza, R. W.; Francis, W. R.;  Bruce, R. D.; Andersen, M. E. (1987)  Physiologically based
pharmacokinetic model for ethylene chloride and its application in risk assessment.  In:
Pharmacokinetics in risk assessment. Drinking Water and Health. Vol 8.  Washington, DC: National
Academy Press.

Fearon, E.; Vogelstein, B. (1990) A genetic model for colorectal tumorigenesis. Cell 61:959-767.

Fidler, I. J.; Radinsky, R. (1990). Genetic control of cancer metastasis. J. Nati. Cancer Inst. 82: 166-168.

Flamm, W. G.; Winbush, J. S.  (1984) Role of mathematical models in assessment of risk and in attempts
to define management strategy. Fundam. Appl. Toxicol. 4: S395-S401.

Flynn, G. L. (1990) Physicochemical determinants of skin absorption.  In:  Gerrit, T. R.; Henry, C.  J.
(eds.) Principles  of route to route extrapolation for risk assessment. New York, NY: Elsevier Science
Publishing Co. pp. 93-127.

                              DRAFT-DO NOT QUOTE OR CITE
 Forsburg, S. L.; Nurse, P. (1991) Identification of a Gl-type cyclin pugl+ in the fission yeast
 Schizosaccharomyces pombe.  Nature 351: 245-248.

 Garfinkel, L; Silverberg, E. (.1991) Lung cancer and smoking trends in the United States over the past 25
 years. Cancer 41: 137-145.

 Garrett, N. E.; Stack, H, F.; Gross, M. R.; Waters, M. D. (1984) An analysis of the spectra of genetic
 activity produced by known or suspected human carcinogens.  Mutat. Res. 134: 89-111.

 Gaylor, D. W.; Kodell, R. L,  (1980) Linear interpolation algorithm for low-dose risk assessment of toxic
 substances.  J. Environ.  Pathol. Toxicol. 4: 305-312.

 Gaylor, D. W. (1988) Quantitative risk estimation. Adv. Mod. Environ. Toxicol. 15: 23-43.

 Gerrity, T. R.; Henry, C., eds. (1990) Principles of route to route extrapolation for risk assessment.  New
 York, NY: Elsevier Science Publishing Co.

 Gillette, J. R. (1983) The use of pharmacokinetics in safety testing.  Safety evaluation and regulation of
 chemicals 2.  2nd Int. Conf., Cambridge, MA: pp. 125-133.

 Greenland, S. (1987) Quantitative methods in the review of epidemiologic literature.  Epidemiol. Rev.

 Hammand, E. C. (1966) Smoking in relation to the death rates of one million men and women.  In:
 Haenxzel, W., ed. Epidemiological approaches to the study of cancer and other chronic  diseases.  National
 Cancer Institute Monograph No.  19, Washington, DC.

 Harris, C. C. (1989) Intel-individual variation among humans in carcinogen metabolism,  DNA adduct
 formation and DNA repair. Carcinogenesis 10:  1563-1566.

 Harris, H. (1990) The role of differentiation in the suppression of malignancy. J. Cell Sci. 97: 5-10.

 Haseman, J. K. (1990) Use of statistical decision rules for evaluating laboratory animal carcinogenicity
 studies. Fundam. Appl. Toxicol. 14: 637-648.

 Haseman, J. K. (1985) Issues  in carcinogenicity testing: Dose selection. Fundam. Appl. Toxicol. 5: 66-78.

 Haseman, J. K. (1984) Statistical issues in the design, analysis and interpretation of animal
 carcinogenicity  studies. Environ. Health Perspect 58: 385-392.

Haseman, J. K.;  Huff, J.; Doorman, G.A. (1984) Use of historical control data in carcinogenicity studies
in rodents. Toxicol. Pathol. 12: 126-135.

Haseman, J. K.  (1983) Issues: A reexamination of false-positive rates for Carcinogenesis studies. Fundam
Appl. Toxicol. 3: 334-339.

                             DRAFT-DO NOT QUOTE OR CITE

Hattis, D. (1990)  Phannaeokinetic principles for dose-rate extrapolation of carcinogenic risk from
genetically active agents.  Risk Anal. 10: 303-316.

Hayward, N. K.; Walker, G. J.; Graham, W.; Cooksley, E. (1991) Hepatocellular carcinoma mutation.
Nature 352: 764.

Herschman, H. R. (1991) Primary response genes induced by growth factors or promoters. Ann. Rev.
Biochem. 60: 281-319.

HSldebrand, B.;  Grasso, P.; Ashby, J.; Chamberlain, M.; Jung, R.; van Kolfschoten, A.; Loeser, E.; Smith,
E.; Bontinck, W. J.  (1991) Validity of considering that early changes may act as indicators for non-
genotoxic carcinogenesis. MutaL Res. 248: 217-220.

Hoel, D. G. (1980) Incorporation of background in dose-response models. Fed. Proc., Fed. Am. Soc. Exp.
Biol. 39: 73-75.

Hoffmann, G. R. (1991) Genetic toxicology. Casarett and Doull's Toxicology: The Basic Science of
Poison. Pergamon Press. Fourth Edition, pp. 201-225

Holliday, R. (1991) Mutations and epimutations in mammalian cells.  Mutat. Res. 250: 351-363.

Hollstein, M.; Sidransky, D.; Vogelstein, B.; Harris,  C. C. (1991) p53 mutations in human cancers.
Science 253: 49-53.

Hsu, I. C.; Metcaff, R. A.; Sun, T.; Welsh, J. A.; Wang, N. J.; Harris, C. C. (1991) Mutational hotspot in
human hepatocellular carcinomas.  Nature  350: 427-428.

Huff, J. E.; McConnell, E. E.; Haseman, J. K.; Boorman, G. A.; Eustis, S. L. et al. (1988) Carcinogenesis
studies: results from 398 experiments on 104 chemicals from the U.S. National Toxicology Program. Ann.
N.Y. Acad. Sci. 534: 1-30.

Hulka, B. S.; Margolin, B. H. (1992) Methodological issues in epidemiologic studies using biological
markers. Am. J.Epidemiol. 135: 122-129.

Hunter, T. (1991) Cooperation between oncogenes. Cell 64: 249-270.

Ingram, A. J.; Grasso, P. (1991) Evidence  for and possible mechanism of non-genotoxic carcinogenesis in
mouse skin. Mutat. Res. 248: 333-340.

Interagency Regulatory Liaison Group  (IRLG). (1979)  Scientific basis  for identification of potential
carcinogens and estimation of risks. J. Nati. Cancer Inst. 63: 245-267.

International Agency for Research on Cancer. (1982) IARC Monographs on the evaluation of the
carcinogenic risk of chemicals to humans,  Suppl. 4.  Lyon, France: IARC.

                              DRAFT-DO NOT QUOTE OR CITE
 International Agency for Research on Cancer (1991) Mechanisms of carcinogenesis in risk identification:
 A consensus report of an IARC Monographs working group.  IARC Internal Technical Report No.
 91/002. Lyon, France: IARC.

 International Life Sciences Institute. (1984) The selection of doses in chronic toxicity/carcinogenic
 studies. In: Grice, H. C, ed. Current issues in toxicology. New York: Springer-Verlag, pp. 6-49.

 Ito, N.; Imaida, K.; Hasegawa, R.; Tsuda, H. (1989) Rapid bioassay methods for carcinogens and
 modifiers of hepatocarcinogenesis; CRC Critical Review in Toxicology. 19(4): 385-415.

 Jackson, M. A.; Stack, H. F.; Waters, M.D.(1992) The genetoxic toxicology of putative nongenotoxic
 carcinogens; Mutat. Res. (in press).                                             .

.Kaplan, D. R.; Hempstead, B. L.; Martin-Zanca, D.; Chao, M. V.; Parada, L. F. (1991) The trkproto-
 oncogene product:a signal transducing receptor for nerve growth factor. Science 252: 554-558.

 Kakizuka, A.; Miller, W. H.; Umesono, K.; Warrell, R. P.; Frankel, S. R.; Marty, V.V.V.S.; Dimitrovsky,
 E.; Evans, R.M. (1991).  Chromosomal translocation t(15;17) in human acute promyelocyte leukemia
 fuses RARoc with a novel putative transcription factor, pml.  Cell 66: 663-674.

 Kelsey, J.  L.; Thompson, W. D.; Evans, A. S. (1986) Methods in observational epidemiology.  New
 York: Oxford University Press.

 Kodell, R. L.; Park, C. N.  Linear extrapolation in cancer risk assessment.  ILSI Risk Science Institute (in
 press).                                   ,.-/'

 Kondo, K. R.; Chilcote, R ; Maurer, H. S.; Rowley, J. D. (1984) Chromosome abnormalities in tumor
 cells from patients with sporadic Wilms'tumor. Cancer Res. 44: 5376-5381

 Krewski, D.; Murdoch, D.J.; Withey, J. R. (1987). The application of pharmacokinetic data in
 carcinogenic risk assessment In: Pharmacokinetics in risk assessment.  Drinking Water and Health.
 Volume 8. Washington, DC: National Academy Press, pp. 441-468

 Krewski, D.; Brown, C.; Murdoch, D. (1984) Determining "safe" levels of exposure: Safety factors of
 mathematical models. Fundam. Appl. Toxicol. 4: S383-S394.

 Kumar, R.; Sukumar, S.; Barbacid, M. (1990) Activation of ras oncogenes preceding the onset of
 neoplasia.  Science 248:  1101-1104.

 Levine, A. J.; Momand, J.; Finlay, C. A. (1991) The p53 tumor suppressor gene. Nature 351: 453-456,

 Lewin, B.  (1991) Oncogenic conversion by regulatory changes in transcription factors. Cell 64: 303-312.

 Lilienfeld, A. M.; Lilienfeld, D. (1979) Foundations of epidemiology. 2nd  ed. New York: Oxford
 University Press.

                             DRAFT-DO NOT QUOTE OR CITE
Lutz, W. K. (1990) Dose-response relationship and low dose extrapolation in chemical carcinogenesis.
Carcinogenesis 11: 1243-1247.

Malkin, D.; Li, F. P.; Strong, L. C.; Fraumeni, J. F., Jr.; Nelson, C. E.; Kim, D. H.; Kassel, J.; Gryka, M.
A.; Bischoff, F. Z.; Tainsky, M. A.; Friend, S. H. (1990) Germ line p53 mutations in a familial syndrome
of breast cancer, sarcomas, and other neoplasms. Science 250: 1233-1238.

Mailer, J. L. (1991) Mitotic control.  Curr. Opin. Cell Biol. 3:269-275.

Mann, C. (1990) Meta-analysis in a breech. Science 249: 476-480.

Marsh, G. M.; Stone, R. A.; Henderson, V. L. (1992) A reanalysis of the National Cancer Institute study
on lung cancer mortality among industrial workers exposed to formaldehyde.  J. Occup. Med. 34: 42-44.

Mantel, N.; Haenszel, W. (1959) Statistical aspects of the analysis of data from retrospective studies of
disease. J. Natl. Cancer Inst. 22: 719-748.

Mantel, N. (1980) Assessing laboratory evidence for neoplastic activity. Biometrics 36: 381-399.

Mausner, J. S.;  Kramer, S. (1985) Epidemiology, 2nd ed.  Philadelphia: W. B. Saunders Company.

McConnell, E. E.; SoUeveld, H. A.; Swenberg, J. A.; Boorman, G. A. (1986) Guidelines for combining
neoplasms for evaluation of rodent carcinogenesis studies. J. Natl. Cancer Inst. 76: 283-289.

McCormick, A.; Campisi, J. (1991)  Cellular aging and senescence.  Curr. Opin. Cell Biol. 3: 230-234.

McMichael, A. J. (1976)  Standardized mortality ratios and the "healthy worker effect": Scratching
beneath the surface.  J. Occup. Med. 18: 165-168.

Moolgavkar, S. H.; Knudson, A. G. (1981) Mutation and  cancer: a model for human carcinogenesis.  J.
Natl. Cancer Inst. 66: 1037-1052.

Monro, A. (1992) What is an appropriate measure of exposure when testing drugs for carcinogenicity in
rodents? Toxicol. Appl. Pharmacol. 112:171-181.

Naeve, G. S.; Sharma, A.; Lee, A. S. (1991) Temporal events regulating the early phases of the
mammalian cell cycle. Curr.  Opin. Cell Biol. 3: 261-268.

National Academy of Sciences (NAS). (1989)  Biological markers in pulmonary toxicology.  Washington,
DC: National Academy Press.

National Research Council. (1983) Risk assessment in  the Federal government: managing the process.
Committee on the Institutional Means for Assessment of Risks to Public Health, Commission on Life
Sciences, NRC. Washington,  DC: National Academy Press.

                             DRAFT-DO NOT QUOTE OR CITE
National Toxicology Program. (1984) Report of the Ad Hoc Panel on Chemical Carcinogenesis Testing
and Evaluation of the National Toxicology Program, Board of Scientific Counselors.  Washington, DC:
U.S. Government Printing Office. 1984-421-132: 4726.

Nebert, D. W. (1991) Polymorphism of human CYP2D genes involved in drug metabolism; possible
relationship to individual cancer risk. Cancer Cells 3: 93-96.

Nebreda, A. R.; Martin-Zanca, D.; Kaplan, D. R.; Parada, L.  F.; Santos, E. (1991) Induction by NGF of
meiotic maturation of xenopus oocytes expressing the trk proto-oncogene product. Science 252: 558-561.

Nowell, P. (1976) The clonal evolution of tumor cell populations. Science 194: 23-28.

Nunez, G.; Hockenberry, D.; McDonnell, J.; Sorenson, C. M.; Korsmeyer, S. J. (1991). Bcl-2 maintains
B cell  memory.  Nature 353: 71-72.                           ,

Office  of Science and Technology Policy (OSTP). (1985) Chemical carcinogens: review of the science
and its associated principles. Federal Register 50: 10372-10442.

Peto, R. (1978) Carcinogenic effects of chronic exposure to very low levels of toxic substances.  Environ.
Health Perspect. 22: 155-161.            '                                 ,

Peto, R.; Pike, M.; Day, N.; Gray, R.; Lee, P.; Parish, S.; Peto, J.; Richard, S.; Wahrendorf, J. (1980)
Guidelines for simple, sensitive, significant tests  for carcinogenic effects in long-term animal experiments.
In: Monographs on the long-term and short-term screening assays for carcinogens: a critical appraisal.
I ARC Monographs, Suppl. 2. Lyon, France:  mternational Agency for Research on Cancer, pp. 311-426.

Pilot, H.; Dragan, Y. P.  (1991) Facts and theories concerning the mechanisms of carcinogenesis. FASEB

Portier, C. (1987) Statistical properties of a two-stage model of carcinogenesis.  Environ. Health
Perspect. 76: 125-131.

Prival,  M. J.; Dunkel, V.C. (1989) Reevaluation of the mutagenicity and carcinogenicity of chemicals
previously identified as "false positives' in the Salmonella typhimurium mutagenicity assay.  Environ.
Molec. Mutag. 13(1): 1-24.                           '3   '•

Raff, M. C. (1992) Social controls on cell survival and cell death. Nature 356: 397-400.

Rail, D. P. (1991) Carcinogens and human health: part 2. Science 251: 10-11.

Rothman, K. T. (1986)  Modern Epidemiology.  Boston:  Little, Brown and Company.

Salomon, D. S.; Kim, N.; Saeki, T.; Ciardiello, F. (1990) Transforming growth factor a - an
oncodevelopmental growth factor. Cancer Cell 2: 389-397.

Sager,  R. (1989) Tumor suppressor genes: the puzzle and the promise.  Science 246: 1406-1412.

                              DRAFT-DO NOT QUOTE OR CITE
Schneider, C.; Gustincich, S.; DelSal, G. (1991) The complexity of cell proliferation control in
mammalian cells. Curr. Opin. Cell Biol. 3: 276-281.

Schuller, H. M. (1991) Receptor-mediated mitogenic signals and lung cancer. Cancer Cells 3: 496-503.

Sidransky, D.; Mikkelsen, T.; Schwechheimer, K.; Rosenblum, M. L.; Cavanee, W.;  Vogelstein, B. (1992)
Clonal expansion of p53 mutant cells is associated with brain tumor progression.  Nature 355: 846-847.

Sidransky, D.; Von Eschenbach, A.; Tsai, Y.C.; Jones,  P.; Summerhayes, I.; Marshall, F.; Paul, M.;
Green, P.; Hamilton, P.F.; Vogelstein, B. (1991) Identification of p53 gene mutations in bladder cancers
and urine samples.  Science 252: 706-710.

Shelby et al. (1989)

Stiltler et al. (1992)

Srivastava, S.; Zou, Z.; Pirollo, K.; Blattner, W.; Chang, E. (1990) Germ-line transmission of a mutated
p53 gene in a cancer-prone family with Li-Fraumeni syndrome. Nature 348(6303): 747-749.

Starr, T.  B. (1990) Quantitative cancer risk estimation for formaldehyde. Risk Anal. 10: 85-91.

Stanbridge, E. J.; Cavenee, W. K. (1989) Heritable cancer and tumor suppressor genes: A tentative
connection. In: Weinberg, R. A. (ed.) Oncogenes and the molecular origins of cancer, p. 281.

Sterling,  T.D.; Weinkam, J.J. (1987) Reanalysis of lung cancer mortality in a National Cancer Institute
study of  "mortality among industrial workers exposed to formaldehyde": additional discussion. J. Occup.
Med. 31: 881-883.

Stewart,  P. A.; Herrick, R. F. (1991) Issues in performing retrospective exposure assessment. Appl.
Occup. Environ. Hygiene.

Strausfeld, U.; Labbe, J. C.; Fesquet, D.; Cavadore, J. C.  Dicard, A.; Sadhu, K.; Russell, P.;  Dor'ee, M.
(1991) Identification of a Gl-type cyclin pucl+ in the fission yeast Schizosaccharomvces pombe.  Nature

Sukumar, S. (1989) ras oncogenes in chemical carcinogenesis.  Curr. Top. MicrobioL Immunol. 148: 93-

Sukumar, S. (1990) An experimental analysis of cancer:  role of ras oncogenes in multistep
carcinogenesis. Cancer Cells 2: 199-204.

Swenberg, J. A.; Richardson, F. C.; Boucheron, J. A.; Deal, F. H.; Belinsky, S. A.; Charbonneau, M.;
Short, B. G. (1987) High to low dose extrapolation:  critical determinants involved in the dose-response
of carcinogenic substances. Environ. Health Perspect. 76: 57-63.

                             DRAFT-DO NOT QUOTE OR CITE
Tanaka, K.; Oshimura, M.; Kikiuchi, R.; Selci, M.; Hayashi, T; Miyaki, M. (1991) Suppression of
tumorigenicity in human colon carcinoma cells by introduction of normal chromosome 5 or 18. Nature
349: 340-342.

Tennant, R. W.; Zeiger, E. (1992) Genetic toxicology: the current status of methods of carcinogen
identification.  Environ. Health Perspect. (in press)

Thompson, T.  C; Southgate, J.; Kitchener, G.; Land, H. (1989) Multistage carcinogenesis induced by ras
and myc oncogenes in a reconstituted organ.  Cell 56: 917-3183.

Tomatis, L.; Aitio, A.; Wilboum, J.; Shuker, L. (1989) Jpn. J. Cancer Res. 80: 795-807.

U.S. Environmental Protection Agency. (1983a) Good laboratory practices standards—toxicology testing.
Federal Register 48: 53922.

U.S. Environmental Protection Agency. (1983b) Hazard evaluations: humans and domestic animals.
Subdivision F.Springfield, VA: NTIS. PB 83-153916.

U.S. Environmental Protection Agency. (1983c) Health effects test guidelines. Springfield, VA: NTIS.
PB 83-232984.

U.S. Environmental Protection Agency. (1984) Estimation of the public health risk from exposure to
gasoline vapor via the gasoline marketing system. Washington, DC: Office of Health and Environmental
Assessment       .

U.S. Environmental Protection Agency. (1986) Health assessment document for beryllium. Washington,
DC:  Office of Health and Environmental Assessment. EPA/600/

U.S. Environmental Protection Agency. (1986a) The risk assessment guidelines of 1986. Washington,
DC:  Office of Health and Environmental Assessment. EPA/600/8-87/045.

U.S. Environmental Protection Agency. (1989a) Interim procedures for estimating risks associated with
exposures to mixtures of chlorinated dibenzo-p-dioxins and -dibenzofurans (CDDs and CDFs) and 1989
update.  Washington, DC: Risk Assessment Forum.  EPA/625/3-89/016.

U.S. Environmental Protection Agency. (1989b) Workshop on EPA guidelines for carcinogen risk
assessment. Washington, DC: Risk Assessment Forum. EPA/625/3-89/015.

U.S.  Environmental Protection Agency. (1989c) Workshop on EPA guidelines for carcinogen risk
assessment: use of human evidence. Washington, DC: Risk Assessment Forum.  EPA/625/3-90/017.

U.S.  Environmental Protection Agency. (1991a) Pesticide Assessment Guidelines: Subdivision F,  hazard
evaluation: human and domestic animals. Series 84, Mutagenicity. PB91-158394, 540/09-91-122.  Office
of Pesticide Programs.

                             DRAFT-DO NOT QUOTE OR CITE

U.S. Environmental Protection Agency. (1991b) Alpha-2u-globulin: association with chemically induced
renzil toxicity and neoplasia in the male rat.  Washington, DC: Risk Assessment Forum. EPA/625/3-

U.S. Environmental Protection Agency. (l?91c) Workshop report on toxicity equivalency factors for
polychlorinated biphenyl congeners.  Washington, DC: Risk Assessment Forum. EPA/625/3-91/020.

U.S. Environmental Protection Agency. (19911) Guidelines for developmental toxicity risk assessment.
Federal Register 56(234): 63798-63826.

U.S. Environmental Protection Agency. (1992a) Guidelines for exposure assessment. Washington, DC:
Federal Register 57(104): 22888-22938

U.S. Environmental Protection Agency. (1992b) Draft Report: A cross-species scaling factor for
carcinogen risk assessment based on equivalence of mg/kg3/4/day. Washington, DC: Federal Register
57(109): 24152-24173.                                              :     .

U.S. Environmental Protection Agency. (1992c, August) Health assessment for 2,3,7,8-tetrachlorodibenzo-
p-dioxin (TCDD) and related  compounds (Chapters 1 through 8). Workshop Review Drafts.
EPA/600/AP-92/001a through OOlh.

U.S. Food and Drag Administration. (December 31, 1987) Federal Register 52: 49577  et. seq.

Vahakangas et al. (1992)

Van Sittert. N.J.; De Jong, G.; Clare, M.G.; Davies, R.; Dean, BJ. Wren, L.R. Wright, A.S. (1985)
Cytogenetic, immunological, and hematological effects in workers in an ethylene oxide manufacturing
plant.  Br. J. Indust. Med. 42:19-26.

Varmus, H. (1989) An historical overview of oncogenes. Cold Spring Harbor Laboratory Press, pp. 3-44.

Waters et al. (1988)

Weinberg, R. A. (1989) Oncogenes,  antioncogenes, and the molecular bases of multistep carcinogenesis.
Cancer Res. 49: 3713-3721.

Wellstein, A.; Lupu,  R.; Zugmaier, G.; Hamm, S. L.; Cheville, A.L.; Bovi, P. D.; Basicico, C.; Lippman,
M. E.; Kern, F.G. (1990) Autocrine growth stimulation by secreted Kaposi fibroblast growth factor but
not by endogenous basic fibroblast growth factor. Cell Growth Differ. 1: 63-71.

Woo, Y. T. and Arcos, J. C. (1989) Role of structure-activity relationship analysis in evaluation of
pesticides for potential carcinogenicity. ACS Symposium Series No. 414. Carcinogenicity and pesticides:
principles, issues, and relationship. Ragsdale, N. N.; Menzer, R. E. eds. pp.  175-200.

Wyzga, R. E. (1988) The role of epidemiology in risk assessments of carcinogens. Adv. Mod. Environ.
Toxicol. 15: 189-208.

                              DRAFT-DO NOT QUOTE OR CITE
Yamasaki, H. (1990) Gap junctional intercellular communication and carcinogenesis. Carcinogenesis

Zeise, L.; Wilson, R.; Crouch, E. A. C. (1987) Dose-response relationships for carcinogens: a review.
Environ. Health Perspect. 73: 259-308.

Zhang, K.; Papageorge, A.G.; Lowry, D.R. (1992) Mechanistic aspects of signalling through ras in NIH
3T3 cells. Science 257: 671-674.
                                                          U.S. -GOVERNMENT PRINTING OFFICE:  -1992 -750-068/60001